patent_number,patent_title,patent_abstract,patent_date,text
4094307,method and apparatus for aiding in the anatomical localization of dysfunction in a brain,a method and apparatus for synthesizing a set of optimal sensory stimuli designed to elicit an optimal response for each particular brain electrode location in a subject whose brain is being examined to anatomically localize brain dysfunction. a pseudorandom input signal having the general characteristics of gaussian white noise is generated and converted into a color video visual stimulus which can be observed by the subject and summed on his retina and associated neural network. a plurality of electrodes are positioned with respect to various different and distinct areas of the brain of the subject to be examined. the subject is shown the color video visual stimulus and the electrical analog response from the electrodes is amplified and stored. the stored analog response signals are cross-correlated with the resynthesized input signal to compute a wiener kernel representation of the response for each electrode. portions of the pseudorandom input signal which resulted in insignificant analog responses are masked out so that the subsequent generation of pseudorandom input signals will be bandwidth-limited. the analog responses to the bandwidth-limited visual stimulus are cross-correlated with the resynthesized masked input signal and new wiener kernel representations are recomputed for each electrode. the recomputed wiener kernel representations of the response from each electrode are then multiplied in an array processor with the resynthesized bandwidth-limited input signal to compute an optimum visual stimulus for each of the electrodes. these optimum visual stimuli may be subsequently displayed to the subject alone or in conjunction with psychophysical tests to aid in anatomically localizing dysfunction in a brain under examination.,1978-06-13,The title of the patent is method and apparatus for aiding in the anatomical localization of dysfunction in a brain and its abstract is a method and apparatus for synthesizing a set of optimal sensory stimuli designed to elicit an optimal response for each particular brain electrode location in a subject whose brain is being examined to anatomically localize brain dysfunction. a pseudorandom input signal having the general characteristics of gaussian white noise is generated and converted into a color video visual stimulus which can be observed by the subject and summed on his retina and associated neural network. a plurality of electrodes are positioned with respect to various different and distinct areas of the brain of the subject to be examined. the subject is shown the color video visual stimulus and the electrical analog response from the electrodes is amplified and stored. the stored analog response signals are cross-correlated with the resynthesized input signal to compute a wiener kernel representation of the response for each electrode. portions of the pseudorandom input signal which resulted in insignificant analog responses are masked out so that the subsequent generation of pseudorandom input signals will be bandwidth-limited. the analog responses to the bandwidth-limited visual stimulus are cross-correlated with the resynthesized masked input signal and new wiener kernel representations are recomputed for each electrode. the recomputed wiener kernel representations of the response from each electrode are then multiplied in an array processor with the resynthesized bandwidth-limited input signal to compute an optimum visual stimulus for each of the electrodes. these optimum visual stimuli may be subsequently displayed to the subject alone or in conjunction with psychophysical tests to aid in anatomically localizing dysfunction in a brain under examination. dated 1978-06-13
4536844,method and apparatus for simulating aural response information,"speech and like signals are analyzed based on a model of the function of the human hearing system. the model of the inner ear is expressed as signal processing operations which map acoustic signals into neural representations. specifically, a high order transfer function is modeled as a cascade/parallel filterbank network of simple linear, time-invariant second-order filter sections. signal transduction and compression are based on a half-wave rectification with a non-linearly coupled, variable time constant automatic gain control network. the result is a simple device which simulates the complex signal transfer function associated with the human ear. the invention lends itself to implementation in digital circuitry for real-time or near real-time processing of speech and other sounds.",1985-08-20,"The title of the patent is method and apparatus for simulating aural response information and its abstract is speech and like signals are analyzed based on a model of the function of the human hearing system. the model of the inner ear is expressed as signal processing operations which map acoustic signals into neural representations. specifically, a high order transfer function is modeled as a cascade/parallel filterbank network of simple linear, time-invariant second-order filter sections. signal transduction and compression are based on a half-wave rectification with a non-linearly coupled, variable time constant automatic gain control network. the result is a simple device which simulates the complex signal transfer function associated with the human ear. the invention lends itself to implementation in digital circuitry for real-time or near real-time processing of speech and other sounds. dated 1985-08-20"
4592359,multi-channel implantable neural stimulator,"a combination of a transmitter and implantable receiver are disclosed wherein data is conveyed from transmitter to receiver utilizing a data format in which each channel to be stimulated is adapted to convey information in monopolar, bipolar or analog form. the data format includes two types of code words: transition words in which one bit is assigned to each channel and can be used to create monopolar pulsatile or bipolar pulsatile waveforms; and amplitude words which can create analog waveforms one channel at a time. an essential element of the output system is a current source digital to analog converter which responds to the code words to form the appropriate output on each channel. each output is composed of a set of eight current sources, four with one polarity of current and the other four with the opposite polarity of current. in each group of four, the current sources are binarily related, i, 2i, 4i and 8i. in this arrangement each channel can supply 16 amplitudes times two polarities of current; meaning 32 current levels. this channel is simply a 5-bit digit to analog converter. the output circuitry contains charge balance switches. these switches are designed to recover residual charge when the current sources are off. they are also designed to current limit during charge recovery if the excess charge is too great so that they do not cause neural damage. each channel charge balances (will not pass dc current or charge) and charge limits to prevent electrode damage and bone growth. the charge balancing is performed by the charge balancing switches and by the blocking capacitor. the charge limiting is performed by the blocking capacitor only. the charge level on each channel is defined using a switch network ladder which combines a plurality of parallel connected switches; closure of each switch doubles the current level handed off from the previous switch.",1986-06-03,"The title of the patent is multi-channel implantable neural stimulator and its abstract is a combination of a transmitter and implantable receiver are disclosed wherein data is conveyed from transmitter to receiver utilizing a data format in which each channel to be stimulated is adapted to convey information in monopolar, bipolar or analog form. the data format includes two types of code words: transition words in which one bit is assigned to each channel and can be used to create monopolar pulsatile or bipolar pulsatile waveforms; and amplitude words which can create analog waveforms one channel at a time. an essential element of the output system is a current source digital to analog converter which responds to the code words to form the appropriate output on each channel. each output is composed of a set of eight current sources, four with one polarity of current and the other four with the opposite polarity of current. in each group of four, the current sources are binarily related, i, 2i, 4i and 8i. in this arrangement each channel can supply 16 amplitudes times two polarities of current; meaning 32 current levels. this channel is simply a 5-bit digit to analog converter. the output circuitry contains charge balance switches. these switches are designed to recover residual charge when the current sources are off. they are also designed to current limit during charge recovery if the excess charge is too great so that they do not cause neural damage. each channel charge balances (will not pass dc current or charge) and charge limits to prevent electrode damage and bone growth. the charge balancing is performed by the charge balancing switches and by the blocking capacitor. the charge limiting is performed by the blocking capacitor only. the charge level on each channel is defined using a switch network ladder which combines a plurality of parallel connected switches; closure of each switch doubles the current level handed off from the previous switch. dated 1986-06-03"
4699875,diagnosis of amyotrophic lateral sclerosis by neurotrophic factors,"the present invention is based on the discovery that amyotrophic lateral sclerosis (als), parkinson disease and alzheimer disease are due to lack of a disorder-specific neurotrophic hormone. diagnosis is accomplished by assaying hormones specific for a particular neuronal network or system: the motor neurotrophic hormones from muscle in the motor neural system are used to diagnose and treat als, dopamine neurotrophic hormones from striatum in the migrostriatal neural system are used to diagnose and treat parkinsonism, and cholinergic neurotrophic hormones released from the cortex and hippocampus which are specific for cholinergic neorons of the nucleus basalis and septal nucleus are used to diagnose and treat alzheimer's disease. with tissue culture, the presence or absence of specific neurotrophic hormones can be assessed in als, parkinsonism, and alzheimer disease. if there is a deficiency, extracted and purified neurotrophic hormones specific to the particular neuronal network or system can be injected in als and alzheimer disease and in parkinsonism.",1987-10-13,"The title of the patent is diagnosis of amyotrophic lateral sclerosis by neurotrophic factors and its abstract is the present invention is based on the discovery that amyotrophic lateral sclerosis (als), parkinson disease and alzheimer disease are due to lack of a disorder-specific neurotrophic hormone. diagnosis is accomplished by assaying hormones specific for a particular neuronal network or system: the motor neurotrophic hormones from muscle in the motor neural system are used to diagnose and treat als, dopamine neurotrophic hormones from striatum in the migrostriatal neural system are used to diagnose and treat parkinsonism, and cholinergic neurotrophic hormones released from the cortex and hippocampus which are specific for cholinergic neorons of the nucleus basalis and septal nucleus are used to diagnose and treat alzheimer's disease. with tissue culture, the presence or absence of specific neurotrophic hormones can be assessed in als, parkinsonism, and alzheimer disease. if there is a deficiency, extracted and purified neurotrophic hormones specific to the particular neuronal network or system can be injected in als and alzheimer disease and in parkinsonism. dated 1987-10-13"
4701407,diagnosis of alzheimer disease,"the present invention is based on the discovery that amyotrophic lateral sclerosis (als), parkinson disease and alzheimer disease are due to lack of a disorder-specific neurotrophic hormone. diagnosis is accomplished by assaying hormones specific for a particular neuronal network or system: the motor neurotrophic hormones from muscle in the motor neural system are used to diagnose and treat als, dopamine neurotrophic hormones from striatum in the nigrostriatal neural system are used to diagnose and treat parkinsonism, and cholinergic neurotrophic hormones released from the cortex and hippocampus which are specific for cholinergic neurons of the nucleus basalis and septal nucleus are used to diagnose and treat alzheimer's disease. with tissue culture, the presence or absence of specific neurotrophic hormones can be assessed in als, parkinsonism, and alzheimer disease. if there is a deficiency, extracted and purified neurotrophic hormones specific to the particular neuronal network or system can be injected in als and alzheimer disease and in parkinsonism.",1987-10-20,"The title of the patent is diagnosis of alzheimer disease and its abstract is the present invention is based on the discovery that amyotrophic lateral sclerosis (als), parkinson disease and alzheimer disease are due to lack of a disorder-specific neurotrophic hormone. diagnosis is accomplished by assaying hormones specific for a particular neuronal network or system: the motor neurotrophic hormones from muscle in the motor neural system are used to diagnose and treat als, dopamine neurotrophic hormones from striatum in the nigrostriatal neural system are used to diagnose and treat parkinsonism, and cholinergic neurotrophic hormones released from the cortex and hippocampus which are specific for cholinergic neurons of the nucleus basalis and septal nucleus are used to diagnose and treat alzheimer's disease. with tissue culture, the presence or absence of specific neurotrophic hormones can be assessed in als, parkinsonism, and alzheimer disease. if there is a deficiency, extracted and purified neurotrophic hormones specific to the particular neuronal network or system can be injected in als and alzheimer disease and in parkinsonism. dated 1987-10-20"
4737929,highly parallel computation network employing a binary-valued t matrix and single output amplifiers,"advantageous neural network realizations are achieved by employing only negative gain amplifiers and a clipped t matrix having conductances t.sub.ij which have only two values. preferably, one of these values is a preselected value set by the value of a fixed resistor, and the other value is zero, created simply with an open circuit. values for the t.sub.ij terms of the clipped t matrix are obtained through an iterative process which operates on the clipped and nonclipped matrices and minimizes the error resulting from the use of the clipped t matrix.",1988-04-12,"The title of the patent is highly parallel computation network employing a binary-valued t matrix and single output amplifiers and its abstract is advantageous neural network realizations are achieved by employing only negative gain amplifiers and a clipped t matrix having conductances t.sub.ij which have only two values. preferably, one of these values is a preselected value set by the value of a fixed resistor, and the other value is zero, created simply with an open circuit. values for the t.sub.ij terms of the clipped t matrix are obtained through an iterative process which operates on the clipped and nonclipped matrices and minimizes the error resulting from the use of the clipped t matrix. dated 1988-04-12"
4752906,temporal sequences with neural networks,"a sequence generator employing a neural network having its output coupled to at least one plurality of delay elements. the delayed outputs are fed back to an input interconnection network, wherein they contribute to the next state transition through an appropriate combination of interconnections.",1988-06-21,"The title of the patent is temporal sequences with neural networks and its abstract is a sequence generator employing a neural network having its output coupled to at least one plurality of delay elements. the delayed outputs are fed back to an input interconnection network, wherein they contribute to the next state transition through an appropriate combination of interconnections. dated 1988-06-21"
4760437,neural networks,"neural network type information processing devices have been proposed. in these devices, a matrix structure is utilized with impedance at the matrix intersection points. it has been found that excellent versatility in design is achieved by utilizing photoconductors at these intersection points and thus affording the possibility of controlling impedance by, in turn, controlling the level of incident light.",1988-07-26,"The title of the patent is neural networks and its abstract is neural network type information processing devices have been proposed. in these devices, a matrix structure is utilized with impedance at the matrix intersection points. it has been found that excellent versatility in design is achieved by utilizing photoconductors at these intersection points and thus affording the possibility of controlling impedance by, in turn, controlling the level of incident light. dated 1988-07-26"
4782460,computing apparatus comprising a programmable resistor,"computing apparatus (e.g., a neural network) advantageously comprises a programmable resistor body comprising typically a multiplicity of resistors r.sub.ij. the resistance of any given r.sub.ij is changeable from a relatively high resistance to a lower resistance by application of an appropriate electrical signal, and can be reset to a higher resistance by application of an appropriate signal of reverse polarity. exemplarily, a programmable resistor body comprises a thin layer of bismuth oxide or strontium barium niobate.",1988-11-01,"The title of the patent is computing apparatus comprising a programmable resistor and its abstract is computing apparatus (e.g., a neural network) advantageously comprises a programmable resistor body comprising typically a multiplicity of resistors r.sub.ij. the resistance of any given r.sub.ij is changeable from a relatively high resistance to a lower resistance by application of an appropriate electrical signal, and can be reset to a higher resistance by application of an appropriate signal of reverse polarity. exemplarily, a programmable resistor body comprises a thin layer of bismuth oxide or strontium barium niobate. dated 1988-11-01"
4807168,hybrid analog-digital associative neural network,"random access memory is used to store synaptic information in the form of a matrix of rows and columns of binary digits. n rows read in sequence are processed through switches and resistors, and a summing amplifier to n neural amplifiers in sequence, one row for each amplifier, using a first array of sample-and-hold devices s/h1 for commutation. the outputs of the neural amplifiers are stored in a second array of sample-and-hold devices s/h2 so that after n rows are processed, all of said second array of sample-and-hold devices are updated. a second memory may be added for binary values of 0 and -1, and processed simultaneously with the first to provide for values of 1, 0, and -1, the results of which are combined in a difference amplifier.",1989-02-21,"The title of the patent is hybrid analog-digital associative neural network and its abstract is random access memory is used to store synaptic information in the form of a matrix of rows and columns of binary digits. n rows read in sequence are processed through switches and resistors, and a summing amplifier to n neural amplifiers in sequence, one row for each amplifier, using a first array of sample-and-hold devices s/h1 for commutation. the outputs of the neural amplifiers are stored in a second array of sample-and-hold devices s/h2 so that after n rows are processed, all of said second array of sample-and-hold devices are updated. a second memory may be added for binary values of 0 and -1, and processed simultaneously with the first to provide for values of 1, 0, and -1, the results of which are combined in a difference amplifier. dated 1989-02-21"
4866645,neural network with dynamic refresh capability,"an analog neural network composed of an array of capacitors for storing weighted electric charges. electric charges, or voltages, on the capacitors control the impedance (resistance) values of a corresponding plurality of mosfets which selectively couple input signals to one input of a summing amplifier. a plurality of semiconductor gating elements (e.g. mosfets) selectively couple to the capacitor's weighted analog voltage values received serially over an input line. the weighted voltage on the input line are periodically applied to the proper capacitors in the neural network via the gating elements so as to refresh the weighted electric charges on the capacitors, and at a multiplex rate that maintains the voltages on the capacitors within acceptable tolerance levels.",1989-09-12,"The title of the patent is neural network with dynamic refresh capability and its abstract is an analog neural network composed of an array of capacitors for storing weighted electric charges. electric charges, or voltages, on the capacitors control the impedance (resistance) values of a corresponding plurality of mosfets which selectively couple input signals to one input of a summing amplifier. a plurality of semiconductor gating elements (e.g. mosfets) selectively couple to the capacitor's weighted analog voltage values received serially over an input line. the weighted voltage on the input line are periodically applied to the proper capacitors in the neural network via the gating elements so as to refresh the weighted electric charges on the capacitors, and at a multiplex rate that maintains the voltages on the capacitors within acceptable tolerance levels. dated 1989-09-12"
4873455,programmable ferroelectric polymer neural network,"the network comprises several memory elements made of ferroelectric polymer, arranged in a matrix organization at the intersections of row and column electrodes. each memory element (mij) memorizes a synaptic coefficient a.sub.ij of the network which may be restored by pyroelectric effect on the corresponding column of the network. amplifier circuits, respectively connected to the columns, give a voltage which is equal to the sum, to which a sign is assigned, of the products of the synaptic coefficients by the voltage components applied to each of the lines of the network.",1989-10-10,"The title of the patent is programmable ferroelectric polymer neural network and its abstract is the network comprises several memory elements made of ferroelectric polymer, arranged in a matrix organization at the intersections of row and column electrodes. each memory element (mij) memorizes a synaptic coefficient a.sub.ij of the network which may be restored by pyroelectric effect on the corresponding column of the network. amplifier circuits, respectively connected to the columns, give a voltage which is equal to the sum, to which a sign is assigned, of the products of the synaptic coefficients by the voltage components applied to each of the lines of the network. dated 1989-10-10"
4876731,neural network model in pattern recognition using probabilistic contextual information,"a pattern recognition system for recognizing an unknown pattern comprised of symbols which are part of a pattern system which is devoid of inherent context such as numbers. artificial contextual information based on other than symbol features and the pattern system and in the form of probability weighted expected interpretations are stored and used in the processing phase of recognition. in the system disclosed, the system comprises a neural network whose forward and feedback paths are controlled by the output cells of the network based, in part, on the contextual information.",1989-10-24,"The title of the patent is neural network model in pattern recognition using probabilistic contextual information and its abstract is a pattern recognition system for recognizing an unknown pattern comprised of symbols which are part of a pattern system which is devoid of inherent context such as numbers. artificial contextual information based on other than symbol features and the pattern system and in the form of probability weighted expected interpretations are stored and used in the processing phase of recognition. in the system disclosed, the system comprises a neural network whose forward and feedback paths are controlled by the output cells of the network based, in part, on the contextual information. dated 1989-10-24"
4884216,neural network system for adaptive sensory-motor coordination of multijoint robots for single postures,"a neural-like network system that adaptively controls a visually guided, two-jointed robot arm to reach spot targets in three dimensions. the system learns and maintains visual-motor calibrations by itself, starting with only loosely defined relationships. the geometry of the system is composed of distributed, interleaved combinations of actuator inputs. it is fault tolerant and uses analog processing. learning is achieved by modifying the distributions of input weights in the system after each arm positioning. modifications of the weights are made incrementally according to errors of consistency between the actuator signals used to orient the cameras and those used to move the arm.",1989-11-28,"The title of the patent is neural network system for adaptive sensory-motor coordination of multijoint robots for single postures and its abstract is a neural-like network system that adaptively controls a visually guided, two-jointed robot arm to reach spot targets in three dimensions. the system learns and maintains visual-motor calibrations by itself, starting with only loosely defined relationships. the geometry of the system is composed of distributed, interleaved combinations of actuator inputs. it is fault tolerant and uses analog processing. learning is achieved by modifying the distributions of input weights in the system after each arm positioning. modifications of the weights are made incrementally according to errors of consistency between the actuator signals used to orient the cameras and those used to move the arm. dated 1989-11-28"
4885757,digital adaptive receiver employing maximum-likelihood sequence estimation with neural networks,"a maximum-likelihood sequence estimator receiver includes a matched filter connected to a digital transmission channel and a sampler for providing sampled signals output by the matched filter. the sampled signals are input to an analog neural network to provide high-speed outputs representative of the transmission channel signals. the neural network outputs are also provided as inputs to a coefficient estimator which produces coefficients for feedback to the matched filter. for time-varying transmission channel characteristics, the coefficient estimator provides a second coefficient output which is utilized for changing the interconnection strengths of the neural network connection matrix to offset the varying transmission channel characteristics.",1989-12-05,"The title of the patent is digital adaptive receiver employing maximum-likelihood sequence estimation with neural networks and its abstract is a maximum-likelihood sequence estimator receiver includes a matched filter connected to a digital transmission channel and a sampler for providing sampled signals output by the matched filter. the sampled signals are input to an analog neural network to provide high-speed outputs representative of the transmission channel signals. the neural network outputs are also provided as inputs to a coefficient estimator which produces coefficients for feedback to the matched filter. for time-varying transmission channel characteristics, the coefficient estimator provides a second coefficient output which is utilized for changing the interconnection strengths of the neural network connection matrix to offset the varying transmission channel characteristics. dated 1989-12-05"
4891782,parallel neural network for a full binary adder,a method for performing the addition of two n-bit binary numbers using palel neural networks. the value of a first register is converted and transferred into a second register in a mathematical fashion so as to add the numbers of the first register into the second register. when the first register contains all zeros then the desired sum is found in the second register.,1990-01-02,The title of the patent is parallel neural network for a full binary adder and its abstract is a method for performing the addition of two n-bit binary numbers using palel neural networks. the value of a first register is converted and transferred into a second register in a mathematical fashion so as to add the numbers of the first register into the second register. when the first register contains all zeros then the desired sum is found in the second register. dated 1990-01-02
4893255,spike transmission for neural networks,"pulse trains are utilized for the transmission of information in a neural network. a squash function is achieved by logically or'ing together pulsed outputs, giving f(x) approximately 1-e.sup.-x. for back propagation, as derived by rumelhart, the derivative of the squash function is available by examining the time when no or'ed together pulses are present, being 1-f(x), or e.sup.-x. logically and'ing of the two signals. mulitplication of input frequencies by weights is accomplished by modulating the width of the output pulses, while keeping the frequency the same.",1990-01-09,"The title of the patent is spike transmission for neural networks and its abstract is pulse trains are utilized for the transmission of information in a neural network. a squash function is achieved by logically or'ing together pulsed outputs, giving f(x) approximately 1-e.sup.-x. for back propagation, as derived by rumelhart, the derivative of the squash function is available by examining the time when no or'ed together pulses are present, being 1-f(x), or e.sup.-x. logically and'ing of the two signals. mulitplication of input frequencies by weights is accomplished by modulating the width of the output pulses, while keeping the frequency the same. dated 1990-01-09"
4896053,solitary wave circuit for neural network emulation,""" a circuit for emulating a nerve cell is used to generate one or more simple neural networks. in the preferred embodiment, the circuit comprises an lc ladder circuit including one or more modules, each of the modules comprising an """"l"""" two-port circuit comprising a first shunt branch having a variable capacitor, a second shunt branch having a series-connected conductance and a variable d.c. bias source, and a third branch connected in series with the first and second branches, the third branch comprising an active inductor. the inductor is formed by one or more operational amplifiers interconnected in a feedback configuration. each of the variable capacitances and the inductances cooperate to emulate a portion of a neuron by receiving a stimulus and generating or propagating a unidirectional solitary wave output representing an action potential. """,1990-01-23,"The title of the patent is solitary wave circuit for neural network emulation and its abstract is "" a circuit for emulating a nerve cell is used to generate one or more simple neural networks. in the preferred embodiment, the circuit comprises an lc ladder circuit including one or more modules, each of the modules comprising an """"l"""" two-port circuit comprising a first shunt branch having a variable capacitor, a second shunt branch having a series-connected conductance and a variable d.c. bias source, and a third branch connected in series with the first and second branches, the third branch comprising an active inductor. the inductor is formed by one or more operational amplifiers interconnected in a feedback configuration. each of the variable capacitances and the inductances cooperate to emulate a portion of a neuron by receiving a stimulus and generating or propagating a unidirectional solitary wave output representing an action potential. "" dated 1990-01-23"
4897811,n-dimensional coulomb neural network which provides for cumulative learning of internal representations,""" a learning algorithm for the n-dimensional coulomb network is disclosed which is applicable to multi-layer networks. the central concept is to define a potential energy of a collection of memory sites. then each memory site is an attractor of other memory sites. with the proper definition of attractive and repulsive potentials between various memory sites, it is possible to minimize the energy of the collection of memories. by this method, internal representations may be """"built-up"""" one layer at a time. following the method of bachmann et al. a system is considered in which memories of events have already been recorded in a layer of cells. a method is found for the consolidation of the number of memories required to correctly represent the pattern environment. this method is shown to be applicable to a supervised or unsupervised learning paradigm in which pairs of input and output patterns are presented sequentially to the network. the resulting learning procedure develops internal representations in an incremental or cumulative fashion, from the layer closest to the input, to the output layer. """,1990-01-30,"The title of the patent is n-dimensional coulomb neural network which provides for cumulative learning of internal representations and its abstract is "" a learning algorithm for the n-dimensional coulomb network is disclosed which is applicable to multi-layer networks. the central concept is to define a potential energy of a collection of memory sites. then each memory site is an attractor of other memory sites. with the proper definition of attractive and repulsive potentials between various memory sites, it is possible to minimize the energy of the collection of memories. by this method, internal representations may be """"built-up"""" one layer at a time. following the method of bachmann et al. a system is considered in which memories of events have already been recorded in a layer of cells. a method is found for the consolidation of the number of memories required to correctly represent the pattern environment. this method is shown to be applicable to a supervised or unsupervised learning paradigm in which pairs of input and output patterns are presented sequentially to the network. the resulting learning procedure develops internal representations in an incremental or cumulative fashion, from the layer closest to the input, to the output layer. "" dated 1990-01-30"
4904881,exclusive-or cell for neural network and the like,a semiconductor cell for producing an output current that is related to the match between an input vector pattern and a weighting pattern is described. the cell is particularly useful as a synapse cell within a neural network to perform pattern recognition tasks. the cell includes a pair of input lines for receiving a differential input vector element value and a pair of output lines for providing a difference current to a current summing neural amplifier. a plurality of floating gate devices each having a floating gate member are employed in the synapse cell to store charge in accordance with a predetermined weight pattern. each of the floating gate devices is uniquely coupled to a combination of an output current line and an input voltage line such that the difference current provided to the neural amplifier is related to the match between the input vector and the stored weight.,1990-02-27,The title of the patent is exclusive-or cell for neural network and the like and its abstract is a semiconductor cell for producing an output current that is related to the match between an input vector pattern and a weighting pattern is described. the cell is particularly useful as a synapse cell within a neural network to perform pattern recognition tasks. the cell includes a pair of input lines for receiving a differential input vector element value and a pair of output lines for providing a difference current to a current summing neural amplifier. a plurality of floating gate devices each having a floating gate member are employed in the synapse cell to store charge in accordance with a predetermined weight pattern. each of the floating gate devices is uniquely coupled to a combination of an output current line and an input voltage line such that the difference current provided to the neural amplifier is related to the match between the input vector and the stored weight. dated 1990-02-27
4904882,superconducting optical switch,""" a combination of optical interconnect technology with superconducting matal to form a superconducting neural network array. superconducting material in a matrix has the superconducting current decreased in one filament of the matrix by interaction of the cooper pairs with radiation controlled by a spatial light modulator. this decrease in current results in a switch of current, in a relative sense, to another filament in the matrix. this """"switching"""" mechanism can be used in a digital or analog fashion in a superconducting computer application. """,1990-02-27,"The title of the patent is superconducting optical switch and its abstract is "" a combination of optical interconnect technology with superconducting matal to form a superconducting neural network array. superconducting material in a matrix has the superconducting current decreased in one filament of the matrix by interaction of the cooper pairs with radiation controlled by a spatial light modulator. this decrease in current results in a switch of current, in a relative sense, to another filament in the matrix. this """"switching"""" mechanism can be used in a digital or analog fashion in a superconducting computer application. "" dated 1990-02-27"
4906865,sample and hold circuit for temporal associations in a neural network,"a sample and hold circuit for introducing delayed feedback into an associative memory is described. the circuit continuously samples an output sequence derived from a neural network; then, in response to a clock signal, it holds that output sequence until the next clock signal. the held sequence is coupled back to the input of the network so that the present output sequence becomes some function of the past output sequence. this delayed feedback enables the associative recall of a memorized sequence from the neural network.",1990-03-06,"The title of the patent is sample and hold circuit for temporal associations in a neural network and its abstract is a sample and hold circuit for introducing delayed feedback into an associative memory is described. the circuit continuously samples an output sequence derived from a neural network; then, in response to a clock signal, it holds that output sequence until the next clock signal. the held sequence is coupled back to the input of the network so that the present output sequence becomes some function of the past output sequence. this delayed feedback enables the associative recall of a memorized sequence from the neural network. dated 1990-03-06"
4912647,neural network training tool,"a method of training an artificial neural network uses a first computer configured as a plurality of interconnected neural units arranged in a network. a neural unit has a first subunit and a second subunit. the first subunit has first inputs and a corresponding first set of variables for operating upon the first inputs to provide a first output during a forward pass. the first set of variables can change in response to feedback representing differences between desired network outputs and actual network outputs. the second subunit has a plurality of second inputs, and a corresponding second set of variables for operating upon the second inputs to provide a second output. the second set of variables can change in response to differences between desired network outputs for selected network inputs and actual network outputs. the computer provides an activating variable representing the difference between current second output and previous second outputs. the activating variable is added to the feedback to accelerate the change of said first set of variables. a second computer is configured as a plurality of interconnected neural units arranged in a network. the network is functionally equivalent to the network of the first computer in a forward pass when provided with sets of values corresponding to each converged first set of variables of the first computer.",1990-03-27,"The title of the patent is neural network training tool and its abstract is a method of training an artificial neural network uses a first computer configured as a plurality of interconnected neural units arranged in a network. a neural unit has a first subunit and a second subunit. the first subunit has first inputs and a corresponding first set of variables for operating upon the first inputs to provide a first output during a forward pass. the first set of variables can change in response to feedback representing differences between desired network outputs and actual network outputs. the second subunit has a plurality of second inputs, and a corresponding second set of variables for operating upon the second inputs to provide a second output. the second set of variables can change in response to differences between desired network outputs for selected network inputs and actual network outputs. the computer provides an activating variable representing the difference between current second output and previous second outputs. the activating variable is added to the feedback to accelerate the change of said first set of variables. a second computer is configured as a plurality of interconnected neural units arranged in a network. the network is functionally equivalent to the network of the first computer in a forward pass when provided with sets of values corresponding to each converged first set of variables of the first computer. dated 1990-03-27"
4912649,accelerating learning in neural networks,"a method of accelerating the training of an artificial neural network uses a computer configured as an artificial neural network with a network input and a network output, and having a plurality of interconnected units arranged in layers including an input layer and an output layer. each unit has a multiplicity of unit inputs and a set of variables for operating upon the unit inputs to provide a unit output. a plurality of examples are serially provided to the network input and the network output is observed. the computer is programmed with a back propagation algorithm for adjusting each set of variables in response to feedback representing differences between the network output for each example and the desired output. the examples are iterated while those values which change are identified. the examples are reiterated and the algorithm is applied to only those values which changed in a previous iteration.",1990-03-27,"The title of the patent is accelerating learning in neural networks and its abstract is a method of accelerating the training of an artificial neural network uses a computer configured as an artificial neural network with a network input and a network output, and having a plurality of interconnected units arranged in layers including an input layer and an output layer. each unit has a multiplicity of unit inputs and a set of variables for operating upon the unit inputs to provide a unit output. a plurality of examples are serially provided to the network input and the network output is observed. the computer is programmed with a back propagation algorithm for adjusting each set of variables in response to feedback representing differences between the network output for each example and the desired output. the examples are iterated while those values which change are identified. the examples are reiterated and the algorithm is applied to only those values which changed in a previous iteration. dated 1990-03-27"
4912651,speeding learning in neural networks,"a method of accelerating the training of an artificial neural network uses a computer configured as an artificial neural network with a network input and a network output, and having a plurality of interconnected units arranged in layers including an input layer and an output layer. each unit has a multiplicity of unit inputs and a set of variables for operating upon a unit inputs to provide a unit output. a plurality of examples are serially provided to the network input and the network output is observed. the computer is programmed with a back propagation algorithm for adjusting each set of variables in response to feedback representing differences between the network output for each example and the desired output. the examples are iterated until the signs of the outputs of the units of the output layer converge. then each set of variables is multiplied by a multiplier. the examples are reiterated until the magnitude of the outputs of the units of the output layer converge.",1990-03-27,"The title of the patent is speeding learning in neural networks and its abstract is a method of accelerating the training of an artificial neural network uses a computer configured as an artificial neural network with a network input and a network output, and having a plurality of interconnected units arranged in layers including an input layer and an output layer. each unit has a multiplicity of unit inputs and a set of variables for operating upon a unit inputs to provide a unit output. a plurality of examples are serially provided to the network input and the network output is observed. the computer is programmed with a back propagation algorithm for adjusting each set of variables in response to feedback representing differences between the network output for each example and the desired output. the examples are iterated until the signs of the outputs of the units of the output layer converge. then each set of variables is multiplied by a multiplier. the examples are reiterated until the magnitude of the outputs of the units of the output layer converge. dated 1990-03-27"
4912652,fast neural network training,a method of accelerating the training of an artificial neural network uses a computer configured as an artificial neural network with a network input and a network output and having a plurality of interconnected units arranged in layers including an input layer and an output layer. each unit has a multiplicity of unit inputs and a set of variables for operating upon a unit inputs to provide a unit output in the range between binary 1 and binary 0. a plurality of training examples is serially provided to the network input and the network output is observed. the computer is programmed with a back propagation algorithm for changing each set of variables in response to feedback representing differences between the network output for each example and the desired output. the examples are iterated while the output of a unit is observed. the feedback to a unit is adjusted so that a larger feedback is obtained when the output of the unit is near binary 1 or binary 0 than when the output is midrange between binary 1 or binary 0.,1990-03-27,The title of the patent is fast neural network training and its abstract is a method of accelerating the training of an artificial neural network uses a computer configured as an artificial neural network with a network input and a network output and having a plurality of interconnected units arranged in layers including an input layer and an output layer. each unit has a multiplicity of unit inputs and a set of variables for operating upon a unit inputs to provide a unit output in the range between binary 1 and binary 0. a plurality of training examples is serially provided to the network input and the network output is observed. the computer is programmed with a back propagation algorithm for changing each set of variables in response to feedback representing differences between the network output for each example and the desired output. the examples are iterated while the output of a unit is observed. the feedback to a unit is adjusted so that a larger feedback is obtained when the output of the unit is near binary 1 or binary 0 than when the output is midrange between binary 1 or binary 0. dated 1990-03-27
4912653,trainable neural network,"a trainable artificial neural network includes a computer configured as a plurality of interconnected neural units arranged in a layered network. an input layer has a network input and an output layer has a network output. a neural unit has a first subunit and a second subunit, with the first subunit having one or more first inputs and a corresponding first set of variables for operating upon the said first inputs to provide a first output. the first set of variables can change in response to feedback representing differences between desired network outputs and actual network outputs. the second subunit has a plurality second inputs, and a corresponding second set of variables for operating upon said second inputs to provide a second output. the second set of variables can change in response to differences between desired network outputs for selected network inputs and actual network outputs. the computer provides an activating variable representing the difference between current second output and previous second outputs, and adds the activating variable to said feedback to accelerate the change of the first set of variables.",1990-03-27,"The title of the patent is trainable neural network and its abstract is a trainable artificial neural network includes a computer configured as a plurality of interconnected neural units arranged in a layered network. an input layer has a network input and an output layer has a network output. a neural unit has a first subunit and a second subunit, with the first subunit having one or more first inputs and a corresponding first set of variables for operating upon the said first inputs to provide a first output. the first set of variables can change in response to feedback representing differences between desired network outputs and actual network outputs. the second subunit has a plurality second inputs, and a corresponding second set of variables for operating upon said second inputs to provide a second output. the second set of variables can change in response to differences between desired network outputs for selected network inputs and actual network outputs. the computer provides an activating variable representing the difference between current second output and previous second outputs, and adds the activating variable to said feedback to accelerate the change of the first set of variables. dated 1990-03-27"
4912654,neural networks learning method,"a method of accelerating the training of an artificial neural network uses a computer configured as an artificial neural network with a network input and a network output, and having a plurality of interconnected units arranged in layers including an input layer and an output layer. each unit has a multiplicity of unit inputs and a set of variables for operating upon a unit inputs to provide a unit output in the range positive 1 and negative 1. a plurality of examples are serially provided to the network input and the network output is observed. the computer is programmed with a back propagation algorithm for calculating changes to the sets of variables in response to feedback representing differences between the network output for each example and the desired output. the absolute magnitude of the product of an input and the corresponding output of a unit is calculated. the feedback to that unit is adjusted in response to absolute magnitude so that said feedback is larger with a larger absolute magnitude than with a smaller absolute magnitude.",1990-03-27,"The title of the patent is neural networks learning method and its abstract is a method of accelerating the training of an artificial neural network uses a computer configured as an artificial neural network with a network input and a network output, and having a plurality of interconnected units arranged in layers including an input layer and an output layer. each unit has a multiplicity of unit inputs and a set of variables for operating upon a unit inputs to provide a unit output in the range positive 1 and negative 1. a plurality of examples are serially provided to the network input and the network output is observed. the computer is programmed with a back propagation algorithm for calculating changes to the sets of variables in response to feedback representing differences between the network output for each example and the desired output. the absolute magnitude of the product of an input and the corresponding output of a unit is calculated. the feedback to that unit is adjusted in response to absolute magnitude so that said feedback is larger with a larger absolute magnitude than with a smaller absolute magnitude. dated 1990-03-27"
4912655,adjusting neural networks,"a method of accelerating the training of an artificial neural network uses a computer configured as an artificial neural network with a network input and a network output, and having a plurality of interconnected units arranged in layers including an input layer and an output layer. each unit has a multiplicity of unit inputs and a set of variables for operating upon a unit inputs to provide a unit output. the computer is programmed with a back propagation algorithm. a plurality of examples are serially provided to the network input and the network output is observed. the examples are iterated and proposed changes to each set of variables are calculated in response to feedback representing differences betwen the network output for each example and the desired output. the proposed changes are accumulated for a predetermined number of iterations, whereupon the accumulated proposed changes are added to the set of variables.",1990-03-27,"The title of the patent is adjusting neural networks and its abstract is a method of accelerating the training of an artificial neural network uses a computer configured as an artificial neural network with a network input and a network output, and having a plurality of interconnected units arranged in layers including an input layer and an output layer. each unit has a multiplicity of unit inputs and a set of variables for operating upon a unit inputs to provide a unit output. the computer is programmed with a back propagation algorithm. a plurality of examples are serially provided to the network input and the network output is observed. the examples are iterated and proposed changes to each set of variables are calculated in response to feedback representing differences betwen the network output for each example and the desired output. the proposed changes are accumulated for a predetermined number of iterations, whereupon the accumulated proposed changes are added to the set of variables. dated 1990-03-27"
4914603,training neural networks,"a method of training an artificial neural network uses a computer configured as a plurality of interconnected neural units arranged in a layered network including an input layer having a network input, and an output layer having a network output. a neural unit has a first subunit and a second subunit. the first subunit having one or more first inputs, and a corresponding first set of variables for operating upon the first inputs to provide a first output. the first set of variables can change in response to feedback representing differences between desired network outputs for selected network inputs and actual network outputs. the second subunit has a plurality of second inputs, and a corresponding second set of variables for operating upon said second inputs to provide a second output. the second set of variables can change in response to differences between desired network outputs for selected network inputs and actual network outputs. the computer provides an activating variable representing the difference between current second output and previous second outputs. a series of examples of data is provided as network input to said network. the activating variable is added to the feedback to accelerate the change of said first set of variables. the actual resulting network outputs are compared to desired outputs corresponding to the examples. the examples are iterated until the network outputs converge to a solution.",1990-04-03,"The title of the patent is training neural networks and its abstract is a method of training an artificial neural network uses a computer configured as a plurality of interconnected neural units arranged in a layered network including an input layer having a network input, and an output layer having a network output. a neural unit has a first subunit and a second subunit. the first subunit having one or more first inputs, and a corresponding first set of variables for operating upon the first inputs to provide a first output. the first set of variables can change in response to feedback representing differences between desired network outputs for selected network inputs and actual network outputs. the second subunit has a plurality of second inputs, and a corresponding second set of variables for operating upon said second inputs to provide a second output. the second set of variables can change in response to differences between desired network outputs for selected network inputs and actual network outputs. the computer provides an activating variable representing the difference between current second output and previous second outputs. a series of examples of data is provided as network input to said network. the activating variable is added to the feedback to accelerate the change of said first set of variables. the actual resulting network outputs are compared to desired outputs corresponding to the examples. the examples are iterated until the network outputs converge to a solution. dated 1990-04-03"
4914708,system for self-organization of stable category recognition codes for analog input patterns,"a neural network includes a feature representation field which receives input patterns. signals from the feature representation field select a category from a category representation field through a first adaptive filter. based on the selected category, a template pattern is applied to the feature representation field, and a match between the template and the input is determined. if the angle between the template vector and a vector within the representation field is too great, the selected category is reset. otherwise the category selection and template pattern are adapted to the input pattern as well as the previously stored template. a complex representation field includes signals normalized relative to signals across the field and feedback for pattern contrast enhancement.",1990-04-03,"The title of the patent is system for self-organization of stable category recognition codes for analog input patterns and its abstract is a neural network includes a feature representation field which receives input patterns. signals from the feature representation field select a category from a category representation field through a first adaptive filter. based on the selected category, a template pattern is applied to the feature representation field, and a match between the template and the input is determined. if the angle between the template vector and a vector within the representation field is too great, the selected category is reset. otherwise the category selection and template pattern are adapted to the input pattern as well as the previously stored template. a complex representation field includes signals normalized relative to signals across the field and feedback for pattern contrast enhancement. dated 1990-04-03"
4918618,discrete weight neural network,"a neural network using interconnecting weights each with two values, one of which is selected for use, can be taught to map a set of input vectors to a set of output vectors. a set of input vectors is applied to the network and in response, a set of output vectors is produced by the network. the error is the difference between desired outputs and actual outputs. the network is trained in the following manner. a set of input vectors is presented to the network, each vector being propogated forward through the network to produce an output vector. a set of error vectors is then presented to the network and propagated backwards. each tensor weight element includes a selective change means which accumulates particular information about the error. after all the input vectors are presented, an update phase is initiated. during the update phase, in accordance with the results of the derived algorithm, the selective change means selects the other weight value if selecting the other weight value will decrease the total error. only one such change is made per set. after the update phase, if a selected value was changed, the entire process is repeated. when no values are switched, the network has adapted as well as it can, and the training is completed.",1990-04-17,"The title of the patent is discrete weight neural network and its abstract is a neural network using interconnecting weights each with two values, one of which is selected for use, can be taught to map a set of input vectors to a set of output vectors. a set of input vectors is applied to the network and in response, a set of output vectors is produced by the network. the error is the difference between desired outputs and actual outputs. the network is trained in the following manner. a set of input vectors is presented to the network, each vector being propogated forward through the network to produce an output vector. a set of error vectors is then presented to the network and propagated backwards. each tensor weight element includes a selective change means which accumulates particular information about the error. after all the input vectors are presented, an update phase is initiated. during the update phase, in accordance with the results of the derived algorithm, the selective change means selects the other weight value if selecting the other weight value will decrease the total error. only one such change is made per set. after the update phase, if a selected value was changed, the entire process is repeated. when no values are switched, the network has adapted as well as it can, and the training is completed. dated 1990-04-17"
4926064,sleep refreshed memory for neural network,"a method and apparatus are disclosed for implementing a neural network having a sleep mode during which capacitively stored synaptic connectivity weights are refreshed. each neuron outputs an analog activity level, represented in a preferred embodiment by the frequency of digital pulses. feed-forward synaptic connection circuits couple the activity level outputs of first level neurons to inputs of second level neurons, and feed-back synaptic connection circuits couple outputs of second level neurons to inputs of first level neurons, the coupling being weighted according to connectivity weights stored on respective storage capacitors in each synaptic connection circuit. the network learns according to a learning algorithm under which the connections in both directions between a particular first level neuron and a particular second level neuron are strengthened to the extent of concurrence of high activity levels in both the first and second level neurons, and weakened to the extent of concurrence of a high activity level in the second level neuron and a low activity level in the first level neuron. the network is put to sleep by disconnecting all environmental inputs and providing a non-specific low activity level signal to each of the first level neurons. this causes the network to randomly traverse its state space with low intensity resonant firings, each state being visited with a probability responsive to the initial connectivity weights of the connections which abut the second level neuron representing such state. refresh is accomplished since the learning algorithm remains active during sleep. thus, the sleep refresh mechanism enhances the contrast in the connectivity terrain and strengthens connections that would otherwise wash out due to lack of visitation while the system is awake. a deep sleep mechanism is also provided for preventing runaway strengthening of favored states, and also to encourage weber law compliance.",1990-05-15,"The title of the patent is sleep refreshed memory for neural network and its abstract is a method and apparatus are disclosed for implementing a neural network having a sleep mode during which capacitively stored synaptic connectivity weights are refreshed. each neuron outputs an analog activity level, represented in a preferred embodiment by the frequency of digital pulses. feed-forward synaptic connection circuits couple the activity level outputs of first level neurons to inputs of second level neurons, and feed-back synaptic connection circuits couple outputs of second level neurons to inputs of first level neurons, the coupling being weighted according to connectivity weights stored on respective storage capacitors in each synaptic connection circuit. the network learns according to a learning algorithm under which the connections in both directions between a particular first level neuron and a particular second level neuron are strengthened to the extent of concurrence of high activity levels in both the first and second level neurons, and weakened to the extent of concurrence of a high activity level in the second level neuron and a low activity level in the first level neuron. the network is put to sleep by disconnecting all environmental inputs and providing a non-specific low activity level signal to each of the first level neurons. this causes the network to randomly traverse its state space with low intensity resonant firings, each state being visited with a probability responsive to the initial connectivity weights of the connections which abut the second level neuron representing such state. refresh is accomplished since the learning algorithm remains active during sleep. thus, the sleep refresh mechanism enhances the contrast in the connectivity terrain and strengthens connections that would otherwise wash out due to lack of visitation while the system is awake. a deep sleep mechanism is also provided for preventing runaway strengthening of favored states, and also to encourage weber law compliance. dated 1990-05-15"
4926180,analog to digital conversion using correlated quantization and collective optimization,"a 1-bit nonstandard a/d converter for converting a block u of n samples of a continuous time analog signal u(t) into n corresponding 1-bit binary values x, such that a distortion measure of the form d(u,x)=(au-bx).sup.t (au-bx) is minimized, is implemented with an n-input parallel sample-and-hold circuit and a neural network having n nonlinear amplifiers, where u and x are n-dimensional vectors, and a and b are n.times.n matrices. minimization of the above distortion measure is equivalent to minimizing the quantity equ 1/2x.sup.t b.sup.t bx-u.sup.t a.sup.t bx, which is achieved to at least a good approximation by the n-amplifier neural network. accordingly, the conductances of the feedback connections among the amplifiers are defined by respective off-diagonal elements of the matrix -b.sup.t b. additionally, each amplifier of the neural network is connected to receive the analog signal samples through respective conductances defined by the matrix b.sup.t. furthermore, each amplifier receives a respective constant signal defined by the diagonal elements of the matrix -b.sup.t b. the stabilized outputs of the n amplifiers are the binary values of the digital signal x. a multiple-bit nonstandard a/d converter based on for foregoing 1-bit a/d converter is also disclosed.",1990-05-15,"The title of the patent is analog to digital conversion using correlated quantization and collective optimization and its abstract is a 1-bit nonstandard a/d converter for converting a block u of n samples of a continuous time analog signal u(t) into n corresponding 1-bit binary values x, such that a distortion measure of the form d(u,x)=(au-bx).sup.t (au-bx) is minimized, is implemented with an n-input parallel sample-and-hold circuit and a neural network having n nonlinear amplifiers, where u and x are n-dimensional vectors, and a and b are n.times.n matrices. minimization of the above distortion measure is equivalent to minimizing the quantity equ 1/2x.sup.t b.sup.t bx-u.sup.t a.sup.t bx, which is achieved to at least a good approximation by the n-amplifier neural network. accordingly, the conductances of the feedback connections among the amplifiers are defined by respective off-diagonal elements of the matrix -b.sup.t b. additionally, each amplifier of the neural network is connected to receive the analog signal samples through respective conductances defined by the matrix b.sup.t. furthermore, each amplifier receives a respective constant signal defined by the diagonal elements of the matrix -b.sup.t b. the stabilized outputs of the n amplifiers are the binary values of the digital signal x. a multiple-bit nonstandard a/d converter based on for foregoing 1-bit a/d converter is also disclosed. dated 1990-05-15"
4931674,programmable analog voltage multiplier circuit means,"an improved programmable analog voltage multiplier circuit means (pavmcm) cluding various embodiments thereof that are operable in linear/nonlinear fashion. the pavmcm is generally made up of multiplier circuit means, at least one switch means and at least one capacitor means. the switch means is connected to a programmable analog voltage (pav) input and the capacitor means. the circuit means is composed of a high impedance analog voltage (hiav) programming input, an analog voltage input and current source output means. the capacitor means is connected to the switch means and the hiav programming input. the capacitor means receives and dynamically stores a pav input when the switch is closed and then applies the dynamically stored pav input to the hiav programming input of the circuit means when the switch is opened. the product of the pav input and the analog voltage input for a circuit means provides the multiplied current output of the output means thereof. because of the high impedance of a fet gate means, it may be used where its gate means is the programming input of the pavmcm means. pavmcm means can be formed using fet multiplier and differential amplifier multiplier circuit means. the pavmcm can be arranged to form embodiments of analog vector-vector and analog vector-matrix multiplier circuit means. one of the advantages of the pavmcm when configured as a vector-matrix multiplier circuit means is that it is useful in an artificial neural network as well as for pattern recognition.",1990-06-05,"The title of the patent is programmable analog voltage multiplier circuit means and its abstract is an improved programmable analog voltage multiplier circuit means (pavmcm) cluding various embodiments thereof that are operable in linear/nonlinear fashion. the pavmcm is generally made up of multiplier circuit means, at least one switch means and at least one capacitor means. the switch means is connected to a programmable analog voltage (pav) input and the capacitor means. the circuit means is composed of a high impedance analog voltage (hiav) programming input, an analog voltage input and current source output means. the capacitor means is connected to the switch means and the hiav programming input. the capacitor means receives and dynamically stores a pav input when the switch is closed and then applies the dynamically stored pav input to the hiav programming input of the circuit means when the switch is opened. the product of the pav input and the analog voltage input for a circuit means provides the multiplied current output of the output means thereof. because of the high impedance of a fet gate means, it may be used where its gate means is the programming input of the pavmcm means. pavmcm means can be formed using fet multiplier and differential amplifier multiplier circuit means. the pavmcm can be arranged to form embodiments of analog vector-vector and analog vector-matrix multiplier circuit means. one of the advantages of the pavmcm when configured as a vector-matrix multiplier circuit means is that it is useful in an artificial neural network as well as for pattern recognition. dated 1990-06-05"
4931763,memory switches based on metal oxide thin films,""" mno.sub.2-x thin films (12) exhibit irreversible memory switching (28) with an """"off/on"""" resistance ratio of at least about 10.sup.3 and the tailorability of """"on"""" state (20) resistance. such films are potentially extremely useful as a """"connection"""" element in a variety of microelectronic circuits and arrays (24). such films provide a pre-tailored, finite, non-volatile resistive element at a desired place in an electric circuit, which can be electrically turned off (22) or """"disconnected"""" as desired, by application of an electrical pulse. microswitch structures (10) constitute the thin film element, contacted by a pair of separate electrodes (16a, 16b) and have a finite, pre-selected on resistance which is ideally suited, for example, as a programmable binary synaptic connection for electronic implementation of neural network architectures. the mno.sub.2-x microswitch is non-volatile, patternable, insensitive to ultraviolet light, and adherent to a variety of insulating substrates (14), such as glass and silicon dioxide-coated silicon substrates. """,1990-06-05,"The title of the patent is memory switches based on metal oxide thin films and its abstract is "" mno.sub.2-x thin films (12) exhibit irreversible memory switching (28) with an """"off/on"""" resistance ratio of at least about 10.sup.3 and the tailorability of """"on"""" state (20) resistance. such films are potentially extremely useful as a """"connection"""" element in a variety of microelectronic circuits and arrays (24). such films provide a pre-tailored, finite, non-volatile resistive element at a desired place in an electric circuit, which can be electrically turned off (22) or """"disconnected"""" as desired, by application of an electrical pulse. microswitch structures (10) constitute the thin film element, contacted by a pair of separate electrodes (16a, 16b) and have a finite, pre-selected on resistance which is ideally suited, for example, as a programmable binary synaptic connection for electronic implementation of neural network architectures. the mno.sub.2-x microswitch is non-volatile, patternable, insensitive to ultraviolet light, and adherent to a variety of insulating substrates (14), such as glass and silicon dioxide-coated silicon substrates. "" dated 1990-06-05"
4937872,neural computation by time concentration,"apparatus that solves the problem of pattern recognition in a temporal signal that is subject to distortions and time warp. the arrangement embodying the invention comprises a neural network, an input interconnection network, and a plurality of signal modification circuits. a plurality of input leads delivers a preselected characteristic stimulus to associated signal modification units, and in response to an applied stimulus, each signal modification unit develops a plurality of output signals that begins at the time of stimulus application, rises to a peak, and decays thereafter. the mean time delay of each output (time to reach the peak) is different for each of the modification unit output signals. the outputs of the signal modification units are applied to the input interconnection unit wherein connections are made in accordance with the sequences that are to be recognized.",1990-06-26,"The title of the patent is neural computation by time concentration and its abstract is apparatus that solves the problem of pattern recognition in a temporal signal that is subject to distortions and time warp. the arrangement embodying the invention comprises a neural network, an input interconnection network, and a plurality of signal modification circuits. a plurality of input leads delivers a preselected characteristic stimulus to associated signal modification units, and in response to an applied stimulus, each signal modification unit develops a plurality of output signals that begins at the time of stimulus application, rises to a peak, and decays thereafter. the mean time delay of each output (time to reach the peak) is different for each of the modification unit output signals. the outputs of the signal modification units are applied to the input interconnection unit wherein connections are made in accordance with the sequences that are to be recognized. dated 1990-06-26"
4941122,neural network image processing system,"a neural-simulating system for an image processing system includes a plurality of networks arranged in a plurality of layers, the output signals of ones of the layers provide input signals to the others of the layers. each of the plurality of layers include a plurality of neurons operating in parallel on the input signals to the layers. the plurality of neurons within a layer are arranged in groups. each of the neurons within a group operates in parallel on the input signals. each neuron within a group of neurons operates to extract a specific feature of an area of the image being processed. each of the neurons derives output signals from the input signals representing the relative weight of the input signal applied thereto based upon a continuously differential transfer function for each function.",1990-07-10,"The title of the patent is neural network image processing system and its abstract is a neural-simulating system for an image processing system includes a plurality of networks arranged in a plurality of layers, the output signals of ones of the layers provide input signals to the others of the layers. each of the plurality of layers include a plurality of neurons operating in parallel on the input signals to the layers. the plurality of neurons within a layer are arranged in groups. each of the neurons within a group operates in parallel on the input signals. each neuron within a group of neurons operates to extract a specific feature of an area of the image being processed. each of the neurons derives output signals from the input signals representing the relative weight of the input signal applied thereto based upon a continuously differential transfer function for each function. dated 1990-07-10"
4943556,superconducting neural network computer and sensor array,""" a combination of optical interconnect technology with superconducting matal to form a superconducting neural network array. superconducting material in a matrix has the superconducting current decreased in one filament of the matrix by interaction of the cooper pairs with radiation controlled by a spatial light modulator. this decrease in current results in a switch of current, in a relative sense, to another filament in the matrix. this """"switching"""" mechanism can be used in a digital or analog fashion in a superconducting computer application. """,1990-07-24,"The title of the patent is superconducting neural network computer and sensor array and its abstract is "" a combination of optical interconnect technology with superconducting matal to form a superconducting neural network array. superconducting material in a matrix has the superconducting current decreased in one filament of the matrix by interaction of the cooper pairs with radiation controlled by a spatial light modulator. this decrease in current results in a switch of current, in a relative sense, to another filament in the matrix. this """"switching"""" mechanism can be used in a digital or analog fashion in a superconducting computer application. "" dated 1990-07-24"
4945494,neural network and system,"neural network systems (100) with learning and recall are applied to clustered multiple-featured data (122, 124, 126) and analog data.",1990-07-31,"The title of the patent is neural network and system and its abstract is neural network systems (100) with learning and recall are applied to clustered multiple-featured data (122, 124, 126) and analog data. dated 1990-07-31"
4947482,state analog neural network and method of implementing same,"a neural network is implemented by discrete-time, continuous voltage state analog device in which neuron, synapse and synaptic strength signals are generated in highly parallel analog circuits in successive states from stored values of the interdependent signals calculated in a previous state. the neuron and synapse signals are refined in a relaxation loop while the synaptic strength signals are held constant. in learning modes, the synaptic strength signals are modified in successive states from stable values of the analog neuron signals. the analog signals are stored for as long as required in master/slaver sample and hold circuits as digitized signals which are periodically refreshed to maintain the stored voltage within a voltage window bracketing the original analog signal.",1990-08-07,"The title of the patent is state analog neural network and method of implementing same and its abstract is a neural network is implemented by discrete-time, continuous voltage state analog device in which neuron, synapse and synaptic strength signals are generated in highly parallel analog circuits in successive states from stored values of the interdependent signals calculated in a previous state. the neuron and synapse signals are refined in a relaxation loop while the synaptic strength signals are held constant. in learning modes, the synaptic strength signals are modified in successive states from stable values of the analog neuron signals. the analog signals are stored for as long as required in master/slaver sample and hold circuits as digitized signals which are periodically refreshed to maintain the stored voltage within a voltage window bracketing the original analog signal. dated 1990-08-07"
4950917,semiconductor cell for neural network employing a four-quadrant multiplier,"a synapse cell for use in providing a weighted connection strength is disclosed. the cell employs a four-quadrant multiplier and a pair of floating gate devices. various charge levels are programmed onto the floating gate devices, establishing weight and reference levels. these levels affect the current flowing through the multiplier. the output of the cell thus becomes a multiple of the input and the programmed charge difference.",1990-08-21,"The title of the patent is semiconductor cell for neural network employing a four-quadrant multiplier and its abstract is a synapse cell for use in providing a weighted connection strength is disclosed. the cell employs a four-quadrant multiplier and a pair of floating gate devices. various charge levels are programmed onto the floating gate devices, establishing weight and reference levels. these levels affect the current flowing through the multiplier. the output of the cell thus becomes a multiple of the input and the programmed charge difference. dated 1990-08-21"
4951239,artificial neural network implementation,"an artificial neural network having analog circuits for simultaneous parallel processing using individually variable synaptic input weights. the processing is implemented with a circuit adapted to vary the weight, which may be stored in a metal oxide field effect transistor, for teaching the network by addressing from outside the network or for hebbian or delta rule learning by the network itself.",1990-08-21,"The title of the patent is artificial neural network implementation and its abstract is an artificial neural network having analog circuits for simultaneous parallel processing using individually variable synaptic input weights. the processing is implemented with a circuit adapted to vary the weight, which may be stored in a metal oxide field effect transistor, for teaching the network by addressing from outside the network or for hebbian or delta rule learning by the network itself. dated 1990-08-21"
4954963,neural network and system,"neural network systems (100) with learning and recall are applied to clustered multiple-featured data (122,124,126) and analog data.",1990-09-04,"The title of the patent is neural network and system and its abstract is neural network systems (100) with learning and recall are applied to clustered multiple-featured data (122,124,126) and analog data. dated 1990-09-04"
4956564,adaptive synapse cell providing both excitatory and inhibitory connections in an associative network,"the present invention covers a synapse cell for providing a weighted connection between an input voltage line and an output summing line having an associated capacitance. connection between input and output lines in the associative network is made using one or more floating-gate transistors which provide both excitatory as well as inhibitory connections. as configured, each transistor's control gate is coupled to an input line and its drain is coupled to an output summing line. the floating-gate of the transistor is used for storing a charge which corresponds to the strength or weight of the neural connection. when a binary voltage pulse having a certain duration is applied to the control gate of the floating-gate transistor, a current is generated which acts to discharge the capacitance associated with the output summing line. the current, and therefore the resulting discharge, is directly proportional to the charge stored on the floating-gate member and the duration of the input pulse.",1990-09-11,"The title of the patent is adaptive synapse cell providing both excitatory and inhibitory connections in an associative network and its abstract is the present invention covers a synapse cell for providing a weighted connection between an input voltage line and an output summing line having an associated capacitance. connection between input and output lines in the associative network is made using one or more floating-gate transistors which provide both excitatory as well as inhibitory connections. as configured, each transistor's control gate is coupled to an input line and its drain is coupled to an output summing line. the floating-gate of the transistor is used for storing a charge which corresponds to the strength or weight of the neural connection. when a binary voltage pulse having a certain duration is applied to the control gate of the floating-gate transistor, a current is generated which acts to discharge the capacitance associated with the output summing line. the current, and therefore the resulting discharge, is directly proportional to the charge stored on the floating-gate member and the duration of the input pulse. dated 1990-09-11"
4958939,centering scheme for pattern recognition,the invention relates to a neural network centering scheme for translation-invariant pattern recognition. the scheme involves the centering of a pattern about its centroid to prepare it for subsequent subjugation to an associative match. the scheme is utilized in a camera assembly of the type used for image acquisition. movement of the camera assembly is controlled in accordance with the scheme to effect the centering of a pattern in the field of view window of the camera assembly.,1990-09-25,The title of the patent is centering scheme for pattern recognition and its abstract is the invention relates to a neural network centering scheme for translation-invariant pattern recognition. the scheme involves the centering of a pattern about its centroid to prepare it for subsequent subjugation to an associative match. the scheme is utilized in a camera assembly of the type used for image acquisition. movement of the camera assembly is controlled in accordance with the scheme to effect the centering of a pattern in the field of view window of the camera assembly. dated 1990-09-25
4959532,optical neural network and method,"an optical neural network stores optical transmission weightings as angularly and spatially distributed gratings within a phase conjugate mirror (pcm), the pcm using a stimulated process to generate a phase conjugated return beam without separate external pump mechanisms. an error signal is generated in response to differences between the actual and a desired output optical pattern, and is used to adjust the pcm gratings toward the desired output. one or more intermediate image planes may be employed along with the input and output planes. the input and intermediate planes, as well as the error signal, are preferably displayed on the surface of a spatial light modulator. the output optical signal is transduced into an electrical format for training the neural network; with the error signal also generated electrically. a significant increase in neuron and interconnection capacity is realized, without cross-talk between neurons, compared to prior optical neural networks.",1990-09-25,"The title of the patent is optical neural network and method and its abstract is an optical neural network stores optical transmission weightings as angularly and spatially distributed gratings within a phase conjugate mirror (pcm), the pcm using a stimulated process to generate a phase conjugated return beam without separate external pump mechanisms. an error signal is generated in response to differences between the actual and a desired output optical pattern, and is used to adjust the pcm gratings toward the desired output. one or more intermediate image planes may be employed along with the input and output planes. the input and intermediate planes, as well as the error signal, are preferably displayed on the surface of a spatial light modulator. the output optical signal is transduced into an electrical format for training the neural network; with the error signal also generated electrically. a significant increase in neuron and interconnection capacity is realized, without cross-talk between neurons, compared to prior optical neural networks. dated 1990-09-25"
4961002,synapse cell employing dual gate transistor structure,"a synapse cell for providing a weighted connection between an input voltage line and an output summing line having an associated capacitance. connection between input and output lines in the associative network is made using a dual-gate transistor. the transistor has a floating gate member for storing electrical charge, a pair of control gates coupled to a pair of input lines, and a drain coupled to an output summing line. the floating gate of the transistor is used for storing a charge which corresponds to the strength or weight of the neural connection. when a binary voltage pulse having a certain duration is applied to either one or both of the control gates of the transistor, a current is generated. this current acts to discharge the capacitance associated with the output summing line. furthermore, by employing a dual-gate structure, programming disturbance of neighboring devices in the network is practically eliminated.",1990-10-02,"The title of the patent is synapse cell employing dual gate transistor structure and its abstract is a synapse cell for providing a weighted connection between an input voltage line and an output summing line having an associated capacitance. connection between input and output lines in the associative network is made using a dual-gate transistor. the transistor has a floating gate member for storing electrical charge, a pair of control gates coupled to a pair of input lines, and a drain coupled to an output summing line. the floating gate of the transistor is used for storing a charge which corresponds to the strength or weight of the neural connection. when a binary voltage pulse having a certain duration is applied to either one or both of the control gates of the transistor, a current is generated. this current acts to discharge the capacitance associated with the output summing line. furthermore, by employing a dual-gate structure, programming disturbance of neighboring devices in the network is practically eliminated. dated 1990-10-02"
4961005,programmable neural circuit implementable in cmos very large scale integration,"the present invention is a neural network circuit including a plurality of neuron circuits. each neuron circuit has an input node for receiving an input signal, an output node for generating an output signal and a self-feedback control node for receiving a self-feedback signal. an interconnection device having an electrically controllable conductance is connected between the input nodes of each pair of neuron circuits. the neural network circuit is consequently programmable via the voltages applied to the self-feedback control nodes and the interconnection devices. such programmability permits the neural network circuit to store certain sets of desirable steady states. in the preferred embodiment the individual neuron circuits and the interconnection devices are constructed in very large scale integration cmos. thus this neural network circuit can be easily constructed with large numbers of neurons.",1990-10-02,"The title of the patent is programmable neural circuit implementable in cmos very large scale integration and its abstract is the present invention is a neural network circuit including a plurality of neuron circuits. each neuron circuit has an input node for receiving an input signal, an output node for generating an output signal and a self-feedback control node for receiving a self-feedback signal. an interconnection device having an electrically controllable conductance is connected between the input nodes of each pair of neuron circuits. the neural network circuit is consequently programmable via the voltages applied to the self-feedback control nodes and the interconnection devices. such programmability permits the neural network circuit to store certain sets of desirable steady states. in the preferred embodiment the individual neuron circuits and the interconnection devices are constructed in very large scale integration cmos. thus this neural network circuit can be easily constructed with large numbers of neurons. dated 1990-10-02"
4962342,dynamic synapse for neural network,an electronic circuit is disclosed having a sample/hold amplifier connected to an adaptive amplifier. a plurality of such electronic circuits may be configured in an array of rows and columns. an input voltage vector may be compared with an analog voltage vector stored in a row or column of the array and the stored vector closest to the applied input vector may be identified and further processed. the stored analog value may be read out of the synapse by applying a voltage to a read line. an array of the readable synapses may be provided and used in conjunction with a dummy synapse to compensate for an error offset introduced by the operating characteristics of the synapses.,1990-10-09,The title of the patent is dynamic synapse for neural network and its abstract is an electronic circuit is disclosed having a sample/hold amplifier connected to an adaptive amplifier. a plurality of such electronic circuits may be configured in an array of rows and columns. an input voltage vector may be compared with an analog voltage vector stored in a row or column of the array and the stored vector closest to the applied input vector may be identified and further processed. the stored analog value may be read out of the synapse by applying a voltage to a read line. an array of the readable synapses may be provided and used in conjunction with a dummy synapse to compensate for an error offset introduced by the operating characteristics of the synapses. dated 1990-10-09
4963725,adaptive optical neural network,"an adaptive optical network is provided for the implementation of learning algorithms. the network comprises a double mach-zehnder interferometer in conjunction with a photorefractive crystal that functions as a holographic medium. light from selectable sources on opposite sides of a beamsplitter is passed through the interferometer, at least one arm of which includes a spatial light modulator for imprinting a data pattern on the light. the light is directed into the holographic medium to develop a refractive index grating corresponding to the data pattern. light from the hologram is sensed by a photodetector that provides a signal to a threshold device. the output of the threshold device is compared with a reference signal to produce an error signal that can be used to select the source of light directed through the network. the interconnections of the optical devices function to compute the inner product between the elements of the data pattern and their weight factors. selecting the light source and changing the data pattern provide additive and subtractive weight change capability for implementing various learning algorithms.",1990-10-16,"The title of the patent is adaptive optical neural network and its abstract is an adaptive optical network is provided for the implementation of learning algorithms. the network comprises a double mach-zehnder interferometer in conjunction with a photorefractive crystal that functions as a holographic medium. light from selectable sources on opposite sides of a beamsplitter is passed through the interferometer, at least one arm of which includes a spatial light modulator for imprinting a data pattern on the light. the light is directed into the holographic medium to develop a refractive index grating corresponding to the data pattern. light from the hologram is sensed by a photodetector that provides a signal to a threshold device. the output of the threshold device is compared with a reference signal to produce an error signal that can be used to select the source of light directed through the network. the interconnections of the optical devices function to compute the inner product between the elements of the data pattern and their weight factors. selecting the light source and changing the data pattern provide additive and subtractive weight change capability for implementing various learning algorithms. dated 1990-10-16"
4965443,focus detection apparatus using neural network means,"an optical image transmitted through a photographing lens is incident on a light-receiving unit of a two-dimensional matrix. an output from the light-receiving unit is input to a first arithmetic logic unit, and the first arithmetic logic unit calculates actual object brightness values in consideration of an aperture value of an aperture. an output from the first arithmetic logic unit is supplied to a multiplexer and a neural network. the neural network determines a main part of the object from a pattern of brightness values of the respective photoelectric transducer elements and outputs a position signal of the main part. the multiplexer selectively passes the brightness value of the photoelectric transducer element corresponding to the main part of the object from the outputs generated by the first arithmetic logic unit. an output from the multiplexer is supplied to a second arithmetic logic unit. the second arithmetic logic unit performs a focus detection calculation based on only the brightness of the main part. the photographing lens is moved along the optical axis, thereby performing a focusing operation.",1990-10-23,"The title of the patent is focus detection apparatus using neural network means and its abstract is an optical image transmitted through a photographing lens is incident on a light-receiving unit of a two-dimensional matrix. an output from the light-receiving unit is input to a first arithmetic logic unit, and the first arithmetic logic unit calculates actual object brightness values in consideration of an aperture value of an aperture. an output from the first arithmetic logic unit is supplied to a multiplexer and a neural network. the neural network determines a main part of the object from a pattern of brightness values of the respective photoelectric transducer elements and outputs a position signal of the main part. the multiplexer selectively passes the brightness value of the photoelectric transducer element corresponding to the main part of the object from the outputs generated by the first arithmetic logic unit. an output from the multiplexer is supplied to a second arithmetic logic unit. the second arithmetic logic unit performs a focus detection calculation based on only the brightness of the main part. the photographing lens is moved along the optical axis, thereby performing a focusing operation. dated 1990-10-23"
4965725,neural network based automated cytological specimen classification system and method,an automated screening system and method for cytological specimen classification in which a neural network is utilized in performance of the classification function. also included is an automated microscope and associated image processing circuitry.,1990-10-23,The title of the patent is neural network based automated cytological specimen classification system and method and its abstract is an automated screening system and method for cytological specimen classification in which a neural network is utilized in performance of the classification function. also included is an automated microscope and associated image processing circuitry. dated 1990-10-23
4970819,firearm safety system and method,actuation of the firing mechanism of a firearm is prevented until grip pattern sensing means on the handgrip of the firearm supply to a microprocessor signals corresponding to a grip pattern stored in a programmed simulated neural network memory. all of these components are contained within the firearm. programming of the neural network memory is accomplished by using a host computer with a simulated neural network to train that network to recognize a particular grip pattern using grip pattern signals generated by the grip pattern sensing means as the sensing means is repeatedly gripped for the person for whom the firearm is to be programmed.,1990-11-20,The title of the patent is firearm safety system and method and its abstract is actuation of the firing mechanism of a firearm is prevented until grip pattern sensing means on the handgrip of the firearm supply to a microprocessor signals corresponding to a grip pattern stored in a programmed simulated neural network memory. all of these components are contained within the firearm. programming of the neural network memory is accomplished by using a host computer with a simulated neural network to train that network to recognize a particular grip pattern using grip pattern signals generated by the grip pattern sensing means as the sensing means is repeatedly gripped for the person for whom the firearm is to be programmed. dated 1990-11-20
4972187,numeric encoding method and apparatus for neural networks,"a numeric encoding method and apparatus for neural networks, encodes numeric input data into a form applicable to an input of a neural network by partitioning a binary input into n-bit input segments, each of which is replaced with a code having m adjacent logic ones and 2.sup.n -1 logic zeros, the bit position of the least significant of the m logic ones corresponding to the binary value of the input segment it replaces. the codes are concatenated to form an encoded input. a decoding method decodes an output from the neural network into a binary form by partitioning the output into output segments having 2.sup.n +m-1 bits each, each of which is replaced with an n-bit binary segment being a bracketed weighted average of the significances of logic ones present in the output segment. the binary segments are concatenated to form a decoded output.",1990-11-20,"The title of the patent is numeric encoding method and apparatus for neural networks and its abstract is a numeric encoding method and apparatus for neural networks, encodes numeric input data into a form applicable to an input of a neural network by partitioning a binary input into n-bit input segments, each of which is replaced with a code having m adjacent logic ones and 2.sup.n -1 logic zeros, the bit position of the least significant of the m logic ones corresponding to the binary value of the input segment it replaces. the codes are concatenated to form an encoded input. a decoding method decodes an output from the neural network into a binary form by partitioning the output into output segments having 2.sup.n +m-1 bits each, each of which is replaced with an n-bit binary segment being a bracketed weighted average of the significances of logic ones present in the output segment. the binary segments are concatenated to form a decoded output. dated 1990-11-20"
4972363,neural network using stochastic processing,"an apparatus and method for implementing a neural network having n nodes coupled to one another by interconnections having interconnect weights t.sub.ij that quantify the influence of node j on node i. the apparatus comprises a node circuit for each node and a data processor. the data processor receives one or more library members, and transmits the interconnect weights to the node circuits. the data processor also stores a current state vector, and receives input data representing a library member to be retrieved. the data processor then performs an iteration in which the current state vector is sent to the node circuits, and an updated state vector is received from the node circuits, the iteration being commenced by setting the current state vector equal to the input data. each node circuit comprises one or more stochastic processors for multiplying the state vector elements by the corresponding interconnect weights, to determine the updated state vector. each stochastic processor preferably includes means for generating a pseudorandom sequence of numbers, and using such sequence to encode the interconnect weights and state vector elements into stochastic input signals that are then multiplied by a stochastic multiplier comprising delay means and an and gate.",1990-11-20,"The title of the patent is neural network using stochastic processing and its abstract is an apparatus and method for implementing a neural network having n nodes coupled to one another by interconnections having interconnect weights t.sub.ij that quantify the influence of node j on node i. the apparatus comprises a node circuit for each node and a data processor. the data processor receives one or more library members, and transmits the interconnect weights to the node circuits. the data processor also stores a current state vector, and receives input data representing a library member to be retrieved. the data processor then performs an iteration in which the current state vector is sent to the node circuits, and an updated state vector is received from the node circuits, the iteration being commenced by setting the current state vector equal to the input data. each node circuit comprises one or more stochastic processors for multiplying the state vector elements by the corresponding interconnect weights, to determine the updated state vector. each stochastic processor preferably includes means for generating a pseudorandom sequence of numbers, and using such sequence to encode the interconnect weights and state vector elements into stochastic input signals that are then multiplied by a stochastic multiplier comprising delay means and an and gate. dated 1990-11-20"
4972473,data communication method and apparatus using neural-networks,"a data communication apparatus comprises: means for dividing data to be transmitted into a plurality of blocks and extracting the data from each block; a first multi-layered neural network of three or more layers which has weighting coefficients to output the same data as the input data for the data extracted from each block and which can output data from an intermediate layer; the transmission data extracted from each block being inputted to the first neural network and outputted from the intermediate layer; means for encoding the transmission data which is outputted from the intermediate layer of the first neural network and, thereafter, transmitting; means for receiving and decoding the transmitted data; a second multi-layered neural network of three or more layers which has the same weight coefficients as those of the first neural network and can input data from an intermediate layer; the decoded data of each block being inputted to the second neural network and outputted from an output layer; and means for restoring the data on the basis of the output data from the output layer of the second neural network.",1990-11-20,"The title of the patent is data communication method and apparatus using neural-networks and its abstract is a data communication apparatus comprises: means for dividing data to be transmitted into a plurality of blocks and extracting the data from each block; a first multi-layered neural network of three or more layers which has weighting coefficients to output the same data as the input data for the data extracted from each block and which can output data from an intermediate layer; the transmission data extracted from each block being inputted to the first neural network and outputted from the intermediate layer; means for encoding the transmission data which is outputted from the intermediate layer of the first neural network and, thereafter, transmitting; means for receiving and decoding the transmitted data; a second multi-layered neural network of three or more layers which has the same weight coefficients as those of the first neural network and can input data from an intermediate layer; the decoded data of each block being inputted to the second neural network and outputted from an output layer; and means for restoring the data on the basis of the output data from the output layer of the second neural network. dated 1990-11-20"
4974169,neural network with memory cycling,"an information processing system and method to calculate output values for a group of neurons. the method comprises transmitting input values for the neurons to a memory unit of a processing section, and then calculating a multitude of series of neuron output values over a multitude of cycles. during a first period of each cycle, a first series of neuron output values are calculated from neuron input values stored in a first memory area of the memory unit; and during a second period of each cycle, a second series of neuron output values are calculated from neuron input values stored in a second memory area of the memory unit. the transmitting step includes the steps of storing in the first memory area of the memory unit, neuron input values transmitted to the memory unit during the period immediately preceding the first period of each cycle; and storing in the second memory area of the memory unit neuron input values transmitted to the memory unit, during the first period of each cycle.",1990-11-27,"The title of the patent is neural network with memory cycling and its abstract is an information processing system and method to calculate output values for a group of neurons. the method comprises transmitting input values for the neurons to a memory unit of a processing section, and then calculating a multitude of series of neuron output values over a multitude of cycles. during a first period of each cycle, a first series of neuron output values are calculated from neuron input values stored in a first memory area of the memory unit; and during a second period of each cycle, a second series of neuron output values are calculated from neuron input values stored in a second memory area of the memory unit. the transmitting step includes the steps of storing in the first memory area of the memory unit, neuron input values transmitted to the memory unit during the period immediately preceding the first period of each cycle; and storing in the second memory area of the memory unit neuron input values transmitted to the memory unit, during the first period of each cycle. dated 1990-11-27"
4975961,multi-layer neural network to which dynamic programming techniques are applicable,"in a neural network, input neuron units of an input layer are grouped into first through j-th input layer frames, where j represents a predetermined natural number. intermediate neuron units of an intermediate layer are grouped into first through j-th intermediate layer frames. an output layer comprises an output neuron unit. each intermediate neuron unit of a j-th intermediate layer frame is connected to the input neuron units of j'-th input layer frames, where j is variable between 1 and j and j' represents at least two consecutive integers, one of which is equal to j and at least one other of which is less than j. each output neuron unit is connected to the intermediate neuron units of the intermediate layer. for recognition of an input pattern represented by a time sequence of feature vectors, each consisting of k vector components, where k represents a predetermined positive integer, each input layer frame consists of k input neuron units. each intermediate layer frame consists of m intermediate neuron units, where m represents a positive integer which is less than k. the vector components of each feature vector are supplied to the respective input neuron units of one of the input layer frames that is preferably selected from three consecutively numbered input layer frames. the neural network is readily trained to make a predetermined one of the output neuron units produce an output signal indicative of the input pattern and can be implemented by a microprocessor.",1990-12-04,"The title of the patent is multi-layer neural network to which dynamic programming techniques are applicable and its abstract is in a neural network, input neuron units of an input layer are grouped into first through j-th input layer frames, where j represents a predetermined natural number. intermediate neuron units of an intermediate layer are grouped into first through j-th intermediate layer frames. an output layer comprises an output neuron unit. each intermediate neuron unit of a j-th intermediate layer frame is connected to the input neuron units of j'-th input layer frames, where j is variable between 1 and j and j' represents at least two consecutive integers, one of which is equal to j and at least one other of which is less than j. each output neuron unit is connected to the intermediate neuron units of the intermediate layer. for recognition of an input pattern represented by a time sequence of feature vectors, each consisting of k vector components, where k represents a predetermined positive integer, each input layer frame consists of k input neuron units. each intermediate layer frame consists of m intermediate neuron units, where m represents a positive integer which is less than k. the vector components of each feature vector are supplied to the respective input neuron units of one of the input layer frames that is preferably selected from three consecutively numbered input layer frames. the neural network is readily trained to make a predetermined one of the output neuron units produce an output signal indicative of the input pattern and can be implemented by a microprocessor. dated 1990-12-04"
4978990,exposure control apparatus for camera,"an optical image of an object is incident on a light-receiving unit of a two-dimensional matrix through a photographing lens. an output from the light-receiving unit is input to a first arithmetic logic unit to calculate an actual object brightness value in consideration of an aperture value of an aperture. an output from the first arithmetic logic unit is input to a multiplexer and a neural network. the neural network determines a main part of the object from brightness value pattern as a set of brightness values of the photoelectric transducer elements and outputs a position signal representing the main part. a multiplexer selectively passes only brightness values of the photoelectric transducer elements corresponding to the main part of the object from the outputs from the first arithmetic logic unit. an output from the multiplexer is supplied to a second arithmetic logic unit, and the second arithmetic logic unit calculates an apex calculation on the basis of a speed value, an aperture value, a time value, and a mode signal representing a shutter or aperture priority operation, thereby determining a shutter speed or an f-number.",1990-12-18,"The title of the patent is exposure control apparatus for camera and its abstract is an optical image of an object is incident on a light-receiving unit of a two-dimensional matrix through a photographing lens. an output from the light-receiving unit is input to a first arithmetic logic unit to calculate an actual object brightness value in consideration of an aperture value of an aperture. an output from the first arithmetic logic unit is input to a multiplexer and a neural network. the neural network determines a main part of the object from brightness value pattern as a set of brightness values of the photoelectric transducer elements and outputs a position signal representing the main part. a multiplexer selectively passes only brightness values of the photoelectric transducer elements corresponding to the main part of the object from the outputs from the first arithmetic logic unit. an output from the multiplexer is supplied to a second arithmetic logic unit, and the second arithmetic logic unit calculates an apex calculation on the basis of a speed value, an aperture value, a time value, and a mode signal representing a shutter or aperture priority operation, thereby determining a shutter speed or an f-number. dated 1990-12-18"
4979126,neural network with non-linear transformations,"a neural network system includes means for accomplishing artificial intelligence functions in three formerly divergent implementations. these functions include: supervised learning, unsupervised learning, and associative memory storage and retrieval. the subject neural network is created by addition of a non-linear layer to a more standard neural network architecture. the non-linear layer functions to expand a functional input space to a signal set including orthonormal elements, when the input signal is visualized as a vector representation. an input signal is selectively passed to a non-linear transform circuit, which outputs a transform signal therefrom. both the input signal and the transform signal are placed in communication with a first layer of a plurality of processing nodes. an improved hardware implementation of the subject system includes a highly parallel, hybrid analog/digital circuitry. included therein is a digitally addressed, random access memory means for storage and retrieval of an analog signal.",1990-12-18,"The title of the patent is neural network with non-linear transformations and its abstract is a neural network system includes means for accomplishing artificial intelligence functions in three formerly divergent implementations. these functions include: supervised learning, unsupervised learning, and associative memory storage and retrieval. the subject neural network is created by addition of a non-linear layer to a more standard neural network architecture. the non-linear layer functions to expand a functional input space to a signal set including orthonormal elements, when the input signal is visualized as a vector representation. an input signal is selectively passed to a non-linear transform circuit, which outputs a transform signal therefrom. both the input signal and the transform signal are placed in communication with a first layer of a plurality of processing nodes. an improved hardware implementation of the subject system includes a highly parallel, hybrid analog/digital circuitry. included therein is a digitally addressed, random access memory means for storage and retrieval of an analog signal. dated 1990-12-18"
4988891,semiconductor neural network including photosensitive coupling elements,"a semiconductor neural network constructed in accordance with models of vital nerve cells has photosensitive elements as coupling elements providing degrees of coupling between neurons which are modeled vital nerve cells. the conductance values of the photosensitive elements can be set by light. due to such structure, not only the degrees of coupling of all the coupling elements can be simultaneously programmed but signal lines for programming the degrees of coupling can be eliminated in the network, whereby a semiconductor neural network having a high degree of integration can be implemented without additional complicating fabrication steps.",1991-01-29,"The title of the patent is semiconductor neural network including photosensitive coupling elements and its abstract is a semiconductor neural network constructed in accordance with models of vital nerve cells has photosensitive elements as coupling elements providing degrees of coupling between neurons which are modeled vital nerve cells. the conductance values of the photosensitive elements can be set by light. due to such structure, not only the degrees of coupling of all the coupling elements can be simultaneously programmed but signal lines for programming the degrees of coupling can be eliminated in the network, whereby a semiconductor neural network having a high degree of integration can be implemented without additional complicating fabrication steps. dated 1991-01-29"
4990838,movement trajectory generating method of a dynamical system,"a movement trajectory generating system of a dynamical system uses neural network units (1, 2, 3) including cascade connection of a first layer (11, 21, 31), a second layer (12, 22, 32), a third layer (13, 23, 33) and a fourth layer (14, 24, 34), to learn a vector field of differential equations indicating forward dynamics of a controlled object (4). conditions concerning trajectories of a final point and a via-point of movement of the controlled object and locations of obstacles are given from a motor center (5). while smoothness of movement is ensured by couplings of electric synapses using errors with respect to those conditions as total energy, least dissipation of energy is attained, whereby trajectory formation and control input for realizing the trajectory are obtained simultaneously.",1991-02-05,"The title of the patent is movement trajectory generating method of a dynamical system and its abstract is a movement trajectory generating system of a dynamical system uses neural network units (1, 2, 3) including cascade connection of a first layer (11, 21, 31), a second layer (12, 22, 32), a third layer (13, 23, 33) and a fourth layer (14, 24, 34), to learn a vector field of differential equations indicating forward dynamics of a controlled object (4). conditions concerning trajectories of a final point and a via-point of movement of the controlled object and locations of obstacles are given from a motor center (5). while smoothness of movement is ensured by couplings of electric synapses using errors with respect to those conditions as total energy, least dissipation of energy is attained, whereby trajectory formation and control input for realizing the trajectory are obtained simultaneously. dated 1991-02-05"
4994982,neural network system and circuit for use therein,a neural network system comprises a memory for storing in binary code the synaptic coefficients indicative of the interconnections among the neurons. means are provided for simultaneously supplying all the synaptic coefficients associated with a given neuron. digital multipliers are provided for determining the product of the supplied synaptic coefficients and the relevant neuron states of the neurons connected to said given neuron. the multipliers deliver their results into an adder tree for determining the sum of the products. as a result of the parallel architecture of the system high operating speeds are attainable. the modular architecture enables extension of the system.,1991-02-19,The title of the patent is neural network system and circuit for use therein and its abstract is a neural network system comprises a memory for storing in binary code the synaptic coefficients indicative of the interconnections among the neurons. means are provided for simultaneously supplying all the synaptic coefficients associated with a given neuron. digital multipliers are provided for determining the product of the supplied synaptic coefficients and the relevant neuron states of the neurons connected to said given neuron. the multipliers deliver their results into an adder tree for determining the sum of the products. as a result of the parallel architecture of the system high operating speeds are attainable. the modular architecture enables extension of the system. dated 1991-02-19
4995088,super resolution,"data analysis systems are provided, especially target imaging and identification systems, which utilize a cam that associatively stores a plurality of known data sets such as target data sets in a synaptic interconnectivity matrix modeled upon the model of learning of neural networks. in accordance with preferred embodiments the systems are able to identify unknown objects when only a partial data set from the object is available. the system is robust and fast, utilizing parallel processing due to the massive interconnectivity of neural elements so that the image produced exhibits the properties of super-resolution. since the system is modeled after a neural network, it is fault tolerant and highly reliable.",1991-02-19,"The title of the patent is super resolution and its abstract is data analysis systems are provided, especially target imaging and identification systems, which utilize a cam that associatively stores a plurality of known data sets such as target data sets in a synaptic interconnectivity matrix modeled upon the model of learning of neural networks. in accordance with preferred embodiments the systems are able to identify unknown objects when only a partial data set from the object is available. the system is robust and fast, utilizing parallel processing due to the massive interconnectivity of neural elements so that the image produced exhibits the properties of super-resolution. since the system is modeled after a neural network, it is fault tolerant and highly reliable. dated 1991-02-19"
4996648,neural network using random binary code,"long and short term memory equations for neural networks are implemented by means of exchange of signals which carry information in the form of both binary and continuously modulated energy emissions. in one embodiment, array of parallel processors exhibits behavior of cooperative-competitive neural networks. parallel bus interconnections and digital and analog processing of analog information contained in the exchanged energy emissions are employed with generally local synchronization of the processors. energy emission and detection is modulated as a function of a random code.",1991-02-26,"The title of the patent is neural network using random binary code and its abstract is long and short term memory equations for neural networks are implemented by means of exchange of signals which carry information in the form of both binary and continuously modulated energy emissions. in one embodiment, array of parallel processors exhibits behavior of cooperative-competitive neural networks. parallel bus interconnections and digital and analog processing of analog information contained in the exchanged energy emissions are employed with generally local synchronization of the processors. energy emission and detection is modulated as a function of a random code. dated 1991-02-26"
4999525,exclusive-or cell for pattern matching employing floating gate devices,a semiconductor cell for producing an output current that is related to the match between an input vector pattern and a weighting pattern is described. the cell is particularly useful as a synapse cell within a neural network to perform pattern recognition tasks. the cell includes a pair of input lines for receiving a differential input vector element value and a pair of output lines for providing a difference current to a current summing neural amplifier. a plurality of floating gate devices each having a floating gate member are employed in the synapse cell to store charge in accordance with a predetermined weight pattern. each of the floating gate devices is uniquely coupled to a combination of an output current line and an input voltage line such that the difference current provided to the neural amplifier is related to the match between the input vector and the stored weight.,1991-03-12,The title of the patent is exclusive-or cell for pattern matching employing floating gate devices and its abstract is a semiconductor cell for producing an output current that is related to the match between an input vector pattern and a weighting pattern is described. the cell is particularly useful as a synapse cell within a neural network to perform pattern recognition tasks. the cell includes a pair of input lines for receiving a differential input vector element value and a pair of output lines for providing a difference current to a current summing neural amplifier. a plurality of floating gate devices each having a floating gate member are employed in the synapse cell to store charge in accordance with a predetermined weight pattern. each of the floating gate devices is uniquely coupled to a combination of an output current line and an input voltage line such that the difference current provided to the neural amplifier is related to the match between the input vector and the stored weight. dated 1991-03-12
5003490,neural network signal processor,""" a neural network signal processor (nsp) (20) that can accept, as input, unprocessed signals (32), such as those directly from a sensor. consecutive portions of the input waveform are directed simultaneously to input processing units, or """"neurons"""" (22). each portion of the input waveform (32) advances through the input neurons (22) until each neuron receives the entire waveform (32). during a training procedure, the nsp 20 receives a training waveform (30) and connective weights, or """"synapses"""" (28) between the neurons are adjusted until a desired output is produced. the nsp (20) is trained to produce a single response while each portion of the input waveform is received by the input neurons (22). once trained, when an unknown waveform (32) is received by the nsp (20), it will respond with the desired output when the unknown waveform (32) contains some form of the training waveform (30). """,1991-03-26,"The title of the patent is neural network signal processor and its abstract is "" a neural network signal processor (nsp) (20) that can accept, as input, unprocessed signals (32), such as those directly from a sensor. consecutive portions of the input waveform are directed simultaneously to input processing units, or """"neurons"""" (22). each portion of the input waveform (32) advances through the input neurons (22) until each neuron receives the entire waveform (32). during a training procedure, the nsp 20 receives a training waveform (30) and connective weights, or """"synapses"""" (28) between the neurons are adjusted until a desired output is produced. the nsp (20) is trained to produce a single response while each portion of the input waveform is received by the input neurons (22). once trained, when an unknown waveform (32) is received by the nsp (20), it will respond with the desired output when the unknown waveform (32) contains some form of the training waveform (30). "" dated 1991-03-26"
5004309,neural processor with holographic optical paths and nonlinear operating means,"an optical apparatus for simulating a highly interconnected neural network is disclosed as including a spatial light modulator (slm), an inputting device, a laser, a detecting device, and a page-oriented hologaphic component. the inputting device applies input signals to the slm. the holographic component optically interconnects n.sup.2 pixels defined on the spatial light modulator to n.sup.2 pixels defined on a detecting surface of the detecting device. the interconnections are made by n.sup.2 patterns of up to n.sup.2 interconnection weight encoded beams projected by n.sup.2 planar, or essentially two-dimensional, holograms arranged in a spatially localized array within the holographic component. the slm modulates the encoded beams and directs them onto the detecting surface wherein a parameter of the beams is evaluated at each pixel thereof. the evaluated parameter is transformed according to a nonlinear threshold function to provide transformed signals which can be fed back to the slm for further iterations.",1991-04-02,"The title of the patent is neural processor with holographic optical paths and nonlinear operating means and its abstract is an optical apparatus for simulating a highly interconnected neural network is disclosed as including a spatial light modulator (slm), an inputting device, a laser, a detecting device, and a page-oriented hologaphic component. the inputting device applies input signals to the slm. the holographic component optically interconnects n.sup.2 pixels defined on the spatial light modulator to n.sup.2 pixels defined on a detecting surface of the detecting device. the interconnections are made by n.sup.2 patterns of up to n.sup.2 interconnection weight encoded beams projected by n.sup.2 planar, or essentially two-dimensional, holograms arranged in a spatially localized array within the holographic component. the slm modulates the encoded beams and directs them onto the detecting surface wherein a parameter of the beams is evaluated at each pixel thereof. the evaluated parameter is transformed according to a nonlinear threshold function to provide transformed signals which can be fed back to the slm for further iterations. dated 1991-04-02"
5004932,unit circuit for constructing a neural network and a semiconductor integrated circuit having the same,""" a semiconductor integrated circuit for constructing a neural network model, comprising a differential amplifier which includes one output terminal and two input terminals, an excitatory synapse circuit which is connected to the noninverting input terminal of said differential amplifier, and an inhibitory synapse circuit which is connected to the inverting input terminal of said differential amplifier, wherein each of said excitatory and inhibitory synapse circuits includes a plurality of current switches, regulated current source ciruits which are equal in number to said current switches and which determine currents to flow through said current switches, and one load resistor which is connected to all of said current switches, input terminals of said each synapse circuit being constructed of terminals which turn """"on"""" and """"off"""" the respective current switches and to which external inputs or outputs of another neural circuit are connected, said each regulated current source circuit being constructed of a circuit whose current value can be increased or decreased by a voltage externally applied separately and as to which a value of the voltage for increasing or decreasing the current value corresponds to a synaptic weight. """,1991-04-02,"The title of the patent is unit circuit for constructing a neural network and a semiconductor integrated circuit having the same and its abstract is "" a semiconductor integrated circuit for constructing a neural network model, comprising a differential amplifier which includes one output terminal and two input terminals, an excitatory synapse circuit which is connected to the noninverting input terminal of said differential amplifier, and an inhibitory synapse circuit which is connected to the inverting input terminal of said differential amplifier, wherein each of said excitatory and inhibitory synapse circuits includes a plurality of current switches, regulated current source ciruits which are equal in number to said current switches and which determine currents to flow through said current switches, and one load resistor which is connected to all of said current switches, input terminals of said each synapse circuit being constructed of terminals which turn """"on"""" and """"off"""" the respective current switches and to which external inputs or outputs of another neural circuit are connected, said each regulated current source circuit being constructed of a circuit whose current value can be increased or decreased by a voltage externally applied separately and as to which a value of the voltage for increasing or decreasing the current value corresponds to a synaptic weight. "" dated 1991-04-02"
5005206,method of and arrangement for image data compression by means of a neural network,"method of and arrangement for image data compression by vector quantization in accordance with a precoding in blocks, thereafter comparing by means of a neural network precoded blocks with reference words stored in the form of a code book so as to transmit selected indices to a receiver. in accordance with the method, the neural network effects a learning phase with prescribed prototypes, thereafter with the aid of test vectors originating from the image generates an adaptive code book which is transmitted to the receiver. this adaptation utilizes attractors, which may be induced metastable states, of the neural network, and which are submitted to an optimizing procedure. the arrangement can process images with a view to their storage. it is also possible to utilize two devices which operate alternately, one device for generating the adaptive code book and the other one to utilize it with the object of processing television pictures in real time.",1991-04-02,"The title of the patent is method of and arrangement for image data compression by means of a neural network and its abstract is method of and arrangement for image data compression by vector quantization in accordance with a precoding in blocks, thereafter comparing by means of a neural network precoded blocks with reference words stored in the form of a code book so as to transmit selected indices to a receiver. in accordance with the method, the neural network effects a learning phase with prescribed prototypes, thereafter with the aid of test vectors originating from the image generates an adaptive code book which is transmitted to the receiver. this adaptation utilizes attractors, which may be induced metastable states, of the neural network, and which are submitted to an optimizing procedure. the arrangement can process images with a view to their storage. it is also possible to utilize two devices which operate alternately, one device for generating the adaptive code book and the other one to utilize it with the object of processing television pictures in real time. dated 1991-04-02"
5008833,parallel optoelectronic neural network processors,"several embodiments of neural processors implemented on a vlsi circuit chip are disclosed, all of which are capable of entering a matrix t into an array of photosensitive devices which may be charge coupled or charge injection devices (ccd or cid). using ccd's to receive and store the synapses of the matrix t from a spatial light modulator, or other optical means of projecting an array of pixels, semiparallel synchronous operation is achieved. using cid's, full parallel synchronous operation is achieved. and using phototransistors to receive the array of pixels, full parallel and asynchronous operation is achieved. in the latter case, the source of the pixel matrix must provide the memory necessary for the matrix t. in the other cases, the source of the pixel matrix may be turned off after the matrix t has been entered and stored by the ccd's or cid's.",1991-04-16,"The title of the patent is parallel optoelectronic neural network processors and its abstract is several embodiments of neural processors implemented on a vlsi circuit chip are disclosed, all of which are capable of entering a matrix t into an array of photosensitive devices which may be charge coupled or charge injection devices (ccd or cid). using ccd's to receive and store the synapses of the matrix t from a spatial light modulator, or other optical means of projecting an array of pixels, semiparallel synchronous operation is achieved. using cid's, full parallel synchronous operation is achieved. and using phototransistors to receive the array of pixels, full parallel and asynchronous operation is achieved. in the latter case, the source of the pixel matrix must provide the memory necessary for the matrix t. in the other cases, the source of the pixel matrix may be turned off after the matrix t has been entered and stored by the ccd's or cid's. dated 1991-04-16"
5010512,neural network having an associative memory that learns by example,"a neural network utilizing the threshold characteristics of a semiconductor device as the various memory elements of the network. each memory element comprises a complementary pair of mosfets in which the threshold voltage is adjusted as a function of the input voltage to the element. the network is able to learn by example using a local learning algorithm. the network includes a series of output amplifiers in which the output is provided by the sum of the outputs of a series of learning elements coupled to the amplifier. the output of each learning element is the difference between the input signal to each learning element and an individual learning threshold at each input. the learning is accomplished by charge trapping in the insulator of each individual input mosfet pair. the thresholds of each transistor automatically adjust to both the input and output voltages to learn the desired state. after input patterns have been learned by the network, the learning functions is set to zero so that the thresholds remain constant and the network will come to an equilibrium state under the influence of a test input pattern thereby providing, as an output, the learned pattern most closely resembling the test input pattern.",1991-04-23,"The title of the patent is neural network having an associative memory that learns by example and its abstract is a neural network utilizing the threshold characteristics of a semiconductor device as the various memory elements of the network. each memory element comprises a complementary pair of mosfets in which the threshold voltage is adjusted as a function of the input voltage to the element. the network is able to learn by example using a local learning algorithm. the network includes a series of output amplifiers in which the output is provided by the sum of the outputs of a series of learning elements coupled to the amplifier. the output of each learning element is the difference between the input signal to each learning element and an individual learning threshold at each input. the learning is accomplished by charge trapping in the insulator of each individual input mosfet pair. the thresholds of each transistor automatically adjust to both the input and output voltages to learn the desired state. after input patterns have been learned by the network, the learning functions is set to zero so that the thresholds remain constant and the network will come to an equilibrium state under the influence of a test input pattern thereby providing, as an output, the learned pattern most closely resembling the test input pattern. dated 1991-04-23"
5014096,optoelectronic integrated circuit with optical gate device and phototransistor,"an optoelectronic integrated circuit including an optical bistable circuit comprises: an optical gate device responsive to a current injected to an active layer thereof and to a first ray transmitted through the active layer for emitting first and second light rays and for controlling intensity of the first light ray in accordance with the current; and a first phototransistor serially connected with the optical gate device so arranged to receive the second light ray for causing the current to flow through the optical gate device in response to the second light ray and a set signal light ray, the first phototransistor holding flowing of the current when the second light ray is emitted. this circuit can control the first light ray incident to the optical gate device in response to a set signal light ray applied to the first phototransistor. a second phototransistor may be included for stopping emission of light by the optical gate device in response to a reset signal light ray. such a circuit can be used in an optical neural network as a light-switching device. the first light ray is applied to an optical gate device perpendicularly or horizontally with respect to the plane of the substrate thereof. the second light ray may be emitted by a light-emitting device serially connected with the optical gate.",1991-05-07,"The title of the patent is optoelectronic integrated circuit with optical gate device and phototransistor and its abstract is an optoelectronic integrated circuit including an optical bistable circuit comprises: an optical gate device responsive to a current injected to an active layer thereof and to a first ray transmitted through the active layer for emitting first and second light rays and for controlling intensity of the first light ray in accordance with the current; and a first phototransistor serially connected with the optical gate device so arranged to receive the second light ray for causing the current to flow through the optical gate device in response to the second light ray and a set signal light ray, the first phototransistor holding flowing of the current when the second light ray is emitted. this circuit can control the first light ray incident to the optical gate device in response to a set signal light ray applied to the first phototransistor. a second phototransistor may be included for stopping emission of light by the optical gate device in response to a reset signal light ray. such a circuit can be used in an optical neural network as a light-switching device. the first light ray is applied to an optical gate device perpendicularly or horizontally with respect to the plane of the substrate thereof. the second light ray may be emitted by a light-emitting device serially connected with the optical gate. dated 1991-05-07"
5014219,mask controled neural networks,"a mask neutral network for processing that allows an external source of control to continuously direct state transition of the neural network toward selected states and away from other states. the network, through externally controlled masking, can focus attention on selected attributes of observed data, solutions or results. the masking is appliciable across three major categories of networks in that it facilitates augmented recall, directed learning and constrained optimization.",1991-05-07,"The title of the patent is mask controled neural networks and its abstract is a mask neutral network for processing that allows an external source of control to continuously direct state transition of the neural network toward selected states and away from other states. the network, through externally controlled masking, can focus attention on selected attributes of observed data, solutions or results. the masking is appliciable across three major categories of networks in that it facilitates augmented recall, directed learning and constrained optimization. dated 1991-05-07"
5016188,discrete-time optimal control by neural network,"a neural network determines optimal control inputs for a linear quadratic discrete-time process at m sampling times, the process being characterized by a quadratic cost function, p state variables, and r control variables. the network includes n=(p+r)m neurons, a distinct neuron being assigned to represent the value of each state variable at each sampling time and a distinct neuron being assigned to represent the value of each control variable at each sampling time. an input bias connected to each neuron has a value determined by the quandratic cost function for the variable represented by the neuron. selected connections are provided between the output of each neuron and the input of selected other neurons in the network, each such connection and the strength of each such connection being determined by the relationship in the cost function between the variable represented by the connected output neuron and the variable represented by the connected input neuron, such that running the neural network for a sufficient time to minimize the cost function will produce optimum values for each control variable at each sampling time.",1991-05-14,"The title of the patent is discrete-time optimal control by neural network and its abstract is a neural network determines optimal control inputs for a linear quadratic discrete-time process at m sampling times, the process being characterized by a quadratic cost function, p state variables, and r control variables. the network includes n=(p+r)m neurons, a distinct neuron being assigned to represent the value of each state variable at each sampling time and a distinct neuron being assigned to represent the value of each control variable at each sampling time. an input bias connected to each neuron has a value determined by the quandratic cost function for the variable represented by the neuron. selected connections are provided between the output of each neuron and the input of selected other neurons in the network, each such connection and the strength of each such connection being determined by the relationship in the cost function between the variable represented by the connected output neuron and the variable represented by the connected input neuron, such that running the neural network for a sufficient time to minimize the cost function will produce optimum values for each control variable at each sampling time. dated 1991-05-14"
5016211,neural network implementation of a binary adder,"a binary adder is provided for adding-processing in a high speed parallel manner two n bit binary digits. the binary adder is implemented using neural network techniques and includes a number of amplifiers corresponding to the n bit output sum and a carry generation from the result of the adding process; an augend input-synapse group, an addend input-synapse group, a carry input-synapse group, a first bias-synapse group a second bias-synapse group an output feedback-synapse group and inverters. the binary adder is efficient and fast compared to conventional techniques.",1991-05-14,"The title of the patent is neural network implementation of a binary adder and its abstract is a binary adder is provided for adding-processing in a high speed parallel manner two n bit binary digits. the binary adder is implemented using neural network techniques and includes a number of amplifiers corresponding to the n bit output sum and a carry generation from the result of the adding process; an augend input-synapse group, an addend input-synapse group, a carry input-synapse group, a first bias-synapse group a second bias-synapse group an output feedback-synapse group and inverters. the binary adder is efficient and fast compared to conventional techniques. dated 1991-05-14"
5017375,method to prepare a neurotrophic composition,"the present invention is based on the discovery that amyotrophic lateral sclerosis (als), parkinson disease and alzheimer disease are due to lack of a disorder-specific neurotrophic hormone or factor. diagnosis is accomplished by assaying factors specific for a particular neuronal network or system; for example, dopamine neutotrophic hormones from striatum or caudate-putamen in the nigrostriatal dopaminergic neural system are used to diagnose and treat parkinsonism. with tissue culture, the presence or absence of spacific neurotrophic factos can be assessed in als, parkinsonism, and alzheimer disease. if there is a deficiency, extracted and purified neurotrophic factors specific to the particular neuronal network or system can be injected into a patient having als, alzheimer disease or parkinsonism for treatment of the disease.",1991-05-21,"The title of the patent is method to prepare a neurotrophic composition and its abstract is the present invention is based on the discovery that amyotrophic lateral sclerosis (als), parkinson disease and alzheimer disease are due to lack of a disorder-specific neurotrophic hormone or factor. diagnosis is accomplished by assaying factors specific for a particular neuronal network or system; for example, dopamine neutotrophic hormones from striatum or caudate-putamen in the nigrostriatal dopaminergic neural system are used to diagnose and treat parkinsonism. with tissue culture, the presence or absence of spacific neurotrophic factos can be assessed in als, parkinsonism, and alzheimer disease. if there is a deficiency, extracted and purified neurotrophic factors specific to the particular neuronal network or system can be injected into a patient having als, alzheimer disease or parkinsonism for treatment of the disease. dated 1991-05-21"
5021988,semiconductor neural network and method of driving the same,"a semiconductor neural network includes a plurality of data input line pairs to which complementary input data pairs are transmitted respectively, data output line pairs respectively deriving complementary output data pairs and a plurality of coupling elements arranged at respective crosspoints of the data input lines and the data output lines. the coupling elements are programmable in states, and couple corresponding data output lines and corresponding data input lines in accordance with the programmed states thereof. differential amplifiers formed by cross-coupled inverting amplifiers are provided in order to detect potentials on the data output lines. the differential amplifiers are provided for respective ones of the data output line pairs.",1991-06-04,"The title of the patent is semiconductor neural network and method of driving the same and its abstract is a semiconductor neural network includes a plurality of data input line pairs to which complementary input data pairs are transmitted respectively, data output line pairs respectively deriving complementary output data pairs and a plurality of coupling elements arranged at respective crosspoints of the data input lines and the data output lines. the coupling elements are programmable in states, and couple corresponding data output lines and corresponding data input lines in accordance with the programmed states thereof. differential amplifiers formed by cross-coupled inverting amplifiers are provided in order to detect potentials on the data output lines. the differential amplifiers are provided for respective ones of the data output line pairs. dated 1991-06-04"
5023045,plant malfunction diagnostic method,"a plant malfunction diagnostic method is characterized by determining by simulation a change in a plant state variable, forming a pattern among plant state variables obtained by autoregressive analysis of the change in plant state variable, inserting the formed pattern among the plant state variables in a neural network, performing learning until a preset precision is obtained, and identifying the cause of the malfunction by inserting, in the neural network, a pattern which indicates the pattern among plant state variables formed by data gathered from the plant. this makes possible early identification of the cause of a malfunction. plant rate of operation and safety are improved by allowing the operator to perform the appropriate recovery operation with a sufficient time margin.",1991-06-11,"The title of the patent is plant malfunction diagnostic method and its abstract is a plant malfunction diagnostic method is characterized by determining by simulation a change in a plant state variable, forming a pattern among plant state variables obtained by autoregressive analysis of the change in plant state variable, inserting the formed pattern among the plant state variables in a neural network, performing learning until a preset precision is obtained, and identifying the cause of the malfunction by inserting, in the neural network, a pattern which indicates the pattern among plant state variables formed by data gathered from the plant. this makes possible early identification of the cause of a malfunction. plant rate of operation and safety are improved by allowing the operator to perform the appropriate recovery operation with a sufficient time margin. dated 1991-06-11"
5023833,feed forward neural network for unary associative memory,"feed forward neural network models for associative content addressable memory utilize a first level matrix of resistor connections to store words and compare addressing cues with the stored words represented by connections of unit resistive value, and a winner-take-all circuit for producing a unary output signal corresponding to the word most closely matched in the first matrix. the unary output signal is converted to a binary output code, such as by a suitable matrix. cues are coded for the address input as binary 1=+v, binary 0=-v, and unknown =0v. two input amplifiers are employed with two input conductors for each input bit position, one noninverting and the other inverting, so that the winner-take-all circuit at the output of the first matrix may be organized to select the highest number of matches with stored words as the unary output signal. by inverting the cues at the input to the first matrix, and inverting the output of the first level matrix, the effect of resistor value imprecision in the first matrix is virtually obviated. by space coding, the first and second matrices may be expanded into multiple sets of matrices, each with its own winner-take-all circuit for producing unary output signals applied from the first set to the second set of matrices. the output conductors of the second set of matrices are grouped to provide a sparse output code that is then converted to a binary code corresponding to the word recalled.",1991-06-11,"The title of the patent is feed forward neural network for unary associative memory and its abstract is feed forward neural network models for associative content addressable memory utilize a first level matrix of resistor connections to store words and compare addressing cues with the stored words represented by connections of unit resistive value, and a winner-take-all circuit for producing a unary output signal corresponding to the word most closely matched in the first matrix. the unary output signal is converted to a binary output code, such as by a suitable matrix. cues are coded for the address input as binary 1=+v, binary 0=-v, and unknown =0v. two input amplifiers are employed with two input conductors for each input bit position, one noninverting and the other inverting, so that the winner-take-all circuit at the output of the first matrix may be organized to select the highest number of matches with stored words as the unary output signal. by inverting the cues at the input to the first matrix, and inverting the output of the first level matrix, the effect of resistor value imprecision in the first matrix is virtually obviated. by space coding, the first and second matrices may be expanded into multiple sets of matrices, each with its own winner-take-all circuit for producing unary output signals applied from the first set to the second set of matrices. the output conductors of the second set of matrices are grouped to provide a sparse output code that is then converted to a binary code corresponding to the word recalled. dated 1991-06-11"
5025282,color image forming apparatus,"the improved color image forming apparatus is so designed that the image forming condition computing means having a learning capability, such as a neural network having a back propagation learning algorithm, is caused to learn preliminarily those image forming conditions which are appropriate for the specific type of a documents (e.g. a reflection-type original or a transmission-type original such as a negative film or a reversal film) or the original image carried on the documents such as sea, mountains or a snow scene, examples of such image forming conditions being exposing conditions (e.g. the balance of three primary colors and their densities) and the conditions of developing, fixing and otherwise processing light-sensitive materials, and image is formed on a particular light-sensitive material under the image forming conditions computed by the computing means which has learned said appropriate conditions. the visible image reproduced with this apparatus always has a good color balance, is free from deterioration of image quality, has none of the unwanted color shades and is optimum for the particular document or original image. as a further advantage, even unskilled users can easily operate this apparatus to reproduce an image that meets the specific preference of the laboratory or the user.",1991-06-18,"The title of the patent is color image forming apparatus and its abstract is the improved color image forming apparatus is so designed that the image forming condition computing means having a learning capability, such as a neural network having a back propagation learning algorithm, is caused to learn preliminarily those image forming conditions which are appropriate for the specific type of a documents (e.g. a reflection-type original or a transmission-type original such as a negative film or a reversal film) or the original image carried on the documents such as sea, mountains or a snow scene, examples of such image forming conditions being exposing conditions (e.g. the balance of three primary colors and their densities) and the conditions of developing, fixing and otherwise processing light-sensitive materials, and image is formed on a particular light-sensitive material under the image forming conditions computed by the computing means which has learned said appropriate conditions. the visible image reproduced with this apparatus always has a good color balance, is free from deterioration of image quality, has none of the unwanted color shades and is optimum for the particular document or original image. as a further advantage, even unskilled users can easily operate this apparatus to reproduce an image that meets the specific preference of the laboratory or the user. dated 1991-06-18"
5027182,high-gain algaas/gaas double heterojunction darlington phototransistors for optical neural networks,"high-gain mocvd-grown (metal-organic chemical vapor deposition) algaas/gaas/algaas n-p-n double heterojunction bipolar transistors (dhbts) (14) and darlington phototransistor pairs (14, 16) are provided for use in optical neural networks and other optoelectronic integrated circuit applications. the reduced base (22) doping level used herein results in effective blockage of zn out-diffusion, enabling a current gain of 500, higher than most previously reported values for zn-diffused-base dhbts. darlington phototransistor pairs of this material can achieve a current gain of over 6,000, which satisfies the gain requirement for optical neural network designs, which advantageously may employ novel neurons (10) comprising the darlington phototransistor pair in series with a light source (12).",1991-06-25,"The title of the patent is high-gain algaas/gaas double heterojunction darlington phototransistors for optical neural networks and its abstract is high-gain mocvd-grown (metal-organic chemical vapor deposition) algaas/gaas/algaas n-p-n double heterojunction bipolar transistors (dhbts) (14) and darlington phototransistor pairs (14, 16) are provided for use in optical neural networks and other optoelectronic integrated circuit applications. the reduced base (22) doping level used herein results in effective blockage of zn out-diffusion, enabling a current gain of 500, higher than most previously reported values for zn-diffused-base dhbts. darlington phototransistor pairs of this material can achieve a current gain of over 6,000, which satisfies the gain requirement for optical neural network designs, which advantageously may employ novel neurons (10) comprising the darlington phototransistor pair in series with a light source (12). dated 1991-06-25"
5033006,self-extending neural-network,a self-extending shape neural-network is capable of a self-extending operation in accordance with the studying results. the self-extending shape neural-network has initially minimum number of the intermediate layers and the number of the nodes (units) within each layer by the self-extension of the network construction so as to shorten the studying time and the discriminating time. this studying may be effected efficiently by the studying being directed towards the focus when the studying is not focused.,1991-07-16,The title of the patent is self-extending neural-network and its abstract is a self-extending shape neural-network is capable of a self-extending operation in accordance with the studying results. the self-extending shape neural-network has initially minimum number of the intermediate layers and the number of the nodes (units) within each layer by the self-extension of the network construction so as to shorten the studying time and the discriminating time. this studying may be effected efficiently by the studying being directed towards the focus when the studying is not focused. dated 1991-07-16
5040134,neural network employing leveled summing scheme with blocked array,"a novel associative network architecture is described in which a neural network is subdivided into a plurality of smaller blocks. each block comprises an array of pattern matching cells which is used for calculating the relative match, or hamming distance, between an input pattern and a stored weight pattern. the cells are arranged in columns along one or more local summing lines. the total current flowing along the local summing lines for a given block corresponds to the match for that block. each of the blocks are coupled together using a plurality of global summing lines. the global summing lines sum the individual current contributions from the local summing lines of each associated block. coupling between the local column lines and the global summing lines is achieved by using a specialized coupling device which permits control of the coupling ratio between the lines. by selectively turning on or off various blocks a measure of the match for individual blocks or for groups of blocks representing a subset of the network, may be calculated. control over the coupling ratio within the blocks also prevents destructive levels of current from building up on the global summing lines.",1991-08-13,"The title of the patent is neural network employing leveled summing scheme with blocked array and its abstract is a novel associative network architecture is described in which a neural network is subdivided into a plurality of smaller blocks. each block comprises an array of pattern matching cells which is used for calculating the relative match, or hamming distance, between an input pattern and a stored weight pattern. the cells are arranged in columns along one or more local summing lines. the total current flowing along the local summing lines for a given block corresponds to the match for that block. each of the blocks are coupled together using a plurality of global summing lines. the global summing lines sum the individual current contributions from the local summing lines of each associated block. coupling between the local column lines and the global summing lines is achieved by using a specialized coupling device which permits control of the coupling ratio between the lines. by selectively turning on or off various blocks a measure of the match for individual blocks or for groups of blocks representing a subset of the network, may be calculated. control over the coupling ratio within the blocks also prevents destructive levels of current from building up on the global summing lines. dated 1991-08-13"
5040215,speech recognition apparatus using neural network and fuzzy logic,"a speech recognition apparatus has a speech input unit for inputting a speech; a speech analysis unit for analyzing the inputted speech to output the time series of a feature vector; a candidates selection unit for inputting the time series of a feature vector from the speech analysis unit to select a plurality of candidates of recognition result from the speech categories; and a discrimination processing unit for discriminating the selected candidates to obtain a final recognition result. the discrimination processing unit includes three components in the form of a pair generation unit for generating all of the two combinations of the n-number of candidates selected by said candidate selection unit, a pair discrimination unit for discriminating which of the candidates of the combinations is more certain for each of all .sub.n c.sub.2 -number of combinations (or pairs) on the basis of the extracted result of the acoustic feature intrinsic to each of said candidate speeches, and a final decision unit for collecting all the pair discrimination results obtained from the pair discrimination unit for each of all the .sub.n c.sub.2 -number of combinations (or pairs) to decide the final result. the pair discrimination unit handles the extracted result of the acoustic feature intrinsic to each of the candidate speeches as fuzzy information and accomplishes the discrimination processing on the basis of fuzzy logic algorithms, and the final decision unit accomplishes its collections on the basis of the fuzzy logic algorithms.",1991-08-13,"The title of the patent is speech recognition apparatus using neural network and fuzzy logic and its abstract is a speech recognition apparatus has a speech input unit for inputting a speech; a speech analysis unit for analyzing the inputted speech to output the time series of a feature vector; a candidates selection unit for inputting the time series of a feature vector from the speech analysis unit to select a plurality of candidates of recognition result from the speech categories; and a discrimination processing unit for discriminating the selected candidates to obtain a final recognition result. the discrimination processing unit includes three components in the form of a pair generation unit for generating all of the two combinations of the n-number of candidates selected by said candidate selection unit, a pair discrimination unit for discriminating which of the candidates of the combinations is more certain for each of all .sub.n c.sub.2 -number of combinations (or pairs) on the basis of the extracted result of the acoustic feature intrinsic to each of said candidate speeches, and a final decision unit for collecting all the pair discrimination results obtained from the pair discrimination unit for each of all the .sub.n c.sub.2 -number of combinations (or pairs) to decide the final result. the pair discrimination unit handles the extracted result of the acoustic feature intrinsic to each of the candidate speeches as fuzzy information and accomplishes the discrimination processing on the basis of fuzzy logic algorithms, and the final decision unit accomplishes its collections on the basis of the fuzzy logic algorithms. dated 1991-08-13"
5040230,associative pattern conversion system and adaptation method thereof,"an associative pattern conversion system is disclosed which may be used for image recognition. the system includes an image input portion, an image processing portion and a recognition portion. the image processing portion includes a process unit for extracting characteristics and a frame memory for holding image data. the recognition portion, which includes a component for the learning of data to be associated, obtains the extracted characteristics from the image processing portion and performs associative pattern conversion from the image input portion. the system of the present invention may be applied to any neutral network, preferably a matrix calculation type neural network.",1991-08-13,"The title of the patent is associative pattern conversion system and adaptation method thereof and its abstract is an associative pattern conversion system is disclosed which may be used for image recognition. the system includes an image input portion, an image processing portion and a recognition portion. the image processing portion includes a process unit for extracting characteristics and a frame memory for holding image data. the recognition portion, which includes a component for the learning of data to be associated, obtains the extracted characteristics from the image processing portion and performs associative pattern conversion from the image input portion. the system of the present invention may be applied to any neutral network, preferably a matrix calculation type neural network. dated 1991-08-13"
5041916,color image data compression and recovery apparatus based on neural networks,"a data compression and recovery apparatus compresses picture element data of a color image by expressing two primary color values of each picture element as a set of parameter values of a neural network in conjunction with reference color data values of a corresponding block of picture elements. date recovery is achieved by inputting each block of reference color values to a neural network while establishing the corresponding set of parameter values in the network, to thereby obtain the original pair of encoded primary color values for each of successive picture elements. the third primary color can be used as the reference color.",1991-08-20,"The title of the patent is color image data compression and recovery apparatus based on neural networks and its abstract is a data compression and recovery apparatus compresses picture element data of a color image by expressing two primary color values of each picture element as a set of parameter values of a neural network in conjunction with reference color data values of a corresponding block of picture elements. date recovery is achieved by inputting each block of reference color values to a neural network while establishing the corresponding set of parameter values in the network, to thereby obtain the original pair of encoded primary color values for each of successive picture elements. the third primary color can be used as the reference color. dated 1991-08-20"
5041976,diagnostic system using pattern recognition for electronic automotive control systems,"a system is disclosed for diagnosing faults in electronic control systems wherein a large volume of information is exchanged between the electronic control processor and a mechanical system under its control. the data is acquired such that parameter vectors describing the system operation are formed. the vectors are provided to a pattern recognition system such as a neural network for classification according to the operating condition of the electronically controlled system. for diagnosis of electronically controlled engine operation, the parameters included in the vectors correspond to individual firing events occurring in the engine operating under a predetermined condition. the diagnostic system can be implemented as a service tool in an automotive service bay or can be implemented within the on-board electronic control system itself.",1991-08-20,"The title of the patent is diagnostic system using pattern recognition for electronic automotive control systems and its abstract is a system is disclosed for diagnosing faults in electronic control systems wherein a large volume of information is exchanged between the electronic control processor and a mechanical system under its control. the data is acquired such that parameter vectors describing the system operation are formed. the vectors are provided to a pattern recognition system such as a neural network for classification according to the operating condition of the electronically controlled system. for diagnosis of electronically controlled engine operation, the parameters included in the vectors correspond to individual firing events occurring in the engine operating under a predetermined condition. the diagnostic system can be implemented as a service tool in an automotive service bay or can be implemented within the on-board electronic control system itself. dated 1991-08-20"
5043913,neural network,"input signals inputted in respective unit circuits forming a synapse array pass through variable connector elements to be integrated into one analog signal, which in turn is converted into a binary associated corresponding signal by an amplifier. two control signals are produced on the basis of the associated corresponding signal and an educator signal. the two control signals are fed back to the respective unit circuits, to control degrees of electrical coupling of the variable connector elements in the respective unit circuits. thus, learning of the respective unit circuits is performed.",1991-08-27,"The title of the patent is neural network and its abstract is input signals inputted in respective unit circuits forming a synapse array pass through variable connector elements to be integrated into one analog signal, which in turn is converted into a binary associated corresponding signal by an amplifier. two control signals are produced on the basis of the associated corresponding signal and an educator signal. the two control signals are fed back to the respective unit circuits, to control degrees of electrical coupling of the variable connector elements in the respective unit circuits. thus, learning of the respective unit circuits is performed. dated 1991-08-27"
5045713,multi-feedback circuit apparatus,a multi-feedback circuit apparatus is provided which can prevent undesired oscillation or chaos phenomena that inevitably arise when the hopfield model is realized by electronic circuits. the apparatus can also reduce the number of synapse nodes in the neural network model.,1991-09-03,The title of the patent is multi-feedback circuit apparatus and its abstract is a multi-feedback circuit apparatus is provided which can prevent undesired oscillation or chaos phenomena that inevitably arise when the hopfield model is realized by electronic circuits. the apparatus can also reduce the number of synapse nodes in the neural network model. dated 1991-09-03
5046019,fuzzy data comparator with neural network postprocessor,"a fuzzy data comparator receives a fuzzy data digital data bit stream and compares each frame thereof with multiple sets of differing known data stored in a plurality of pattern memories, using a selected comparison metric. the results of the comparisons are accumulated as error values. a first neural postprocessing network ranks error values less than a preselected threshold. a second neural network receives the first neural network solutions and provides an expansion bus for interconnecting to additional comparators.",1991-09-03,"The title of the patent is fuzzy data comparator with neural network postprocessor and its abstract is a fuzzy data comparator receives a fuzzy data digital data bit stream and compares each frame thereof with multiple sets of differing known data stored in a plurality of pattern memories, using a selected comparison metric. the results of the comparisons are accumulated as error values. a first neural postprocessing network ranks error values less than a preselected threshold. a second neural network receives the first neural network solutions and provides an expansion bus for interconnecting to additional comparators. dated 1991-09-03"
5047655,programmable analog neural network,"the neural network of the invention, of the type with a cartesian matrix, has a first column of addition of the input signals, and at each intersection of the m lines and n columns it comprises a synapse constituted of a simple logic gate.",1991-09-10,"The title of the patent is programmable analog neural network and its abstract is the neural network of the invention, of the type with a cartesian matrix, has a first column of addition of the input signals, and at each intersection of the m lines and n columns it comprises a synapse constituted of a simple logic gate. dated 1991-09-10"
5048097,optical character recognition neural network system for machine-printed characters,"character images which are to be sent to a neural network trained to recognize a predetermined set of symbols are first processed by an optical character recognition pre-processor which normalizes the character images. the output of the neural network is processed by an optical character recognition post-processor. the post-processor corrects erroneous symbol identifications made by the neural network. the post-processor identifies special symbols and symbol cases not identifiable by the neural network following character normalization. for characters identified by the neural network with low scores, the post-processor attempts to find and separate adjacent characters which are kerned and characters which are touching. the touching characters are separated in one of nine successively initiated processes depending upon the geometric parameters of the image. when all else fails, the post-processor selects either the second or third highest scoring symbol identified by the neural network based upon the likelihood of the second or third highest scoring symbol being confused with the highest scoring symbol.",1991-09-10,"The title of the patent is optical character recognition neural network system for machine-printed characters and its abstract is character images which are to be sent to a neural network trained to recognize a predetermined set of symbols are first processed by an optical character recognition pre-processor which normalizes the character images. the output of the neural network is processed by an optical character recognition post-processor. the post-processor corrects erroneous symbol identifications made by the neural network. the post-processor identifies special symbols and symbol cases not identifiable by the neural network following character normalization. for characters identified by the neural network with low scores, the post-processor attempts to find and separate adjacent characters which are kerned and characters which are touching. the touching characters are separated in one of nine successively initiated processes depending upon the geometric parameters of the image. when all else fails, the post-processor selects either the second or third highest scoring symbol identified by the neural network based upon the likelihood of the second or third highest scoring symbol being confused with the highest scoring symbol. dated 1991-09-10"
5048100,self organizing neural network method and system for general classification of patterns,a neural network system and method that can adaptively recognize each of many pattern configurations from a set. the system learns and maintains accurate associations between signal pattern configurations and pattern classes with training from a teaching mechanism. the classifying system consists of a distributed input processor and an adaptive association processor. the input processor decomposes an input pattern into modules of localized contextual elements. these elements in turn are mapped onto pattern classes using a self-organizing associative neural scheme. the associative mapping determines which pattern class best represents the input pattern. the computation is done through gating elements that correspond to the contextual elements. learning is achieved by modifying the gating elements from a true/false response to the computed probabilities for all classes in the set. the system is a parallel and fault tolerant process. it can easily be extended to accommodate an arbitrary number of patterns at an arbitrary degree of precision. the classifier can be applied to automated recognition and inspection of many different types of signals and patterns.,1991-09-10,The title of the patent is self organizing neural network method and system for general classification of patterns and its abstract is a neural network system and method that can adaptively recognize each of many pattern configurations from a set. the system learns and maintains accurate associations between signal pattern configurations and pattern classes with training from a teaching mechanism. the classifying system consists of a distributed input processor and an adaptive association processor. the input processor decomposes an input pattern into modules of localized contextual elements. these elements in turn are mapped onto pattern classes using a self-organizing associative neural scheme. the associative mapping determines which pattern class best represents the input pattern. the computation is done through gating elements that correspond to the contextual elements. learning is achieved by modifying the gating elements from a true/false response to the computed probabilities for all classes in the set. the system is a parallel and fault tolerant process. it can easily be extended to accommodate an arbitrary number of patterns at an arbitrary degree of precision. the classifier can be applied to automated recognition and inspection of many different types of signals and patterns. dated 1991-09-10
5050095,neural network auto-associative memory with two rules for varying the weights,"a neural network associative memory which has a single layer of primatives and which utilizes a variant of the generalized delta for calculating the connection weights between the primatives. the delta rule is characterized by its utilization of predetermined values for the primitive and an error index which compares, during iterations, the predetermined primative values with actual primative values until the delta factor becomes a predetermined minimum value.",1991-09-17,"The title of the patent is neural network auto-associative memory with two rules for varying the weights and its abstract is a neural network associative memory which has a single layer of primatives and which utilizes a variant of the generalized delta for calculating the connection weights between the primatives. the delta rule is characterized by its utilization of predetermined values for the primitive and an error index which compares, during iterations, the predetermined primative values with actual primative values until the delta factor becomes a predetermined minimum value. dated 1991-09-17"
5050096,path cost computing neural network,""" the operation of an electronic neural computer is described. this electronic neural computer solves for the optimal path in a space of """"cost functions"""" which are represented as delays at the nodes of a grid (in two, three, four, or more dimensions). time gating by delays lets the optimal solution thread the maze of the network first. the neural computer starts to compute all possible paths through the cost function field and shuts down after the first (optimal solution) emerges at the target node. the cost function delays are set from outside the neural computer architecture. """,1991-09-17,"The title of the patent is path cost computing neural network and its abstract is "" the operation of an electronic neural computer is described. this electronic neural computer solves for the optimal path in a space of """"cost functions"""" which are represented as delays at the nodes of a grid (in two, three, four, or more dimensions). time gating by delays lets the optimal solution thread the maze of the network first. the neural computer starts to compute all possible paths through the cost function field and shuts down after the first (optimal solution) emerges at the target node. the cost function delays are set from outside the neural computer architecture. "" dated 1991-09-17"
5052043,neural network with back propagation controlled through an output confidence measure,"apparatus, and an accompanying method, for a neural network, particularly one suited for use in optical character recognition (ocr) systems, which through controlling back propagation and adjustment of neural weight and bias values through an output confidence measure, smoothly, rapidly and accurately adapts its response to actual changing input data (characters). specifically, the results of appropriate actual unknown input characters, which have been recognized with an output confidence measure that lies within a pre-defined range, are used to adaptively re-train the network during pattern recognition. by limiting the maximum value of the output confidence measure at which this re-training will occur, the network re-trains itself only when the input characters have changed by a sufficient margin from initial training data such that this re-training is likely to produce a subsequent noticeable increase in the recognition accuracy provided by the network. output confidence is measured as a ratio between the highest and next highest values produced by output neurons in the network. by broadening the entire base of training data to include actual dynamically changing input characters, the inventive neural network provides more robust performance than which heretofore occurs in neural networks known in the art.",1991-09-24,"The title of the patent is neural network with back propagation controlled through an output confidence measure and its abstract is apparatus, and an accompanying method, for a neural network, particularly one suited for use in optical character recognition (ocr) systems, which through controlling back propagation and adjustment of neural weight and bias values through an output confidence measure, smoothly, rapidly and accurately adapts its response to actual changing input data (characters). specifically, the results of appropriate actual unknown input characters, which have been recognized with an output confidence measure that lies within a pre-defined range, are used to adaptively re-train the network during pattern recognition. by limiting the maximum value of the output confidence measure at which this re-training will occur, the network re-trains itself only when the input characters have changed by a sufficient margin from initial training data such that this re-training is likely to produce a subsequent noticeable increase in the recognition accuracy provided by the network. output confidence is measured as a ratio between the highest and next highest values produced by output neurons in the network. by broadening the entire base of training data to include actual dynamically changing input characters, the inventive neural network provides more robust performance than which heretofore occurs in neural networks known in the art. dated 1991-09-24"
5054094,rotationally impervious feature extraction for optical character recognition,"a feature-based optical character recognition system, employing a feature-based recognition device such as a neural network or an absolute distance measure device, extracts a set of features from segmented character images in a document, at least some of the extracted features being at least nearly impervious to rotation or skew of the document image, so as to enhance the reliability of the system. one rotationally invariant feature extracted by the system is the number of intercepts between boundary transitions in the image with at least a selected one of a plurality of radii centered at the centroid of the character in the image.",1991-10-01,"The title of the patent is rotationally impervious feature extraction for optical character recognition and its abstract is a feature-based optical character recognition system, employing a feature-based recognition device such as a neural network or an absolute distance measure device, extracts a set of features from segmented character images in a document, at least some of the extracted features being at least nearly impervious to rotation or skew of the document image, so as to enhance the reliability of the system. one rotationally invariant feature extracted by the system is the number of intercepts between boundary transitions in the image with at least a selected one of a plurality of radii centered at the centroid of the character in the image. dated 1991-10-01"
5055897,semiconductor cell for neural network and the like,"a cell employing floating gate storage device particularly suited for neural networks. the floating gate from the floating gate device extends to and becomes part of a second, field effect device. current through the second device is affected by the charge on the floating gate. the weighting factor for the cell is determined by the amount of charge on the floating gate. by charging the floating gate to various levels, a continuum of weighting factors is obtained. multiplication is obtained since the current through the second device is a function of the weighting factor.",1991-10-08,"The title of the patent is semiconductor cell for neural network and the like and its abstract is a cell employing floating gate storage device particularly suited for neural networks. the floating gate from the floating gate device extends to and becomes part of a second, field effect device. current through the second device is affected by the charge on the floating gate. the weighting factor for the cell is determined by the amount of charge on the floating gate. by charging the floating gate to various levels, a continuum of weighting factors is obtained. multiplication is obtained since the current through the second device is a function of the weighting factor. dated 1991-10-08"
5056037,analog hardware for learning neural networks,"this is a recurrent or feedforward analog neural network processor having a multi-level neuron array and a synaptic matrix for storing weighted analog values of synaptic connection strengths which is characterized by temporarily changing one connection strength at a time to determine its effect on system output relative to the desired target. that connection strength is then adjusted based on the effect, whereby the processor is taught the correct response to training examples connection by connection.",1991-10-08,"The title of the patent is analog hardware for learning neural networks and its abstract is this is a recurrent or feedforward analog neural network processor having a multi-level neuron array and a synaptic matrix for storing weighted analog values of synaptic connection strengths which is characterized by temporarily changing one connection strength at a time to determine its effect on system output relative to the desired target. that connection strength is then adjusted based on the effect, whereby the processor is taught the correct response to training examples connection by connection. dated 1991-10-08"
5056897,spatial light modulating element and neural network circuit,"a spatial light modulator and a neural network circuit are disclosed. the modulator is used in pattern recognition and has an arrangement in which a photoconductive layer held between conductive electrodes is connected in series to a liquid crystal cell including a liquid crystal layer held between two opposite electrodes. setting the rate between the area of the photoconductive layer and the area of at least one of the opposite electrodes between which the liquid crystal layer is disposed, provides a highly efficient reflective and transmissive spatial light modulator of a simple structure. both reflective and transmissive spatial light modulating elements are applied to a neurocomputer or the like.",1991-10-15,"The title of the patent is spatial light modulating element and neural network circuit and its abstract is a spatial light modulator and a neural network circuit are disclosed. the modulator is used in pattern recognition and has an arrangement in which a photoconductive layer held between conductive electrodes is connected in series to a liquid crystal cell including a liquid crystal layer held between two opposite electrodes. setting the rate between the area of the photoconductive layer and the area of at least one of the opposite electrodes between which the liquid crystal layer is disposed, provides a highly efficient reflective and transmissive spatial light modulator of a simple structure. both reflective and transmissive spatial light modulating elements are applied to a neurocomputer or the like. dated 1991-10-15"
5058034,digital neural network with discrete point rule space,this application discloses a system that optimizes a neural network by generating all of the discrete weights for a given neural node by creating a normalized weight vector for each possible weight combination. the normalized vectors for each node define the weight space for that node. this complete set of weight vectors for each node is searched using a direct search method during the learning phase to optimize the network. the search evaluates a node cost function to determine a base point from which a pattern more within the weight space is made. around the pattern mode point exploratory moves are made which are cost function evaluated. the pattern move is performed by eliminating from the search vectors with lower commonality.,1991-10-15,The title of the patent is digital neural network with discrete point rule space and its abstract is this application discloses a system that optimizes a neural network by generating all of the discrete weights for a given neural node by creating a normalized weight vector for each possible weight combination. the normalized vectors for each node define the weight space for that node. this complete set of weight vectors for each node is searched using a direct search method during the learning phase to optimize the network. the search evaluates a node cost function to determine a base point from which a pattern more within the weight space is made. around the pattern mode point exploratory moves are made which are cost function evaluated. the pattern move is performed by eliminating from the search vectors with lower commonality. dated 1991-10-15
5058180,neural network apparatus and method for pattern recognition,"a self-organizing neural network having input and output neurons mutually coupled via bottom-up and top-down adaptive weight matrics performs pattern recognition while using substantially fewer neurons and being substantially immune from pattern distortion or rotation. the network is first trained in accordance with the adaptive resonance theory by inputting reference pattern data into the input neurons for clustering within the output neurons. the input neurons then receive subject pattern data which are transferred via a bottom-up adaptive weight matrix to a set of output neurons. vigilance testing is performed and multiple computed vigilance parameters are generated. a predetermined, but selectively variable, reference vigilance parameter is compared individually against each computed vigilance parameter and adjusted with each comparison until each computed vigilance parameter equals or exceeds the adjusted reference vigilance parameter, thereby producing an adjusted reference vigilance parameter for each output neuron. the input pattern is classified according to the output neuron corresponding to the maximum adjusted reference vigilance parameter. alternatively, the original computed vigilance parameters can be used by classifying the input pattern according to the output neuron corresponding to the maximum computer vigilance parameter.",1991-10-15,"The title of the patent is neural network apparatus and method for pattern recognition and its abstract is a self-organizing neural network having input and output neurons mutually coupled via bottom-up and top-down adaptive weight matrics performs pattern recognition while using substantially fewer neurons and being substantially immune from pattern distortion or rotation. the network is first trained in accordance with the adaptive resonance theory by inputting reference pattern data into the input neurons for clustering within the output neurons. the input neurons then receive subject pattern data which are transferred via a bottom-up adaptive weight matrix to a set of output neurons. vigilance testing is performed and multiple computed vigilance parameters are generated. a predetermined, but selectively variable, reference vigilance parameter is compared individually against each computed vigilance parameter and adjusted with each comparison until each computed vigilance parameter equals or exceeds the adjusted reference vigilance parameter, thereby producing an adjusted reference vigilance parameter for each output neuron. the input pattern is classified according to the output neuron corresponding to the maximum adjusted reference vigilance parameter. alternatively, the original computed vigilance parameters can be used by classifying the input pattern according to the output neuron corresponding to the maximum computer vigilance parameter. dated 1991-10-15"
5058184,hierachical information processing system,"plural efferent signal paths paired with plural conventional afferent signal paths respectively are provided between lower order cell-layers and higher order cell-layers of a neural network model. once an output response has been derived from the higher order cell-layer, an efferent signal is transmitted through the efferent signal path paired with the afferent signal path concerned in the output response. under the control of which efferent signal, the afferent signal path contributing to the output response of the higher order cell-layer is affected by an excitatory effect, while the afferent signal path not contributing to the same is affected by an inhibitory effect. hence the information processing consisting of both the associative memory and the pattern recognition provided with the faculty of segmentation can be attained despite deformation and positional error of the input pattern.",1991-10-15,"The title of the patent is hierachical information processing system and its abstract is plural efferent signal paths paired with plural conventional afferent signal paths respectively are provided between lower order cell-layers and higher order cell-layers of a neural network model. once an output response has been derived from the higher order cell-layer, an efferent signal is transmitted through the efferent signal path paired with the afferent signal path concerned in the output response. under the control of which efferent signal, the afferent signal path contributing to the output response of the higher order cell-layer is affected by an excitatory effect, while the afferent signal path not contributing to the same is affected by an inhibitory effect. hence the information processing consisting of both the associative memory and the pattern recognition provided with the faculty of segmentation can be attained despite deformation and positional error of the input pattern. dated 1991-10-15"
5060276,technique for object orientation detection using a feed-forward neural network,""" the present invention relates to a technique in the form of an exemplary computer vision system for detecting the orientation of text or features on an object of manufacture. in the present system, an image of the features or text is used to extract lines using horizontal bitmap sums, and then individual symbols using vertical bitmap sums, using thresholds with each of the sums. the separated symbols are then appropriately trimmed and sealed to provide individual normalized symbols. a decision module comprising a feed-forward neural network and a sequential decision arrangement determines the """"up"""", """"down"""" or """"indeterminate"""" orientation of the text after a variable number of symbols have been processed. the system can then compare the determined orientation with a database to further determine if the object is in the """"right-side up"""" """"upside down"""" or """"indeterminate"""" orientation. """,1991-10-22,"The title of the patent is technique for object orientation detection using a feed-forward neural network and its abstract is "" the present invention relates to a technique in the form of an exemplary computer vision system for detecting the orientation of text or features on an object of manufacture. in the present system, an image of the features or text is used to extract lines using horizontal bitmap sums, and then individual symbols using vertical bitmap sums, using thresholds with each of the sums. the separated symbols are then appropriately trimmed and sealed to provide individual normalized symbols. a decision module comprising a feed-forward neural network and a sequential decision arrangement determines the """"up"""", """"down"""" or """"indeterminate"""" orientation of the text after a variable number of symbols have been processed. the system can then compare the determined orientation with a database to further determine if the object is in the """"right-side up"""" """"upside down"""" or """"indeterminate"""" orientation. "" dated 1991-10-22"
5060278,pattern recognition apparatus using a neural network system,"a pattern recognition apparatus includes a pattern input unit inputting pattern data and learning data, and a neural network system including a plurality of neural networks, each of the plurality of neural networks being assigned a corresponding one of a plurality of identification classes and having only two output units of a first unit (uo1) and a second unit (uo2). learning for each of the plurality of neural networks is performed by using the learning data. the image recognition apparatus also includes judgment unit judging which one of the identification classes the pattern data input from the image reading unit belongs to on the basis of output values a and b from the two output units (uo1) and (uo2) of all neural networks.",1991-10-22,"The title of the patent is pattern recognition apparatus using a neural network system and its abstract is a pattern recognition apparatus includes a pattern input unit inputting pattern data and learning data, and a neural network system including a plurality of neural networks, each of the plurality of neural networks being assigned a corresponding one of a plurality of identification classes and having only two output units of a first unit (uo1) and a second unit (uo2). learning for each of the plurality of neural networks is performed by using the learning data. the image recognition apparatus also includes judgment unit judging which one of the identification classes the pattern data input from the image reading unit belongs to on the basis of output values a and b from the two output units (uo1) and (uo2) of all neural networks. dated 1991-10-22"
5061866,"analog, continuous time vector scalar multiplier circuits and programmable feedback neural network using them","a four quadrant, analog multiplier circuit useful for mos implementation of feedback/feedforward neural networks. the multiplier circuit uses only one op-amp and one pair of input mos fets. it becomes a multiplier/summer by the addition of only one additional pair of input fets for each additional product to be summed and achieves the vector scalar product of 2 n-tuple vector inputs using only 2(n+1) mos transistors.",1991-10-29,"The title of the patent is analog, continuous time vector scalar multiplier circuits and programmable feedback neural network using them and its abstract is a four quadrant, analog multiplier circuit useful for mos implementation of feedback/feedforward neural networks. the multiplier circuit uses only one op-amp and one pair of input mos fets. it becomes a multiplier/summer by the addition of only one additional pair of input fets for each additional product to be summed and achieves the vector scalar product of 2 n-tuple vector inputs using only 2(n+1) mos transistors. dated 1991-10-29"
5063521,neuram: neural network with ram,"a random access memory (ram) circuit is provided wherein an input signal matrix forming an identifiable original pattern is learned and stored such that a distorted facsimile thereof may be applied to generate an output signal matrix forming a replication of the original pattern having improved recognizable features over the distorted facsimile. the input signal matrix is logically divided into a plurality of predetermined subsets comprising a unique element of the input signal matrix and the elements in the neighborhood thereof. each predetermined subset is quantized into a first digital address and applied at the address inputs of a memory circuit for retrieving data stored in the addressed memory location, while one signal of the predetermined subset is digitized and weighted and combined with the data retrieved from the addressed memory location for storage in the same addressed memory location. next, a plurality of second digital addresses is generated including predetermined combinations of the first digital address perturbed at least one bit and sequentially applied at the address inputs of the memory circuit whereby the steps of digitizing and weighting one signal of the predetermined subset of the input signal matrix, combining the digitized and weighted signal with the data retrieved from the addressed memory location, and storing the combination back into the addressed memory location are repeated for the second digital addresses.",1991-11-05,"The title of the patent is neuram: neural network with ram and its abstract is a random access memory (ram) circuit is provided wherein an input signal matrix forming an identifiable original pattern is learned and stored such that a distorted facsimile thereof may be applied to generate an output signal matrix forming a replication of the original pattern having improved recognizable features over the distorted facsimile. the input signal matrix is logically divided into a plurality of predetermined subsets comprising a unique element of the input signal matrix and the elements in the neighborhood thereof. each predetermined subset is quantized into a first digital address and applied at the address inputs of a memory circuit for retrieving data stored in the addressed memory location, while one signal of the predetermined subset is digitized and weighted and combined with the data retrieved from the addressed memory location for storage in the same addressed memory location. next, a plurality of second digital addresses is generated including predetermined combinations of the first digital address perturbed at least one bit and sequentially applied at the address inputs of the memory circuit whereby the steps of digitizing and weighting one signal of the predetermined subset of the input signal matrix, combining the digitized and weighted signal with the data retrieved from the addressed memory location, and storing the combination back into the addressed memory location are repeated for the second digital addresses. dated 1991-11-05"
5063531,optical neural net trainable in rapid time,"among light emitting and sensitive element pairs arranged along rows and columns of a matrix in each of first and second layers of an optical computer operable as a neural network with one-to-one correspondence kept between the pairs in the first layer and the pairs in the second layer, the light emitting elements and the light sensitive elements are connected along the rows in the first layer and along the columns in the second layer. optical intensity controlling elements of a panel are placed in optical paths defined by the pairs in the first layer and the pairs which correspond in the second layer to the pairs of the first layer, respectively. when the light emitting element rows are driven, optical beams are emitted by the light emitting elements of the first layer and controlled by the respective controlling elements to have first-layer controlled amounts of light, respectively. in response to the controlled amounts of light, the light sensitive element columns of the second layer produce second-layer output signals. it is possible to use the second-layer output signals in controlling the controlling elements and thereby to train the optical computer. if desired, the light emitting element columns of the second layer are driven by the second-layer output signals to make the light sensitive element rows of the first layer produce first-layer output signals and to use the first-layer output signals in controlling the controlling elements.",1991-11-05,"The title of the patent is optical neural net trainable in rapid time and its abstract is among light emitting and sensitive element pairs arranged along rows and columns of a matrix in each of first and second layers of an optical computer operable as a neural network with one-to-one correspondence kept between the pairs in the first layer and the pairs in the second layer, the light emitting elements and the light sensitive elements are connected along the rows in the first layer and along the columns in the second layer. optical intensity controlling elements of a panel are placed in optical paths defined by the pairs in the first layer and the pairs which correspond in the second layer to the pairs of the first layer, respectively. when the light emitting element rows are driven, optical beams are emitted by the light emitting elements of the first layer and controlled by the respective controlling elements to have first-layer controlled amounts of light, respectively. in response to the controlled amounts of light, the light sensitive element columns of the second layer produce second-layer output signals. it is possible to use the second-layer output signals in controlling the controlling elements and thereby to train the optical computer. if desired, the light emitting element columns of the second layer are driven by the second-layer output signals to make the light sensitive element rows of the first layer produce first-layer output signals and to use the first-layer output signals in controlling the controlling elements. dated 1991-11-05"
5063601,fast-learning neural network system for adaptive pattern recognition apparatus,a neural network for an adaptive pattern recognition apparatus includes a plurality of comparators coupled to an input signal. each comparators compares the input to a different offset voltage. the comparator output is fed to scaling multipliers and then summed to generate an output. the scaling multipliers receive weighing factors generated by using a specific equation selected to insure a fat-learning neural network.,1991-11-05,The title of the patent is fast-learning neural network system for adaptive pattern recognition apparatus and its abstract is a neural network for an adaptive pattern recognition apparatus includes a plurality of comparators coupled to an input signal. each comparators compares the input to a different offset voltage. the comparator output is fed to scaling multipliers and then summed to generate an output. the scaling multipliers receive weighing factors generated by using a specific equation selected to insure a fat-learning neural network. dated 1991-11-05
5065040,reverse flow neuron,"a neural network is provided for performing bi-directional signal transformations through a matrix of synapses by alternately sending and receiving signal vectors therethrough via switchable driver circuits. in the forward direction, the input signal is transformed according to the weighting elements of the synapses for providing an output signal. the drive direction of the switchable driver circuits may be reversed allowing the output signal to flow back through the same synapses thereby performing a reverse transformation, which may actually be an improved estimate of the original input signal. sample and hold circuits are provided for latching the output signals of the switchable driver circuits back to the inputs thereof for repeated forward and reverse signal transformations until an acceptable transformation of the original input signal is realized, thereby achieving an improved estimate of the input signal and corresponding output transformation. more generally, a first input signal may be transformed in one direction through the synapses, while a second input signal, possibly independent and unrelated to the first input signal, may be reverse transformed in the opposite direction using the same synapses as the first direction.",1991-11-12,"The title of the patent is reverse flow neuron and its abstract is a neural network is provided for performing bi-directional signal transformations through a matrix of synapses by alternately sending and receiving signal vectors therethrough via switchable driver circuits. in the forward direction, the input signal is transformed according to the weighting elements of the synapses for providing an output signal. the drive direction of the switchable driver circuits may be reversed allowing the output signal to flow back through the same synapses thereby performing a reverse transformation, which may actually be an improved estimate of the original input signal. sample and hold circuits are provided for latching the output signals of the switchable driver circuits back to the inputs thereof for repeated forward and reverse signal transformations until an acceptable transformation of the original input signal is realized, thereby achieving an improved estimate of the input signal and corresponding output transformation. more generally, a first input signal may be transformed in one direction through the synapses, while a second input signal, possibly independent and unrelated to the first input signal, may be reverse transformed in the opposite direction using the same synapses as the first direction. dated 1991-11-12"
5065339,orthogonal row-column neural processor,"the neural computing paradigm is characterized as a dynamic and highly parallel computationally intensive system typically consisting of input weight multiplications, product summation, neural state calculations, and complete connectivity among the neurons. herein is described neural network architecture called snap which uses a unique intercommunication scheme within an array structure that provides high performance for completely connected network models such as the hopfield model. snap's packaging and expansion capabilities are addressed, demonstrating snap's scalability to larger networks. each neuron generating a neuron value from a selected set of input function elements and communicating said neuron value back to said set of input function elements. the total connectivity of each neuron to all neurons is accomplished by an orthogonal row-column relationship of neurons where a given multiplier element operates during a first cycle as a row element within an input function to a column neuron, and during a second cycle as a column element within an input function to a row neuron.",1991-11-12,"The title of the patent is orthogonal row-column neural processor and its abstract is the neural computing paradigm is characterized as a dynamic and highly parallel computationally intensive system typically consisting of input weight multiplications, product summation, neural state calculations, and complete connectivity among the neurons. herein is described neural network architecture called snap which uses a unique intercommunication scheme within an array structure that provides high performance for completely connected network models such as the hopfield model. snap's packaging and expansion capabilities are addressed, demonstrating snap's scalability to larger networks. each neuron generating a neuron value from a selected set of input function elements and communicating said neuron value back to said set of input function elements. the total connectivity of each neuron to all neurons is accomplished by an orthogonal row-column relationship of neurons where a given multiplier element operates during a first cycle as a row element within an input function to a column neuron, and during a second cycle as a column element within an input function to a row neuron. dated 1991-11-12"
5067095,spann: sequence processing artificial neural network,"an artificial neural network is provided using a modular, self-organizing approach wherein a separate neural field is contained within each module for recognition and synthesis of particular characteristics of respective input and output signals thereby allowing several of these modules to be interconnected to perform a variety of operations. the first output and second input of one module is respectively coupled to the first input and second output of a second module allowing each module to perform a bi-directional transformation of the information content of the first and second input signals for creating first and second output signals having different levels of information content with respect thereto. in the upward direction, the first low-level input signal of each module is systematically delayed to create a temporal spatial vector from which a lower frequency, high-level first output signal is provided symbolic of the incoming information content. since the first output signal contains the same relevant information as the first input signal while operating at a lower frequency, the information content of the latter is said to be compressed into a first high-level output signal. in the downward direction, a second output signal having a low-level of information content is synthesized from a second input signal having a high-level of information content. the second input signal is the best prediction of the first output signal available from the knowledge base of the module, while similarly the second output signal is the prediction of the first input signal.",1991-11-19,"The title of the patent is spann: sequence processing artificial neural network and its abstract is an artificial neural network is provided using a modular, self-organizing approach wherein a separate neural field is contained within each module for recognition and synthesis of particular characteristics of respective input and output signals thereby allowing several of these modules to be interconnected to perform a variety of operations. the first output and second input of one module is respectively coupled to the first input and second output of a second module allowing each module to perform a bi-directional transformation of the information content of the first and second input signals for creating first and second output signals having different levels of information content with respect thereto. in the upward direction, the first low-level input signal of each module is systematically delayed to create a temporal spatial vector from which a lower frequency, high-level first output signal is provided symbolic of the incoming information content. since the first output signal contains the same relevant information as the first input signal while operating at a lower frequency, the information content of the latter is said to be compressed into a first high-level output signal. in the downward direction, a second output signal having a low-level of information content is synthesized from a second input signal having a high-level of information content. the second input signal is the best prediction of the first output signal available from the knowledge base of the module, while similarly the second output signal is the prediction of the first input signal. dated 1991-11-19"
5067164,hierarchical constrained automatic learning neural network for character recognition,"highly accurate, reliable optical character recognition is afforded by a layered network having several layers of constrained feature detection wherein each layer of constrained feature detection includes a plurality of constrained feature maps and a corresponding plurality of feature reduction maps. each feature reduction map is connected to only one constrained feature map in the same layer for undersampling that constrained feature map. units in each constrained feature map of the first constrained feature detection layer respond as a function of a corresponding kernel and of different portions of the pixel image of the character captured in a receptive field associated with the unit. units in each feature map of the second constrained feature detection layer respond as a function of a corresponding kernel and of different portions of an individual feature reduction map or a combination of several feature reduction maps in the first constrained feature detection layer as captured in a receptive field of the unit. the feature reduction maps of the second constrained feature detection layer are fully connected to each unit in the final character classification layer. kernels are automatically learned by constrained back propagation during network initialization or training.",1991-11-19,"The title of the patent is hierarchical constrained automatic learning neural network for character recognition and its abstract is highly accurate, reliable optical character recognition is afforded by a layered network having several layers of constrained feature detection wherein each layer of constrained feature detection includes a plurality of constrained feature maps and a corresponding plurality of feature reduction maps. each feature reduction map is connected to only one constrained feature map in the same layer for undersampling that constrained feature map. units in each constrained feature map of the first constrained feature detection layer respond as a function of a corresponding kernel and of different portions of the pixel image of the character captured in a receptive field associated with the unit. units in each feature map of the second constrained feature detection layer respond as a function of a corresponding kernel and of different portions of an individual feature reduction map or a combination of several feature reduction maps in the first constrained feature detection layer as captured in a receptive field of the unit. the feature reduction maps of the second constrained feature detection layer are fully connected to each unit in the final character classification layer. kernels are automatically learned by constrained back propagation during network initialization or training. dated 1991-11-19"
5068662,neural network analog-to-digital converter,"an asynchronous, rapid, neural network analog-to-digital converter. this converter requires only two different resistance values in r2r resistor ladders, and does not require both positive and negative biases. an average of n/2 steps is required for an n-bit conversion.",1991-11-26,"The title of the patent is neural network analog-to-digital converter and its abstract is an asynchronous, rapid, neural network analog-to-digital converter. this converter requires only two different resistance values in r2r resistor ladders, and does not require both positive and negative biases. an average of n/2 steps is required for an n-bit conversion. dated 1991-11-26"
5068801,"optical interconnector and highly interconnected, learning neural network incorporating optical interconnector therein","a variable weight optical interconnector is disclosed to include a projecting device and an interconnection weighting device remote from the projecting device. the projecting device projects a distribution of interconnecting light beams when illuminated by a spatially-modulated light pattern. the weighting device includes a photosensitive screen provided in optical alignment with the projecting device to independently control the intensity of each projected interconnecting beam to thereby assign an interconnection weight to each such beam. further in accordance with the present invention, a highly-interconnected optical neural network having learning capability is disclosed as including a spatial light modulator, a detecting device, an interconnector according to the present invention, and a device responsive to detection signals generated by the detecting device to modify the interconnection weights assigned by the photosensitive screen of the interconnector.",1991-11-26,"The title of the patent is optical interconnector and highly interconnected, learning neural network incorporating optical interconnector therein and its abstract is a variable weight optical interconnector is disclosed to include a projecting device and an interconnection weighting device remote from the projecting device. the projecting device projects a distribution of interconnecting light beams when illuminated by a spatially-modulated light pattern. the weighting device includes a photosensitive screen provided in optical alignment with the projecting device to independently control the intensity of each projected interconnecting beam to thereby assign an interconnection weight to each such beam. further in accordance with the present invention, a highly-interconnected optical neural network having learning capability is disclosed as including a spatial light modulator, a detecting device, an interconnector according to the present invention, and a device responsive to detection signals generated by the detecting device to modify the interconnection weights assigned by the photosensitive screen of the interconnector. dated 1991-11-26"
5071231,bidirectional spatial light modulator for neural network computers,"digital data processing unit includes two slms assembled back-to-back with a common photoreceptor, to form a bidirectional spatial light modulator (bslm) which facilitates the flow of data in the forward and reverse directions. an image can be written from the left side of the bslm and read from the left or right side of the unit. an image can also be written from the right side and read from the right or left or both sides of the unit. the photoreceptor sums the light image intensities when data is concurrently written from both sides into the photoreceptor.",1991-12-10,"The title of the patent is bidirectional spatial light modulator for neural network computers and its abstract is digital data processing unit includes two slms assembled back-to-back with a common photoreceptor, to form a bidirectional spatial light modulator (bslm) which facilitates the flow of data in the forward and reverse directions. an image can be written from the left side of the bslm and read from the left or right side of the unit. an image can also be written from the right side and read from the right or left or both sides of the unit. the photoreceptor sums the light image intensities when data is concurrently written from both sides into the photoreceptor. dated 1991-12-10"
5073867,digital neural network processing elements,"a preprocessing device is disclosed which performs a linear transformation or power series expansion transformation on the input signals to a neural network node. the outputs of the preprocessing device are combined as a product of these linear transformations and compared to a threshold. this processing element configuration, combining a transformation with a product and threshold comparison, performs non-linear transformations between input data and output results. as a result, this processing element will, by itself, produce both linearly and non-linearly separable boolean logic functions. when this processing element is configured in a network, a two layer neural network can be created which will solve any arbitrary decision making function. this element can be configured in a probability based binary tree neural network which is validatable and verifiable in which the threshold comparison operation can be eliminated. the element can also be implemented in binary logic for ultra high speed. if the linkage element performs the power series expansion, a universal or general purpose element is created.",1991-12-17,"The title of the patent is digital neural network processing elements and its abstract is a preprocessing device is disclosed which performs a linear transformation or power series expansion transformation on the input signals to a neural network node. the outputs of the preprocessing device are combined as a product of these linear transformations and compared to a threshold. this processing element configuration, combining a transformation with a product and threshold comparison, performs non-linear transformations between input data and output results. as a result, this processing element will, by itself, produce both linearly and non-linearly separable boolean logic functions. when this processing element is configured in a network, a two layer neural network can be created which will solve any arbitrary decision making function. this element can be configured in a probability based binary tree neural network which is validatable and verifiable in which the threshold comparison operation can be eliminated. the element can also be implemented in binary logic for ultra high speed. if the linkage element performs the power series expansion, a universal or general purpose element is created. dated 1991-12-17"
5075868,memory modification of artificial neural networks,"an artificial neural network, which has a plurality of neurons each receiving a plurality of inputs whose effect is determined by adjust able weights at synapses individually connecting the inputs to the neuron to provide a sum signal to a sigmoidal function generator determining the output of the neuron, undergoes memory modification by a steepest-descent method in which individual variations in the outputs of the neurons are successively generated by small perturbations imposed on the sum signals. as each variation is generated on the output of a neuron, an overall error of all the neuron outputs in relation to their desired values is measured and compared to this error prior to the perturbation. the difference in these errors, with adjustments which may be changed as the neuron outputs converge toward their desired values, is used to modify each weight of the neuron presently subjected to the perturbation.",1991-12-24,"The title of the patent is memory modification of artificial neural networks and its abstract is an artificial neural network, which has a plurality of neurons each receiving a plurality of inputs whose effect is determined by adjust able weights at synapses individually connecting the inputs to the neuron to provide a sum signal to a sigmoidal function generator determining the output of the neuron, undergoes memory modification by a steepest-descent method in which individual variations in the outputs of the neurons are successively generated by small perturbations imposed on the sum signals. as each variation is generated on the output of a neuron, an overall error of all the neuron outputs in relation to their desired values is measured and compared to this error prior to the perturbation. the difference in these errors, with adjustments which may be changed as the neuron outputs converge toward their desired values, is used to modify each weight of the neuron presently subjected to the perturbation. dated 1991-12-24"
5075869,neural network exhibiting improved tolerance to temperature and power supply variations,"an analog neural network is described which provides a means for reducing the sensitivity of the network to temperature and power supply variations. a first circuit is utilized for generating a signal which exhibits a dependence on temperature corresponding to the variation normally experienced by the network in response to a change in temperature. a second circuit is employed to generate another signal which exhibits a similar dependence, except on power supply variations. by coupling these signals as inputs to the neural network the sensitivity of the network to temperature and power supply fluctuations is essentially nulified.",1991-12-24,"The title of the patent is neural network exhibiting improved tolerance to temperature and power supply variations and its abstract is an analog neural network is described which provides a means for reducing the sensitivity of the network to temperature and power supply variations. a first circuit is utilized for generating a signal which exhibits a dependence on temperature corresponding to the variation normally experienced by the network in response to a change in temperature. a second circuit is employed to generate another signal which exhibits a similar dependence, except on power supply variations. by coupling these signals as inputs to the neural network the sensitivity of the network to temperature and power supply fluctuations is essentially nulified. dated 1991-12-24"
5075871,variable gain neural network image processing system,"a neural-simulating system for an image processing system includes a plurality of networks arranged in a plurality of layers, the output signals of ones of the layers provide input signals to the others of the layers. each of the plurality of layers include a plurality of neurons operating in parallel on the input signals to the layers. the plurality of neurons within a layer are arrange in groups. each of the neurons within a group operate in parallel on the input signals. each neuron within a group of neuron operates to extract a specific feature of an area of the image being processed. each of the neutrons derives output signals from the input signals representing the relative weight of the input signal and a gain weight associated with each of the neurons applied thereto based upon a continuously differential transfer function for each function.",1991-12-24,"The title of the patent is variable gain neural network image processing system and its abstract is a neural-simulating system for an image processing system includes a plurality of networks arranged in a plurality of layers, the output signals of ones of the layers provide input signals to the others of the layers. each of the plurality of layers include a plurality of neurons operating in parallel on the input signals to the layers. the plurality of neurons within a layer are arrange in groups. each of the neurons within a group operate in parallel on the input signals. each neuron within a group of neuron operates to extract a specific feature of an area of the image being processed. each of the neutrons derives output signals from the input signals representing the relative weight of the input signal and a gain weight associated with each of the neurons applied thereto based upon a continuously differential transfer function for each function. dated 1991-12-24"
5075889,arrangement of data cells and neural network system utilizing such an arrangement,"an arrangement of data cells which stores at least one matrix of data words which are arranged in rows and columns, the matrix being distributed in the arrangement in order to deliver/receive, via a single bus, permuted data words which correspond either to a row or to a column of the matrix. each data cell is connected to the single bus via series-connected switches which are associated with a respective addressing mode, the switches which address a same word of a same mode being directly controlled by a same selection signal. circulation members enable the original order of the data on the bus to be restored. an arrangement of this kind is used in a layered neural network system for executing the error backpropagation algorithm. application: calculator, microprocessors, processor, neural network system. reference: fig. 4.",1991-12-24,"The title of the patent is arrangement of data cells and neural network system utilizing such an arrangement and its abstract is an arrangement of data cells which stores at least one matrix of data words which are arranged in rows and columns, the matrix being distributed in the arrangement in order to deliver/receive, via a single bus, permuted data words which correspond either to a row or to a column of the matrix. each data cell is connected to the single bus via series-connected switches which are associated with a respective addressing mode, the switches which address a same word of a same mode being directly controlled by a same selection signal. circulation members enable the original order of the data on the bus to be restored. an arrangement of this kind is used in a layered neural network system for executing the error backpropagation algorithm. application: calculator, microprocessors, processor, neural network system. reference: fig. 4. dated 1991-12-24"
5077677,probabilistic inference gate,"the present system performs linear transformations on input probabilities and produces outputs which indicate the likelihood of one or more events. the transformation performed is a product of linear transforms such as p.sub.o =[a.sub.j p.sub.j +b.sub.j ].multidot.[a.sub.k p.sub.k +b.sub.k ] where p.sub.j and p.sub.k are input probabilities, p.sub.o is an output event probability and a.sub.j, b.sub.j, a.sub.k and b.sub.k are transformation constants. the system includes a basic processing unit or computational unit which performs a probabilistic gate operation to convert two input probability signals into one output probability signal where the output probability is equal to the product of linear transformations of the input probabilities. by appropriate selection of transformation constants logical and probabilistic gates performing the functions of and, nand, or, nor, xor, not, implies and not implies can be created. the basic unit can include three multipliers and two adders if a discrete component hardwired version is needed for speed or a single multiplier/adder, associated storage and multiplex circuits can be used to accomplish the functions of the hardwired version for economy. this basic unit can also be provided as a software implementation, can be implemented as a hardwired decision tree element implementation or implemented as a universal probabilistic processor and provided with a bus communication structure to create expert systems or neural networks suitable for specific tasks. the basic units can be combined to produce a virtual basic building block which has more virtual processors than physical processors to improve processor utilization. the building blocks can be combined into an array to produce either a high speed expert system or a high speed neural network.",1991-12-31,"The title of the patent is probabilistic inference gate and its abstract is the present system performs linear transformations on input probabilities and produces outputs which indicate the likelihood of one or more events. the transformation performed is a product of linear transforms such as p.sub.o =[a.sub.j p.sub.j +b.sub.j ].multidot.[a.sub.k p.sub.k +b.sub.k ] where p.sub.j and p.sub.k are input probabilities, p.sub.o is an output event probability and a.sub.j, b.sub.j, a.sub.k and b.sub.k are transformation constants. the system includes a basic processing unit or computational unit which performs a probabilistic gate operation to convert two input probability signals into one output probability signal where the output probability is equal to the product of linear transformations of the input probabilities. by appropriate selection of transformation constants logical and probabilistic gates performing the functions of and, nand, or, nor, xor, not, implies and not implies can be created. the basic unit can include three multipliers and two adders if a discrete component hardwired version is needed for speed or a single multiplier/adder, associated storage and multiplex circuits can be used to accomplish the functions of the hardwired version for economy. this basic unit can also be provided as a software implementation, can be implemented as a hardwired decision tree element implementation or implemented as a universal probabilistic processor and provided with a bus communication structure to create expert systems or neural networks suitable for specific tasks. the basic units can be combined to produce a virtual basic building block which has more virtual processors than physical processors to improve processor utilization. the building blocks can be combined into an array to produce either a high speed expert system or a high speed neural network. dated 1991-12-31"
5080464,optical neural network apparatus using primary processing,"for inputting a two-dimensional image into an optical neural network apparatus, a primary processing device is used to extract the characteristic feature of an object pattern. thereafter, compressed information as a result of the above processing is inputted into the input of the all-optical type optical neural network apparatus, that implements parallel processings adaptively through optical computing, at individual points on the input of the same. therefore, the primary processing device that was capable of dealing with only logical input information until now can process even vague input information by the use of the optical neural network apparatus located on the later stage. on the other hand, the use of the primary processing device on the previous stage of the optical neural network apparatus enables a limited input range of the optical neural network apparatus to be expanded together with the assurance of higher degree processing by inputting into the optical neural network apparatus results of the characteristic feature extraction from an original image.",1992-01-14,"The title of the patent is optical neural network apparatus using primary processing and its abstract is for inputting a two-dimensional image into an optical neural network apparatus, a primary processing device is used to extract the characteristic feature of an object pattern. thereafter, compressed information as a result of the above processing is inputted into the input of the all-optical type optical neural network apparatus, that implements parallel processings adaptively through optical computing, at individual points on the input of the same. therefore, the primary processing device that was capable of dealing with only logical input information until now can process even vague input information by the use of the optical neural network apparatus located on the later stage. on the other hand, the use of the primary processing device on the previous stage of the optical neural network apparatus enables a limited input range of the optical neural network apparatus to be expanded together with the assurance of higher degree processing by inputting into the optical neural network apparatus results of the characteristic feature extraction from an original image. dated 1992-01-14"
5083285,matrix-structured neural network with learning circuitry,"a multi-layer perceptron circuit device using integrated configuration which is capable of incorporating self-learning function and which is easily extendable. the device includes: at least one synapse blocks containing: a plurality of synapses for performing weight calculation on input signals to obtain output signals, which are arranged in planar array defined by a first and a second directions; input signal lines for transmitting the input signals to the synapses, arranged along the first direction; and output signal lines for transmitting the output signal from the synapses, arranged along the second direction not identical to the first direction; at least one input neuron blocks containing a plurality of neurons to be connected with the input signal lines; and at least one output neuron blocks containing a plurality of neurons to be connected with the output signal lines.",1992-01-21,"The title of the patent is matrix-structured neural network with learning circuitry and its abstract is a multi-layer perceptron circuit device using integrated configuration which is capable of incorporating self-learning function and which is easily extendable. the device includes: at least one synapse blocks containing: a plurality of synapses for performing weight calculation on input signals to obtain output signals, which are arranged in planar array defined by a first and a second directions; input signal lines for transmitting the input signals to the synapses, arranged along the first direction; and output signal lines for transmitting the output signal from the synapses, arranged along the second direction not identical to the first direction; at least one input neuron blocks containing a plurality of neurons to be connected with the input signal lines; and at least one output neuron blocks containing a plurality of neurons to be connected with the output signal lines. dated 1992-01-21"
5086405,floating point adder circuit using neural network,a floating point adder circuit using neural network concepts and having high speed operation is obtained by a controlling circuit using a comparator and an operating circuit using an adder and a subtractor.,1992-02-04,The title of the patent is floating point adder circuit using neural network and its abstract is a floating point adder circuit using neural network concepts and having high speed operation is obtained by a controlling circuit using a comparator and an operating circuit using an adder and a subtractor. dated 1992-02-04
5086479,information processing system using neural network learning function,"an information processing apparatus using a neural network learning function has, in one embodiment, a computer system and a pattern recognition apparatus associated with each other via a communication cable. the computer system includes a learning section having a first neural network and serves to adjust the weights of connection therein as a result of learning with a learning data signal supplied thereto from the pattern recognition apparatus via the communication cable. the pattern recognition apparatus includes an associative output section having a second neural network and receives data on the adjusted weights from the learning section via the communication cable to reconstruct the second neural network with the data on the adjusted weights. the pattern recognition apparatus with the associative output section having the reconstructed second neural network performs pattern recognition independently of the computer system with the communication cable being brought into an electrical isolation mode.",1992-02-04,"The title of the patent is information processing system using neural network learning function and its abstract is an information processing apparatus using a neural network learning function has, in one embodiment, a computer system and a pattern recognition apparatus associated with each other via a communication cable. the computer system includes a learning section having a first neural network and serves to adjust the weights of connection therein as a result of learning with a learning data signal supplied thereto from the pattern recognition apparatus via the communication cable. the pattern recognition apparatus includes an associative output section having a second neural network and receives data on the adjusted weights from the learning section via the communication cable to reconstruct the second neural network with the data on the adjusted weights. the pattern recognition apparatus with the associative output section having the reconstructed second neural network performs pattern recognition independently of the computer system with the communication cable being brought into an electrical isolation mode. dated 1992-02-04"
5087826,multi-layer neural network employing multiplexed output neurons,"a multi-layer electrically trainable analog neural network employing multiplexed output neurons having inputs organized into two groups, external and recurrent (i.e., feedback). each layer of the network comprises a matrix of synapse cells which implement a matrix multiplication between an input vector and a weight matrix. in normal operation, an external input vector coupled to the first synaptic array generates a sigmoid response at the output of a set of neurons. this output is then fed back to the next and subsequent layers of the network as a recurrent input vector. the output of second layer processing is generated by the same neurons used in first layer processing. thus, the neural network of the present invention can handle n-layer operation by using recurrent connections and a single set of multiplexed output neurons.",1992-02-11,"The title of the patent is multi-layer neural network employing multiplexed output neurons and its abstract is a multi-layer electrically trainable analog neural network employing multiplexed output neurons having inputs organized into two groups, external and recurrent (i.e., feedback). each layer of the network comprises a matrix of synapse cells which implement a matrix multiplication between an input vector and a weight matrix. in normal operation, an external input vector coupled to the first synaptic array generates a sigmoid response at the output of a set of neurons. this output is then fed back to the next and subsequent layers of the network as a recurrent input vector. the output of second layer processing is generated by the same neurons used in first layer processing. thus, the neural network of the present invention can handle n-layer operation by using recurrent connections and a single set of multiplexed output neurons. dated 1992-02-11"
5089862,monocrystalline three-dimensional integrated circuit,""" a monocrystalline monolith contains a 3-d array of interconnected lattice-matched devices (which may be of one kind exclusively, or that kind in combination with one or more other kinds) performing digital, analog, image-processing, or neural-network functions, singly or in combination. localized inclusions of lattice-matched metal and (or) insulator can exist in the monolith, but monolith-wide layers of insulator are avoided. the devices may be self-isolated, junction-isolated, or insulator-isolated, and may include but not be limited to mosfets, bjts, jfets, mfets, ccds, resistors, and capacitors. the monolith is fabricated in a single apparatus using a process such as mbe or sputter epitaxy executed in a continuous or quasicontinuous manner under automatic control, and supplanting hundreds of discrete steps with handling and storage steps interpolated. """"writing"""" on the growing crystal is done during crystal growth by methods that may include but not be limited to ion beams, laser beams, patterned light exposures, and physical masks. the interior volume of the fabrication apparatus is far cleaner and more highly controlled than that of a clean room. the apparatus is highly replicated and is amenable to mass production. the product has unprecedented volumetric function density, and high performance stems from short signal paths, low parasitic loading, and 3-d architecture. high reliability stems from contamination-free fabrication, small signal-arrival skew, and generous noise margins. economy stems from mass-produced factory apparatus, automatic ic manufacture, and high ic yield. among the ic products are fast and efficient memories with equally fast and efficient error-correction abilities, crosstalk-free operational amplifiers, and highly paralleled and copiously interconnected neural networks. """,1992-02-18,"The title of the patent is monocrystalline three-dimensional integrated circuit and its abstract is "" a monocrystalline monolith contains a 3-d array of interconnected lattice-matched devices (which may be of one kind exclusively, or that kind in combination with one or more other kinds) performing digital, analog, image-processing, or neural-network functions, singly or in combination. localized inclusions of lattice-matched metal and (or) insulator can exist in the monolith, but monolith-wide layers of insulator are avoided. the devices may be self-isolated, junction-isolated, or insulator-isolated, and may include but not be limited to mosfets, bjts, jfets, mfets, ccds, resistors, and capacitors. the monolith is fabricated in a single apparatus using a process such as mbe or sputter epitaxy executed in a continuous or quasicontinuous manner under automatic control, and supplanting hundreds of discrete steps with handling and storage steps interpolated. """"writing"""" on the growing crystal is done during crystal growth by methods that may include but not be limited to ion beams, laser beams, patterned light exposures, and physical masks. the interior volume of the fabrication apparatus is far cleaner and more highly controlled than that of a clean room. the apparatus is highly replicated and is amenable to mass production. the product has unprecedented volumetric function density, and high performance stems from short signal paths, low parasitic loading, and 3-d architecture. high reliability stems from contamination-free fabrication, small signal-arrival skew, and generous noise margins. economy stems from mass-produced factory apparatus, automatic ic manufacture, and high ic yield. among the ic products are fast and efficient memories with equally fast and efficient error-correction abilities, crosstalk-free operational amplifiers, and highly paralleled and copiously interconnected neural networks. "" dated 1992-02-18"
5091780,a trainable security system emthod for the same,"a security system comprised of a device for monitoring an area under surveillance. the monitoring device produces images of the area. the security system is also comprised of a device for processing the images to determine whether the area is in a desired state or an undesired state. the processing device is trainable to learn the difference between the desired state and the undesired state. in a preferred embodiment, the monitoring device includes a video camera which produces video images of the area and the processing device includes a computer simulating a neural network. a method for determining whether an area under surveillance is in a desired state or an undesired state. the method comprises the steps of collecting data in a computer about the area which defines when the area is in the desired state or the undesired state. next, training the computer from the collected data to essentially correctly identify when the area is in the desired state or in the undesired state while the area is under surveillance. next, performing surveillance of the area with a computer such that the computer determines whether the area is in a desired state or the undesired state.",1992-02-25,"The title of the patent is a trainable security system emthod for the same and its abstract is a security system comprised of a device for monitoring an area under surveillance. the monitoring device produces images of the area. the security system is also comprised of a device for processing the images to determine whether the area is in a desired state or an undesired state. the processing device is trainable to learn the difference between the desired state and the undesired state. in a preferred embodiment, the monitoring device includes a video camera which produces video images of the area and the processing device includes a computer simulating a neural network. a method for determining whether an area under surveillance is in a desired state or an undesired state. the method comprises the steps of collecting data in a computer about the area which defines when the area is in the desired state or the undesired state. next, training the computer from the collected data to essentially correctly identify when the area is in the desired state or in the undesired state while the area is under surveillance. next, performing surveillance of the area with a computer such that the computer determines whether the area is in a desired state or the undesired state. dated 1992-02-25"
5091864,systolic processor elements for a neural network,"a neural net signal processor provided with a single layer neural net constituted of n neuron circuits which sums the results of the multiplication of each of n input signals xj(j=1 to n) by a coefficient mij to produce a multiply-accumulate value ##equ1## thereof, in which input signals xj(j=1 to n) for input to the single layer neural net are input as serial input data, comprising: a multiplicity of systolic processor elements spe-1(i=1 to m), each comprised of a two-state input data delay latch; a coefficient memory; means for multiplying and summing for multiply-accumulate output operations; an accumulator; a multiplexor for selecting a preceding stage multiply-accumulate output sk(k=1 to i-1) and the multiply-accumulate product si computed by the said circuit; wherein the multiplicity of systolic processor elements are serially connected to form an element array and element multiply-accumulate output operations are executed sequentially to obtain the serial multiply-accumulate outputs si(i=1 to m) of one layer from the element array.",1992-02-25,"The title of the patent is systolic processor elements for a neural network and its abstract is a neural net signal processor provided with a single layer neural net constituted of n neuron circuits which sums the results of the multiplication of each of n input signals xj(j=1 to n) by a coefficient mij to produce a multiply-accumulate value ##equ1## thereof, in which input signals xj(j=1 to n) for input to the single layer neural net are input as serial input data, comprising: a multiplicity of systolic processor elements spe-1(i=1 to m), each comprised of a two-state input data delay latch; a coefficient memory; means for multiplying and summing for multiply-accumulate output operations; an accumulator; a multiplexor for selecting a preceding stage multiply-accumulate output sk(k=1 to i-1) and the multiply-accumulate product si computed by the said circuit; wherein the multiplicity of systolic processor elements are serially connected to form an element array and element multiply-accumulate output operations are executed sequentially to obtain the serial multiply-accumulate outputs si(i=1 to m) of one layer from the element array. dated 1992-02-25"
5091965,video image processing apparatus,"a video image processing apparatus in which an analog video image can be satisfactorily converted to a binary value by calculating theshold values of respective neurons and coupling coefficients of respective synapses of a neural network circuit on the basis of the input analog video image and a pre-determined function. by arranging so a difference component e between the input analog video image and the binary value video image is defined as ##equ1## where .alpha. is the coefficient, u.sub.(i) is the value which results from converting the input analog video image into the binary value, and p.sub.(i,j) is the value obtained from the function and g.sub.(i) is the value which is obtained from the function and the input analog video image, it is possible to convert the video image into the binary value by the use of the neural network circuit. further, by setting that the function to have a frequency characteristic of a human's eyes, it is possible to obtain a binary value video image which is excellent from a human's visual standpoint. furthermore, if the input analog video image is a computer hologram and the function is a window function which indicates a range of a desired video image in a reproduced image which results from fourier-transforming the computer hologram, the noise in the range of the desired video image in the reproduced video image can be reduced to provide an excellent reproduced image.",1992-02-25,"The title of the patent is video image processing apparatus and its abstract is a video image processing apparatus in which an analog video image can be satisfactorily converted to a binary value by calculating theshold values of respective neurons and coupling coefficients of respective synapses of a neural network circuit on the basis of the input analog video image and a pre-determined function. by arranging so a difference component e between the input analog video image and the binary value video image is defined as ##equ1## where .alpha. is the coefficient, u.sub.(i) is the value which results from converting the input analog video image into the binary value, and p.sub.(i,j) is the value obtained from the function and g.sub.(i) is the value which is obtained from the function and the input analog video image, it is possible to convert the video image into the binary value by the use of the neural network circuit. further, by setting that the function to have a frequency characteristic of a human's eyes, it is possible to obtain a binary value video image which is excellent from a human's visual standpoint. furthermore, if the input analog video image is a computer hologram and the function is a window function which indicates a range of a desired video image in a reproduced image which results from fourier-transforming the computer hologram, the noise in the range of the desired video image in the reproduced video image can be reduced to provide an excellent reproduced image. dated 1992-02-25"
5092343,waveform analysis apparatus and method using neural network techniques,"a waveform analysis assembly (10) includes a sensor (12) for detecting physiological electrical and mechanical signals produced by the body. an extraction neural network (22, 22') will learn a repetitive waveform of the electrical signal, store the waveform in memory (18), extract the waveform from the electrical signal, store the location times of occurrences of the waveform, and subtract the waveform from the electrical signal. each significantly different waveform in the electrical signal is learned and extracted. a single or multilayer layer neural network (22, 22') accomplishes the learning and extraction with either multiple passes over the electrical signal or accomplishes the learning and extraction of all waveforms in a single pass over the electrical signal. a reducer (20) receives the stored waveforms and times and reduces them into features characterizing the waveforms. a classifier neural network (36) analyzes the features by classifying them through nonliner mapping techniques within the network representing diseased states and produces results of diseased states based on learned features of the normal and patient groups.",1992-03-03,"The title of the patent is waveform analysis apparatus and method using neural network techniques and its abstract is a waveform analysis assembly (10) includes a sensor (12) for detecting physiological electrical and mechanical signals produced by the body. an extraction neural network (22, 22') will learn a repetitive waveform of the electrical signal, store the waveform in memory (18), extract the waveform from the electrical signal, store the location times of occurrences of the waveform, and subtract the waveform from the electrical signal. each significantly different waveform in the electrical signal is learned and extracted. a single or multilayer layer neural network (22, 22') accomplishes the learning and extraction with either multiple passes over the electrical signal or accomplishes the learning and extraction of all waveforms in a single pass over the electrical signal. a reducer (20) receives the stored waveforms and times and reduces them into features characterizing the waveforms. a classifier neural network (36) analyzes the features by classifying them through nonliner mapping techniques within the network representing diseased states and produces results of diseased states based on learned features of the normal and patient groups. dated 1992-03-03"
5093781,cellular network assignment processor using minimum/maximum convergence technique,"a cellular network assignment processor (10) for solving optimization problems utilizing a neural network architecture having a matrix of simple processing cells (12) that are highly interconnected in a regular structure. the cells (12) accept as input, costs in an assignment problem. the position of each cell (12) corresponds to the position of the cost in the associated constraint space of the assignment problem. each cell (12) is capable of receiving, storing and transmitting cost values and is also capable of determining if it is the maximum or the minimum of cells (12) to which it's connected. operating on one row of cells (12) at a time the processor (10) determines if a conflict exists between selected connected cells (12) until a cell (12) with no conflict is found in each row. the end result is a chosen cell (12), in each row, the chosen cells (12) together representing a valid solution to the assignment problem.",1992-03-03,"The title of the patent is cellular network assignment processor using minimum/maximum convergence technique and its abstract is a cellular network assignment processor (10) for solving optimization problems utilizing a neural network architecture having a matrix of simple processing cells (12) that are highly interconnected in a regular structure. the cells (12) accept as input, costs in an assignment problem. the position of each cell (12) corresponds to the position of the cost in the associated constraint space of the assignment problem. each cell (12) is capable of receiving, storing and transmitting cost values and is also capable of determining if it is the maximum or the minimum of cells (12) to which it's connected. operating on one row of cells (12) at a time the processor (10) determines if a conflict exists between selected connected cells (12) until a cell (12) with no conflict is found in each row. the end result is a chosen cell (12), in each row, the chosen cells (12) together representing a valid solution to the assignment problem. dated 1992-03-03"
5093792,combustion prediction and discrimination apparatus for an internal combustion engine and control apparatus therefor,"an apparatus for predicting and discriminating whether or not misfire, knocking and the like will occur from the cylinder pressure before the occurrence of the misfire, the knocking and the like by the use of a three layered neural network. the cylinder pressure signal detected by a cylinder pressure sensor is sampled and input to each of the elements of the input layer. the signal then is modulated corresponding to the strength (weight) of the connection between each of the elements, and transmitted to the hidden and output layers. the magnitude of signal from the elements of the output layer represents the prediction and discrimination results. the weight is learned and determined by a back propagation method.",1992-03-03,"The title of the patent is combustion prediction and discrimination apparatus for an internal combustion engine and control apparatus therefor and its abstract is an apparatus for predicting and discriminating whether or not misfire, knocking and the like will occur from the cylinder pressure before the occurrence of the misfire, the knocking and the like by the use of a three layered neural network. the cylinder pressure signal detected by a cylinder pressure sensor is sampled and input to each of the elements of the input layer. the signal then is modulated corresponding to the strength (weight) of the connection between each of the elements, and transmitted to the hidden and output layers. the magnitude of signal from the elements of the output layer represents the prediction and discrimination results. the weight is learned and determined by a back propagation method. dated 1992-03-03"
5093899,neural network with normalized learning constant for high-speed stable learning,"the present invention is concerned with a signal processing system having a learning function pursuant to the back-propagation learning rule by the neural network, in which the learning rate is dynamically changed as a function of input values to effect high-speed stable learning. the signal processing system of the present invention is so arranged that, by executing signal processing for the input signals by the recurrent network formed by units each corresponding to a neuron, the features of the sequential time series pattern such as voice signals fluctuating on the time axis can be extracted through learning the coupling state of the recurrent network. the present invention modifies the prior art weight change algorithm .delta.w.sub.ji(n+1) =.eta...delta..sub.pi.+.alpha..w.sub.ji(n) into .delta.w.sub.ji(n+1) =.eta...beta.(.alpha..sub.pj o.sub.pi)+.alpha..w.sub.ji(n) where .beta..sub.j =1/(.sigma..sub.i o.sub.pi.sup.2 +1) is used to normalize the learning constant.",1992-03-03,"The title of the patent is neural network with normalized learning constant for high-speed stable learning and its abstract is the present invention is concerned with a signal processing system having a learning function pursuant to the back-propagation learning rule by the neural network, in which the learning rate is dynamically changed as a function of input values to effect high-speed stable learning. the signal processing system of the present invention is so arranged that, by executing signal processing for the input signals by the recurrent network formed by units each corresponding to a neuron, the features of the sequential time series pattern such as voice signals fluctuating on the time axis can be extracted through learning the coupling state of the recurrent network. the present invention modifies the prior art weight change algorithm .delta.w.sub.ji(n+1) =.eta...delta..sub.pi.+.alpha..w.sub.ji(n) into .delta.w.sub.ji(n+1) =.eta...beta.(.alpha..sub.pj o.sub.pi)+.alpha..w.sub.ji(n) where .beta..sub.j =1/(.sigma..sub.i o.sub.pi.sup.2 +1) is used to normalize the learning constant. dated 1992-03-03"
5093900,reconfigurable neural network,"realization of a reconfigurable neuron for use in a neural network has been achieved using analog techniques. in the reconfigurable neuron, digital input data are multiplied by programmable digital weights in a novel connection structure whose output permits straightforward summation of the products thereby forming a sum signal. the sum signal is multiplied by a programmable scalar, in general, 1, when the input data and the digital weights are binary. when the digital input data and the digital weights are multilevel, the scalar in each reconfigurable neuron is programmed to be a fraction which corresponds to the bit position in the digital data representation, that is, a programmable scalar of 1/2, 1/4, 1/8, and so on. the signal formed by scalar multiplication is passed through a programmable build out circuit which permits neural network reconfiguration by interconnection of one neuron to one or more other neurons. following the build out circuit, the output signal therefrom is supplied to one input of a differential comparator for the reconfigurable neuron. the differential comparator receives its other input from a supplied reference potential. in general, the comparator and reference potential level are designed to generate the nonlinearity for the neuron. one common nonlinearity is a hard limiter function. the present neuron offers the capability of synthesizing other nonlinear transfer functions by utilizing several reference potential levels connected through a controllable switching circuit.",1992-03-03,"The title of the patent is reconfigurable neural network and its abstract is realization of a reconfigurable neuron for use in a neural network has been achieved using analog techniques. in the reconfigurable neuron, digital input data are multiplied by programmable digital weights in a novel connection structure whose output permits straightforward summation of the products thereby forming a sum signal. the sum signal is multiplied by a programmable scalar, in general, 1, when the input data and the digital weights are binary. when the digital input data and the digital weights are multilevel, the scalar in each reconfigurable neuron is programmed to be a fraction which corresponds to the bit position in the digital data representation, that is, a programmable scalar of 1/2, 1/4, 1/8, and so on. the signal formed by scalar multiplication is passed through a programmable build out circuit which permits neural network reconfiguration by interconnection of one neuron to one or more other neurons. following the build out circuit, the output signal therefrom is supplied to one input of a differential comparator for the reconfigurable neuron. the differential comparator receives its other input from a supplied reference potential. in general, the comparator and reference potential level are designed to generate the nonlinearity for the neuron. one common nonlinearity is a hard limiter function. the present neuron offers the capability of synthesizing other nonlinear transfer functions by utilizing several reference potential levels connected through a controllable switching circuit. dated 1992-03-03"
5095443,plural neural network system having a successive approximation learning method,"a neural network structure includes input units for receiving input data, and a plurality of neural networks connected in parallel and connected to the input units. the plurality of neural networks learn in turn correspondence between the input data and teacher data so that the difference between the input data and the teacher becomes small. the neural network structure further includes output units connected to the plurality of neural networks, for outputting a result of learning on the basis of the results of learning in the plurality of neural networks.",1992-03-10,"The title of the patent is plural neural network system having a successive approximation learning method and its abstract is a neural network structure includes input units for receiving input data, and a plurality of neural networks connected in parallel and connected to the input units. the plurality of neural networks learn in turn correspondence between the input data and teacher data so that the difference between the input data and the teacher becomes small. the neural network structure further includes output units connected to the plurality of neural networks, for outputting a result of learning on the basis of the results of learning in the plurality of neural networks. dated 1992-03-10"
5095459,optical neural network,"an optical neural network which imitates a biological neural network, to provide an associative and/or pattern recognition function, is made of light emitting elements to represent an input neuron state vector, a correlation matrix which modulates light according to stored vector information, light receiving elements, an accumulator and a comparator to perform a threshold function. a stored vector closest to an input vector can be found from a large amount of information without increasing the system size by dividing the correlation matrix and the input neuron state vector with time division techniques, frequency modulation or phase modulation techniques. positive and negative valves can also be provided with similar techniques.",1992-03-10,"The title of the patent is optical neural network and its abstract is an optical neural network which imitates a biological neural network, to provide an associative and/or pattern recognition function, is made of light emitting elements to represent an input neuron state vector, a correlation matrix which modulates light according to stored vector information, light receiving elements, an accumulator and a comparator to perform a threshold function. a stored vector closest to an input vector can be found from a large amount of information without increasing the system size by dividing the correlation matrix and the input neuron state vector with time division techniques, frequency modulation or phase modulation techniques. positive and negative valves can also be provided with similar techniques. dated 1992-03-10"
5099114,optical wavelength demultiplexer,"an optical wavelength demultiplexer including an optical conversion device which converts a difference in wavelengths of a plurality of input signals into a difference in spatial power distribution of the input light signals, and a pattern recognition element for recognizing patterns of the spatial power distribution and taking out output signals. at the output portion of the optical conversion device, spatial power distributions are formed which are different for different wavelengths. after converting the spatial power distributions by the pattern recognition element into electrical signals, pattern recognition of the signals is performed to regenerate the original input signals with their respective wavelengths. the optical conversion device uses a diffractive grating or a combination of an optical multimode circuit, an optical multimode fiber, and a plurality of optical wavelengths. the pattern recognition element is constructed by a combination of a photo-detector array and a neural network, or a combination of a hologram element, a photo-detector array and a neural network.",1992-03-24,"The title of the patent is optical wavelength demultiplexer and its abstract is an optical wavelength demultiplexer including an optical conversion device which converts a difference in wavelengths of a plurality of input signals into a difference in spatial power distribution of the input light signals, and a pattern recognition element for recognizing patterns of the spatial power distribution and taking out output signals. at the output portion of the optical conversion device, spatial power distributions are formed which are different for different wavelengths. after converting the spatial power distributions by the pattern recognition element into electrical signals, pattern recognition of the signals is performed to regenerate the original input signals with their respective wavelengths. the optical conversion device uses a diffractive grating or a combination of an optical multimode circuit, an optical multimode fiber, and a plurality of optical wavelengths. the pattern recognition element is constructed by a combination of a photo-detector array and a neural network, or a combination of a hologram element, a photo-detector array and a neural network. dated 1992-03-24"
5099434,continuous-time optical neural network,"an all-optical, continuous-time, recurrent neural network is disclosed which is capable of executing a broad class of energy-minimizing neural net algorithms. the network is a resonator which contains a saturable, two-beam amplifier; two volume holograms; and a linear, two-beam amplifier. the saturable amplifier permits, through the use of a spatially patterned signal beam, the realization of a two-dimensional optical neuron array; the two volume holograms provide adaptive, global network interconnectivity; and the linear amplifier supplies sufficient resonator gain to permit convergent operation of the network.",1992-03-24,"The title of the patent is continuous-time optical neural network and its abstract is an all-optical, continuous-time, recurrent neural network is disclosed which is capable of executing a broad class of energy-minimizing neural net algorithms. the network is a resonator which contains a saturable, two-beam amplifier; two volume holograms; and a linear, two-beam amplifier. the saturable amplifier permits, through the use of a spatially patterned signal beam, the realization of a two-dimensional optical neuron array; the two volume holograms provide adaptive, global network interconnectivity; and the linear amplifier supplies sufficient resonator gain to permit convergent operation of the network. dated 1992-03-24"
5103431,apparatus for detecting sonar signals embedded in noise,"apparatus for detecting sonar signals embedded in noise includes a neural network trained to detect signals in response to the slope of amplitude rank ordered noise corrected powers. a detector detects an analog waveform. means samples and digitizes the analog waveform to obtain digital samples which in turn are passed through a cosine window. the digital samples are fourier transformed into conjugate sets of complex numbers representing amplitude and phase. one conjugate set of the complex numbers are discarded, and the remaining complex numbers ranked according to frequency. the sum of the square of the real and imaginary component of each of the remaining complex numbers in a frequency band are provided to obtain a corresponding series of representing estimated power ranked by frequency over the band. the noise contained in subbands of the band is estimated. each estimated power is then divided by the estimated noise of the subband containing the estimated power to obtain corresponding noise corrected powers, which are ranked ordered according to amplitude. the amplitude rank ordered noise powers are provided to corresponding inputs of the neural network.",1992-04-07,"The title of the patent is apparatus for detecting sonar signals embedded in noise and its abstract is apparatus for detecting sonar signals embedded in noise includes a neural network trained to detect signals in response to the slope of amplitude rank ordered noise corrected powers. a detector detects an analog waveform. means samples and digitizes the analog waveform to obtain digital samples which in turn are passed through a cosine window. the digital samples are fourier transformed into conjugate sets of complex numbers representing amplitude and phase. one conjugate set of the complex numbers are discarded, and the remaining complex numbers ranked according to frequency. the sum of the square of the real and imaginary component of each of the remaining complex numbers in a frequency band are provided to obtain a corresponding series of representing estimated power ranked by frequency over the band. the noise contained in subbands of the band is estimated. each estimated power is then divided by the estimated noise of the subband containing the estimated power to obtain corresponding noise corrected powers, which are ranked ordered according to amplitude. the amplitude rank ordered noise powers are provided to corresponding inputs of the neural network. dated 1992-04-07"
5103488,method of and device for moving image contour recognition,the recognition method is applied to visual telephony image coding. matrices of digital samples relevant to the individual frames of the video transmission are submitted to a first processing whereby the foreground region containing the figure is identified. the information concerning the elements of such a region is then processed by edge recognition algorithms to detect a group of elements possibly belonging to the contour. the group of elements is analyzed to select a sequence of elements distributed on the average along a line. the sequency of elements is processed by a neural network to build up the continuous contour which is then coded.,1992-04-07,The title of the patent is method of and device for moving image contour recognition and its abstract is the recognition method is applied to visual telephony image coding. matrices of digital samples relevant to the individual frames of the video transmission are submitted to a first processing whereby the foreground region containing the figure is identified. the information concerning the elements of such a region is then processed by edge recognition algorithms to detect a group of elements possibly belonging to the contour. the group of elements is analyzed to select a sequence of elements distributed on the average along a line. the sequency of elements is processed by a neural network to build up the continuous contour which is then coded. dated 1992-04-07
5103496,artificial neural network system for memory modification,"an artificial neural network, which has a plurality of neurons each receiving a plurality of inputs whose effect is determined by adjust able weights at synapses individually connecting the inputs to the neuron to provide a sum signal to a sigmoidal function generator determining the output of the neuron, undergoes memory modification by a steepest-descent method in which individual variations in the outputs of the neurons are successively generated by small perturbations imposed on the sum signals. as each variation is generated on the output of a neuron, an overall error of all the neuron outputs in relation to their desired values is measured and compared to this error prior to the perturbation. the difference in these errors, with adjustments which may be changed as the neuron outputs converge toward their desired values, is used to modify each weight of the neuron presently subjected to the perturbation.",1992-04-07,"The title of the patent is artificial neural network system for memory modification and its abstract is an artificial neural network, which has a plurality of neurons each receiving a plurality of inputs whose effect is determined by adjust able weights at synapses individually connecting the inputs to the neuron to provide a sum signal to a sigmoidal function generator determining the output of the neuron, undergoes memory modification by a steepest-descent method in which individual variations in the outputs of the neurons are successively generated by small perturbations imposed on the sum signals. as each variation is generated on the output of a neuron, an overall error of all the neuron outputs in relation to their desired values is measured and compared to this error prior to the perturbation. the difference in these errors, with adjustments which may be changed as the neuron outputs converge toward their desired values, is used to modify each weight of the neuron presently subjected to the perturbation. dated 1992-04-07"
5105468,time delay neural network for printed and cursive handwritten character recognition,"a time delay neural network is defined having feature detection layers which are constrained for extracting features and subsampling a sequence of feature vectors input to the particular feature detection layer. output from the network for both digit and uppercase letters is provided by an output classification layer which is fully connected to the final feature detection layer. each feature vector relates to coordinate information about the original character preserved in a temporal order together with additional information related to the original character at the particular coordinate point. such additional information may include local geometric information, local pen information, and phantom stroke coordinate information relating to connecting segments between the end point of one stroke and the beginning point of another stroke. the network is also defined to increase the number of feature elements in each feature vector from one feature detection layer to the next. that is, as the network is reducing its dependence on temporally related features, it is increasing its dependence on more features and more complex features.",1992-04-14,"The title of the patent is time delay neural network for printed and cursive handwritten character recognition and its abstract is a time delay neural network is defined having feature detection layers which are constrained for extracting features and subsampling a sequence of feature vectors input to the particular feature detection layer. output from the network for both digit and uppercase letters is provided by an output classification layer which is fully connected to the final feature detection layer. each feature vector relates to coordinate information about the original character preserved in a temporal order together with additional information related to the original character at the particular coordinate point. such additional information may include local geometric information, local pen information, and phantom stroke coordinate information relating to connecting segments between the end point of one stroke and the beginning point of another stroke. the network is also defined to increase the number of feature elements in each feature vector from one feature detection layer to the next. that is, as the network is reducing its dependence on temporally related features, it is increasing its dependence on more features and more complex features. dated 1992-04-14"
5107442,adaptive neural network image processing system,"a neural-simulating system for processing input stimuli includes a plurality of layers, each layer receives layer input signals and generates layer output signals, the layer input signals include signals from the input stimuli and ones of the layer output signals from only previous layers within the plurality of layers. each of the plurality of layers includes a plurality of neurons operating in parallel on the layer input signals applied to the plurality of layers. each of the neurons derives neuron output signals from a continuously differentiable transfer function for each of the neurons based upon a combination of sets of weights associated with the neurons and the layer input signals. an adaptive network is associated with each neuron for generating weight correction signals based upon gradient estimate signals and convergence factors signals of each neuron and for processing the weight correction signals to thereby modify the weights associated with each neuron. an error measuring circuit generates relative powered error signals for use in generating the gradient estimate signals and the convergence factors signals.",1992-04-21,"The title of the patent is adaptive neural network image processing system and its abstract is a neural-simulating system for processing input stimuli includes a plurality of layers, each layer receives layer input signals and generates layer output signals, the layer input signals include signals from the input stimuli and ones of the layer output signals from only previous layers within the plurality of layers. each of the plurality of layers includes a plurality of neurons operating in parallel on the layer input signals applied to the plurality of layers. each of the neurons derives neuron output signals from a continuously differentiable transfer function for each of the neurons based upon a combination of sets of weights associated with the neurons and the layer input signals. an adaptive network is associated with each neuron for generating weight correction signals based upon gradient estimate signals and convergence factors signals of each neuron and for processing the weight correction signals to thereby modify the weights associated with each neuron. an error measuring circuit generates relative powered error signals for use in generating the gradient estimate signals and the convergence factors signals. dated 1992-04-21"
5107454,pattern associative memory system,"in a pattern associative memory system, an error correcting circuit is constructed in the form of a neural network. a memory condition of the error correcting circuit is established according to a back propagation method. if a memory pattern is recollected, an output from the error correcting circuit is again inputted to the error correcting circuit for feedback, thereby repeatedly performing error correction calculations of a pattern as the basis of recollection.",1992-04-21,"The title of the patent is pattern associative memory system and its abstract is in a pattern associative memory system, an error correcting circuit is constructed in the form of a neural network. a memory condition of the error correcting circuit is established according to a back propagation method. if a memory pattern is recollected, an output from the error correcting circuit is again inputted to the error correcting circuit for feedback, thereby repeatedly performing error correction calculations of a pattern as the basis of recollection. dated 1992-04-21"
5108170,perimetric instrument,"the present invention relates to a perimetric instrument for measuring a range of a visual field of an eye of a man, and more particularly to a perimetric instrument which includes an abnormal visual field pattern analogical inferring section using a multilayer neural network and can analogically infer an abnormal visual field pattern of a measurement object person. further, the present invention provides a perimetric instrument which can automatically make a determination of an additional target. in particular, according to the present invention, a multilayer neural network including an input layer, hidden layers and an output layer introduces a neural weight ratio which is determined based on responses when the visual field is normal and abnormal, and as a response from a responding section is inputted to the input layer while an output from the output layer is sent out to an analogical inferring section, the analogical inferring section can analogically infer an abnormal visual field pattern of the measurement object person. accordingly, it is possible to help judgment of an abnormal visual field pattern by a measurer, and since also labor, time and so forth of the measurer are reduced, burdens to the measurer and them easurement object person can be reduced remarkably.",1992-04-28,"The title of the patent is perimetric instrument and its abstract is the present invention relates to a perimetric instrument for measuring a range of a visual field of an eye of a man, and more particularly to a perimetric instrument which includes an abnormal visual field pattern analogical inferring section using a multilayer neural network and can analogically infer an abnormal visual field pattern of a measurement object person. further, the present invention provides a perimetric instrument which can automatically make a determination of an additional target. in particular, according to the present invention, a multilayer neural network including an input layer, hidden layers and an output layer introduces a neural weight ratio which is determined based on responses when the visual field is normal and abnormal, and as a response from a responding section is inputted to the input layer while an output from the output layer is sent out to an analogical inferring section, the analogical inferring section can analogically infer an abnormal visual field pattern of the measurement object person. accordingly, it is possible to help judgment of an abnormal visual field pattern by a measurer, and since also labor, time and so forth of the measurer are reduced, burdens to the measurer and them easurement object person can be reduced remarkably. dated 1992-04-28"
5109275,printing signal correction and printer operation control apparatus utilizing neural network,"an apparatus for printing signal correction and printer operation control, for use in applications such as color copiers, utilizes a neural network to convert input image signals, derived for example by scanning and analyzing an original image, into printing density signals which are supplied to a printer. in addition, a detection signal expressing at least one internal environmental condition of the printer, such as temperature, is inputted to the neural network, so that the output printing density signals are automatically compensated for changes in internal environment of the printer.",1992-04-28,"The title of the patent is printing signal correction and printer operation control apparatus utilizing neural network and its abstract is an apparatus for printing signal correction and printer operation control, for use in applications such as color copiers, utilizes a neural network to convert input image signals, derived for example by scanning and analyzing an original image, into printing density signals which are supplied to a printer. in addition, a detection signal expressing at least one internal environmental condition of the printer, such as temperature, is inputted to the neural network, so that the output printing density signals are automatically compensated for changes in internal environment of the printer. dated 1992-04-28"
5111516,apparatus for visual recognition,"a basic image of objects is extracted from a two-dimensional image of objects. geometrical elements of the objects are extracted from the extracted basic image. the objects to be recognized are identified by searching a combination of the geometrical elements which match a geometrical model and then utilizing candidate position/orientation of the objects to be recognized, said candidate position/orientation being determined from a relationship in relative position between the combination of geometrical elements and the geometrical model. mesh cells fixed to the geometrical model are mapped on the basic image based on the candidate position/orientation. in addition, verification is made as to whether an image of the geometrical model mapped by the candidate position/orientation is accurately matched with an image of one of the objects to be recognized, through a neural network to which values got from the basic image included in the individual mesh cells are to be applied as input values. combination weight factors employed in the neural network are learned according to the verified results. it is also possible to recognize the multi-purpose objects according to how to learn the combination weight factors.",1992-05-05,"The title of the patent is apparatus for visual recognition and its abstract is a basic image of objects is extracted from a two-dimensional image of objects. geometrical elements of the objects are extracted from the extracted basic image. the objects to be recognized are identified by searching a combination of the geometrical elements which match a geometrical model and then utilizing candidate position/orientation of the objects to be recognized, said candidate position/orientation being determined from a relationship in relative position between the combination of geometrical elements and the geometrical model. mesh cells fixed to the geometrical model are mapped on the basic image based on the candidate position/orientation. in addition, verification is made as to whether an image of the geometrical model mapped by the candidate position/orientation is accurately matched with an image of one of the objects to be recognized, through a neural network to which values got from the basic image included in the individual mesh cells are to be applied as input values. combination weight factors employed in the neural network are learned according to the verified results. it is also possible to recognize the multi-purpose objects according to how to learn the combination weight factors. dated 1992-05-05"
5111531,process control using neural network,a control system and method for a continuous process in which a trained neural network predicts the value of an indirectly controlled process variable and the values of directly controlled process variables are changed to cause the predicted value to approach a desired value.,1992-05-05,The title of the patent is process control using neural network and its abstract is a control system and method for a continuous process in which a trained neural network predicts the value of an indirectly controlled process variable and the values of directly controlled process variables are changed to cause the predicted value to approach a desired value. dated 1992-05-05
5113482,neural network model for reaching a goal state,""" an object, such as a robot, is located at an initial state in a finite state space area and moves under the control of the unsupervised neural network model of the invention. the network instructs the object to move in one of several directions from the initial state. upon reaching another state, the model again instructs the object to move in one of several directions. these instructions continue until either: a) the object has completed a cycle by ending up back at a state it has been to previously during this cycle, or b) the object has completed a cycle by reaching the goal state. if the object ends up back at a state it has been to previously during this cycle, the neural network model ends the cycle and immediately begins a new cycle from the present location. when the object reaches the goal state, the neural network model learns that this path is productive towards reaching the goal state, and is given delayed reinforcement in the form of a """"reward"""". upon reaching a state, the neural network model calculates a level of satisfaction with its progress towards reaching the goal state. if the level of satisfaction is low, the neural network model is more likely to override what has been learned thus far and deviate from a path known to lead to the goal state to experiment with new and possibly better paths. if the level of satisfaction is high, the neural network model is much less likely to experiment with new paths. the object is guaranteed to eventually find the best path to the goal state from any starting location, assuming that the level of satisfaction does not exceed a threshold point where learning ceases. """,1992-05-12,"The title of the patent is neural network model for reaching a goal state and its abstract is "" an object, such as a robot, is located at an initial state in a finite state space area and moves under the control of the unsupervised neural network model of the invention. the network instructs the object to move in one of several directions from the initial state. upon reaching another state, the model again instructs the object to move in one of several directions. these instructions continue until either: a) the object has completed a cycle by ending up back at a state it has been to previously during this cycle, or b) the object has completed a cycle by reaching the goal state. if the object ends up back at a state it has been to previously during this cycle, the neural network model ends the cycle and immediately begins a new cycle from the present location. when the object reaches the goal state, the neural network model learns that this path is productive towards reaching the goal state, and is given delayed reinforcement in the form of a """"reward"""". upon reaching a state, the neural network model calculates a level of satisfaction with its progress towards reaching the goal state. if the level of satisfaction is low, the neural network model is more likely to override what has been learned thus far and deviate from a path known to lead to the goal state to experiment with new and possibly better paths. if the level of satisfaction is high, the neural network model is much less likely to experiment with new paths. the object is guaranteed to eventually find the best path to the goal state from any starting location, assuming that the level of satisfaction does not exceed a threshold point where learning ceases. "" dated 1992-05-12"
5113483,neural network with semi-localized non-linear mapping of the input space,a neural network includes an input layer comprising a plurality of input units (24) interconnected to a hidden layer with a plurality of hidden units (26) disposed therein through an interconnection matrix (28). each of the hidden units (26) is a single output that is connected to output units (32) in an output layer through an interconnection matrix (30). each of the interconnections between one of the hidden units (26) to one of the output units (32) has a weight associated therewith. each of the hidden units (26) has an activation in the i'th dimension and extending across all the other dimensions in a non-localized manner in accordance with the following equation: ##equ1## that the network learns by the back propagation method to vary the output weights and the parameters of the activation function .mu..sub.hi and .sigma..sub.hi.,1992-05-12,The title of the patent is neural network with semi-localized non-linear mapping of the input space and its abstract is a neural network includes an input layer comprising a plurality of input units (24) interconnected to a hidden layer with a plurality of hidden units (26) disposed therein through an interconnection matrix (28). each of the hidden units (26) is a single output that is connected to output units (32) in an output layer through an interconnection matrix (30). each of the interconnections between one of the hidden units (26) to one of the output units (32) has a weight associated therewith. each of the hidden units (26) has an activation in the i'th dimension and extending across all the other dimensions in a non-localized manner in accordance with the following equation: ##equ1## that the network learns by the back propagation method to vary the output weights and the parameters of the activation function .mu..sub.hi and .sigma..sub.hi. dated 1992-05-12
5113484,rank filter using neural newwork,"a rank filter is provided which can be used for improving an image signal degraded by noise, while at the same time maintaining edge information. the rank filter is implemented by using a neural network and obtains a high processing speed with a simple circuit arrangement, as compared to conventional rank filters, hpfs, lpfs and average filters. the rank filter using the concept of a neural network includes decoder devices, a comparison device and a counter.",1992-05-12,"The title of the patent is rank filter using neural newwork and its abstract is a rank filter is provided which can be used for improving an image signal degraded by noise, while at the same time maintaining edge information. the rank filter is implemented by using a neural network and obtains a high processing speed with a simple circuit arrangement, as compared to conventional rank filters, hpfs, lpfs and average filters. the rank filter using the concept of a neural network includes decoder devices, a comparison device and a counter. dated 1992-05-12"
5113485,optical neural network system,"an optical system of an optical neural network model for parallel data processing is disclosed. taking advantage of the fact that an auto-correlation matrix is symmetric with respect to a main diagonal and the weights for modulating the values of diagonals of the auto-correlation matrix are equal to each other, the configuration of an optical modulation unit is simplified by a one-dimensional modulation array on the one hand, and both positive and negative weights are capable of being computed at the same time on the other hand. in particular, the optical system makes up a second-order neural network exhibiting invariant characteristics against the translation and scale.",1992-05-12,"The title of the patent is optical neural network system and its abstract is an optical system of an optical neural network model for parallel data processing is disclosed. taking advantage of the fact that an auto-correlation matrix is symmetric with respect to a main diagonal and the weights for modulating the values of diagonals of the auto-correlation matrix are equal to each other, the configuration of an optical modulation unit is simplified by a one-dimensional modulation array on the one hand, and both positive and negative weights are capable of being computed at the same time on the other hand. in particular, the optical system makes up a second-order neural network exhibiting invariant characteristics against the translation and scale. dated 1992-05-12"
5115492,digital correlators incorporating analog neural network structures operated on a bit-sliced basis,"plural-bit digital input signals to be subjected to weighted summation are bit-sliced; and a number n of respective first through n.sup.th weighted summations of the bits of the digital input signals in each bit slice are performed, resulting in a respective set of first through n.sup.th partial weighted summation results. each weighted summation of a bit slice of the digital input signals is performed using a capacitive network that generates partial weighted summation results in the analog regime; and analog-to-digital conversion circuitry digitizes the partial weighted summation results. weighted summations of the digitized partial weighted summation results of similar ordinal number are then performed to generate first through n.sup.th final weighted summation results in digital form, which results are respective correlations of the pattern of the digital input signals with the patterns of weights established by the capacitive networks. a neural net layer can be formed by combining such weighted summation circuitry with digital circuits processing each final weighted summation result non-linearly, with a system function that is sigmoidal.",1992-05-19,"The title of the patent is digital correlators incorporating analog neural network structures operated on a bit-sliced basis and its abstract is plural-bit digital input signals to be subjected to weighted summation are bit-sliced; and a number n of respective first through n.sup.th weighted summations of the bits of the digital input signals in each bit slice are performed, resulting in a respective set of first through n.sup.th partial weighted summation results. each weighted summation of a bit slice of the digital input signals is performed using a capacitive network that generates partial weighted summation results in the analog regime; and analog-to-digital conversion circuitry digitizes the partial weighted summation results. weighted summations of the digitized partial weighted summation results of similar ordinal number are then performed to generate first through n.sup.th final weighted summation results in digital form, which results are respective correlations of the pattern of the digital input signals with the patterns of weights established by the capacitive networks. a neural net layer can be formed by combining such weighted summation circuitry with digital circuits processing each final weighted summation result non-linearly, with a system function that is sigmoidal. dated 1992-05-19"
5119438,recognizing apparatus,"a recognizing apparatus is provided for recognizing a class to which an inputted characteristic pattern belongs from among a plurality of classes to be discriminated using a neural network. the classes are classified into a plurality of categories. the apparatus includes a network selecting portion for selecting a category to which the inputted characteristic pattern belongs and for selecting a neural network for use in discriminating the class to which the inputted characteristic pattern belongs in the selected category. the apparatus further includes a network memory portion, a network setting portion and a details discriminating portion. the network memory portion stores structures of a plurality of neural networks which have finished learning for respective categories, weights of the neural networks set by the learning and a plurality of discriminating algorithms to be used when the classes are discriminated by the neural networks. the network setting portion sets the structure and weights of a neural network selected by the network selecting portion and a discriminating alogrithm appropriate to the selected category. the details discriminating portion recognizes the class to which the inputted characteristic pattern belongs by performing the details discriminating operation using the neural network set by the neural network setting portion.",1992-06-02,"The title of the patent is recognizing apparatus and its abstract is a recognizing apparatus is provided for recognizing a class to which an inputted characteristic pattern belongs from among a plurality of classes to be discriminated using a neural network. the classes are classified into a plurality of categories. the apparatus includes a network selecting portion for selecting a category to which the inputted characteristic pattern belongs and for selecting a neural network for use in discriminating the class to which the inputted characteristic pattern belongs in the selected category. the apparatus further includes a network memory portion, a network setting portion and a details discriminating portion. the network memory portion stores structures of a plurality of neural networks which have finished learning for respective categories, weights of the neural networks set by the learning and a plurality of discriminating algorithms to be used when the classes are discriminated by the neural networks. the network setting portion sets the structure and weights of a neural network selected by the network selecting portion and a discriminating alogrithm appropriate to the selected category. the details discriminating portion recognizes the class to which the inputted characteristic pattern belongs by performing the details discriminating operation using the neural network set by the neural network setting portion. dated 1992-06-02"
5119469,neural network with weight adjustment based on prior history of input signals,a dynamically stable associative learning neural network system include a plurality of synapses and a non-linear function circuit and includes an adaptive weight circuit for adjusting the weight of each synapse based upon the present signal and the prior history of signals applied to the input of the particular synapse and the present signal and the prior history of signals applied to the input of a predetermined set of other collateral synapses. a flow-through neuron circuit embodiment includes a flow-through synapse having a predetermined fixed weight. a neural network is formed employing neuron circuits of both the above types. a set of flow-through neuron circuits are connected by flow-through synapses to form separate paths between each input terminal and a corresponding output terminal. other neuron circuits having only adjustable weight synapses are included within the network. this neuron network is initialized by setting the adjustable synapses at some value near the minimum weight. the neural network is taught by successively application of sets of inputs signals to the input terminals until a dynamic equilibrium is reached.,1992-06-02,The title of the patent is neural network with weight adjustment based on prior history of input signals and its abstract is a dynamically stable associative learning neural network system include a plurality of synapses and a non-linear function circuit and includes an adaptive weight circuit for adjusting the weight of each synapse based upon the present signal and the prior history of signals applied to the input of the particular synapse and the present signal and the prior history of signals applied to the input of a predetermined set of other collateral synapses. a flow-through neuron circuit embodiment includes a flow-through synapse having a predetermined fixed weight. a neural network is formed employing neuron circuits of both the above types. a set of flow-through neuron circuits are connected by flow-through synapses to form separate paths between each input terminal and a corresponding output terminal. other neuron circuits having only adjustable weight synapses are included within the network. this neuron network is initialized by setting the adjustable synapses at some value near the minimum weight. the neural network is taught by successively application of sets of inputs signals to the input terminals until a dynamic equilibrium is reached. dated 1992-06-02
5121467,neural network/expert system process control system and method,"a neural network/expert system process control system and method combines the decision-making capabilities of expert systems with the predictive capabilities of neural networks for improved process control. neural networks provide predictions of measurements which are difficult to make, or supervisory or regulatory control changes which are difficult to implement using classical control techniques. expert systems make decisions automatically based on knowledge which is well-known and can be expressed in rules or other knowledge representation forms. sensor and laboratory data is effictively used. in one approach, the output data from the neural network can be used by the controller in controlling the process, and the expert system can make a decision using sensor or lab data to control the controller(s). in another approach, the output data of the neural network can be used by the expert system in making its decision, and control of the process carried out using lab or sensor data. in another approach, the output data can be used both to control the process and to make decisions.",1992-06-09,"The title of the patent is neural network/expert system process control system and method and its abstract is a neural network/expert system process control system and method combines the decision-making capabilities of expert systems with the predictive capabilities of neural networks for improved process control. neural networks provide predictions of measurements which are difficult to make, or supervisory or regulatory control changes which are difficult to implement using classical control techniques. expert systems make decisions automatically based on knowledge which is well-known and can be expressed in rules or other knowledge representation forms. sensor and laboratory data is effictively used. in one approach, the output data from the neural network can be used by the controller in controlling the process, and the expert system can make a decision using sensor or lab data to control the controller(s). in another approach, the output data of the neural network can be used by the expert system in making its decision, and control of the process carried out using lab or sensor data. in another approach, the output data can be used both to control the process and to make decisions. dated 1992-06-09"
5124918,neural-based autonomous robotic system,"a system for achieving ambulatory control of a multi-legged system employs stimulus and response-based modeling. a adapted neural network-based system is employed for dictating motion characteristics of a plurality of leg members. rhythmic movements necessary to accomplish motion are provided by a series of signal generators. a first signal generator functions as a pacemaker governing overall system characteristics. one or more axis control signals are provided to a plurality of leg controllers, which axis control signals work in concert with a system coordination signal from the pacemaker. a sensory mechanism is also employed to govern ambulatory system responses.",1992-06-23,"The title of the patent is neural-based autonomous robotic system and its abstract is a system for achieving ambulatory control of a multi-legged system employs stimulus and response-based modeling. a adapted neural network-based system is employed for dictating motion characteristics of a plurality of leg members. rhythmic movements necessary to accomplish motion are provided by a series of signal generators. a first signal generator functions as a pacemaker governing overall system characteristics. one or more axis control signals are provided to a plurality of leg controllers, which axis control signals work in concert with a system coordination signal from the pacemaker. a sensory mechanism is also employed to govern ambulatory system responses. dated 1992-06-23"
5129037,neural network for performing beta-token partitioning in a rete network,"a method and system for beta-token partitioning a target expert system program. the target expert system program is first compiled to form a rete network for execution on a single processor, the compilation including directives for collecting selected processing statistics. the target expert system program is then executed on a single processor, generating during execution processing statistics in connection with each node of the rete network. the processing statistics are then applied to a programmed neural network to identify nodes in the rete network for beta-token partitioning, and the target expert system program is then recompiled to form a rete network for execution on multiple processors, the rete network being beta-token partitioned at nodes identified by the neural network.",1992-07-07,"The title of the patent is neural network for performing beta-token partitioning in a rete network and its abstract is a method and system for beta-token partitioning a target expert system program. the target expert system program is first compiled to form a rete network for execution on a single processor, the compilation including directives for collecting selected processing statistics. the target expert system program is then executed on a single processor, generating during execution processing statistics in connection with each node of the rete network. the processing statistics are then applied to a programmed neural network to identify nodes in the rete network for beta-token partitioning, and the target expert system program is then recompiled to form a rete network for execution on multiple processors, the rete network being beta-token partitioned at nodes identified by the neural network. dated 1992-07-07"
5129038,neural network with selective error reduction to increase learning speed,"an improved iterative learning machine having a plurality of multi-input/single-output signal processing units connected in a hierarchical structure includes a weight coefficient change control unit which controls weight change quantities for those multi-input/single-output signal processing units having iteratively reduced errors thereby increasing the learning speed, contrary to conventional learning machines which perform a learning operation in order to minimize a square error of multi-input/single-output signal processing units in the highest hierarchy of the hierarchical structure.",1992-07-07,"The title of the patent is neural network with selective error reduction to increase learning speed and its abstract is an improved iterative learning machine having a plurality of multi-input/single-output signal processing units connected in a hierarchical structure includes a weight coefficient change control unit which controls weight change quantities for those multi-input/single-output signal processing units having iteratively reduced errors thereby increasing the learning speed, contrary to conventional learning machines which perform a learning operation in order to minimize a square error of multi-input/single-output signal processing units in the highest hierarchy of the hierarchical structure. dated 1992-07-07"
5129039,recurrent neural network with variable size intermediate layer,"the present invention is concerned with a signal processing system having a learning function pursuant to the back-propagation learning rule by the neural network, in which the learning rate is dynamically changed as a function of input values to effect high-speed stable learning. the signal processing system of the present invention is so arranged that, by executing signal processing for the input signals by the recurrent network formed by units each corresponding to a neuron, the features of the sequential time series pattern such as voice signals fluctuating on the time axis can be extracted through learning the coupling state of the recurrent network. the present invention is also concerned with a learning processing system adapted to cause the signal processing section formed by a neural network to undergo signal processing pursuant to the back-propagation learning rule, wherein the local minimum state in the course of the learning processing may be avoided by learning the coefficient of coupling strength while simultaneously increasing the number of the unit of the intermediate layer.",1992-07-07,"The title of the patent is recurrent neural network with variable size intermediate layer and its abstract is the present invention is concerned with a signal processing system having a learning function pursuant to the back-propagation learning rule by the neural network, in which the learning rate is dynamically changed as a function of input values to effect high-speed stable learning. the signal processing system of the present invention is so arranged that, by executing signal processing for the input signals by the recurrent network formed by units each corresponding to a neuron, the features of the sequential time series pattern such as voice signals fluctuating on the time axis can be extracted through learning the coupling state of the recurrent network. the present invention is also concerned with a learning processing system adapted to cause the signal processing section formed by a neural network to undergo signal processing pursuant to the back-propagation learning rule, wherein the local minimum state in the course of the learning processing may be avoided by learning the coefficient of coupling strength while simultaneously increasing the number of the unit of the intermediate layer. dated 1992-07-07"
5129040,neural network system for image processing,"a visual information processing device has a pair of neural networks which respectively comprise an upper layer and a lower layer of the device. each of the pair of neural networks comprises a semiconductor integrated circuit having a plurality of neuron circuit regions which are disposed in a matrix form, each of the neuron circuit regions performing a neuron function; a molecule film having a photoelectric function and provided on the semiconductor integrated circuit, the molecule film having (i) a plurality of t.sub.ij signal input sections each performing a wiring function among the plurality of neuron circuit regions, in each of which a t.sub.ij signal representing the bonding strength among the plurality of neuron circuit regions is optically written, and (ii) a plurality of video input sections each performing a sensor function of sensing a visual image in which one pixel corresponds to one neuron circuit region; and a wiring for electrically connecting the semiconductor integrated circuit and the molecule film. each of the plurality of neuron circuit regions is bonded with the neighboring neuron circuit regions in each of the pair of neural networks comprising the upper and lower layers, and each of the plurality of neuron circuit regions is bonded with the corresponding one between the pair of neural networks.",1992-07-07,"The title of the patent is neural network system for image processing and its abstract is a visual information processing device has a pair of neural networks which respectively comprise an upper layer and a lower layer of the device. each of the pair of neural networks comprises a semiconductor integrated circuit having a plurality of neuron circuit regions which are disposed in a matrix form, each of the neuron circuit regions performing a neuron function; a molecule film having a photoelectric function and provided on the semiconductor integrated circuit, the molecule film having (i) a plurality of t.sub.ij signal input sections each performing a wiring function among the plurality of neuron circuit regions, in each of which a t.sub.ij signal representing the bonding strength among the plurality of neuron circuit regions is optically written, and (ii) a plurality of video input sections each performing a sensor function of sensing a visual image in which one pixel corresponds to one neuron circuit region; and a wiring for electrically connecting the semiconductor integrated circuit and the molecule film. each of the plurality of neuron circuit regions is bonded with the neighboring neuron circuit regions in each of the pair of neural networks comprising the upper and lower layers, and each of the plurality of neuron circuit regions is bonded with the corresponding one between the pair of neural networks. dated 1992-07-07"
5129041,optical neural network processing element with multiple holographic element interconnects,"a neural network processing element uses primarily optical components to model a biological neuron having both spatial and temporal dependence. the neural network processing element includes a switch-controlled laser source, a multiple holographic lens, a spatial/temporal light modulator, and a photodetector array. laser beam control may be optical, electrical or acoustical, or a combination of these.",1992-07-07,"The title of the patent is optical neural network processing element with multiple holographic element interconnects and its abstract is a neural network processing element uses primarily optical components to model a biological neuron having both spatial and temporal dependence. the neural network processing element includes a switch-controlled laser source, a multiple holographic lens, a spatial/temporal light modulator, and a photodetector array. laser beam control may be optical, electrical or acoustical, or a combination of these. dated 1992-07-07"
5129042,sorting circuit using neural network,"a sorting circuit for arranging data in sequence according to the magnitudes of the data values, uses the concept of a neural network. the sorting circuit is constructed of shift registers, magnitude comparators, binary counters, binary bit separators and registers.",1992-07-07,"The title of the patent is sorting circuit using neural network and its abstract is a sorting circuit for arranging data in sequence according to the magnitudes of the data values, uses the concept of a neural network. the sorting circuit is constructed of shift registers, magnitude comparators, binary counters, binary bit separators and registers. dated 1992-07-07"
5130563,optoelectronic sensory neural network,"a neural network for processing sensory information. the network comprise one or more layers including interconnecting cells having individual states. each cell is connected to one or more neighboring cells. sensory signals and signals from interconnected neighboring cells control a current or a conductance within a cell to influence the cell's state. in some embodiments, the current or conductance of a cell can be controlled by a signal arising externally of the layer. each cell can comprise an electrical circuit which receives an input signal and causes a current corresponding to the signal to pass through a variable conductance. the conductance is a function of the states of the one or more interconnecting neighboring cells. proper interconnection of the cells on a layer can produce a neural network which is sensitive to predetermined patterns or the passage of such patterns across a sensor array whose signals are input into the network. the layers in the network can be made sensitive to distinct sensory parameters, so that networks which are sensitive to different wavelengths or polarizations of light energy can be produced.",1992-07-14,"The title of the patent is optoelectronic sensory neural network and its abstract is a neural network for processing sensory information. the network comprise one or more layers including interconnecting cells having individual states. each cell is connected to one or more neighboring cells. sensory signals and signals from interconnected neighboring cells control a current or a conductance within a cell to influence the cell's state. in some embodiments, the current or conductance of a cell can be controlled by a signal arising externally of the layer. each cell can comprise an electrical circuit which receives an input signal and causes a current corresponding to the signal to pass through a variable conductance. the conductance is a function of the states of the one or more interconnecting neighboring cells. proper interconnection of the cells on a layer can produce a neural network which is sensitive to predetermined patterns or the passage of such patterns across a sensor array whose signals are input into the network. the layers in the network can be made sensitive to distinct sensory parameters, so that networks which are sensitive to different wavelengths or polarizations of light energy can be produced. dated 1992-07-14"
5130936,method and apparatus for diagnostic testing including a neural network for determining testing sufficiency,"a diagnostic tester evaluates at least one inputted test signal corresponding to test data relating to at least one predetermined parameter of a system being tested, to produce first and second candidate signals corresponding respectively to first and second possible diagnoses of the condition of the system respectively having the first and second highest levels of certainty of being valid, and first and second certainty signals corresponding respectively to values of the first and second highest levels of certainty. the diagnostic tester further determines the sufficiency of the testing that has taken place responsive to the first and second certainty signals, and produces an output signal indicative of whether sufficient test data has been evaluated to declare a diagnosis. preferably, an uncertainty signal corresponding to a measure of the uncertainty that the evaluated at least one test signal can be validly evaluated is also produced and used to produce the output signal.",1992-07-14,"The title of the patent is method and apparatus for diagnostic testing including a neural network for determining testing sufficiency and its abstract is a diagnostic tester evaluates at least one inputted test signal corresponding to test data relating to at least one predetermined parameter of a system being tested, to produce first and second candidate signals corresponding respectively to first and second possible diagnoses of the condition of the system respectively having the first and second highest levels of certainty of being valid, and first and second certainty signals corresponding respectively to values of the first and second highest levels of certainty. the diagnostic tester further determines the sufficiency of the testing that has taken place responsive to the first and second certainty signals, and produces an output signal indicative of whether sufficient test data has been evaluated to declare a diagnosis. preferably, an uncertainty signal corresponding to a measure of the uncertainty that the evaluated at least one test signal can be validly evaluated is also produced and used to produce the output signal. dated 1992-07-14"
5130944,divider circuit adopting a neural network architecture to increase division processing speed and reduce hardware components,"a divider circuit for efficiently and quickly performing a hardware implemented division by adopting a neural network architecture. the circuit includes a series of cascaded subtracter components that complement the divisor input and effectively perform an adder function. the subtracters include a synaptic configuration consisting of pmos transistors, nmos transistors, and cmos inverters. the components are arranged in accordance with the predetermined connection strength assigned to each of the transistors and its respective position in the neural type network arrangement.",1992-07-14,"The title of the patent is divider circuit adopting a neural network architecture to increase division processing speed and reduce hardware components and its abstract is a divider circuit for efficiently and quickly performing a hardware implemented division by adopting a neural network architecture. the circuit includes a series of cascaded subtracter components that complement the divisor input and effectively perform an adder function. the subtracters include a synaptic configuration consisting of pmos transistors, nmos transistors, and cmos inverters. the components are arranged in accordance with the predetermined connection strength assigned to each of the transistors and its respective position in the neural type network arrangement. dated 1992-07-14"
5131072,neurocomputer with analog signal bus,"an analogue neuron processor (anp) performs an operation of sum-of-products of a time divisional analog input signal sequentially input from an analog signal bus and weight data and output an analog signal to an analog signal bus through a nonlinear circuit. a layered type or a feedback type neural network is formed of anps. the neural network reads necessary control data from a control pattern memory under the control of micro sequencer and reads the necessary weight data from the weight memory thereby realizing a neuron computer. the neuron computer connects a plurality of anps by using a single analog bus, thereby greatly decreasing the number of the wires used for the neural network and also decreasing the size of the circuit. a plurality of anps in a single layer simultaneously receives analog signal from an analog bus and carries out a parallel operation in the same time period and anps in different layers perform a parallel operation in a pipeline manner, thereby increasing a speed of an operation. accordingly, the prsent invention can provide a neuron computer with a high practicality.",1992-07-14,"The title of the patent is neurocomputer with analog signal bus and its abstract is an analogue neuron processor (anp) performs an operation of sum-of-products of a time divisional analog input signal sequentially input from an analog signal bus and weight data and output an analog signal to an analog signal bus through a nonlinear circuit. a layered type or a feedback type neural network is formed of anps. the neural network reads necessary control data from a control pattern memory under the control of micro sequencer and reads the necessary weight data from the weight memory thereby realizing a neuron computer. the neuron computer connects a plurality of anps by using a single analog bus, thereby greatly decreasing the number of the wires used for the neural network and also decreasing the size of the circuit. a plurality of anps in a single layer simultaneously receives analog signal from an analog bus and carries out a parallel operation in the same time period and anps in different layers perform a parallel operation in a pipeline manner, thereby increasing a speed of an operation. accordingly, the prsent invention can provide a neuron computer with a high practicality. dated 1992-07-14"
5132813,neural processor with holographic optical paths and nonlinear operating means,"an optical apparatus for simulating a highly interconnected neural network is disclosed as including a spatial light modulator (slm), an inputting device, a laser, a detecting device, and a page-oriented holographic component. the inputting device applies input signals to the slm. the holographic component optically interconnects n.sup.2 pixels defined on the spatial light modulator to n.sup.2 pixels defined on a detecting surface of the detecting device. the interconnections are made by n.sup.2 patterns of up to n.sup.2 interconnection weight encoded beams projected by n.sup.2 planar, or essentially two-dimensional, holograms arranged in a spatially localized array within the holographic component. the slm modulates the encoded beams and directs them onto the detecting surface wherein a parameter of the beams is evaluated at each pixel thereof. the evaluated parameter is transformed according to a nonlinear threshold function to provide transformed signals which can be fed back to the slm for further iterations.",1992-07-21,"The title of the patent is neural processor with holographic optical paths and nonlinear operating means and its abstract is an optical apparatus for simulating a highly interconnected neural network is disclosed as including a spatial light modulator (slm), an inputting device, a laser, a detecting device, and a page-oriented holographic component. the inputting device applies input signals to the slm. the holographic component optically interconnects n.sup.2 pixels defined on the spatial light modulator to n.sup.2 pixels defined on a detecting surface of the detecting device. the interconnections are made by n.sup.2 patterns of up to n.sup.2 interconnection weight encoded beams projected by n.sup.2 planar, or essentially two-dimensional, holograms arranged in a spatially localized array within the holographic component. the slm modulates the encoded beams and directs them onto the detecting surface wherein a parameter of the beams is evaluated at each pixel thereof. the evaluated parameter is transformed according to a nonlinear threshold function to provide transformed signals which can be fed back to the slm for further iterations. dated 1992-07-21"
5132835,continuous-time optical neural network process,"an all-optical, continuous-time, recurrent neural network is disclosed which is capable of executing a broad class of energy-minimizing neural net algorithms. the network is a resonator which contains a saturable, two-beam amplifier; two volume holograms; and a linear, two-beam amplifier. the saturable amplifier permits, through the use of a spatially patterned signal beam, the realization of a two-dimensional optical neuron array; the two volume holograms provide adaptive, global network interconnectivity; and the linear amplifier supplies sufficient resonator gain to permit convergent operation of the network.",1992-07-21,"The title of the patent is continuous-time optical neural network process and its abstract is an all-optical, continuous-time, recurrent neural network is disclosed which is capable of executing a broad class of energy-minimizing neural net algorithms. the network is a resonator which contains a saturable, two-beam amplifier; two volume holograms; and a linear, two-beam amplifier. the saturable amplifier permits, through the use of a spatially patterned signal beam, the realization of a two-dimensional optical neuron array; the two volume holograms provide adaptive, global network interconnectivity; and the linear amplifier supplies sufficient resonator gain to permit convergent operation of the network. dated 1992-07-21"
5133021,system for self-organization of stable category recognition codes for analog input patterns,"a neural network includes a feature representation field which receives input patterns. signals from the feature representative field select a category from a category representation field through a first adaptive filter. based on the selected category, a template pattern is applied to the feature representation field, and a match between the template and the input is determined. if the angle between the template vector and a vector within the representation field is too great, the selected category is reset. otherwise the category selection and template pattern are adapted to the input pattern as well as the previously stored template. a complex representation field includes signals normalized relative to signals across the field and feedback for pattern contrast enhancement.",1992-07-21,"The title of the patent is system for self-organization of stable category recognition codes for analog input patterns and its abstract is a neural network includes a feature representation field which receives input patterns. signals from the feature representative field select a category from a category representation field through a first adaptive filter. based on the selected category, a template pattern is applied to the feature representation field, and a match between the template and the input is determined. if the angle between the template vector and a vector within the representation field is too great, the selected category is reset. otherwise the category selection and template pattern are adapted to the input pattern as well as the previously stored template. a complex representation field includes signals normalized relative to signals across the field and feedback for pattern contrast enhancement. dated 1992-07-21"
5134396,method and apparatus for encoding and decoding data utilizing data compression and neural networks,"a method structure for the compression of data utilizes an encoder which effects a transform with the aid of a coding neural network, and a decoder which includes a matched decoding neural network with effects almost the inverse transform of the encoder. the method puts in competition m coding neural networks (30.sub.1 to 30.sub.m) wherein m>1 positioned at the transmission end which effects a same type of transform and the encoded data of one of which are transmitted, after selection (32, 33) at a given instant, towards a matched decoding neural network which forms part of a set of several matched neural networks (60.sub.1 to 60.sub.q) provided at the receiver end. learning is effected on the basis of predetermined samples. the encoder may comprise, in addition to the coding neural network (30.sub.1 to 30.sub.m), a matched decoding neural network (35.sub.1 to 35.sub.m) so as to effect the selection (32, 33) of the best coding neural network in accordance with an error criterion.",1992-07-28,"The title of the patent is method and apparatus for encoding and decoding data utilizing data compression and neural networks and its abstract is a method structure for the compression of data utilizes an encoder which effects a transform with the aid of a coding neural network, and a decoder which includes a matched decoding neural network with effects almost the inverse transform of the encoder. the method puts in competition m coding neural networks (30.sub.1 to 30.sub.m) wherein m>1 positioned at the transmission end which effects a same type of transform and the encoded data of one of which are transmitted, after selection (32, 33) at a given instant, towards a matched decoding neural network which forms part of a set of several matched neural networks (60.sub.1 to 60.sub.q) provided at the receiver end. learning is effected on the basis of predetermined samples. the encoder may comprise, in addition to the coding neural network (30.sub.1 to 30.sub.m), a matched decoding neural network (35.sub.1 to 35.sub.m) so as to effect the selection (32, 33) of the best coding neural network in accordance with an error criterion. dated 1992-07-28"
5134685,"neural node, a netowrk and a chaotic annealing optimization method for the network","the present invention is a node for a network that combines a hopfield and tank type neuron, having a sigmoid type transfer function, with a nonmonotonic neuron, having a transfer function such as a parabolic transfer function, to produce a neural node with a deterministic chaotic response suitable for quickly and globally solving optimizatioin problems and avoiding local minima. the node can be included in a completely connected single layer network. the hopfield neuron operates continuously while the nonmonotonic neuron operates periodically to prevent the network from getting stuck in a local optimum solution. the node can also be included in a local area architecture where local areas can be linked together in a hierarchy of nonmonotonic neurons.",1992-07-28,"The title of the patent is neural node, a netowrk and a chaotic annealing optimization method for the network and its abstract is the present invention is a node for a network that combines a hopfield and tank type neuron, having a sigmoid type transfer function, with a nonmonotonic neuron, having a transfer function such as a parabolic transfer function, to produce a neural node with a deterministic chaotic response suitable for quickly and globally solving optimizatioin problems and avoiding local minima. the node can be included in a completely connected single layer network. the hopfield neuron operates continuously while the nonmonotonic neuron operates periodically to prevent the network from getting stuck in a local optimum solution. the node can also be included in a local area architecture where local areas can be linked together in a hierarchy of nonmonotonic neurons. dated 1992-07-28"
5138695,systolic array image processing system,"a systolic array of processing elements is connected to receive weight inputs and multiplexed data inputs for operation in feedforward, partially-- or fully-connected neural network mode or in cooperative, competitive neural network mode. feature vector or two-dimensional image data is retrieved from external data memory and is transformed via input look-up table to input data for the systolic array that performs a convolution with kernal values as weight inputs. the convoluted image or neuron outputs from the systolic array are scaled and transformed via output look-up table for storage in the external data memory.",1992-08-11,"The title of the patent is systolic array image processing system and its abstract is a systolic array of processing elements is connected to receive weight inputs and multiplexed data inputs for operation in feedforward, partially-- or fully-connected neural network mode or in cooperative, competitive neural network mode. feature vector or two-dimensional image data is retrieved from external data memory and is transformed via input look-up table to input data for the systolic array that performs a convolution with kernal values as weight inputs. the convoluted image or neuron outputs from the systolic array are scaled and transformed via output look-up table for storage in the external data memory. dated 1992-08-11"
5138924,electronic musical instrument utilizing a neural network,"a musical tone parameter generating method and a musical tone generating device of this invention feature that when data inputted by a player is inputted into a neural network as input pattern, the neural network infers the parameters necessary to specify a musical tone wave form to be formed. this makes it possible to get parameters other than those stored in a memory by inferring, which increases variation of the musical tone to be generated.",1992-08-18,"The title of the patent is electronic musical instrument utilizing a neural network and its abstract is a musical tone parameter generating method and a musical tone generating device of this invention feature that when data inputted by a player is inputted into a neural network as input pattern, the neural network infers the parameters necessary to specify a musical tone wave form to be formed. this makes it possible to get parameters other than those stored in a memory by inferring, which increases variation of the musical tone to be generated. dated 1992-08-18"
5138928,rhythm pattern learning apparatus,"a rhythm pattern generating apparatus is provided having a layered neural network to perform learning with feedback to generate an output pattern signal indicative of a musical sound pattern. the output pattern signal is generated by the layered neural network with feedback in response to a performance operation of a player. the layered neural network generates the output pattern signal indicative of the musical sound pattern based on both an input pattern signal and a weight signal. the output pattern signal is fed back by the feedback circuit to the layered neural network to perform the learning process. a drum pad can be used to provide an input to the rhythm pattern generating apparatus or, specifically, to gate an input pattern selector for selecting input pattern signals. the layered neural network with the feedback can perform the learning process using a back propagation method. in the present invention, when a new rhythm pattern is input by a musician, an output pattern signal is generated through an analogy with the rhythm style of the musician.",1992-08-18,"The title of the patent is rhythm pattern learning apparatus and its abstract is a rhythm pattern generating apparatus is provided having a layered neural network to perform learning with feedback to generate an output pattern signal indicative of a musical sound pattern. the output pattern signal is generated by the layered neural network with feedback in response to a performance operation of a player. the layered neural network generates the output pattern signal indicative of the musical sound pattern based on both an input pattern signal and a weight signal. the output pattern signal is fed back by the feedback circuit to the layered neural network to perform the learning process. a drum pad can be used to provide an input to the rhythm pattern generating apparatus or, specifically, to gate an input pattern selector for selecting input pattern signals. the layered neural network with the feedback can perform the learning process using a back propagation method. in the present invention, when a new rhythm pattern is input by a musician, an output pattern signal is generated through an analogy with the rhythm style of the musician. dated 1992-08-18"
5140523,neural network for predicting lightning,"a system and method are provided for the automated prediction of lightning strikes in a set of different spatial regions for different times in the future. in a preferred embodiment, the system utilizes measurements of many weather phenomena. the types of measurements that can be utilized in approximately the same geographical region as that for which the strike predictions are made. this embodiment utilizes a correlation network to relate these weather measurements to future lightning strikes.",1992-08-18,"The title of the patent is neural network for predicting lightning and its abstract is a system and method are provided for the automated prediction of lightning strikes in a set of different spatial regions for different times in the future. in a preferred embodiment, the system utilizes measurements of many weather phenomena. the types of measurements that can be utilized in approximately the same geographical region as that for which the strike predictions are made. this embodiment utilizes a correlation network to relate these weather measurements to future lightning strikes. dated 1992-08-18"
5140530,genetic algorithm synthesis of neural networks,the disclosure relates to the use of genetic learning techniques to evolve neural network architectures for specific applications in which a general representation of neural network architecture is linked with a genetic learning strategy to create a very flexible environment for the construction of custom neural networks.,1992-08-18,The title of the patent is genetic algorithm synthesis of neural networks and its abstract is the disclosure relates to the use of genetic learning techniques to evolve neural network architectures for specific applications in which a general representation of neural network architecture is linked with a genetic learning strategy to create a very flexible environment for the construction of custom neural networks. dated 1992-08-18
5140531,analog neural nets supplied digital synapse signals on a bit-slice basis,"plural-bit digital input signals to be subjected to weighted summation in a neural net layer are bit-sliced; and a number n of respective first through n.sup.th weighted summations of the bits of the digital input signals in each bit slice are performed, resulting in a respective set of first through n.sup.th partial weighted summation results. weighted summations of the partial weighted summation results of similar ordinal number are then performed to generate first through n.sup.th final weighted summation results. each weighted summation of a bit slice of the digital input signals is performed using a capacitive network that generates partial weighted summation results in the analog regime. in this capacitive network each weight is determined by the difference in the capacitances of a respective pair of capacitive elements. the weighted summation to generate a final weighted summation result also is advantageously done in the analog regime, since this facilitates the analog final weighted summation result being non-linearly processed in an analog amplifier with sigmoidal response. this non-linear processing generates an analog axonal output response for a neural net layer, which analog axonal output response can then be digitized.",1992-08-18,"The title of the patent is analog neural nets supplied digital synapse signals on a bit-slice basis and its abstract is plural-bit digital input signals to be subjected to weighted summation in a neural net layer are bit-sliced; and a number n of respective first through n.sup.th weighted summations of the bits of the digital input signals in each bit slice are performed, resulting in a respective set of first through n.sup.th partial weighted summation results. weighted summations of the partial weighted summation results of similar ordinal number are then performed to generate first through n.sup.th final weighted summation results. each weighted summation of a bit slice of the digital input signals is performed using a capacitive network that generates partial weighted summation results in the analog regime. in this capacitive network each weight is determined by the difference in the capacitances of a respective pair of capacitive elements. the weighted summation to generate a final weighted summation result also is advantageously done in the analog regime, since this facilitates the analog final weighted summation result being non-linearly processed in an analog amplifier with sigmoidal response. this non-linear processing generates an analog axonal output response for a neural net layer, which analog axonal output response can then be digitized. dated 1992-08-18"
5140670,cellular neural network,"a novel class of information-processing systems called a cellular neural network is discussed. like a neural network, it is a large-scale nonlinear analog circuit which processes signals in real time. like cellular automata, it is made of a massive aggregate of regularly spaced circuit clones, called cells, which communicate with each other directly only through its nearest neighbors. each cell is made of a linear capacitor, a nonlinear voltage-controlled current source, and a few resistive linear circuit elements. cellular neural networks share the best features of both worlds; its continuous time feature allows real-time signal processing found within the digital domain and its local interconnection feature makes it tailor made for vlsi implementation. cellular neural networks are uniquely suited for high-speed parallel signal processing.",1992-08-18,"The title of the patent is cellular neural network and its abstract is a novel class of information-processing systems called a cellular neural network is discussed. like a neural network, it is a large-scale nonlinear analog circuit which processes signals in real time. like cellular automata, it is made of a massive aggregate of regularly spaced circuit clones, called cells, which communicate with each other directly only through its nearest neighbors. each cell is made of a linear capacitor, a nonlinear voltage-controlled current source, and a few resistive linear circuit elements. cellular neural networks share the best features of both worlds; its continuous time feature allows real-time signal processing found within the digital domain and its local interconnection feature makes it tailor made for vlsi implementation. cellular neural networks are uniquely suited for high-speed parallel signal processing. dated 1992-08-18"
5142612,computer neural network supervisory process control system and method,"a neural network for adjusting a setpoint in process control replaces a human operator. the neural network operates in three modes: training, operation, and retraining. in operation, the neural network is trained using training input data along with input data. the input data is from the sensor(s) monitoring the process. the input data is used by the neural network to develop output data. the training input data are the setpoint adjustments made by a human operator. the output data is compared with the training input data to produce error data, which is used to adjust the weights of the neural network so as to train it. after training has been completed, the neural network enters the operation mode. in this mode, the present invention uses the input data to predict output data used to adjust the setpoint supplied to the regulatory controller. thus, the operator is effectively replaced. the present invention in the retraining mode utilizes new training input data to retrain the neural network by adjusting the weight(s).",1992-08-25,"The title of the patent is computer neural network supervisory process control system and method and its abstract is a neural network for adjusting a setpoint in process control replaces a human operator. the neural network operates in three modes: training, operation, and retraining. in operation, the neural network is trained using training input data along with input data. the input data is from the sensor(s) monitoring the process. the input data is used by the neural network to develop output data. the training input data are the setpoint adjustments made by a human operator. the output data is compared with the training input data to produce error data, which is used to adjust the weights of the neural network so as to train it. after training has been completed, the neural network enters the operation mode. in this mode, the present invention uses the input data to predict output data used to adjust the setpoint supplied to the regulatory controller. thus, the operator is effectively replaced. the present invention in the retraining mode utilizes new training input data to retrain the neural network by adjusting the weight(s). dated 1992-08-25"
5142665,neural network shell for application programs,"a neural network shell has a defined interface to an application program. by interfacing with the neural network shell, any application program becomes a neural network application program. the neural network shell contains a set of utility programs that transfers data into and out of a neural network data structure. this set of utility programs allows an application program to define a new neural network model, create a neural network data structure, train a neural network, and run a neural network. once trained, the neural network data structure can be transported to other computer systems or to application programs written in different computing languages running on similar or different computer systems.",1992-08-25,"The title of the patent is neural network shell for application programs and its abstract is a neural network shell has a defined interface to an application program. by interfacing with the neural network shell, any application program becomes a neural network application program. the neural network shell contains a set of utility programs that transfers data into and out of a neural network data structure. this set of utility programs allows an application program to define a new neural network model, create a neural network data structure, train a neural network, and run a neural network. once trained, the neural network data structure can be transported to other computer systems or to application programs written in different computing languages running on similar or different computer systems. dated 1992-08-25"
5142666,learning system in a neuron computer,"a learning system in a neuron computer includes a neural network for receiving an analog signal from a first analog bus through an analog input port in a time divisional manner and performing a sum-of-the-products operation, and outputting an analog output signal to a second analog bus. a control pattern memory stores a pattern of a signal for controlling the neural network. a sequencer produces an address of the control pattern memory and a weight memory. the weight memory stores weight data of the neural network. a digital control unit controls the neural network, control pattern memory, sequencer, and weight memory, and executes a learning algorithm. the learning system further includes an input control unit provided on the input side of the neural network for selecting an input signal for executing the learning algorithm input from the digital control unit or an analog input signal input from the analog input port.",1992-08-25,"The title of the patent is learning system in a neuron computer and its abstract is a learning system in a neuron computer includes a neural network for receiving an analog signal from a first analog bus through an analog input port in a time divisional manner and performing a sum-of-the-products operation, and outputting an analog output signal to a second analog bus. a control pattern memory stores a pattern of a signal for controlling the neural network. a sequencer produces an address of the control pattern memory and a weight memory. the weight memory stores weight data of the neural network. a digital control unit controls the neural network, control pattern memory, sequencer, and weight memory, and executes a learning algorithm. the learning system further includes an input control unit provided on the input side of the neural network for selecting an input signal for executing the learning algorithm input from the digital control unit or an analog input signal input from the analog input port. dated 1992-08-25"
5144642,interference detection and characterization method and apparatus,"the interference detection and characterization system of this invention supports modern communication systems which have to contend with a variety of intentional and unintentional interference souces. as an add-on to existing communications equipment, the invention employs novel signal processing techniques to automatically detect the presence of communications channel irregularities in near real-time and alert the attending operator. information provided to the operator through a user-friendly interface is used to characterize the type of interference and its degree of severity. once characterized, the information is used by the operator to take corrective actions including the activation of alternative communication plans or, in some instances, mitigation of the interference. since output from the system of this invention lends itself well to expert system and neural network environments, such systems could be employed to further aid the operator. the unique interference signal measurements provided by the system makes it useful in applications well beyond those for which it was originally intended. other uses for which the invention has shown great potential include bit error rate estimators, communication channel scanners, and as laboratory test equipment to support receiver development and performance verification.",1992-09-01,"The title of the patent is interference detection and characterization method and apparatus and its abstract is the interference detection and characterization system of this invention supports modern communication systems which have to contend with a variety of intentional and unintentional interference souces. as an add-on to existing communications equipment, the invention employs novel signal processing techniques to automatically detect the presence of communications channel irregularities in near real-time and alert the attending operator. information provided to the operator through a user-friendly interface is used to characterize the type of interference and its degree of severity. once characterized, the information is used by the operator to take corrective actions including the activation of alternative communication plans or, in some instances, mitigation of the interference. since output from the system of this invention lends itself well to expert system and neural network environments, such systems could be employed to further aid the operator. the unique interference signal measurements provided by the system makes it useful in applications well beyond those for which it was originally intended. other uses for which the invention has shown great potential include bit error rate estimators, communication channel scanners, and as laboratory test equipment to support receiver development and performance verification. dated 1992-09-01"
5146420,communicating adder tree system for neural array processor,"the neural computing paradigm is characterized as a dynamic and highly computationally intensive system typically consisting of input weight multiplications, product summation, neural state calculations, and complete connectivity among the neurons. herein is described neural network architecture for a scalable neural array process (snap) which uses a unique intercommunication scheme within an array structure that provides high performance for completely connected network models such as the hopfield model. snap's packaging and expansion capabilities are addressed, demonstrating snap's scalability to larger networks. the array processor uses a special type of adder tree which computes in a first direction and communicates in a second direction. the adder tree is thus responsive to a compute state and a communication state. the adder tree has the ability to provide a first driver responsive to a compute state for communicating an adder output to a data path and a second driver responsive to the communication state for connecting the data path to the neuron inputs.",1992-09-08,"The title of the patent is communicating adder tree system for neural array processor and its abstract is the neural computing paradigm is characterized as a dynamic and highly computationally intensive system typically consisting of input weight multiplications, product summation, neural state calculations, and complete connectivity among the neurons. herein is described neural network architecture for a scalable neural array process (snap) which uses a unique intercommunication scheme within an array structure that provides high performance for completely connected network models such as the hopfield model. snap's packaging and expansion capabilities are addressed, demonstrating snap's scalability to larger networks. the array processor uses a special type of adder tree which computes in a first direction and communicates in a second direction. the adder tree is thus responsive to a compute state and a communication state. the adder tree has the ability to provide a first driver responsive to a compute state for communicating an adder output to a data path and a second driver responsive to the communication state for connecting the data path to the neuron inputs. dated 1992-09-08"
5146541,signal phase pattern sensitive neural network system and method,"a signal phase pattern sensitive neural network system can discern persist patterns of phase in a time varying or oscillatory signal. the system employs duplicate inputs from each of its sensors to the processing elements of a first layer of its neural network, with the exception that one input is phase shifted relative to the other. the system also employs a modification of a conventional kohonen competitive learning rule which is applied by the processing and learning elements of a second layer of its neural network.",1992-09-08,"The title of the patent is signal phase pattern sensitive neural network system and method and its abstract is a signal phase pattern sensitive neural network system can discern persist patterns of phase in a time varying or oscillatory signal. the system employs duplicate inputs from each of its sensors to the processing elements of a first layer of its neural network, with the exception that one input is phase shifted relative to the other. the system also employs a modification of a conventional kohonen competitive learning rule which is applied by the processing and learning elements of a second layer of its neural network. dated 1992-09-08"
5146543,scalable neural array processor,"the neural computing paradigm is characterized as a dynamic and highly computationally intensive system typically consisting of input weight multiplications, product summation, neural state calculations, and complete connectivity among the neurons. herein is described neural network architecture for a scalable neural array process (snap) which uses a unique intercommunication scheme within an array structure that provides high performance for completely connected network models such as the hopfield model. snap's packaging and expansion capabilities are addressed, demonstrating snap's scalability to larger networks. each neuron of the processor has an input function element, an activity function element, and a communicating adder. the neuron functions with two state modes, a compute state and a communications state. in response to a compute state, the input function element and said activity function generate a neuron value, and the communicating adder is placed in a compute mode and is responsive to the processor compute state. in a communications state a neuron is responsive to a communications state for operating the communicating adder for communicating a neuron value to an input function element.",1992-09-08,"The title of the patent is scalable neural array processor and its abstract is the neural computing paradigm is characterized as a dynamic and highly computationally intensive system typically consisting of input weight multiplications, product summation, neural state calculations, and complete connectivity among the neurons. herein is described neural network architecture for a scalable neural array process (snap) which uses a unique intercommunication scheme within an array structure that provides high performance for completely connected network models such as the hopfield model. snap's packaging and expansion capabilities are addressed, demonstrating snap's scalability to larger networks. each neuron of the processor has an input function element, an activity function element, and a communicating adder. the neuron functions with two state modes, a compute state and a communications state. in response to a compute state, the input function element and said activity function generate a neuron value, and the communicating adder is placed in a compute mode and is responsive to the processor compute state. in a communications state a neuron is responsive to a communications state for operating the communicating adder for communicating a neuron value to an input function element. dated 1992-09-08"
5146602,method of increasing the accuracy of an analog neural network and the like,"a method for increasing the accuracy of an analog neural network which computers a sum-of-products between an input vector and a stored weight pattern is described. in one embodiment of the present invention, the method comprises initially training the network by programming the synapses with a certain weight pattern. the training may be carried out using any standard learning algorithm. preferably, a back-propagation learning algorithm is employed. next, network is baked at an elevated temperature to effectuate a change in the weight pattern previously programmed during initial training. this change results from a charge redistribution which occurs within each of the synapses of the network. after baking, the network is then retrained to compensate for the change resulting from the charge redistribution. the baking and retraining steps may be successively repeated to increase the accuracy of the neural network to any desired level.",1992-09-08,"The title of the patent is method of increasing the accuracy of an analog neural network and the like and its abstract is a method for increasing the accuracy of an analog neural network which computers a sum-of-products between an input vector and a stored weight pattern is described. in one embodiment of the present invention, the method comprises initially training the network by programming the synapses with a certain weight pattern. the training may be carried out using any standard learning algorithm. preferably, a back-propagation learning algorithm is employed. next, network is baked at an elevated temperature to effectuate a change in the weight pattern previously programmed during initial training. this change results from a charge redistribution which occurs within each of the synapses of the network. after baking, the network is then retrained to compensate for the change resulting from the charge redistribution. the baking and retraining steps may be successively repeated to increase the accuracy of the neural network to any desired level. dated 1992-09-08"
5148385,serial systolic processor,"a serial systolic processor for performing neural network functions. a serial processor (90) provides the digital processing circuits for processing an input serial data stream applied to a serial input (20). a memory (29) stores digital signals representative of interconnection strengths or coefficient data corresponding to autocorrelation matrix elements. plural outputs (a.sub.o -a.sub.n) of the memory (29) are connected respectively to each of the processor neurons (p.sub.o -p.sub.n) of the serial processor (90). the digital stream is output, unchanged, on processor output bus (22), while a processed data stream is output on bus (30).",1992-09-15,"The title of the patent is serial systolic processor and its abstract is a serial systolic processor for performing neural network functions. a serial processor (90) provides the digital processing circuits for processing an input serial data stream applied to a serial input (20). a memory (29) stores digital signals representative of interconnection strengths or coefficient data corresponding to autocorrelation matrix elements. plural outputs (a.sub.o -a.sub.n) of the memory (29) are connected respectively to each of the processor neurons (p.sub.o -p.sub.n) of the serial processor (90). the digital stream is output, unchanged, on processor output bus (22), while a processed data stream is output on bus (30). dated 1992-09-15"
5148514,neural network integrated circuit device having self-organizing function,"an extension directed integrated circuit device having a learning function on a boltzmann model, includes a plurality of synapse representing units arrayed in a matrix to form a rectangle including a first and second triangles on a semiconductor chip, a plurality of neuron representing units and a plurality of educator signal control circuits which are arranged along first and second sides of the rectangle, and a plurality of buffer circuits arranged along third and fourth sides of the rectangle. the first side is opposite to the third side, and the second side is opposite to the fourth side. axon signal transfer lines and dendrite signal lines are so arranged that the neuron representing units are full-connected in each of the first right triangle the second right triangle. alternatively, axon signal lines and dendrite signal ines are arranged in parallel with rows and columns of the synapse representing unit matrix, so that the neuron representing units are full-connected in the rectangle. each synapse representing unit is connected to a pair of axon signal transfer lines and a pair of dendrite signal transfer lines.",1992-09-15,"The title of the patent is neural network integrated circuit device having self-organizing function and its abstract is an extension directed integrated circuit device having a learning function on a boltzmann model, includes a plurality of synapse representing units arrayed in a matrix to form a rectangle including a first and second triangles on a semiconductor chip, a plurality of neuron representing units and a plurality of educator signal control circuits which are arranged along first and second sides of the rectangle, and a plurality of buffer circuits arranged along third and fourth sides of the rectangle. the first side is opposite to the third side, and the second side is opposite to the fourth side. axon signal transfer lines and dendrite signal lines are so arranged that the neuron representing units are full-connected in each of the first right triangle the second right triangle. alternatively, axon signal lines and dendrite signal ines are arranged in parallel with rows and columns of the synapse representing unit matrix, so that the neuron representing units are full-connected in the rectangle. each synapse representing unit is connected to a pair of axon signal transfer lines and a pair of dendrite signal transfer lines. dated 1992-09-15"
5148515,scalable neural array processor and method,"an array processor and method for a scalable array neural processor (snap) permits computing as a dynamic and highly parallel computationally intensive system typically consisting of input weight multiplications, product summation, neural state calculations, and complete connectivity among the neurons. the scalable neural array processor (snap) uses a unique intercommunication scheme within an array structure that provides high performance for completely connected network models such as the hopfield model. snap's packaging and expansion capabilities are addressed, demonstrating snap's scalability to larger networks. the array processor is scalable. it has an array of function elements and a plurality of orthogonal horizontal and vertical processing elements for communication, computation and reduction. this structure permits in a first computation state the generation of a set of output values and in the first communication state the processing elements produce, responsive to the output values, first reduction values. in a second computation state processing elements, responsive to the first reduction values, generate vertical output values, and in a second computation state the vertical output values are communicated back to the inputs of the function elements. responsive to a third computation state responsive to the vertical output values, a second set of output values is generated by said function elements, and in a third communication state the horizontal processing elements produce second reduction values. in a fourth computation state the horizontal processing elements generate horizontal output values, and responsive to a fourth communication state the horizontal processing elements communicate the horizontal output values back to the inputs of the function elements.",1992-09-15,"The title of the patent is scalable neural array processor and method and its abstract is an array processor and method for a scalable array neural processor (snap) permits computing as a dynamic and highly parallel computationally intensive system typically consisting of input weight multiplications, product summation, neural state calculations, and complete connectivity among the neurons. the scalable neural array processor (snap) uses a unique intercommunication scheme within an array structure that provides high performance for completely connected network models such as the hopfield model. snap's packaging and expansion capabilities are addressed, demonstrating snap's scalability to larger networks. the array processor is scalable. it has an array of function elements and a plurality of orthogonal horizontal and vertical processing elements for communication, computation and reduction. this structure permits in a first computation state the generation of a set of output values and in the first communication state the processing elements produce, responsive to the output values, first reduction values. in a second computation state processing elements, responsive to the first reduction values, generate vertical output values, and in a second computation state the vertical output values are communicated back to the inputs of the function elements. responsive to a third computation state responsive to the vertical output values, a second set of output values is generated by said function elements, and in a third communication state the horizontal processing elements produce second reduction values. in a fourth computation state the horizontal processing elements generate horizontal output values, and responsive to a fourth communication state the horizontal processing elements communicate the horizontal output values back to the inputs of the function elements. dated 1992-09-15"
5150323,adaptive network for in-band signal separation,"an adaptive network for in-band signal separation (26) and method for providing in-band separation of a composite signal (32) into its constituent signals (28), (30). the input to the network (26) is a series of sampled portions of the composite signal (32). the network (26) is trained with at least one of said composite signals (28) (30) using a neural network training paradigm by presenting one or more of the constituent signals (28) (30) to said network (28). the network (26) may be used to separate multiple speech signals from a composite signal from a single sensor such as a microphone.",1992-09-22,"The title of the patent is adaptive network for in-band signal separation and its abstract is an adaptive network for in-band signal separation (26) and method for providing in-band separation of a composite signal (32) into its constituent signals (28), (30). the input to the network (26) is a series of sampled portions of the composite signal (32). the network (26) is trained with at least one of said composite signals (28) (30) using a neural network training paradigm by presenting one or more of the constituent signals (28) (30) to said network (28). the network (26) may be used to separate multiple speech signals from a composite signal from a single sensor such as a microphone. dated 1992-09-22"
5150449,speech recognition apparatus of speaker adaptation type,"a speech recognition apparatus of the speaker adaptation type operates to recognize an inputted speech pattern produced by a particular speaker by using a reference pattern produced by a voice of a standard speaker. the speech recognition apparatus is adapted to the speech of the particular speaker by converting the reference pattern into a normalized pattern by a neural network unit, internal parameters of which are modified through a learning operation using a normalized feature vector of the training pattern produced by the voice of the particular speaker and normalized on the basis of the reference pattern, so that the neural netowrk unit provides an optimum output similar to the corresponding normalized feature vector of the training pattern. in the alternative, the speech recognition apparatus operates to recognize an inputted speech pattern by converting the inputted speech pattern into a normalized speech pattern by the neural network unit, internal parameters of which are modified through a learning operation using a feature vector of the reference pattern normalized on the basis of the training pattern, so that the neural network unit provides an optimum output similar to the corresponding normalized feature vector of the reference pattern and recognizing the normalized speech pattern according to the reference pattern.",1992-09-22,"The title of the patent is speech recognition apparatus of speaker adaptation type and its abstract is a speech recognition apparatus of the speaker adaptation type operates to recognize an inputted speech pattern produced by a particular speaker by using a reference pattern produced by a voice of a standard speaker. the speech recognition apparatus is adapted to the speech of the particular speaker by converting the reference pattern into a normalized pattern by a neural network unit, internal parameters of which are modified through a learning operation using a normalized feature vector of the training pattern produced by the voice of the particular speaker and normalized on the basis of the reference pattern, so that the neural netowrk unit provides an optimum output similar to the corresponding normalized feature vector of the training pattern. in the alternative, the speech recognition apparatus operates to recognize an inputted speech pattern by converting the inputted speech pattern into a normalized speech pattern by the neural network unit, internal parameters of which are modified through a learning operation using a feature vector of the reference pattern normalized on the basis of the training pattern, so that the neural network unit provides an optimum output similar to the corresponding normalized feature vector of the reference pattern and recognizing the normalized speech pattern according to the reference pattern. dated 1992-09-22"
5150450,method and circuits for neuron perturbation in artificial neural network memory modification,"an artificial neural network has a plurality of output circuits individually perturbable for memory modification or learning by the network. the network has a plurality of synapses individually connecting each of a plurality of inputs to each output circuit. each synapse has a weight determining the effect on the associated output circuit of a signal provided on the associated input, and the synapse is addressable for selective variation of the weight. a perturbation signal is provided to one input, while data signals are provided to others of the inputs, so that perturbation of each output circuit may be controlled by varying the weights of a set of the synapses connecting the perturbation signal to the output circuits. an output circuit may be selected for perturbation by loading an appropriate weight in the synapse connecting the perturbation signal to the output circuit while zeroing the weights of the synapses connecting the perturbation signal to other output circuits. where the weights are provided by devices incapable of repeated cycles of zeroing and reloading, each synapse connecting the perturbation intput to an output circuit has an addressable switch which is closed for perturbation of this output circuit and which is open at other times. perturbations of different output circuits may be balanced by varying the weights of the set of synapses connected to the perturbation input or by varying the weights of another set of the synapses connected to one of inputs which receives a balancing signal.",1992-09-22,"The title of the patent is method and circuits for neuron perturbation in artificial neural network memory modification and its abstract is an artificial neural network has a plurality of output circuits individually perturbable for memory modification or learning by the network. the network has a plurality of synapses individually connecting each of a plurality of inputs to each output circuit. each synapse has a weight determining the effect on the associated output circuit of a signal provided on the associated input, and the synapse is addressable for selective variation of the weight. a perturbation signal is provided to one input, while data signals are provided to others of the inputs, so that perturbation of each output circuit may be controlled by varying the weights of a set of the synapses connecting the perturbation signal to the output circuits. an output circuit may be selected for perturbation by loading an appropriate weight in the synapse connecting the perturbation signal to the output circuit while zeroing the weights of the synapses connecting the perturbation signal to other output circuits. where the weights are provided by devices incapable of repeated cycles of zeroing and reloading, each synapse connecting the perturbation intput to an output circuit has an addressable switch which is closed for perturbation of this output circuit and which is open at other times. perturbations of different output circuits may be balanced by varying the weights of the set of synapses connected to the perturbation input or by varying the weights of another set of the synapses connected to one of inputs which receives a balancing signal. dated 1992-09-22"
5151822,transform digital/optical processing system including wedge/ring accumulator,"a transform digital optical processing system generates a transform signal of an image. fourier or other well-known transforms may be employed. the transform signal may be generated in one of two ways: optically or electronically. in optical generation a two dimensional object is generated by modulating a beam of coherent light with an image of the object. a transform image of the modulated coherent light beam is formed, using an optical transform element. the optical transform is then stored in a two dimensional buffer. the transform signal may also be generated electronically by storing a digital video image of an object and generating a fourier or other transform of the digital video image using vector processing chips or other commercially available digital transform generating computers. this digitally generated information may be analyzed and classified through a neural network type processor. the two-dimensional transform data is then processed to obtain the inspection or other characteristics for comparison against predetermined characteristics. the two dimensional transform is divided into two types of zones, namely wedges and rings. the transform data is then mapped into a corresponding wedge and ring, and the data for each wedge and ring is accumulated or summed to obtain data values. it has been found that the summed wedge and ring data values can accurately characterize an image for inspection or other comparison purposes.",1992-09-29,"The title of the patent is transform digital/optical processing system including wedge/ring accumulator and its abstract is a transform digital optical processing system generates a transform signal of an image. fourier or other well-known transforms may be employed. the transform signal may be generated in one of two ways: optically or electronically. in optical generation a two dimensional object is generated by modulating a beam of coherent light with an image of the object. a transform image of the modulated coherent light beam is formed, using an optical transform element. the optical transform is then stored in a two dimensional buffer. the transform signal may also be generated electronically by storing a digital video image of an object and generating a fourier or other transform of the digital video image using vector processing chips or other commercially available digital transform generating computers. this digitally generated information may be analyzed and classified through a neural network type processor. the two-dimensional transform data is then processed to obtain the inspection or other characteristics for comparison against predetermined characteristics. the two dimensional transform is divided into two types of zones, namely wedges and rings. the transform data is then mapped into a corresponding wedge and ring, and the data for each wedge and ring is accumulated or summed to obtain data values. it has been found that the summed wedge and ring data values can accurately characterize an image for inspection or other comparison purposes. dated 1992-09-29"
5151874,integrated circuit for square root operation using neural network,"an integrated circuit for performing a square root operation uses adders made in accordance with neural network concepts. the integrated circuit includes an exponent part, a first mantissa part, a second mantissa part and a control part. the exponent part computes an exponent of the square root of an input operand; the first mantissa part preprocesses the mantissa of the input operand; the second mantissa part computes the square root of the output from the first mantissa part; and the control part controls interaction of input and output among various components of the integrated circuits. because the adders used in integrated circuit are composed of neural network circuits having a short propagation time for carry bits, the integrated circuit can computer a square root fast and efficiently.",1992-09-29,"The title of the patent is integrated circuit for square root operation using neural network and its abstract is an integrated circuit for performing a square root operation uses adders made in accordance with neural network concepts. the integrated circuit includes an exponent part, a first mantissa part, a second mantissa part and a control part. the exponent part computes an exponent of the square root of an input operand; the first mantissa part preprocesses the mantissa of the input operand; the second mantissa part computes the square root of the output from the first mantissa part; and the control part controls interaction of input and output among various components of the integrated circuits. because the adders used in integrated circuit are composed of neural network circuits having a short propagation time for carry bits, the integrated circuit can computer a square root fast and efficiently. dated 1992-09-29"
5151971,arrangement of data cells and neural network system utilizing such an arrangement,"an arrangement of data cells which stores at least one matrix of data words which are arranged in rows and columns, the matrix being distributed in the arrangement in order to deliver/receive, via a single bus, permuted data words which correspond either to a row or to a column of the matrix. each data cell is connected to the single bus via series-connected switches which are associated with a respective addressing mode, the switches which address a same word of a same mode being directly controlled by a same selection signal. circulation members enable the original order of the data on the bus to be restored. an arrangement of this kind is used in a layered neural network system for executing the error backpropagation algorithm.",1992-09-29,"The title of the patent is arrangement of data cells and neural network system utilizing such an arrangement and its abstract is an arrangement of data cells which stores at least one matrix of data words which are arranged in rows and columns, the matrix being distributed in the arrangement in order to deliver/receive, via a single bus, permuted data words which correspond either to a row or to a column of the matrix. each data cell is connected to the single bus via series-connected switches which are associated with a respective addressing mode, the switches which address a same word of a same mode being directly controlled by a same selection signal. circulation members enable the original order of the data on the bus to be restored. an arrangement of this kind is used in a layered neural network system for executing the error backpropagation algorithm. dated 1992-09-29"
5153923,high order information processing method by means of a neural network and minimum and maximum searching method therefor,"in order to improve the problematical points concerning the structure and the processing speed of a prior art neural network, the optimum structure of the neural network, in which a synapse structure constructed on the basis of living body physiological knowledge or presumed therefrom is determined to make it possible to realize high level information processing functions such as feature extraction, feature unification, memory, etc. applications to an image recognition, a movement control, etc. making the most of the robust recognizing power thereof, or application to an optimum problem, a large scale numerical analysis, etc. making the most of the parallel processing power thereof are made possible.",1992-10-06,"The title of the patent is high order information processing method by means of a neural network and minimum and maximum searching method therefor and its abstract is in order to improve the problematical points concerning the structure and the processing speed of a prior art neural network, the optimum structure of the neural network, in which a synapse structure constructed on the basis of living body physiological knowledge or presumed therefrom is determined to make it possible to realize high level information processing functions such as feature extraction, feature unification, memory, etc. applications to an image recognition, a movement control, etc. making the most of the robust recognizing power thereof, or application to an optimum problem, a large scale numerical analysis, etc. making the most of the parallel processing power thereof are made possible. dated 1992-10-06"
5155699,divider using neural network,"a divider using neural network configurations comprises a subtractor, a selecting means, a first latch means, a second latch means, a shift register and a control means. the subtractor of the divider comprises plural inverters and plural 3-bit full-adders which are composed of four output lines, an input synapse group, a first bias synapse group, a second bias synapse group, a feedback synapse group, a neuron group and an inverter group.",1992-10-13,"The title of the patent is divider using neural network and its abstract is a divider using neural network configurations comprises a subtractor, a selecting means, a first latch means, a second latch means, a shift register and a control means. the subtractor of the divider comprises plural inverters and plural 3-bit full-adders which are composed of four output lines, an input synapse group, a first bias synapse group, a second bias synapse group, a feedback synapse group, a neuron group and an inverter group. dated 1992-10-13"
5155763,look ahead method and apparatus for predictive dialing using a neural network,"a predictive dialing system having a computer connected to a telephone switch stores a group of call records in its internal storage. each call record contains a group of input parameters, including the date, the time, and one or more workload factors. workload factors can indicate the number of pending calls, the number of available operators, the average idle time, the connection delay, the completion rate, and the nuisance call rate, among other things. in the preferred embodiment, each call record also contains a dial action, which indicates whether a call was initiated or not. these call records are analyzed by a neutral network to determine a relationship between the input parameters and the dial action stored in each call record. this analysis is done as part of the training process for the neutral network. after this relationship is determined, the computer system sends a current group of input parameters to the neural network, and, based on the analysis of the previous call records, the neural network determines whether a call should be intiated or not. the neural network bases its decision on the complex relationship it has learned from its training data--perhaps several thousand call records spanning several days, months, or even years. the neural network is able to automatically adjust--in a look ahead, proactive manner--for slow and fast periods of the day, week, month, and year.",1992-10-13,"The title of the patent is look ahead method and apparatus for predictive dialing using a neural network and its abstract is a predictive dialing system having a computer connected to a telephone switch stores a group of call records in its internal storage. each call record contains a group of input parameters, including the date, the time, and one or more workload factors. workload factors can indicate the number of pending calls, the number of available operators, the average idle time, the connection delay, the completion rate, and the nuisance call rate, among other things. in the preferred embodiment, each call record also contains a dial action, which indicates whether a call was initiated or not. these call records are analyzed by a neutral network to determine a relationship between the input parameters and the dial action stored in each call record. this analysis is done as part of the training process for the neutral network. after this relationship is determined, the computer system sends a current group of input parameters to the neural network, and, based on the analysis of the previous call records, the neural network determines whether a call should be intiated or not. the neural network bases its decision on the complex relationship it has learned from its training data--perhaps several thousand call records spanning several days, months, or even years. the neural network is able to automatically adjust--in a look ahead, proactive manner--for slow and fast periods of the day, week, month, and year. dated 1992-10-13"
5155801,clustered neural networks,""" a plurality of neural networks are coupled to an output neural network, or judge network, to form a clustered neural network. each of the plurality of clustered networks comprises a supervised learning rule back-propagated neural network. each of the clustered neural networks are trained to perform substantially the same mapping function before they are clustered. following training, the clustered neural network computes its output by taking an """"average"""" of the outputs of the individual neural networks that make up the cluster. the judge network combines the outputs of the plurality of individual neural networks to provide the output from the entire clustered network. in addition, the output of the judge network may be fed back to each of the individual neural networks and used as a training input thereto, in order to provide for continuous training. the use of the clustered network increases the speed of learning and results in better generalization. in addition, clustering multiple back-propagation networks provides for increased performance and fault tolerance when compared to a single unclustered network having substantially the same computational complexity. the present invention may be used in applications that are amenable to neural network solutions, including control and image processing applications. clustering of the networks also permits the use of smaller networks and provides for improved performance. the clustering of multiple back-propagation networks provides for synergy that improves the properties of the clustered network over a comparably complex non-clustered network. """,1992-10-13,"The title of the patent is clustered neural networks and its abstract is "" a plurality of neural networks are coupled to an output neural network, or judge network, to form a clustered neural network. each of the plurality of clustered networks comprises a supervised learning rule back-propagated neural network. each of the clustered neural networks are trained to perform substantially the same mapping function before they are clustered. following training, the clustered neural network computes its output by taking an """"average"""" of the outputs of the individual neural networks that make up the cluster. the judge network combines the outputs of the plurality of individual neural networks to provide the output from the entire clustered network. in addition, the output of the judge network may be fed back to each of the individual neural networks and used as a training input thereto, in order to provide for continuous training. the use of the clustered network increases the speed of learning and results in better generalization. in addition, clustering multiple back-propagation networks provides for increased performance and fault tolerance when compared to a single unclustered network having substantially the same computational complexity. the present invention may be used in applications that are amenable to neural network solutions, including control and image processing applications. clustering of the networks also permits the use of smaller networks and provides for improved performance. the clustering of multiple back-propagation networks provides for synergy that improves the properties of the clustered network over a comparably complex non-clustered network. "" dated 1992-10-13"
5157399,neural network quantizer,"a neural network quantizer for quantizing input analog signals includes a plurality of multi-level neurons. the input analog signals are sampled and supplied to respective ones of the multi-level neurons. output values of the multi-level neurons are converted into analog values, weighted by weighting coefficients determined in accordance with a frequency band of at least one frequency component of the input analog signals and fed back to the respective one of the multi-level neurons and to the other multi-level neurons. the weighted analog values fed back are compared with the respective ones of the sampled input analog signals. the output values of the multi-level neurons are corrected in response to the compared results, and when the compared results are converged within a predetermined range, the output values of the multi-level neurons are produced to quantize the input analog signals.",1992-10-20,"The title of the patent is neural network quantizer and its abstract is a neural network quantizer for quantizing input analog signals includes a plurality of multi-level neurons. the input analog signals are sampled and supplied to respective ones of the multi-level neurons. output values of the multi-level neurons are converted into analog values, weighted by weighting coefficients determined in accordance with a frequency band of at least one frequency component of the input analog signals and fed back to the respective one of the multi-level neurons and to the other multi-level neurons. the weighted analog values fed back are compared with the respective ones of the sampled input analog signals. the output values of the multi-level neurons are corrected in response to the compared results, and when the compared results are converged within a predetermined range, the output values of the multi-level neurons are produced to quantize the input analog signals. dated 1992-10-20"
5157733,"radiation image processing apparatus, determination apparatus, and radiation image read-out apparatus","in a radiation image processing apparatus, signal processing for determining the shape and location of an irradiation field, adjusting read-out conditions for a final readout from a preliminary read-out image signal, adjusting image processing conditions, and/or detecting an abnormal pattern is carried out on an image signal representing a radiation image by using a neural network. after the neural network, the learning operations of which have been carried out, is incorporated into the radiation image processing apparatus, modifying information is entered from an input device into the neural network. the modifying information is used to modify the signal processing carried out by the neural network and thereby to carry out re-learning operations of the neural network.",1992-10-20,"The title of the patent is radiation image processing apparatus, determination apparatus, and radiation image read-out apparatus and its abstract is in a radiation image processing apparatus, signal processing for determining the shape and location of an irradiation field, adjusting read-out conditions for a final readout from a preliminary read-out image signal, adjusting image processing conditions, and/or detecting an abnormal pattern is carried out on an image signal representing a radiation image by using a neural network. after the neural network, the learning operations of which have been carried out, is incorporated into the radiation image processing apparatus, modifying information is entered from an input device into the neural network. the modifying information is used to modify the signal processing carried out by the neural network and thereby to carry out re-learning operations of the neural network. dated 1992-10-20"
5159590,multi-slot call relocation control method and system,"a multi-slot call relocation control method and system having a multi-slot call switching system and/or transmission equipment constituted by an address control memory, address controller and address location changing circuit whereby address write and read information for a channel memory is controlled. where unoccupied circuits are 2.sup.n (n: natural number) times the basic switching unit, incoming calls with a maximum of 2.sup.n basic switching units in capacity may not be switched or transmitted by the unoccupied circuits depending on their status involving the presence of other calls. in that case, calls are relocated within a frame using the fewest steps possible. this is achieved by a neural network in the address control memory of multi-slot call switching system a, the neural network learning to output a call allocation pattern such that the number of times calls are relocated becomes minimal. the information from the network makes it possible to relocate the least number of times the calls whose capacity is not more than 2.sup.n basic switching unit in the channel memory. the relocation information is transmitted from switching system a to another system b, connected oppositely to system a. using the relocation information received, system b relocates calls within a channel memory of its own.",1992-10-27,"The title of the patent is multi-slot call relocation control method and system and its abstract is a multi-slot call relocation control method and system having a multi-slot call switching system and/or transmission equipment constituted by an address control memory, address controller and address location changing circuit whereby address write and read information for a channel memory is controlled. where unoccupied circuits are 2.sup.n (n: natural number) times the basic switching unit, incoming calls with a maximum of 2.sup.n basic switching units in capacity may not be switched or transmitted by the unoccupied circuits depending on their status involving the presence of other calls. in that case, calls are relocated within a frame using the fewest steps possible. this is achieved by a neural network in the address control memory of multi-slot call switching system a, the neural network learning to output a call allocation pattern such that the number of times calls are relocated becomes minimal. the information from the network makes it possible to relocate the least number of times the calls whose capacity is not more than 2.sup.n basic switching unit in the channel memory. the relocation information is transmitted from switching system a to another system b, connected oppositely to system a. using the relocation information received, system b relocates calls within a channel memory of its own. dated 1992-10-27"
5159661,vertically interconnected parallel distributed processor,a parallel distributed processor comprises matrices of unit cells arranged in a stacked configuration. each unit cell includes a chalcogenide body which may be set and reset to a plurality of values of a given physical property. interconnections between the unit cells are established via the chalcogenide materials and the pattern and strength of the interconnections is determined by the set values of the chalcogenide. the processor is readily adapted to the construction of neural network computing systems.,1992-10-27,The title of the patent is vertically interconnected parallel distributed processor and its abstract is a parallel distributed processor comprises matrices of unit cells arranged in a stacked configuration. each unit cell includes a chalcogenide body which may be set and reset to a plurality of values of a given physical property. interconnections between the unit cells are established via the chalcogenide materials and the pattern and strength of the interconnections is determined by the set values of the chalcogenide. the processor is readily adapted to the construction of neural network computing systems. dated 1992-10-27
5161014,neural networks as for video signal processing,"a television signal processing apparatus includes at least one neural network for processing a signal representing an image. the neural network includes a plurality of perceptrons each of which includes circuitry for weighting a plurality of delayed representations of said signal, circuitry for providing sums of weighted signals provided by said weighting circuitry, and circuitry for processing said sums with a sigmoidal transfer function. the neural network also includes circuitry for combining output signals provided by ones of said perceptrons for providing a processed signal.",1992-11-03,"The title of the patent is neural networks as for video signal processing and its abstract is a television signal processing apparatus includes at least one neural network for processing a signal representing an image. the neural network includes a plurality of perceptrons each of which includes circuitry for weighting a plurality of delayed representations of said signal, circuitry for providing sums of weighted signals provided by said weighting circuitry, and circuitry for processing said sums with a sigmoidal transfer function. the neural network also includes circuitry for combining output signals provided by ones of said perceptrons for providing a processed signal. dated 1992-11-03"
5161204,apparatus for generating a feature matrix based on normalized out-class and in-class variation matrices,"a method and apparatus under software control for pattern recognition utilizes a neural network implementation to recognize two dimensional input images which are sufficiently similar to a database of previously stored two dimensional images. images are first image processed and subjected to a fourier transform which yields a power spectrum. an in-class to out-of-class study is performed on a typical collection of images in order to determine the most discriminatory regions of the fourier transform. a feature vector consisting of the highest order (most discriminatory) magnitude information from the power spectrum of the fourier transform of the image is formed. feature vectors are input to a neural network having preferably two hidden layers, input dimensionality of the number of elements in the feature vector and output dimensionality of the number of data elements stored in the database. unique identifier numbers are preferably stored along with the feature vector. application of a query feature vector to the neural network will result in an output vector. the output vector is subjected to statistical analysis to determine if a sufficiently high confidence level exists to indicate that a successful identification has been made. where a successful identification has occurred, the unique identifier number may be displayed.",1992-11-03,"The title of the patent is apparatus for generating a feature matrix based on normalized out-class and in-class variation matrices and its abstract is a method and apparatus under software control for pattern recognition utilizes a neural network implementation to recognize two dimensional input images which are sufficiently similar to a database of previously stored two dimensional images. images are first image processed and subjected to a fourier transform which yields a power spectrum. an in-class to out-of-class study is performed on a typical collection of images in order to determine the most discriminatory regions of the fourier transform. a feature vector consisting of the highest order (most discriminatory) magnitude information from the power spectrum of the fourier transform of the image is formed. feature vectors are input to a neural network having preferably two hidden layers, input dimensionality of the number of elements in the feature vector and output dimensionality of the number of data elements stored in the database. unique identifier numbers are preferably stored along with the feature vector. application of a query feature vector to the neural network will result in an output vector. the output vector is subjected to statistical analysis to determine if a sufficiently high confidence level exists to indicate that a successful identification has been made. where a successful identification has occurred, the unique identifier number may be displayed. dated 1992-11-03"
5162899,color data correction apparatus ultilizing neural network,"a color correction apparatus for use in an apparatus such as a color copier for operating on data obtained by scanning and color analysis of a source image to obtain color density data for use in printing a copy of the source image, the correction apparatus containing a neural network. parameter values of the neural network are established by repetitive computations based on amounts of difference between color density data previously used to print a plurality of color samples and color density data produced by the neural network in response to color analysis data obtained by analyzing these color samples.",1992-11-10,"The title of the patent is color data correction apparatus ultilizing neural network and its abstract is a color correction apparatus for use in an apparatus such as a color copier for operating on data obtained by scanning and color analysis of a source image to obtain color density data for use in printing a copy of the source image, the correction apparatus containing a neural network. parameter values of the neural network are established by repetitive computations based on amounts of difference between color density data previously used to print a plurality of color samples and color density data produced by the neural network in response to color analysis data obtained by analyzing these color samples. dated 1992-11-10"
5164837,method of correcting setup parameter decision characteristics and automatic setup apparatus using a neural network,"image data of an original read by an image reader is analyzed by analyzer means to be supplied to a neural network. an operator inputs scene information and desired finish information to the neural network with data input means. the neural network calculates setup parameter values in compliance with a conversion rule specified by predetermined weighting values and functional forms, and sets these values in an image data converter. then, the operator corrects the setup parameter values on the basis of finish condition of the produced color separation films. the corrected setup parameter values are inputted to leaning means. the leaning means computes proper weighting values with which the neural network calculates setup parameter values equal to or approximate to the corrected setup parameter values. such proper weighting values are supplied to the neural network as new weighting values.",1992-11-17,"The title of the patent is method of correcting setup parameter decision characteristics and automatic setup apparatus using a neural network and its abstract is image data of an original read by an image reader is analyzed by analyzer means to be supplied to a neural network. an operator inputs scene information and desired finish information to the neural network with data input means. the neural network calculates setup parameter values in compliance with a conversion rule specified by predetermined weighting values and functional forms, and sets these values in an image data converter. then, the operator corrects the setup parameter values on the basis of finish condition of the produced color separation films. the corrected setup parameter values are inputted to leaning means. the leaning means computes proper weighting values with which the neural network calculates setup parameter values equal to or approximate to the corrected setup parameter values. such proper weighting values are supplied to the neural network as new weighting values. dated 1992-11-17"
5165009,neural network processing system using semiconductor memories,"herein disclosed is a data processing system having a memory packaged therein for realizing a large-scale and high-speed parallel distributed processing and, especially, a data processing system for the neural network processing. the neural network processing system according to the present invention comprises: a memory circuit for storing neuron output values, connection weights, the desired values of outputs, and data necessary for learning; and input/output circuit for writing or reading data in or out of said memory circuit; a processing circuit for performing a processing for determining the neuron outputs such as the product, sum and nonlinear conversion of the data stored in said memory circuit, a comparison of the output value and its desired value, and a processing necessary for learning; and a control circuit for controlling the operations of said memory circuit, said input/output circuit and said processing circuit. the processing circuit is constructed to include at least one of an adder, a multiplier, a nonlinear transfer function circuit and a comparator so that at least a portion of the processing necessary for determining the neutron output values such as the product or sum may be accomplished in parallel. moreover, these circuits are shared among a plurality of neutrons and are operated in a time sharing manner to determine the plural neuron output values. still moreover, the aforementioned comparator compares the neuron output value determined and the desired value of the otuput in parallel.",1992-11-17,"The title of the patent is neural network processing system using semiconductor memories and its abstract is herein disclosed is a data processing system having a memory packaged therein for realizing a large-scale and high-speed parallel distributed processing and, especially, a data processing system for the neural network processing. the neural network processing system according to the present invention comprises: a memory circuit for storing neuron output values, connection weights, the desired values of outputs, and data necessary for learning; and input/output circuit for writing or reading data in or out of said memory circuit; a processing circuit for performing a processing for determining the neuron outputs such as the product, sum and nonlinear conversion of the data stored in said memory circuit, a comparison of the output value and its desired value, and a processing necessary for learning; and a control circuit for controlling the operations of said memory circuit, said input/output circuit and said processing circuit. the processing circuit is constructed to include at least one of an adder, a multiplier, a nonlinear transfer function circuit and a comparator so that at least a portion of the processing necessary for determining the neutron output values such as the product or sum may be accomplished in parallel. moreover, these circuits are shared among a plurality of neutrons and are operated in a time sharing manner to determine the plural neuron output values. still moreover, the aforementioned comparator compares the neuron output value determined and the desired value of the otuput in parallel. dated 1992-11-17"
5165069,method and system for non-invasively identifying the operational status of a vcr,a method and apparatus are provided for identifying one of a plurality of operational modes of a monitored video cassette recorder (vcr). a sensor is positioned near the monitored vcr for detecting a radiated signal of the monitored vcr. the detected signal is appplied to a filter for filtering the detected signal and for providing a plurality of predetermined band-pass filtered signals. a neural network is used for processing the plurality of predetermined band-pass filtered signals to identify the operational mode of the vcr.,1992-11-17,The title of the patent is method and system for non-invasively identifying the operational status of a vcr and its abstract is a method and apparatus are provided for identifying one of a plurality of operational modes of a monitored video cassette recorder (vcr). a sensor is positioned near the monitored vcr for detecting a radiated signal of the monitored vcr. the detected signal is appplied to a filter for filtering the detected signal and for providing a plurality of predetermined band-pass filtered signals. a neural network is used for processing the plurality of predetermined band-pass filtered signals to identify the operational mode of the vcr. dated 1992-11-17
5165270,non-destructive materials testing apparatus and technique for use in the field,"the present invention features an apparatus and method for impact-echo testing of structures in situ, in the field. the impact-echo testing method provides a non-invasive, non-destructive way of determining the defects in the structure. the method is both uniform and reliable, the test procedure being substantially identical every time. the apparatus of the invention comprises a portable, hand-held unit having a plurality of impactors disposed therein. the plurality of impactors comprise a number of differently weighted spheres that are each designed to impart a different impact energy into the structure to be tested. each sphere is disposed on a distal end of a spring-steel rod. a particular weighted sphere is chosen by a selector disposed on the testing unit. the sphere is withdrawn from the rest position by a pair of jaws to a given height above the structure. at a predetermined release point, the sphere is released, causing it to impact the structure with a specific duration and impart a given energy thereto. the impact produces stress waves that are reflected from the internal flaws and external surfaces of the structure. the reflected waves are detected by a transducer that converts the stress waves into an electrical signal (displacement waveform). the waveform is then processed to provide an amplitude spectrum, and in the case of plates, a reflection spectrum. for plates, the reflection spectrum can be interpreted by a neural network provides results that are indicative of either the thickness of the structure or of the defects disposed therein.",1992-11-24,"The title of the patent is non-destructive materials testing apparatus and technique for use in the field and its abstract is the present invention features an apparatus and method for impact-echo testing of structures in situ, in the field. the impact-echo testing method provides a non-invasive, non-destructive way of determining the defects in the structure. the method is both uniform and reliable, the test procedure being substantially identical every time. the apparatus of the invention comprises a portable, hand-held unit having a plurality of impactors disposed therein. the plurality of impactors comprise a number of differently weighted spheres that are each designed to impart a different impact energy into the structure to be tested. each sphere is disposed on a distal end of a spring-steel rod. a particular weighted sphere is chosen by a selector disposed on the testing unit. the sphere is withdrawn from the rest position by a pair of jaws to a given height above the structure. at a predetermined release point, the sphere is released, causing it to impact the structure with a specific duration and impart a given energy thereto. the impact produces stress waves that are reflected from the internal flaws and external surfaces of the structure. the reflected waves are detected by a transducer that converts the stress waves into an electrical signal (displacement waveform). the waveform is then processed to provide an amplitude spectrum, and in the case of plates, a reflection spectrum. for plates, the reflection spectrum can be interpreted by a neural network provides results that are indicative of either the thickness of the structure or of the defects disposed therein. dated 1992-11-24"
5166539,neural network circuit,"a neural network circuit, in which a number n of weight coefficients (wl-wn) corresponding to a number n of inputs are provided, subtraction circuits determine the difference between inputs and the weight coefficients in each input terminal, the result thereof is inputted into absolute value circuits, all calculation results of the absolute value circuts corresponding to the inputs and the weight coefficients are inputted into an addition circuit and accumulated, and this accumulation result determines the output value. the threshold value circuit, which determines the final output value, has characteristics of a step function pattern, a polygonal line pattern, or a sigmoid function pattern, depending on the object. in the case in which a neural network circuit is realized by means of digital circuits, the absolute value circuits can comprise simply ex-or logic (exclusive or) gates. furthermore, in the case in which the input terminals have two input paths and two weight coefficients corresponding to each input path, the neuron circuits form a recognition area having a flexible shape which is controlled by the weight coefficients. neuron circuits are widely used in pattern recognition; neuron circuits react to a pattern inputted into the input layer and recognition is thereby conducted.",1992-11-24,"The title of the patent is neural network circuit and its abstract is a neural network circuit, in which a number n of weight coefficients (wl-wn) corresponding to a number n of inputs are provided, subtraction circuits determine the difference between inputs and the weight coefficients in each input terminal, the result thereof is inputted into absolute value circuits, all calculation results of the absolute value circuts corresponding to the inputs and the weight coefficients are inputted into an addition circuit and accumulated, and this accumulation result determines the output value. the threshold value circuit, which determines the final output value, has characteristics of a step function pattern, a polygonal line pattern, or a sigmoid function pattern, depending on the object. in the case in which a neural network circuit is realized by means of digital circuits, the absolute value circuits can comprise simply ex-or logic (exclusive or) gates. furthermore, in the case in which the input terminals have two input paths and two weight coefficients corresponding to each input path, the neuron circuits form a recognition area having a flexible shape which is controlled by the weight coefficients. neuron circuits are widely used in pattern recognition; neuron circuits react to a pattern inputted into the input layer and recognition is thereby conducted. dated 1992-11-24"
5166896,discrete cosine transform chip using neural network concepts for calculating values of a discrete cosine transform function,"a discrete cosine transform chip includes circuits using neural network concepts that have parallel processing capability as well as conventional digital logic circuits. in particular, the discrete cosine transform chip includes a cosine term processing portion, a multiplier, an adder, a subtractor, and two groups of latches. the multiplier, the adder and the subtractor incorporated in the discrete cosine transform chip use unidirectional feed back neural network models.",1992-11-24,"The title of the patent is discrete cosine transform chip using neural network concepts for calculating values of a discrete cosine transform function and its abstract is a discrete cosine transform chip includes circuits using neural network concepts that have parallel processing capability as well as conventional digital logic circuits. in particular, the discrete cosine transform chip includes a cosine term processing portion, a multiplier, an adder, a subtractor, and two groups of latches. the multiplier, the adder and the subtractor incorporated in the discrete cosine transform chip use unidirectional feed back neural network models. dated 1992-11-24"
5166938,error correction circuit using a design based on a neural network model comprising an encoder portion and a decoder portion,an error correction circuit is provided which uses nmos and pmos synapses to form network type responses to a coded multi-bit input. use of mos technology logic in error correction circuits allows such devices to be easily interfaced with other like technology circuits without the need to use distinct interface logic as with conventional error correction circuitry.,1992-11-24,The title of the patent is error correction circuit using a design based on a neural network model comprising an encoder portion and a decoder portion and its abstract is an error correction circuit is provided which uses nmos and pmos synapses to form network type responses to a coded multi-bit input. use of mos technology logic in error correction circuits allows such devices to be easily interfaced with other like technology circuits without the need to use distinct interface logic as with conventional error correction circuitry. dated 1992-11-24
5167006,"neuron unit, neural network and signal processing method","a neuron unit processes a plurality of input signals and outputs an output signal which is indicative of a result of the processing. the neuron unit includes input lines for receiving the input signals, a forward process part including a supplying part for supplying weight functions and an operation part for carrying out an operation on each of the input signals using one of the weight functions and for outputting the output signal, and a self-learning part including a generating part for generating new weight functions based on errors between the output signal of the forward process part and teaching signals and a varying part for varying the weight functions supplied by the supplying part of the forward process part to the new weight functions generated by the generating part.",1992-11-24,"The title of the patent is neuron unit, neural network and signal processing method and its abstract is a neuron unit processes a plurality of input signals and outputs an output signal which is indicative of a result of the processing. the neuron unit includes input lines for receiving the input signals, a forward process part including a supplying part for supplying weight functions and an operation part for carrying out an operation on each of the input signals using one of the weight functions and for outputting the output signal, and a self-learning part including a generating part for generating new weight functions based on errors between the output signal of the forward process part and teaching signals and a varying part for varying the weight functions supplied by the supplying part of the forward process part to the new weight functions generated by the generating part. dated 1992-11-24"
5167007,multilayered optical neural network system,"a multilayered optical neural network system comprise an input layer, an output layer, at least one hidden layer provided between the input layer and the output layer, a memory matrix holding device provided between the respective layers for holding weighted couplings between the layers, a correlation operating device for optically computing a correlation between an output optical pattern from the previous layer and the memory matrix pattern, an output function operating device for implementing optical computing of an output function corresponding to a result of the correlation operation, and a memory matrix correcting device provided between the respective layers for optically correcting a memory matrix held in the memory matrix holding device by a learning operation, whereby the system is capable of two-dimensional optical computing of all data transfers and operations and executing a great amount of computing without use of holograms.",1992-11-24,"The title of the patent is multilayered optical neural network system and its abstract is a multilayered optical neural network system comprise an input layer, an output layer, at least one hidden layer provided between the input layer and the output layer, a memory matrix holding device provided between the respective layers for holding weighted couplings between the layers, a correlation operating device for optically computing a correlation between an output optical pattern from the previous layer and the memory matrix pattern, an output function operating device for implementing optical computing of an output function corresponding to a result of the correlation operation, and a memory matrix correcting device provided between the respective layers for optically correcting a memory matrix held in the memory matrix holding device by a learning operation, whereby the system is capable of two-dimensional optical computing of all data transfers and operations and executing a great amount of computing without use of holograms. dated 1992-11-24"
5167008,digital circuitry for approximating sigmoidal response in a neural network layer,"a plurality of neural circuits are connected in a neural network layer for generating their respective digital axonal responses to the same plurality of synapse input signals. each neural circuit includes digital circuitry for approximating a sigmoidal response connected after respective circuitry for performing a weighted summation of the synapse input signals to generate a weighted summation result in digital form. in this digital circuitry the absolute value of the digital weighted summation result is first determined. then, a window comparator determines into which of a plurality of amplitude ranges the absolute value of the weighted summation result falls. a digital intercept value and a digital slope value are selected in accordance with the range into which the absolute value of the weighted summation result falls. the absolute value of the digital weighted summation result is multiplied by the selected digital slope value to generate a digital product; and the digital intercept value is added to the digital product to generate an absolute value representation of a digital axonal response. the polarity of the weighted summation result is determined, and the same polarity is assigned to the absolute value representation of the digital axonal response, thereby to generate the digital axonal response.",1992-11-24,"The title of the patent is digital circuitry for approximating sigmoidal response in a neural network layer and its abstract is a plurality of neural circuits are connected in a neural network layer for generating their respective digital axonal responses to the same plurality of synapse input signals. each neural circuit includes digital circuitry for approximating a sigmoidal response connected after respective circuitry for performing a weighted summation of the synapse input signals to generate a weighted summation result in digital form. in this digital circuitry the absolute value of the digital weighted summation result is first determined. then, a window comparator determines into which of a plurality of amplitude ranges the absolute value of the weighted summation result falls. a digital intercept value and a digital slope value are selected in accordance with the range into which the absolute value of the weighted summation result falls. the absolute value of the digital weighted summation result is multiplied by the selected digital slope value to generate a digital product; and the digital intercept value is added to the digital product to generate an absolute value representation of a digital axonal response. the polarity of the weighted summation result is determined, and the same polarity is assigned to the absolute value representation of the digital axonal response, thereby to generate the digital axonal response. dated 1992-11-24"
5167009,on-line process control neural network using data pointers,"an on-line process control neural network using data pointers allows the neural network to be easily configured to use data in a process control environment. the inputs, outputs, training inputs and errors can be retrieved and/or stored from any available data source without programming. the user of the neural network specifies data pointers indicating the particular computer system in which the data resides or will be stored; the type of data to be retrieved and/or stored; and the specific data value or storage location to be used. the data pointers include maximum, minimum, and maximum change limits, which can also serve as scaling limits for the neural network. data pointers indicating time-dependent data, such as time averages, also include time boundary specifiers. the data pointers are entered by the user of the neural network using pop-up menus and by completing fields in a template. an historical database provides both a source of input data and a storage function for output and error data.",1992-11-24,"The title of the patent is on-line process control neural network using data pointers and its abstract is an on-line process control neural network using data pointers allows the neural network to be easily configured to use data in a process control environment. the inputs, outputs, training inputs and errors can be retrieved and/or stored from any available data source without programming. the user of the neural network specifies data pointers indicating the particular computer system in which the data resides or will be stored; the type of data to be retrieved and/or stored; and the specific data value or storage location to be used. the data pointers include maximum, minimum, and maximum change limits, which can also serve as scaling limits for the neural network. data pointers indicating time-dependent data, such as time averages, also include time boundary specifiers. the data pointers are entered by the user of the neural network using pop-up menus and by completing fields in a template. an historical database provides both a source of input data and a storage function for output and error data. dated 1992-11-24"
5168262,fire alarm system,a fire alarm system employs a neural network for obtaining one or more types of fire related information values. a plurality of detection information values are time-serially collected from plural fire phenomenon detectors. the detection information values are signal processed such that a weighting coefficient is assigned thereto in accordance with a relative significance of the detection information value to the desired fire related information value. the various weighting coefficients are stored in advance in a memory. the weighting coefficients stored are established so that the fire related information value for a particular set of detection information values approximates a desired fire related information value.,1992-12-01,The title of the patent is fire alarm system and its abstract is a fire alarm system employs a neural network for obtaining one or more types of fire related information values. a plurality of detection information values are time-serially collected from plural fire phenomenon detectors. the detection information values are signal processed such that a weighting coefficient is assigned thereto in accordance with a relative significance of the detection information value to the desired fire related information value. the various weighting coefficients are stored in advance in a memory. the weighting coefficients stored are established so that the fire related information value for a particular set of detection information values approximates a desired fire related information value. dated 1992-12-01
5168352,coloring device for performing adaptive coloring of a monochromatic image,"a coloring device includes an image sampling device for sampling an input signal block representing a group of n.times.m pixels of a monochromatic image and for outputting first signals representing the sampled pixels of the input signal block of the monochromatic image; and artificial neural network, a connection for providing to the artificial neural network, substantially simultaneously, pattern information on patterns to be contained in the monochromatic image and color information on first data indicating colors given to the patterns indicated by the pattern information prior to generation of a color image signal, the artificial neural network having internal state parameters which are adaptively optimized by using a learning algorithm prior to the generation of a color image, the artificial neural network operating for receiving data representing the first signal, for determining which of colors preliminarily and respectively assigned to patterns to be contained in the group of pixels of the monochromatic image represented by the input signal block is given to a pattern actually contained in the group of pixels represented by the input signal block and for outputting second signals representing second data on three primary colors which are used to represent the determined colors given to the patterns actually contained in the group of pixels represented by the input signal block; and a color image storing device for receiving the second signals outputted from the artificial neural network, for storing the received second signals in locations thereof corresponding to the positions of the pixels represented by the input signal block and for outputting third signals representing the three primary color component images of the pixels represented by the input signal block; wherein the image sampling device further functions for scanning the whole of the monochromatic image by generating successive input signal blocks representing successive groups n.times.m pixels to be sampled, thereby outputting third signals for all pixels of the monochromatic image.",1992-12-01,"The title of the patent is coloring device for performing adaptive coloring of a monochromatic image and its abstract is a coloring device includes an image sampling device for sampling an input signal block representing a group of n.times.m pixels of a monochromatic image and for outputting first signals representing the sampled pixels of the input signal block of the monochromatic image; and artificial neural network, a connection for providing to the artificial neural network, substantially simultaneously, pattern information on patterns to be contained in the monochromatic image and color information on first data indicating colors given to the patterns indicated by the pattern information prior to generation of a color image signal, the artificial neural network having internal state parameters which are adaptively optimized by using a learning algorithm prior to the generation of a color image, the artificial neural network operating for receiving data representing the first signal, for determining which of colors preliminarily and respectively assigned to patterns to be contained in the group of pixels of the monochromatic image represented by the input signal block is given to a pattern actually contained in the group of pixels represented by the input signal block and for outputting second signals representing second data on three primary colors which are used to represent the determined colors given to the patterns actually contained in the group of pixels represented by the input signal block; and a color image storing device for receiving the second signals outputted from the artificial neural network, for storing the received second signals in locations thereof corresponding to the positions of the pixels represented by the input signal block and for outputting third signals representing the three primary color component images of the pixels represented by the input signal block; wherein the image sampling device further functions for scanning the whole of the monochromatic image by generating successive input signal blocks representing successive groups n.times.m pixels to be sampled, thereby outputting third signals for all pixels of the monochromatic image. dated 1992-12-01"
5168549,inference rule determining method and inference device,""" an inference rule determining process according to the present invention sequentially determines, using a learning function of a neural network model, a membership function representing a degree which the conditions of the if part of each inference rule is satisfied when input data is received to thereby obtain an optimal inference result without using experience rules. the inventive inference device uses an inference rule of the type """"if . . . then . . . """" and includes a membership value determiner (1) which includes all of if part and has a neural network; individual inference quantity determiners (21)-(2r) which correspond to the respective then parts of the inference rules and determine the corresponding inference quantities for the inference rules; and a final inference quantity determiner which determines these inference quantities synthetically to obtain the final results of the inference. if the individual inference quantity determiners (2) each has a neural network structure, the non-linearity of the neural network models is used to obtain the result of the inference with high inference accuracy even if an object to be inferred is non-linear. """,1992-12-01,"The title of the patent is inference rule determining method and inference device and its abstract is "" an inference rule determining process according to the present invention sequentially determines, using a learning function of a neural network model, a membership function representing a degree which the conditions of the if part of each inference rule is satisfied when input data is received to thereby obtain an optimal inference result without using experience rules. the inventive inference device uses an inference rule of the type """"if . . . then . . . """" and includes a membership value determiner (1) which includes all of if part and has a neural network; individual inference quantity determiners (21)-(2r) which correspond to the respective then parts of the inference rules and determine the corresponding inference quantities for the inference rules; and a final inference quantity determiner which determines these inference quantities synthetically to obtain the final results of the inference. if the individual inference quantity determiners (2) each has a neural network structure, the non-linearity of the neural network models is used to obtain the result of the inference with high inference accuracy even if an object to be inferred is non-linear. "" dated 1992-12-01"
5168551,mos decoder circuit implemented using a neural network architecture,"a decoder circuit based on the concept of a neural network architecture has a unique configuration using a connection structure having cmos inverters, and pmos and nmos bias and synapse transistors. the decoder circuit consists of m parallel inverter input circuit corresponding to an m-bit digital signal and forming an input neuron group, a 2.sup.m parallel inverter output circuit corresponding to 2.sup.m decoded outputs and forming an output neuron group, and a synapse group connected between the input neuron group and the output neuron group responsive to a bias group and the m-bit digital original for providing a decoded output signal to one of the 2.sup.m outputs of the output neuron group when a match is detected. hence, only one of the 2.sup.m outputs will be active at any one time.",1992-12-01,"The title of the patent is mos decoder circuit implemented using a neural network architecture and its abstract is a decoder circuit based on the concept of a neural network architecture has a unique configuration using a connection structure having cmos inverters, and pmos and nmos bias and synapse transistors. the decoder circuit consists of m parallel inverter input circuit corresponding to an m-bit digital signal and forming an input neuron group, a 2.sup.m parallel inverter output circuit corresponding to 2.sup.m decoded outputs and forming an output neuron group, and a synapse group connected between the input neuron group and the output neuron group responsive to a bias group and the m-bit digital original for providing a decoded output signal to one of the 2.sup.m outputs of the output neuron group when a match is detected. hence, only one of the 2.sup.m outputs will be active at any one time. dated 1992-12-01"
5170071,stochastic artifical neuron with multilayer training capability,"a probabilistic or stochastic artificial neuron in which the inputs and synaptic weights are represented as probabilistic or stochastic functions of time, thus providing efficient implementations of the synapses. stochastic processing removes both the time criticality and the discrete symbol nature of traditional digital processing, while retaining the basic digital processing technology. this provides large gains in relaxed timing design constraints and fault tolerance, while the simplicity of stochastic arithmetic allows for the fabrication of very high densities of neurons. the synaptic weights are individually controlled by a backward error propagation which provides the capability to train multiple layers of neurons in a neural network.",1992-12-08,"The title of the patent is stochastic artifical neuron with multilayer training capability and its abstract is a probabilistic or stochastic artificial neuron in which the inputs and synaptic weights are represented as probabilistic or stochastic functions of time, thus providing efficient implementations of the synapses. stochastic processing removes both the time criticality and the discrete symbol nature of traditional digital processing, while retaining the basic digital processing technology. this provides large gains in relaxed timing design constraints and fault tolerance, while the simplicity of stochastic arithmetic allows for the fabrication of very high densities of neurons. the synaptic weights are individually controlled by a backward error propagation which provides the capability to train multiple layers of neurons in a neural network. dated 1992-12-08"
5172204,artificial ionic synapse,"an artificial neural synapse (10) is constructed to function as a modifiable excitatory synapse. in accordance with an embodiment of the invention the synapse is fabricated as a silicon mosfet that is modified to have ions within a gate oxide. the ions, such as lithium, sodium, potassium or fluoride ions, are selected for their ability to drift within the gate oxide under the influence of an applied electric field. in response to a positive voltage applied to a gate terminal of the device, positively charged ions, such as sodium or potassium ions, drift to a silicon/silicon dioxide interface, causing an increase in current flow through the device. the invention also pertains to assemblages of such devices that are interconnected to form an artificial neuron and to assemblages of such artificial neurons that form an artificial neural network.",1992-12-15,"The title of the patent is artificial ionic synapse and its abstract is an artificial neural synapse (10) is constructed to function as a modifiable excitatory synapse. in accordance with an embodiment of the invention the synapse is fabricated as a silicon mosfet that is modified to have ions within a gate oxide. the ions, such as lithium, sodium, potassium or fluoride ions, are selected for their ability to drift within the gate oxide under the influence of an applied electric field. in response to a positive voltage applied to a gate terminal of the device, positively charged ions, such as sodium or potassium ions, drift to a silicon/silicon dioxide interface, causing an increase in current flow through the device. the invention also pertains to assemblages of such devices that are interconnected to form an artificial neuron and to assemblages of such artificial neurons that form an artificial neural network. dated 1992-12-15"
5172253,neural network model for reaching a goal state,""" an object, such as a robot, is located at an initial state in a finite state space area and moves under the control of the unsupervised neural network model of the invention. the network instructs the object to move in one of several directions from the initial state. upon reaching another state, the model again instructs the object to move in one of several directions. these instructions continue until either: a) the object has completed a cycle by ending up back at a state it has been to previously during this cycle, or b) the object has completed a cycle by reaching the goal state. if the object ends up back at a state it has been to previously during this cycle, the neural network model ends the cycle and immediately begins a new cycle from the present location. when the object reaches the goal state, the neural network model learns that this path is productive towards reaching the goal state, and is given delayed reinforcement in the form of a """"reward"""". upon reaching a state, the neural network model calculates a level of satisfaction with its progress towards reaching the goal state. if the level of satisfaction is low, the neural network model is more likely to override what has been learned thus far and deviate from a path known to lead to the goal state to experiment with new and possibly better paths. """,1992-12-15,"The title of the patent is neural network model for reaching a goal state and its abstract is "" an object, such as a robot, is located at an initial state in a finite state space area and moves under the control of the unsupervised neural network model of the invention. the network instructs the object to move in one of several directions from the initial state. upon reaching another state, the model again instructs the object to move in one of several directions. these instructions continue until either: a) the object has completed a cycle by ending up back at a state it has been to previously during this cycle, or b) the object has completed a cycle by reaching the goal state. if the object ends up back at a state it has been to previously during this cycle, the neural network model ends the cycle and immediately begins a new cycle from the present location. when the object reaches the goal state, the neural network model learns that this path is productive towards reaching the goal state, and is given delayed reinforcement in the form of a """"reward"""". upon reaching a state, the neural network model calculates a level of satisfaction with its progress towards reaching the goal state. if the level of satisfaction is low, the neural network model is more likely to override what has been learned thus far and deviate from a path known to lead to the goal state to experiment with new and possibly better paths. "" dated 1992-12-15"
5172490,clothes dryer with neurocontrol device,"a clothes dryer of the dehumidification type is disclosed in which hot air induced by a heater is circulated from a drying compartment through a heat exchanger. a volume, wetness, wetness unevenness, temperature, temperature unevenness of clothes to be dried and the temperature of the hot air blown out of the drying compartment are detected by respective detectors. results of detection are input to a control device incorporating a neural network. the control device operates in the manner of neurocontrol to control a volume of outside air supplied to the heat exchanger and a heating value of the heater.",1992-12-22,"The title of the patent is clothes dryer with neurocontrol device and its abstract is a clothes dryer of the dehumidification type is disclosed in which hot air induced by a heater is circulated from a drying compartment through a heat exchanger. a volume, wetness, wetness unevenness, temperature, temperature unevenness of clothes to be dried and the temperature of the hot air blown out of the drying compartment are detected by respective detectors. results of detection are input to a control device incorporating a neural network. the control device operates in the manner of neurocontrol to control a volume of outside air supplied to the heat exchanger and a heating value of the heater. dated 1992-12-22"
5175678,method and procedure for neural control of dynamic processes,a neural network control based on a general multi-variable nonlinear dynamic model incorporating time delays is disclosed. the inverse dynamics of the process being controlled is learned represented by a multi-layer neural network which is used as a feedforward control to achieve a specified closed loop response under varying conditions. the weights between the layers in the neural network are adjusted during the learning process. the learning process is based on minimizing the combined error between the desired process value and the actual process output and the error between the desired process value and the inverse process neural network output.,1992-12-29,The title of the patent is method and procedure for neural control of dynamic processes and its abstract is a neural network control based on a general multi-variable nonlinear dynamic model incorporating time delays is disclosed. the inverse dynamics of the process being controlled is learned represented by a multi-layer neural network which is used as a feedforward control to achieve a specified closed loop response under varying conditions. the weights between the layers in the neural network are adjusted during the learning process. the learning process is based on minimizing the combined error between the desired process value and the actual process output and the error between the desired process value and the inverse process neural network output. dated 1992-12-29
5175793,recognition apparatus using articulation positions for recognizing a voice,"a first voice recognition apparatus includes a device for analyzing frequencies of the input voice and a device coupled to the analyzing unit for determining vowel zones and consonant zones of the analyzed input voice. the apparatus further includes a device for determining positions of articulation of an input voice determined from the vowel zones by calculating from frequency components of the input voice in accordance with a predetermined algorithm based on frequency components of monophthongs having known phonation contents and positions of articulation. a second voice recognition apparatus includes a device for analyzing frequencies of the input voice so as to derive acoustic parameters from the input voice. a pattern converting unit is coupled to the analyzing unit and uses a neural network for converting the acoustic parameters to articulartory vectors. the neural network is capable of learning, by the error back propagation method using target data produced by a predetermined sequence based on the acoustic parameters, to create rules for converting the acoustic parameters of the input voice to articulatory vectors having at least two vector elements. a recognizing unit is coupled to the pattern converting unit for recognizing the input voice by comparing a feature pattern of the analyzed input voice having the articulatory vector with reference feature patterns in a predetermined sequence. a storage unit is coupled to the recognizing unit for storing the reference feature patterns having the articulatory vectors created by the pattern converting unit.",1992-12-29,"The title of the patent is recognition apparatus using articulation positions for recognizing a voice and its abstract is a first voice recognition apparatus includes a device for analyzing frequencies of the input voice and a device coupled to the analyzing unit for determining vowel zones and consonant zones of the analyzed input voice. the apparatus further includes a device for determining positions of articulation of an input voice determined from the vowel zones by calculating from frequency components of the input voice in accordance with a predetermined algorithm based on frequency components of monophthongs having known phonation contents and positions of articulation. a second voice recognition apparatus includes a device for analyzing frequencies of the input voice so as to derive acoustic parameters from the input voice. a pattern converting unit is coupled to the analyzing unit and uses a neural network for converting the acoustic parameters to articulartory vectors. the neural network is capable of learning, by the error back propagation method using target data produced by a predetermined sequence based on the acoustic parameters, to create rules for converting the acoustic parameters of the input voice to articulatory vectors having at least two vector elements. a recognizing unit is coupled to the pattern converting unit for recognizing the input voice by comparing a feature pattern of the analyzed input voice having the articulatory vector with reference feature patterns in a predetermined sequence. a storage unit is coupled to the recognizing unit for storing the reference feature patterns having the articulatory vectors created by the pattern converting unit. dated 1992-12-29"
5175798,digital artificial neuron based on a probabilistic ram,"a neuron for use in a neural processing network, comprises a memory having a plurality of storage locations at each of which a number representing a probability is stored, each of the storage locations being selectively addressable to cause the contents of the location to be read to an input of a comparator. a noise generator inputs to the comparator a random number representing noise. at an output of the comparator an output signal appears having a first or second value depending on the values of the numbers received from the addressed storage location and the noise generator, the probability of the output signal having a given one of the first and second values being determined by the number at the addressed location. preferably the neuron receives from the environment signals representing success or failure of the network, the value of the number stored at the addressed location being changed in such a way as to increase the probability of the successful action if a success signal is received, and to decrease the probability of the unsuccessful action if a failure signal is received.",1992-12-29,"The title of the patent is digital artificial neuron based on a probabilistic ram and its abstract is a neuron for use in a neural processing network, comprises a memory having a plurality of storage locations at each of which a number representing a probability is stored, each of the storage locations being selectively addressable to cause the contents of the location to be read to an input of a comparator. a noise generator inputs to the comparator a random number representing noise. at an output of the comparator an output signal appears having a first or second value depending on the values of the numbers received from the addressed storage location and the noise generator, the probability of the output signal having a given one of the first and second values being determined by the number at the addressed location. preferably the neuron receives from the environment signals representing success or failure of the network, the value of the number stored at the addressed location being changed in such a way as to increase the probability of the successful action if a success signal is received, and to decrease the probability of the unsuccessful action if a failure signal is received. dated 1992-12-29"
5177746,error correction circuit using a design based on a neural network model,an error correction circuit is provided which uses nmos and pmos synapses to form neural network type responses to a coded multi-bit input. use of mos technology logic in error correction circuits allows such devices to be easily interfaced with other like technology circuits without the need to use distinct interface logic as with conventional error correction circuitry.,1993-01-05,The title of the patent is error correction circuit using a design based on a neural network model and its abstract is an error correction circuit is provided which uses nmos and pmos synapses to form neural network type responses to a coded multi-bit input. use of mos technology logic in error correction circuits allows such devices to be easily interfaced with other like technology circuits without the need to use distinct interface logic as with conventional error correction circuitry. dated 1993-01-05
5177994,odor sensing system,"an odor sensing system is comprised of a sensor cell including a plurality of quartz resonator sensors aligned therein to detect odor by variation of resonance frequencies derived from weight loading on surfaces thereof, a recognition line including a neural network which recognizes data obtained by subtraction between an output signal of the sensor as frequency variation and, a reference signal selected by one of the output signals of the sensor. the sensor cell is thermostatically regulated by circulating thermostatic water therein to maintain the temperature higher than an advance line of the system. a sample to be recognized is supplied to the sensor cell in a form of vapor generated by blowing a standard gas onto the surface of the sample.",1993-01-12,"The title of the patent is odor sensing system and its abstract is an odor sensing system is comprised of a sensor cell including a plurality of quartz resonator sensors aligned therein to detect odor by variation of resonance frequencies derived from weight loading on surfaces thereof, a recognition line including a neural network which recognizes data obtained by subtraction between an output signal of the sensor as frequency variation and, a reference signal selected by one of the output signals of the sensor. the sensor cell is thermostatically regulated by circulating thermostatic water therein to maintain the temperature higher than an advance line of the system. a sample to be recognized is supplied to the sensor cell in a form of vapor generated by blowing a standard gas onto the surface of the sample. dated 1993-01-12"
5179624,speech recognition apparatus using neural network and fuzzy logic,"a speech recognition apparatus has: a speech input unit for inputting a speech; a speech analysis unit for analyzing the inputted speech to output the time series of a feature vector; a candidates selection unit for inputting the time series of a feature vector from the speech analysis unit to select a plurality of candidates of recognition result from the speech categories; and a discrimination processing unit for discriminating the selected candidates to obtain a final recognition result. the discrimination processing unit includes three components in the form of a pair generation unit for generating all of the two combinations of the n-number of candidates selected by said candidate selection unit a pair discrimination unit for discriminating which of the candidates of the combinations is more certain for each of all .sub.n c.sub.2 -number of combinations (or pairs) on the basis of the extracted result of the acoustic feature intrinsic to each of said candidate speeches and a final decision unit for collecting all the pair discrimination results obtained from the pair discrimination unit for each of all the .sub.n c.sub.2 -number of combinations (or pairs) to decide the final result. the pair discrimination unit handles the extracted result of the acoustic feature intrinsic to each of the candidate speeches as fuzzy information and accomplishes the discrimination processing on the basis of fuzzy logic algorithms, and the final decision unit accomplishes its collections on the basis of the fuzzy logic algorithms.",1993-01-12,"The title of the patent is speech recognition apparatus using neural network and fuzzy logic and its abstract is a speech recognition apparatus has: a speech input unit for inputting a speech; a speech analysis unit for analyzing the inputted speech to output the time series of a feature vector; a candidates selection unit for inputting the time series of a feature vector from the speech analysis unit to select a plurality of candidates of recognition result from the speech categories; and a discrimination processing unit for discriminating the selected candidates to obtain a final recognition result. the discrimination processing unit includes three components in the form of a pair generation unit for generating all of the two combinations of the n-number of candidates selected by said candidate selection unit a pair discrimination unit for discriminating which of the candidates of the combinations is more certain for each of all .sub.n c.sub.2 -number of combinations (or pairs) on the basis of the extracted result of the acoustic feature intrinsic to each of said candidate speeches and a final decision unit for collecting all the pair discrimination results obtained from the pair discrimination unit for each of all the .sub.n c.sub.2 -number of combinations (or pairs) to decide the final result. the pair discrimination unit handles the extracted result of the acoustic feature intrinsic to each of the candidate speeches as fuzzy information and accomplishes the discrimination processing on the basis of fuzzy logic algorithms, and the final decision unit accomplishes its collections on the basis of the fuzzy logic algorithms. dated 1993-01-12"
5179631,neural network logic system,""" a novel neural network implementation for logic systems has been developed. the neural network can determine whether a particular logic system and knowledge base are self-consistent, which can be a difficult problem for more complex systems. through neural network hardware using parallel computation, valid solutions may be found more rapidly than could be done with previous, software-based implementations. this neural network is particularly suited for use in large, real-time problems, such as in a real-time expert system for testing the consistency of a programmable process controller, for testing the consistency of an integrated circuit design, or for testing the consistency of an """"expert system."""" this neural network may also be used as an """"inference engine,"""" i.e., to test the validity of a particular logical expression in the context of a given logic system and knowledge base, or to search for all valid solutions, or to search for valid solutions consistent with given truth values which have been """"clamped"""" as true or false. the neural network may be used with many different types of logic systems: those based on conventional """"truth table"""" logic, those based on a truth maintenance system, or many other types of logic systems. the """"justifications"""" corresponding to a particular logic system and knowledge base may be permanently hard-wired by the manufacturer, or may be supplied by the user, either reversibly or irreversibly. """,1993-01-12,"The title of the patent is neural network logic system and its abstract is "" a novel neural network implementation for logic systems has been developed. the neural network can determine whether a particular logic system and knowledge base are self-consistent, which can be a difficult problem for more complex systems. through neural network hardware using parallel computation, valid solutions may be found more rapidly than could be done with previous, software-based implementations. this neural network is particularly suited for use in large, real-time problems, such as in a real-time expert system for testing the consistency of a programmable process controller, for testing the consistency of an integrated circuit design, or for testing the consistency of an """"expert system."""" this neural network may also be used as an """"inference engine,"""" i.e., to test the validity of a particular logical expression in the context of a given logic system and knowledge base, or to search for all valid solutions, or to search for valid solutions consistent with given truth values which have been """"clamped"""" as true or false. the neural network may be used with many different types of logic systems: those based on conventional """"truth table"""" logic, those based on a truth maintenance system, or many other types of logic systems. the """"justifications"""" corresponding to a particular logic system and knowledge base may be permanently hard-wired by the manufacturer, or may be supplied by the user, either reversibly or irreversibly. "" dated 1993-01-12"
5180911,parameter measurement systems and methods having a neural network comprising parameter output means,"a system for measuring the value of a parameter, e.g., structural strain, includes an optical waveguide, a laser or equivalent light source for launching coherent light into the waveguide to propagate therein as multi modes, an array of a plurality of spaced apart photodetectors each comprising a light receptor surface and signal output, said array being arranged to have light emitted from said waveguide output portion irradiate said light receptor surfaces, an artificial neural network formed of a plurality of spaced apart neurons, connectors to impose weighted portions of signal outputs from the photodetectors upon the neurons which register the parameter value on a meter or like output device.",1993-01-19,"The title of the patent is parameter measurement systems and methods having a neural network comprising parameter output means and its abstract is a system for measuring the value of a parameter, e.g., structural strain, includes an optical waveguide, a laser or equivalent light source for launching coherent light into the waveguide to propagate therein as multi modes, an array of a plurality of spaced apart photodetectors each comprising a light receptor surface and signal output, said array being arranged to have light emitted from said waveguide output portion irradiate said light receptor surfaces, an artificial neural network formed of a plurality of spaced apart neurons, connectors to impose weighted portions of signal outputs from the photodetectors upon the neurons which register the parameter value on a meter or like output device. dated 1993-01-19"
5181171,adaptive network for automated first break picking of seismic refraction events and method of operating the same,"an adaptive, or neural, network and a method of operating the same is disclosed which is particularly adapted for performing first break analysis for seismic shot records. the adaptive network is first trained according to the generalized delta rule. the disclosed training method includes selection of the seismic trace with the highest error, where the backpropagation is performed according to the error of this worst trace. the learning and momentum factors in the generalized delta rule are adjusted according to the value of the worst error, so that the learning and momentum factors increase as the error decreases. the training method further includes detection of slow convergence regions, and methods for escaping such regions including restoration of previously trimmed dormant links, renormalization of the weighting factor values, and the addition of new layers to the network. the network, after the addition of a new layer, includes links between nodes which skip the hidden layer. the error value used in the backpropagation is reduced from that actually calculated, by adjusting the desired output value, in order to reduce the growth of the weighting factors. after the training of the network, data corresponding to an average of the graphical display of a portion of the shot record, including multiple traces over a period of time, is provided to the network. the time of interest of the data is incremented until such time as the network indicates that the time of interest equals the first break time. the analysis may be repeated for all of the traces in the shot record.",1993-01-19,"The title of the patent is adaptive network for automated first break picking of seismic refraction events and method of operating the same and its abstract is an adaptive, or neural, network and a method of operating the same is disclosed which is particularly adapted for performing first break analysis for seismic shot records. the adaptive network is first trained according to the generalized delta rule. the disclosed training method includes selection of the seismic trace with the highest error, where the backpropagation is performed according to the error of this worst trace. the learning and momentum factors in the generalized delta rule are adjusted according to the value of the worst error, so that the learning and momentum factors increase as the error decreases. the training method further includes detection of slow convergence regions, and methods for escaping such regions including restoration of previously trimmed dormant links, renormalization of the weighting factor values, and the addition of new layers to the network. the network, after the addition of a new layer, includes links between nodes which skip the hidden layer. the error value used in the backpropagation is reduced from that actually calculated, by adjusting the desired output value, in order to reduce the growth of the weighting factors. after the training of the network, data corresponding to an average of the graphical display of a portion of the shot record, including multiple traces over a period of time, is provided to the network. the time of interest of the data is incremented until such time as the network indicates that the time of interest equals the first break time. the analysis may be repeated for all of the traces in the shot record. dated 1993-01-19"
5181256,pattern recognition device using a neural network,"a pattern recognition device has a dp matching section. the dp matching section performs frequency expansion dp matching to a standard pattern and a characteristic pattern obtained from input voice waveform to obtain a dp score and dp path pattern. it is determined by means of a category identification neural network using the dp path pattern obtained from the dp matching section whether a category of the standard pattern and a category of the characteristic pattern are the same, and a determination result corresponding to the degree of identification is obtained. a normalized dp score, which is the dp score normalized for individual differences within a required range, is then obtained in a divider by compensating the dp score using the determination result.",1993-01-19,"The title of the patent is pattern recognition device using a neural network and its abstract is a pattern recognition device has a dp matching section. the dp matching section performs frequency expansion dp matching to a standard pattern and a characteristic pattern obtained from input voice waveform to obtain a dp score and dp path pattern. it is determined by means of a category identification neural network using the dp path pattern obtained from the dp matching section whether a category of the standard pattern and a category of the characteristic pattern are the same, and a determination result corresponding to the degree of identification is obtained. a normalized dp score, which is the dp score normalized for individual differences within a required range, is then obtained in a divider by compensating the dp score using the determination result. dated 1993-01-19"
5182794,recurrent neural networks teaching system,"a teaching method for a recurrent neural network having hidden, output and input neurons calculates weighting errors over a limited number of propagations of the network. this process permits the use of conventional teaching sets, such as are used with feedforward networks, to be used with recurrent networks. the teaching outputs are substituted for the computed activations of the output neurons in the forward propagation and error correction stages. back propagated error from the last propagation is assumed to be zero for the hidden neurons. a method of reducing drift of the network with respect to a modeled process is also described and a forced cycling method to eliminate the time lag between network input and output.",1993-01-26,"The title of the patent is recurrent neural networks teaching system and its abstract is a teaching method for a recurrent neural network having hidden, output and input neurons calculates weighting errors over a limited number of propagations of the network. this process permits the use of conventional teaching sets, such as are used with feedforward networks, to be used with recurrent networks. the teaching outputs are substituted for the computed activations of the output neurons in the forward propagation and error correction stages. back propagated error from the last propagation is assumed to be zero for the hidden neurons. a method of reducing drift of the network with respect to a modeled process is also described and a forced cycling method to eliminate the time lag between network input and output. dated 1993-01-26"
5184218,bandwidth compression and expansion system,a bandwidth compression and expansion system is provided in which analog data is processed in real time using a sub-sampling technique in which pixels or other data values within a sub-sampling region determine the value of a corresponding signal which also denotes trends or patterns in accordance with the other pixels or signal values within a sampling region encompassing the sub-sampling region. neural networks are used to implement the sub-sampling process both during bandwidth compression and during bandwidth expansion in which interpolation and extrapolation are employed to reverse the sub-sampling process used during compression. the neural network forms part of an arrangement in which analog input signals are converted to digital signals that are then stored in a random access memory which operates in conjunction with an address generator for identifying a succession of sampling and sub-sampling regions within the memory. the output of the memory is converted to an analog signal before being held in a sample and hold memory for use in the neural network.,1993-02-02,The title of the patent is bandwidth compression and expansion system and its abstract is a bandwidth compression and expansion system is provided in which analog data is processed in real time using a sub-sampling technique in which pixels or other data values within a sub-sampling region determine the value of a corresponding signal which also denotes trends or patterns in accordance with the other pixels or signal values within a sampling region encompassing the sub-sampling region. neural networks are used to implement the sub-sampling process both during bandwidth compression and during bandwidth expansion in which interpolation and extrapolation are employed to reverse the sub-sampling process used during compression. the neural network forms part of an arrangement in which analog input signals are converted to digital signals that are then stored in a random access memory which operates in conjunction with an address generator for identifying a succession of sampling and sub-sampling regions within the memory. the output of the memory is converted to an analog signal before being held in a sample and hold memory for use in the neural network. dated 1993-02-02
5185816,method of selecting characeteristics data for a data processing system,"a method of selecting characteristics data for a data processing system from a group of input data for reducing data volume of each input data by said data processing system having a neural network structure or a structure equivalent thereto, where each input data consists of a plurality of said characteristics data. the method includes the steps of storing outputs of said input data; selecting a specific characteristics data of a pair of different input data; exchanging said characteristics data of said input data pair with each other; comparing outputs from said data processing system in response to said input data before and after the exchange of said characteristics data; and removing said characteristics data from said input data in said group, when a difference between said outputs before and after is comparatively small.",1993-02-09,"The title of the patent is method of selecting characeteristics data for a data processing system and its abstract is a method of selecting characteristics data for a data processing system from a group of input data for reducing data volume of each input data by said data processing system having a neural network structure or a structure equivalent thereto, where each input data consists of a plurality of said characteristics data. the method includes the steps of storing outputs of said input data; selecting a specific characteristics data of a pair of different input data; exchanging said characteristics data of said input data pair with each other; comparing outputs from said data processing system in response to said input data before and after the exchange of said characteristics data; and removing said characteristics data from said input data in said group, when a difference between said outputs before and after is comparatively small. dated 1993-02-09"
5185848,noise reduction system using neural network,"a noise reduction system used for transmission and/or recognition of speech includes a speech analyzer for analyzing a noisy speech input signal thereby converting the speech signal into feature vectors such as autocorrelation coefficients, and a neural network for receiving the feature vectors of the noisy speech signal as its input. the neural network extracts from a codebook an index of prototype vectors corresponding to a noise-free equivalent to the noisy speech input signal. feature vectors of speech are read out from the codebook on the basis of the index delivered as an output from the neural network, thereby causing the speech input to be reproduced on the basis of the feature vectors of speech read out from the codebook.",1993-02-09,"The title of the patent is noise reduction system using neural network and its abstract is a noise reduction system used for transmission and/or recognition of speech includes a speech analyzer for analyzing a noisy speech input signal thereby converting the speech signal into feature vectors such as autocorrelation coefficients, and a neural network for receiving the feature vectors of the noisy speech signal as its input. the neural network extracts from a codebook an index of prototype vectors corresponding to a noise-free equivalent to the noisy speech input signal. feature vectors of speech are read out from the codebook on the basis of the index delivered as an output from the neural network, thereby causing the speech input to be reproduced on the basis of the feature vectors of speech read out from the codebook. dated 1993-02-09"
5185850,color transformation method and apparatus for transforming physical to psychological attribute using a neural network,"to practice a method of transforming color sensation informations such that multidimensional physical informations and color sensation informations sensed by living bodies in response to the physical informations are non-linearly transformed therebetween, a multilayer feedforward type neural network is used for the purpose of accomplishing the foregoing transformation. the physical informations are provided in the form of data derived from multidimensional spectral distribution of light and the color sensation informations are provided in the form of sensitive colors each sensed by the living bodies as a psychological quantity relative to a certain color. an apparatus for carrying out the method includes an input section into which a physical quantity is inputted as an electrical signal, an information transforming section in which the inputted signal is transformed into a color sensation information representing psychological quantity of color and an output section from which the transformed color information is outputted. the information transforming section includes a multilayer feedforward type neural network.",1993-02-09,"The title of the patent is color transformation method and apparatus for transforming physical to psychological attribute using a neural network and its abstract is to practice a method of transforming color sensation informations such that multidimensional physical informations and color sensation informations sensed by living bodies in response to the physical informations are non-linearly transformed therebetween, a multilayer feedforward type neural network is used for the purpose of accomplishing the foregoing transformation. the physical informations are provided in the form of data derived from multidimensional spectral distribution of light and the color sensation informations are provided in the form of sensitive colors each sensed by the living bodies as a psychological quantity relative to a certain color. an apparatus for carrying out the method includes an input section into which a physical quantity is inputted as an electrical signal, an information transforming section in which the inputted signal is transformed into a color sensation information representing psychological quantity of color and an output section from which the transformed color information is outputted. the information transforming section includes a multilayer feedforward type neural network. dated 1993-02-09"
5195169,control device for controlling learning of a neural network,"a control device for controlling the learning of a neural netowrk includes a monitor for monitoring weight values of synapse connections between units of the neural netowrk during learning of the neural network so as to update these weight values. when one of the weight values satisfies a preset condition, the weight value is updated to a predetermined value such that configuration of the neural network is determined in an optimum manner.",1993-03-16,"The title of the patent is control device for controlling learning of a neural network and its abstract is a control device for controlling the learning of a neural netowrk includes a monitor for monitoring weight values of synapse connections between units of the neural netowrk during learning of the neural network so as to update these weight values. when one of the weight values satisfies a preset condition, the weight value is updated to a predetermined value such that configuration of the neural network is determined in an optimum manner. dated 1993-03-16"
5195170,neural-network dedicated processor for solving assignment problems,"a neural network processor for solving first-order competitive assignment problems consists of a matrix of n.times.m processing units, each of which corresponds to the pairing of a first number of elements of {r.sub.i } with a second number of elements {c.sub.j }, wherein limits of the first number are programmed in row control superneurons, and limits of the second number are programmed in column superneurons as min and max values. the cost (weight) w.sub.ij of the pairings is programmed separately into each pu. for each row and column of pus, a dedicated constraint superneuron insures that the number of active neurons within the associated row or column fall within a specified range. annealing is provided by gradually increasing the pu gain for each row and column or increasing positive feedback to each pu, the latter being effective to increase hysteresis of each pu or by combining both of these techniques.",1993-03-16,"The title of the patent is neural-network dedicated processor for solving assignment problems and its abstract is a neural network processor for solving first-order competitive assignment problems consists of a matrix of n.times.m processing units, each of which corresponds to the pairing of a first number of elements of {r.sub.i } with a second number of elements {c.sub.j }, wherein limits of the first number are programmed in row control superneurons, and limits of the second number are programmed in column superneurons as min and max values. the cost (weight) w.sub.ij of the pairings is programmed separately into each pu. for each row and column of pus, a dedicated constraint superneuron insures that the number of active neurons within the associated row or column fall within a specified range. annealing is provided by gradually increasing the pu gain for each row and column or increasing positive feedback to each pu, the latter being effective to increase hysteresis of each pu or by combining both of these techniques. dated 1993-03-16"
5197114,computer neural network regulatory process control system and method,"a computer neural network regulatory process control system and method allows for the elimination of a human operator from real time control of the process. the present invention operates in three modes: training, operation (prediction), and retraining. in the training mode, training input data is produced by the control adjustment made to the process by the human operator. the neural network of the present invention is trained by producing output data using input data for prediction. the output data is compared with the training input data to produce error data, which is used to adjust the weight(s) of the neural network. when the error data is less than a preselected criterion, training has been completed. in the operation mode, the neutral network of the present invention provides output data based upon predictions using the input data. the output data is used to control a state of the process via an actuator. in the retraining mode, retraining data is supplied by monitoring the supplemental actions of the human operator. the retraining data is used by the neural network for adjusting the weight(s) of the neural network.",1993-03-23,"The title of the patent is computer neural network regulatory process control system and method and its abstract is a computer neural network regulatory process control system and method allows for the elimination of a human operator from real time control of the process. the present invention operates in three modes: training, operation (prediction), and retraining. in the training mode, training input data is produced by the control adjustment made to the process by the human operator. the neural network of the present invention is trained by producing output data using input data for prediction. the output data is compared with the training input data to produce error data, which is used to adjust the weight(s) of the neural network. when the error data is less than a preselected criterion, training has been completed. in the operation mode, the neutral network of the present invention provides output data based upon predictions using the input data. the output data is used to control a state of the process via an actuator. in the retraining mode, retraining data is supplied by monitoring the supplemental actions of the human operator. the retraining data is used by the neural network for adjusting the weight(s) of the neural network. dated 1993-03-23"
5200816,method and apparatus for color processing with neural networks,""" a method and apparatus for constructing, training and utilizing an artificial neural network (also termed herein a """"neural network"""", an ann, or an nn) in order to transform a first color value in a first color coordinate system into a second color value in a second color coordinate system. """,1993-04-06,"The title of the patent is method and apparatus for color processing with neural networks and its abstract is "" a method and apparatus for constructing, training and utilizing an artificial neural network (also termed herein a """"neural network"""", an ann, or an nn) in order to transform a first color value in a first color coordinate system into a second color value in a second color coordinate system. "" dated 1993-04-06"
5200898,method of controlling motor vehicle,"a motor vehicle is controlled with a neural network which has a data learning capability. a present value of the throttle valve opening of the engine on the motor vehicle and a rate of change of the present value of the throttle valve opening are periodically supplied to the neural network. the neural network is controlled to learn the present value of the throttle valve opening when the rate of change of the present value of the throttle valve opening becomes zero so that a predicted value of the throttle valve opening approaches the actual value of the throttle valve opening at the time the rate of change thereof becomes zero. an operating condition of the motor vehicle is controlled based on the predicted value of the throttle valve opening, which is represented by a periodically produced output signal from the neural network.",1993-04-06,"The title of the patent is method of controlling motor vehicle and its abstract is a motor vehicle is controlled with a neural network which has a data learning capability. a present value of the throttle valve opening of the engine on the motor vehicle and a rate of change of the present value of the throttle valve opening are periodically supplied to the neural network. the neural network is controlled to learn the present value of the throttle valve opening when the rate of change of the present value of the throttle valve opening becomes zero so that a predicted value of the throttle valve opening approaches the actual value of the throttle valve opening at the time the rate of change thereof becomes zero. an operating condition of the motor vehicle is controlled based on the predicted value of the throttle valve opening, which is represented by a periodically produced output signal from the neural network. dated 1993-04-06"
5200908,placement optimizing method/apparatus and apparatus for designing semiconductor devices,"a method of finding the optimal placement of circuit elements is disclosed in which the optimal position of each circuit element is determined from the results of arithmetic operations performed by a processor network where a plurality of processors are interconnected so as to form a neural network, and each processor takes in its own output and the outputs of all other processors to solve a problem.",1993-04-06,"The title of the patent is placement optimizing method/apparatus and apparatus for designing semiconductor devices and its abstract is a method of finding the optimal placement of circuit elements is disclosed in which the optimal position of each circuit element is determined from the results of arithmetic operations performed by a processor network where a plurality of processors are interconnected so as to form a neural network, and each processor takes in its own output and the outputs of all other processors to solve a problem. dated 1993-04-06"
5201026,method of architecting multiple neural network and system therefor,"to facilitate architecting of a multiple neural network, irrespective of the quantity of cases and the complexity of case dependence relationship, sets of input instances and the desirable outputs corresponding thereto are stored; the stored sets are read in sequence to discriminate whether all variables included in the input instances of the read set are included in the outputs of any given stored set, to mark variables not included in the outputs of any sets; the sets whose input instances include only the marked variables are selected from among the read sets; unit neural networks for learning the selected sets are formed and simultaneously variables included in the outputs of the formed unit neural networks are marked; a unit neural network for learning any given set is formed; and the formed unit neural networks are connected to each other to architect a multiple neural network.",1993-04-06,"The title of the patent is method of architecting multiple neural network and system therefor and its abstract is to facilitate architecting of a multiple neural network, irrespective of the quantity of cases and the complexity of case dependence relationship, sets of input instances and the desirable outputs corresponding thereto are stored; the stored sets are read in sequence to discriminate whether all variables included in the input instances of the read set are included in the outputs of any given stored set, to mark variables not included in the outputs of any sets; the sets whose input instances include only the marked variables are selected from among the read sets; unit neural networks for learning the selected sets are formed and simultaneously variables included in the outputs of the formed unit neural networks are marked; a unit neural network for learning any given set is formed; and the formed unit neural networks are connected to each other to architect a multiple neural network. dated 1993-04-06"
5202956,semiconductor neural network and operating method thereof,"a semiconductor neural network includes a coupling matrix having coupling elements arranged in a matrix which couple with specific coupling strengths internal data input lines to internal data output lines. the internal data output lines are divided into groups. the neural network further comprises weighting addition circuits provided corresponding to the groups of the internal data output lines. a weighting addition circuit includes weighing elements for adding weights to signals on the internal data output lines in the corresponding group and outputting the weighted signals, and an addition circuit for outputting a total sum of the outputs of those weighting elements. the internal data output lines are arranged to form pairs and the addition circuit has a first input terminal for receiving one weighting element output of each of the pairs in common, a second input terminal for receiving the other weighting element output of each of the pairs in common, and sense amplifier for differentially amplifying signals at the first and second input terminals. the neural network further includes a circuit for detecting a change time of an input signals, a circuit responsive to an input signal change for equalizing the first and second input terminals for a predetermined period, and a circuit for activating the sense amplifier after the equalization is completed. the information retention capability of each coupling element is set according to the weight of an associated weighting element. this neural network can provide multi-valued expression of coupling strength with fewer coupling elements.",1993-04-13,"The title of the patent is semiconductor neural network and operating method thereof and its abstract is a semiconductor neural network includes a coupling matrix having coupling elements arranged in a matrix which couple with specific coupling strengths internal data input lines to internal data output lines. the internal data output lines are divided into groups. the neural network further comprises weighting addition circuits provided corresponding to the groups of the internal data output lines. a weighting addition circuit includes weighing elements for adding weights to signals on the internal data output lines in the corresponding group and outputting the weighted signals, and an addition circuit for outputting a total sum of the outputs of those weighting elements. the internal data output lines are arranged to form pairs and the addition circuit has a first input terminal for receiving one weighting element output of each of the pairs in common, a second input terminal for receiving the other weighting element output of each of the pairs in common, and sense amplifier for differentially amplifying signals at the first and second input terminals. the neural network further includes a circuit for detecting a change time of an input signals, a circuit responsive to an input signal change for equalizing the first and second input terminals for a predetermined period, and a circuit for activating the sense amplifier after the equalization is completed. the information retention capability of each coupling element is set according to the weight of an associated weighting element. this neural network can provide multi-valued expression of coupling strength with fewer coupling elements. dated 1993-04-13"
5203984,monitoring system for plant operation condition and its in-situ electrochemical electrode,"a plant operational status monitoring supervisory system comprising; means for extracting information directly relating to water quality of an objective portion consecutively for a period of time by means of an electrochemical water quality sensor installed in an objective portion to monitor in-situ in a plant; means for evaluating water quality based on thus extracted information; means for comparing an obtained water quality evaluation result with a reference value for a predetermined plant operation procedure; and means for displaying or storing necessary portion out of said comparison results; is disclosed. an electrochemical reference electrode used in this system being provided with an electrolyte layer containing ion of the electrode member; a porous ceramic layer surrounding the same without permeating liquid; and electrode member electrochemically contacting with said elec-trolyte layer; and a terminal electrically contacting with said electrode member; and further having a long life in high temperature water, various status of high temperature water in objective portions and that of nearby constituent members in a plant are possible to be monitored online by means of this reference electrode. further, because monitored data are processed by means of a neural network, the higher precision level of monitoring has been achieved.",1993-04-20,"The title of the patent is monitoring system for plant operation condition and its in-situ electrochemical electrode and its abstract is a plant operational status monitoring supervisory system comprising; means for extracting information directly relating to water quality of an objective portion consecutively for a period of time by means of an electrochemical water quality sensor installed in an objective portion to monitor in-situ in a plant; means for evaluating water quality based on thus extracted information; means for comparing an obtained water quality evaluation result with a reference value for a predetermined plant operation procedure; and means for displaying or storing necessary portion out of said comparison results; is disclosed. an electrochemical reference electrode used in this system being provided with an electrolyte layer containing ion of the electrode member; a porous ceramic layer surrounding the same without permeating liquid; and electrode member electrochemically contacting with said elec-trolyte layer; and a terminal electrically contacting with said electrode member; and further having a long life in high temperature water, various status of high temperature water in objective portions and that of nearby constituent members in a plant are possible to be monitored online by means of this reference electrode. further, because monitored data are processed by means of a neural network, the higher precision level of monitoring has been achieved. dated 1993-04-20"
5204872,control system for electric arc furnace,"an improved arc furnace regulator employs neural circuits connected in a multi-layer network configuration with various weighted relationships between the successive layers which are automatically changed over time as a function of an error signal by means of the back-propagation method so that the regulator gradually improves its control algorithm as a result of accumulated experience. the network is implemented in software which can be developed and run on a pc with extra co-computing capability for greater execution speed. a second trainable neural network which emulates the arc furnace is used to develop the error signal, and is trained in mutually exclusive time periods with the training of the regular network.",1993-04-20,"The title of the patent is control system for electric arc furnace and its abstract is an improved arc furnace regulator employs neural circuits connected in a multi-layer network configuration with various weighted relationships between the successive layers which are automatically changed over time as a function of an error signal by means of the back-propagation method so that the regulator gradually improves its control algorithm as a result of accumulated experience. the network is implemented in software which can be developed and run on a pc with extra co-computing capability for greater execution speed. a second trainable neural network which emulates the arc furnace is used to develop the error signal, and is trained in mutually exclusive time periods with the training of the regular network. dated 1993-04-20"
5204938,method of implementing a neural network on a digital computer,"a digital computer architecture specifically tailored for implementing a neural network. several simultaneously operable processors (10) each have their own local memory (17) for storing weight and connectivity information corresponding to nodes of the neural network whose output values will be calculated by said processor (10). a global memory (55,56) is coupled to each of the processors (10) via a common data bus (30). output values corresponding to a first layer of the neural network are broadcast from the global memory (55,56) into each of the processors (10). the processors (10) calculate output values for a set of nodes of the next higher-ordered layer of the neural network. said newly-calculated output values are broadcast from each processor (10) to the global memory (55,56) and to all the other processors (10), which use the output values as a head start in calculating a new set of output values corresponding to the next layer of the neural network.",1993-04-20,"The title of the patent is method of implementing a neural network on a digital computer and its abstract is a digital computer architecture specifically tailored for implementing a neural network. several simultaneously operable processors (10) each have their own local memory (17) for storing weight and connectivity information corresponding to nodes of the neural network whose output values will be calculated by said processor (10). a global memory (55,56) is coupled to each of the processors (10) via a common data bus (30). output values corresponding to a first layer of the neural network are broadcast from the global memory (55,56) into each of the processors (10). the processors (10) calculate output values for a set of nodes of the next higher-ordered layer of the neural network. said newly-calculated output values are broadcast from each processor (10) to the global memory (55,56) and to all the other processors (10), which use the output values as a head start in calculating a new set of output values corresponding to the next layer of the neural network. dated 1993-04-20"
5208900,digital neural network computation ring,"an artificial neural network is provided using a digital architecture having feedforward and feedback processors interconnected with a digital computation ring or data bus to handle complex neural feedback arrangements. the feedforward processor receives a sequence of digital input signals and multiplies each by a weight in a predetermined manner and stores the results in an accumulator. the accumulated values may be shifted around the computation ring and read from a tap point thereof, or reprocessed through the feedback processor with predetermined scaling factors and combined with the feedforward outcomes for providing various types neural network feedback computations. alternately, the feedforward outcomes may be placed sequentially on a data bus for feedback processing through the network. the digital architecture includes a predetermined number of data input terminals for the digital input signal irrespective of the number of synapses per neuron and the number of neurons per neural network, and allows the synapses to share a common multiplier and thereby reduce the physical area of the neural network. a learning circuit may be utilized in the feedforward processor for real-time updating the weights thereof to reflect changes in the environement.",1993-05-04,"The title of the patent is digital neural network computation ring and its abstract is an artificial neural network is provided using a digital architecture having feedforward and feedback processors interconnected with a digital computation ring or data bus to handle complex neural feedback arrangements. the feedforward processor receives a sequence of digital input signals and multiplies each by a weight in a predetermined manner and stores the results in an accumulator. the accumulated values may be shifted around the computation ring and read from a tap point thereof, or reprocessed through the feedback processor with predetermined scaling factors and combined with the feedforward outcomes for providing various types neural network feedback computations. alternately, the feedforward outcomes may be placed sequentially on a data bus for feedback processing through the network. the digital architecture includes a predetermined number of data input terminals for the digital input signal irrespective of the number of synapses per neuron and the number of neurons per neural network, and allows the synapses to share a common multiplier and thereby reduce the physical area of the neural network. a learning circuit may be utilized in the feedforward processor for real-time updating the weights thereof to reflect changes in the environement. dated 1993-05-04"
5210798,vector neural network for low signal-to-noise ratio detection of a target,"a vector neural network (vnn) of interconnected neurons is provided in transition mappings of potential targets wherein the threshold (energy) of a single frame does not provide adequate information (energy) to declare a target position. the vnn enhances the signal-to-noise ratio (snr) by integrating target energy over multiple frames including the steps of postulating massive numbers of target tracks (the hypotheses), propagating these target tracks over multiple frames, and accommodating different velocity target by pixel quantization. the vnn then defers thresholding to subsequent target stages when higher snr's are prevalent so that the loss of target information is minimized, and the vnn can declare both target location and velocity. the vnn can further include target maneuver detection by a process of energy balancing hypotheses.",1993-05-11,"The title of the patent is vector neural network for low signal-to-noise ratio detection of a target and its abstract is a vector neural network (vnn) of interconnected neurons is provided in transition mappings of potential targets wherein the threshold (energy) of a single frame does not provide adequate information (energy) to declare a target position. the vnn enhances the signal-to-noise ratio (snr) by integrating target energy over multiple frames including the steps of postulating massive numbers of target tracks (the hypotheses), propagating these target tracks over multiple frames, and accommodating different velocity target by pixel quantization. the vnn then defers thresholding to subsequent target stages when higher snr's are prevalent so that the loss of target information is minimized, and the vnn can declare both target location and velocity. the vnn can further include target maneuver detection by a process of energy balancing hypotheses. dated 1993-05-11"
5212741,preprocessing of dot-matrix/ink-jet printed text for optical character recognition,"method and apparatus are disclosed for processing image data of dot-matrix/ink-jet printed text to perform optical character recognition (ocr) of such image data. in the method and apparatus, the image data is viewed for detecting if dot-matrix/ink-jet printed text is present. any detected dot-matrix/ink-jet produced text is then pre-processed by determining the image characteristic thereof by forming a histogram of pixel density values in the image data. a 2-d spatial averaging operation as a second pre-processing step smooths the dots of the characters into strokes and reduces the dynamic range of the image data. the resultant spatially averaged image data is then contrast stretched in a third pre-processing step to darken dark regions of the image data and lighten light regions of the image data. edge enhancement is then applied to the contrast stretched image data in a fourth pre-processing step to bring out higher frequency line details. the edge enhanced image data is then binarized and applied to a dot-matrix/ink jet neural network classifier for recognizing characters in the binarized image data from a predetermined set of symbols prior to ocr.",1993-05-18,"The title of the patent is preprocessing of dot-matrix/ink-jet printed text for optical character recognition and its abstract is method and apparatus are disclosed for processing image data of dot-matrix/ink-jet printed text to perform optical character recognition (ocr) of such image data. in the method and apparatus, the image data is viewed for detecting if dot-matrix/ink-jet printed text is present. any detected dot-matrix/ink-jet produced text is then pre-processed by determining the image characteristic thereof by forming a histogram of pixel density values in the image data. a 2-d spatial averaging operation as a second pre-processing step smooths the dots of the characters into strokes and reduces the dynamic range of the image data. the resultant spatially averaged image data is then contrast stretched in a third pre-processing step to darken dark regions of the image data and lighten light regions of the image data. edge enhancement is then applied to the contrast stretched image data in a fourth pre-processing step to bring out higher frequency line details. the edge enhanced image data is then binarized and applied to a dot-matrix/ink jet neural network classifier for recognizing characters in the binarized image data from a predetermined set of symbols prior to ocr. dated 1993-05-18"
5212765,on-line training neural network system for process control,"an on-line training neural network for process control system and method trains by retrieving training sets from the stream of process data. the neural network detects the availability of new training data, and constructs a training set by retrieving the corresponding input data. the neural network is trained using the training set. over time, many training sets are presented to the neural network. when multiple presentations are needed to effectively train, a buffer of training sets is filled and updated as new training data becomes available. the size of the buffer is selected in accordance with the training needs of the neural network. once the buffer is full, a new training set bumps the oldest training set off the top of the buffer stack. the training sets in the buffer stack can be presented one or more times each time a new training set is constructed. an historical database of timestamped data can be used to construct training sets when training input data has a time delay from sample time to availability for the neural network. the timestamps of the training input data are used to select the appropriate timestamp at which input data is retrieved for use in the training set. using the historical database, the neural network can be trained retrospectively by searching the historical database and constructing training sets based on past data.",1993-05-18,"The title of the patent is on-line training neural network system for process control and its abstract is an on-line training neural network for process control system and method trains by retrieving training sets from the stream of process data. the neural network detects the availability of new training data, and constructs a training set by retrieving the corresponding input data. the neural network is trained using the training set. over time, many training sets are presented to the neural network. when multiple presentations are needed to effectively train, a buffer of training sets is filled and updated as new training data becomes available. the size of the buffer is selected in accordance with the training needs of the neural network. once the buffer is full, a new training set bumps the oldest training set off the top of the buffer stack. the training sets in the buffer stack can be presented one or more times each time a new training set is constructed. an historical database of timestamped data can be used to construct training sets when training input data has a time delay from sample time to availability for the neural network. the timestamps of the training input data are used to select the appropriate timestamp at which input data is retrieved for use in the training set. using the historical database, the neural network can be trained retrospectively by searching the historical database and constructing training sets based on past data. dated 1993-05-18"
5212766,neural network representing apparatus having self-organizing function,a neutral network representing apparatus includes a plurality of neuron expressing units and a plurality of synapse load expressing units. each of the synapse load expressing units couples two neuron expressing units through a synapse load which is specific thereto. the synapse load of the synapse load expressing unit is adjusted in accordance with a prescribed learning rule in learning of the neural network representing apparatus. this learning rule includes a learning coefficient which defines the amount of a synapse load to be changed in a single learning cycle. this learning coefficient is set according to a spatial or physical distance between two neurons expressed by two neuron expressing units which are coupled by a synapse load expressing unit. the learning coefficient is provided by a monotone decreasing function of the distance between the two neurons.,1993-05-18,The title of the patent is neural network representing apparatus having self-organizing function and its abstract is a neutral network representing apparatus includes a plurality of neuron expressing units and a plurality of synapse load expressing units. each of the synapse load expressing units couples two neuron expressing units through a synapse load which is specific thereto. the synapse load of the synapse load expressing unit is adjusted in accordance with a prescribed learning rule in learning of the neural network representing apparatus. this learning rule includes a learning coefficient which defines the amount of a synapse load to be changed in a single learning cycle. this learning coefficient is set according to a spatial or physical distance between two neurons expressed by two neuron expressing units which are coupled by a synapse load expressing unit. the learning coefficient is provided by a monotone decreasing function of the distance between the two neurons. dated 1993-05-18
5212767,multi-layer network and learning method therefor,"a multi-layer neural network comprising an input layer, a hidden layer and an output layer and a learning method for such a network are disclosed. a processor belonging to the hidden layer stores both the factors of multiplication or weights of link for a successive layer nearer to the input layer and the factors of multiplication or weights of link for a preceding layer nearer to the output layer. namely, the weight for a certain connection is doubly stored in processors which are at opposite ends of that connection. upon forward calculation, the access to the weights for the successive layer among the weights stored in the processors of the hidden layer can be made by the processors independently from each other. similarly, upon backward calculation, the access to weights for the preceding layer can be made by the processors independently from each other.",1993-05-18,"The title of the patent is multi-layer network and learning method therefor and its abstract is a multi-layer neural network comprising an input layer, a hidden layer and an output layer and a learning method for such a network are disclosed. a processor belonging to the hidden layer stores both the factors of multiplication or weights of link for a successive layer nearer to the input layer and the factors of multiplication or weights of link for a preceding layer nearer to the output layer. namely, the weight for a certain connection is doubly stored in processors which are at opposite ends of that connection. upon forward calculation, the access to the weights for the successive layer among the weights stored in the processors of the hidden layer can be made by the processors independently from each other. similarly, upon backward calculation, the access to weights for the preceding layer can be made by the processors independently from each other. dated 1993-05-18"
5214715,predictive self-organizing neural network,"an a pattern recognition subsystem responds to an a feature representation input to select a-category-representation and predict a b-category-representation and its associated b feature representation input. during learning trials, a predicted b-category-representation is compared to that obtained through a b pattern recognition subsystem. with mismatch, a vigilance parameter of the a-pattern-recognition subsystem is increased to cause reset of the first-category-representation selection. inputs to the pattern recognition subsystems may be preprocessed to complement code the inputs.",1993-05-25,"The title of the patent is predictive self-organizing neural network and its abstract is an a pattern recognition subsystem responds to an a feature representation input to select a-category-representation and predict a b-category-representation and its associated b feature representation input. during learning trials, a predicted b-category-representation is compared to that obtained through a b pattern recognition subsystem. with mismatch, a vigilance parameter of the a-pattern-recognition subsystem is increased to cause reset of the first-category-representation selection. inputs to the pattern recognition subsystems may be preprocessed to complement code the inputs. dated 1993-05-25"
5214744,method and apparatus for automatically identifying targets in sonar images,"a method and apparatus for automatically identifying targets in sonar images utilizes three processing systems which preferably operate simultaneously. after the image has been filtered and fourier transformed, a highlight-shadow detector classifies portions of the image as a highlight, a shadow or background according to greyness levels of pixels in such portions. a statistical cuer selects those portions which have been classified as a highlight or a shadow. a neural network then classifies the sets of highlight and shadow reports as targets or background.",1993-05-25,"The title of the patent is method and apparatus for automatically identifying targets in sonar images and its abstract is a method and apparatus for automatically identifying targets in sonar images utilizes three processing systems which preferably operate simultaneously. after the image has been filtered and fourier transformed, a highlight-shadow detector classifies portions of the image as a highlight, a shadow or background according to greyness levels of pixels in such portions. a statistical cuer selects those portions which have been classified as a highlight or a shadow. a neural network then classifies the sets of highlight and shadow reports as targets or background. dated 1993-05-25"
5214746,method and apparatus for training a neural network using evolutionary programming,"a method and apparatus for training neural networks using evolutionary programming. a network is adjusted to operate in a weighted configuration defined by a set of weight values and a plurality of training patterns are input to the network to generate evaluations of the training patterns as network outputs. each evaluation is compared to a desired output to obtain a corresponding error. from all of the errors, an overall error value corresponding to the set of weight values is determined. the above steps are repeated with different weighted configurations to obtain a plurality of overall error values. then, for each set of weight values, a score is determined by selecting error comparison values from a predetermined variable probability distribution and comparing them to the corresponding overall error value. a predetermined number of the sets of weight values determined to have the best scores are selected and copies are made. the copies are mutated by adding random numbers to their weights and the above steps are repeated with the best sets and the mutated copies defining the weighted configurations. this procedure is repeated until the overall error values diminish to below an acceptable threshold. the random numbers added to the weight values of copies are obtained from a continuous random distribution of numbers having zero mean and variance determined such that it would be expected to converge to zero as the different sets of weight values in successive iterations converge toward sets of weight values yielding the desired neural network performance.",1993-05-25,"The title of the patent is method and apparatus for training a neural network using evolutionary programming and its abstract is a method and apparatus for training neural networks using evolutionary programming. a network is adjusted to operate in a weighted configuration defined by a set of weight values and a plurality of training patterns are input to the network to generate evaluations of the training patterns as network outputs. each evaluation is compared to a desired output to obtain a corresponding error. from all of the errors, an overall error value corresponding to the set of weight values is determined. the above steps are repeated with different weighted configurations to obtain a plurality of overall error values. then, for each set of weight values, a score is determined by selecting error comparison values from a predetermined variable probability distribution and comparing them to the corresponding overall error value. a predetermined number of the sets of weight values determined to have the best scores are selected and copies are made. the copies are mutated by adding random numbers to their weights and the above steps are repeated with the best sets and the mutated copies defining the weighted configurations. this procedure is repeated until the overall error values diminish to below an acceptable threshold. the random numbers added to the weight values of copies are obtained from a continuous random distribution of numbers having zero mean and variance determined such that it would be expected to converge to zero as the different sets of weight values in successive iterations converge toward sets of weight values yielding the desired neural network performance. dated 1993-05-25"
5214747,segmented neural network with daisy chain control,"the present invention is a direct digitally implemented network system in which neural nodes 24, 26 and 28 which output to the same destination node 22 in the network share the same channel 30. if a set of nodes does not output any data to any node to which a second set of nodes outputs data (the two sets of nodes to not overlap or intersect), the two sets of nodes are independent and do not share a channel and have separate channels 120 and 122. the network is configured as parallel operating non-intersecting segments or independent sets where each segment has a segment communication channel or bus 30. each node in the independent set or segment is sequentially activated to produce an output by a daisy chain control signal. the outputs are thereby time division multiplexed over the channel 30 to the destination node 22. the nodes are implemented on integrated circuits 158 with multiple nodes per circuit. the outputs of the nodes on the circuits in a segment are connected to the segment channel. each node includes a memory array 136 that stores the weights applied to each input via a multiplier 152. the multiplied inputs are accumulated and applied to a lookup table 132 that performs any threshold comparison operation. the output of the lookup table 134 is placed on a common bus serving as the channel for the independent set of nodes by a tristate driver 44 controlled by the daisy chain control signal.",1993-05-25,"The title of the patent is segmented neural network with daisy chain control and its abstract is the present invention is a direct digitally implemented network system in which neural nodes 24, 26 and 28 which output to the same destination node 22 in the network share the same channel 30. if a set of nodes does not output any data to any node to which a second set of nodes outputs data (the two sets of nodes to not overlap or intersect), the two sets of nodes are independent and do not share a channel and have separate channels 120 and 122. the network is configured as parallel operating non-intersecting segments or independent sets where each segment has a segment communication channel or bus 30. each node in the independent set or segment is sequentially activated to produce an output by a daisy chain control signal. the outputs are thereby time division multiplexed over the channel 30 to the destination node 22. the nodes are implemented on integrated circuits 158 with multiple nodes per circuit. the outputs of the nodes on the circuits in a segment are connected to the segment channel. each node includes a memory array 136 that stores the weights applied to each input via a multiplier 152. the multiplied inputs are accumulated and applied to a lookup table 132 that performs any threshold comparison operation. the output of the lookup table 134 is placed on a common bus serving as the channel for the independent set of nodes by a tristate driver 44 controlled by the daisy chain control signal. dated 1993-05-25"
5216463,electrophotographic process control device using a neural network to control an amount of exposure,"an electrophotographic process control device capable of controlling the supply of a toner in such a manner as to stabilize an image against changes in the characteristics of a photoconductive element and in toner density. at the learning stage of a neural network, data from sensors are applied to the input layer of the network while a latent image gamma characteristic indicative of a relation between the amount of exposure and the potential of an image area is used as learning data to be given via the output layer of the network. at a control stage, the data from the sensors are applied to the input layer of the network, as at the learning stage, and the amount of exposure is so controlled as to set up a desired potential in an image area on the basis of a latent image gamma characteristic obtainable from the output layer of the network.",1993-06-01,"The title of the patent is electrophotographic process control device using a neural network to control an amount of exposure and its abstract is an electrophotographic process control device capable of controlling the supply of a toner in such a manner as to stabilize an image against changes in the characteristics of a photoconductive element and in toner density. at the learning stage of a neural network, data from sensors are applied to the input layer of the network while a latent image gamma characteristic indicative of a relation between the amount of exposure and the potential of an image area is used as learning data to be given via the output layer of the network. at a control stage, the data from the sensors are applied to the input layer of the network, as at the learning stage, and the amount of exposure is so controlled as to set up a desired potential in an image area on the basis of a latent image gamma characteristic obtainable from the output layer of the network. dated 1993-06-01"
5216746,error absorbing system in a neuron computer,"an error absorbing system for absorbing errors through a weight correction is provided in a neuron computer for receiving an analog input signal through a first analog bus in a time divisional manner, performing a sum-of-the-products operation, and outputting an analog output signal to a second analog bus. the error absorbing system includes a dummy node for producing a fixed voltage to an analog bus in a test mode. the dummy node is connected to the analog bus of the neural network. an error measuring unit compulsorily inputs 0 volts to the first analog bus through the dummy node in a first state of a test mode and detects an offset voltage produced in an analog neuron processor through the second analog bus. a weight correcting unit, in a second state of the test mode, determines a temporary weight between the dummy node and the neuron processor. the temporary weight is multiplied by the fixed voltage produced by the dummy node, based on an offset voltage of respective neuron processors. the weight correcting unit calculates a correct weight using a gain based on the detection output voltage output from the second analog bus. a weight memory stores the weight corrected by the weight correcting unit.",1993-06-01,"The title of the patent is error absorbing system in a neuron computer and its abstract is an error absorbing system for absorbing errors through a weight correction is provided in a neuron computer for receiving an analog input signal through a first analog bus in a time divisional manner, performing a sum-of-the-products operation, and outputting an analog output signal to a second analog bus. the error absorbing system includes a dummy node for producing a fixed voltage to an analog bus in a test mode. the dummy node is connected to the analog bus of the neural network. an error measuring unit compulsorily inputs 0 volts to the first analog bus through the dummy node in a first state of a test mode and detects an offset voltage produced in an analog neuron processor through the second analog bus. a weight correcting unit, in a second state of the test mode, determines a temporary weight between the dummy node and the neuron processor. the temporary weight is multiplied by the fixed voltage produced by the dummy node, based on an offset voltage of respective neuron processors. the weight correcting unit calculates a correct weight using a gain based on the detection output voltage output from the second analog bus. a weight memory stores the weight corrected by the weight correcting unit. dated 1993-06-01"
5216750,computation system and method using hamming distance,preferred embodiments include systems with neural network processors (58) having input encoders (56) that encode integers as binary vectors so that close integers encode as close binary vectors by requiring adjacent integers have encoded binary vectors that differ in a fixed fraction of their bits.,1993-06-01,The title of the patent is computation system and method using hamming distance and its abstract is preferred embodiments include systems with neural network processors (58) having input encoders (56) that encode integers as binary vectors so that close integers encode as close binary vectors by requiring adjacent integers have encoded binary vectors that differ in a fixed fraction of their bits. dated 1993-06-01
5216751,digital processing element in an artificial neural network,"an artificial neural network is provided using a digital architecture having feedforward and feedback processors interconnected with a digital computation ring or data bus to handle complex neural feedback arrangements. the feedforward processor receives a sequence of digital input signals and multiplies each by a weight in a predetermined manner and stores the results in an accumulator. the accumulated values may be shifted around the computation ring and read from a tap point thereof, or reprocessed through the feedback processor with predetermined scaling factors and combined with the feedforward outcomes for providing various types neural network feedback computations. alternately, the feedforward outcomes may be placed sequentially on a data bus for feedback processing through the network. the digital architecture includes a predetermined number of data input terminals for the digital input signal irrespective of the number of synapses per neuron and the number of neurons per neural network, and allows the synapses to share a common multiplier and thereby reduce the physical area of the neural network. a learning circuit may be utilized in the feedforward processor for real-time updating the weights thereof to reflect changes in the environment.",1993-06-01,"The title of the patent is digital processing element in an artificial neural network and its abstract is an artificial neural network is provided using a digital architecture having feedforward and feedback processors interconnected with a digital computation ring or data bus to handle complex neural feedback arrangements. the feedforward processor receives a sequence of digital input signals and multiplies each by a weight in a predetermined manner and stores the results in an accumulator. the accumulated values may be shifted around the computation ring and read from a tap point thereof, or reprocessed through the feedback processor with predetermined scaling factors and combined with the feedforward outcomes for providing various types neural network feedback computations. alternately, the feedforward outcomes may be placed sequentially on a data bus for feedback processing through the network. the digital architecture includes a predetermined number of data input terminals for the digital input signal irrespective of the number of synapses per neuron and the number of neurons per neural network, and allows the synapses to share a common multiplier and thereby reduce the physical area of the neural network. a learning circuit may be utilized in the feedforward processor for real-time updating the weights thereof to reflect changes in the environment. dated 1993-06-01"
5216752,interspike interval decoding neural network,a multi-layered neural network is disclosed that converts an incoming temporally coded spike train into a spatially distributed topographical map from which interspike-interval and bandwidth information may be extracted. this neural network may be used to decode multiplexed pulse-coded signals embedded serially in an incoming spike train into parallel distributed topographically mapped channels. a signal processing and code conversion algorithm not requiring learning is provided.,1993-06-01,The title of the patent is interspike interval decoding neural network and its abstract is a multi-layered neural network is disclosed that converts an incoming temporally coded spike train into a spatially distributed topographical map from which interspike-interval and bandwidth information may be extracted. this neural network may be used to decode multiplexed pulse-coded signals embedded serially in an incoming spike train into parallel distributed topographically mapped channels. a signal processing and code conversion algorithm not requiring learning is provided. dated 1993-06-01
5218245,programmable neural logic device,"a programmable logic cell, compatible with lssd (level sensitive scan design) technique, is described whose internal logic function can be initially loaded from an eprom or external processor. the output or contents of one cell can be connected to another cell to alter the logic operation of the second cell even while this second cell is in operation. the cells can be connected together to form a neural network.",1993-06-08,"The title of the patent is programmable neural logic device and its abstract is a programmable logic cell, compatible with lssd (level sensitive scan design) technique, is described whose internal logic function can be initially loaded from an eprom or external processor. the output or contents of one cell can be connected to another cell to alter the logic operation of the second cell even while this second cell is in operation. the cells can be connected together to form a neural network. dated 1993-06-08"
5218440,switched resistive neural network for sensor fusion,"an electronic image processing system uses data provided by one or more sensors to perform cooperative computations and improve image recognition performance. a smoothing resistive network, which may comprise an integrated circuit chip, has switching elements connected to each node. the system uses a first sensory output comprising primitives, such as discontinuities or object boundaries, detected by at least a first sensor to define a region for smoothing of a second sensory output comprising at least a second, distinct output of the first sensor or a distinct output of at least a second sensor. a bit pattern for controlling the switches is generated from the detected image discontinuities in the first sensory output. the second sensory output is applied to the resistive network for data smoothing. the switches turned off by the data from the first sensory output define regional boundaries for smoothing of the data provided by the second sensory output. smoothing operations based on this sensor fusion can proceed without spreading object characteristics beyond the object boundaries.",1993-06-08,"The title of the patent is switched resistive neural network for sensor fusion and its abstract is an electronic image processing system uses data provided by one or more sensors to perform cooperative computations and improve image recognition performance. a smoothing resistive network, which may comprise an integrated circuit chip, has switching elements connected to each node. the system uses a first sensory output comprising primitives, such as discontinuities or object boundaries, detected by at least a first sensor to define a region for smoothing of a second sensory output comprising at least a second, distinct output of the first sensor or a distinct output of at least a second sensor. a bit pattern for controlling the switches is generated from the detected image discontinuities in the first sensory output. the second sensory output is applied to the resistive network for data smoothing. the switches turned off by the data from the first sensory output define regional boundaries for smoothing of the data provided by the second sensory output. smoothing operations based on this sensor fusion can proceed without spreading object characteristics beyond the object boundaries. dated 1993-06-08"
5218529,neural network system and methods for analysis of organic materials and structures using spectral data,"apparatus and processes for recognizing and identifying materials. characteristic spectra are obtained for the materials via spectroscopy techniques including nuclear magnetic resonance spectroscopy, infrared absorption analysis, x-ray analysis, mass spectroscopy and gas chromatography. desired portions of the spectra may be selected and then placed in proper form and format for presentation to a number of input layer neurons in an offline neural network. the network is first trained according to a predetermined training process; it may then be employed to identify particular materials. such apparatus and processes are particularly useful for recognizing and identifying organic compounds such as complex carbohydrates, whose spectra conventionally require a high level of training and many hours of hard work to identify, and are frequently indistinguishable from one another by human interpretation.",1993-06-08,"The title of the patent is neural network system and methods for analysis of organic materials and structures using spectral data and its abstract is apparatus and processes for recognizing and identifying materials. characteristic spectra are obtained for the materials via spectroscopy techniques including nuclear magnetic resonance spectroscopy, infrared absorption analysis, x-ray analysis, mass spectroscopy and gas chromatography. desired portions of the spectra may be selected and then placed in proper form and format for presentation to a number of input layer neurons in an offline neural network. the network is first trained according to a predetermined training process; it may then be employed to identify particular materials. such apparatus and processes are particularly useful for recognizing and identifying organic compounds such as complex carbohydrates, whose spectra conventionally require a high level of training and many hours of hard work to identify, and are frequently indistinguishable from one another by human interpretation. dated 1993-06-08"
5218646,"classification procedure implemented in a hierarchical neural network, and hierarchical neural network","classification procedure implemented in a tree-like neural network which, in the course of learning steps, determines with the aid of a tree-like structure the number of neurons and their synaptic coefficients required for the processing of problems of classification of multi-class examples. each neuron tends to distinguish, from the examples, two groups of examples approximating as well as possible to a division into two predetermined groups of classes. this division can be obtained through a principal component analysis of the distribution of examples. the neural network comprises a directory of addresses of successor neurons which is loaded in learning mode then read in exploitation mode. a memory stores example classes associated with the ends of the branches of the tree.",1993-06-08,"The title of the patent is classification procedure implemented in a hierarchical neural network, and hierarchical neural network and its abstract is classification procedure implemented in a tree-like neural network which, in the course of learning steps, determines with the aid of a tree-like structure the number of neurons and their synaptic coefficients required for the processing of problems of classification of multi-class examples. each neuron tends to distinguish, from the examples, two groups of examples approximating as well as possible to a division into two predetermined groups of classes. this division can be obtained through a principal component analysis of the distribution of examples. the neural network comprises a directory of addresses of successor neurons which is loaded in learning mode then read in exploitation mode. a memory stores example classes associated with the ends of the branches of the tree. dated 1993-06-08"
5220202,memory device and memory apparatus using the same suitable for neural network,"a memory device includes a nonlinear electric conductivity element, a charge accumulation element, and a switching element. the nonlinear electric conductivity element has an insulating layer having opposite surfaces, and first and second conductive layers respectively formed on the opposite surfaces of the insulating layer. the nonlinear electric conductivity element receives an external write signal applied to one of the first and second conductive layers, and outputs a signal having nonlinear electric conductivity characteristics from the other of the first and second conductive layers. the charge accumulation element has charge accumulation characteristics and is connected to receive and store the signal output from the other of the first and second conductive layers. the switching element is on/off-controlled upon reception of the signal charge stored in the charge accumulation element. the switching element receives an external read voltage to read out the signal charge stored in the charge accumulation element as storage data. a memory apparatus includes a plurality of memory devices each having the nonlinear electric conductivity element the charge accumulation element and the switching element. the plurality of memory devices are connected in a matrix form such that the switching elements in at least two memory devices can commonly receive the read voltage and can commonly read out the storage data.",1993-06-15,"The title of the patent is memory device and memory apparatus using the same suitable for neural network and its abstract is a memory device includes a nonlinear electric conductivity element, a charge accumulation element, and a switching element. the nonlinear electric conductivity element has an insulating layer having opposite surfaces, and first and second conductive layers respectively formed on the opposite surfaces of the insulating layer. the nonlinear electric conductivity element receives an external write signal applied to one of the first and second conductive layers, and outputs a signal having nonlinear electric conductivity characteristics from the other of the first and second conductive layers. the charge accumulation element has charge accumulation characteristics and is connected to receive and store the signal output from the other of the first and second conductive layers. the switching element is on/off-controlled upon reception of the signal charge stored in the charge accumulation element. the switching element receives an external read voltage to read out the signal charge stored in the charge accumulation element as storage data. a memory apparatus includes a plurality of memory devices each having the nonlinear electric conductivity element the charge accumulation element and the switching element. the plurality of memory devices are connected in a matrix form such that the switching elements in at least two memory devices can commonly receive the read voltage and can commonly read out the storage data. dated 1993-06-15"
5220373,electrophotographic process control device using a neural network for estimating states of the device,"an electrophotographic process control device for an electrophotographic image forming apparatus. a neural network is incorporated in the control device for estimating the state of the image forming unit. parameters of the kind which should not be frequently measured, e.g., the surface potential of a photoconductive element and the amount of toner deposition thereon and parameters which are not easy to measure are determined by inference so as to control each section of the apparatus in an optimum way.",1993-06-15,"The title of the patent is electrophotographic process control device using a neural network for estimating states of the device and its abstract is an electrophotographic process control device for an electrophotographic image forming apparatus. a neural network is incorporated in the control device for estimating the state of the image forming unit. parameters of the kind which should not be frequently measured, e.g., the surface potential of a photoconductive element and the amount of toner deposition thereon and parameters which are not easy to measure are determined by inference so as to control each section of the apparatus in an optimum way. dated 1993-06-15"
5220618,classification method implemented in a layered neural network for multiclass classification and layered neural network,"classification method implemented in a layered neural network, comprising learning steps during which at least one layer is constructed by the addition of the successive neurons necessary for operating, by successive dichotomies, a classification of examples distributed over classes. in order to create at least one layer starting with a group of examples distributed over more than two classes, each successive neuron tends to distinguish its input data according to two predetermined sub-groups of classes peculiar to the said neuron according to a principal components analysis of the distribution of the said input data subjected to the learning of the neuron of the layer in question.",1993-06-15,"The title of the patent is classification method implemented in a layered neural network for multiclass classification and layered neural network and its abstract is classification method implemented in a layered neural network, comprising learning steps during which at least one layer is constructed by the addition of the successive neurons necessary for operating, by successive dichotomies, a classification of examples distributed over classes. in order to create at least one layer starting with a group of examples distributed over more than two classes, each successive neuron tends to distinguish its input data according to two predetermined sub-groups of classes peculiar to the said neuron according to a principal components analysis of the distribution of the said input data subjected to the learning of the neuron of the layer in question. dated 1993-06-15"
5220640,neural net architecture for rate-varying inputs,"a neural net architecture provides for the recognition of an input signal which is a rate variant of a learned signal pattern, reducing the neural net training requirements. the duration of a digital sampling of the input signal is scaled by a time-scaling network, creating a multiplicity of scaled signals which are then compared to memorized signal patterns contained in a self-organizing feature map. the feature map outputs values which indicate how well the scaled input signals match various learned signal patterns. a comparator determines which one of the values is greatest, thus indicating a best match between the input signal and one of the learned signal patterns.",1993-06-15,"The title of the patent is neural net architecture for rate-varying inputs and its abstract is a neural net architecture provides for the recognition of an input signal which is a rate variant of a learned signal pattern, reducing the neural net training requirements. the duration of a digital sampling of the input signal is scaled by a time-scaling network, creating a multiplicity of scaled signals which are then compared to memorized signal patterns contained in a self-organizing feature map. the feature map outputs values which indicate how well the scaled input signals match various learned signal patterns. a comparator determines which one of the values is greatest, thus indicating a best match between the input signal and one of the learned signal patterns. dated 1993-06-15"
5220643,monolithic neural network element,"a neural plane, which can form the basis of a neural network or a component thereof, is comprised by an optical modulator, an electrical non-linearity circuit and an optical detector interconnected whereby in use the non-linearity circuit controls the modulator in dependence on the detector output. there are parallel arrays (10, 11, 12) of such modulators, non-linearity circuits and detectors (m, t, d, 30, 33, 34). the modulator, non-linearity circuits and detectors have components formed in a common semiconductor substrate (20), for example by vlsi techniques with a silicon substrate, the modulators (30) may be comprised by liquid crystal on silicon in that case (figs. 4, 7).",1993-06-15,"The title of the patent is monolithic neural network element and its abstract is a neural plane, which can form the basis of a neural network or a component thereof, is comprised by an optical modulator, an electrical non-linearity circuit and an optical detector interconnected whereby in use the non-linearity circuit controls the modulator in dependence on the detector output. there are parallel arrays (10, 11, 12) of such modulators, non-linearity circuits and detectors (m, t, d, 30, 33, 34). the modulator, non-linearity circuits and detectors have components formed in a common semiconductor substrate (20), for example by vlsi techniques with a silicon substrate, the modulators (30) may be comprised by liquid crystal on silicon in that case (figs. 4, 7). dated 1993-06-15"
5220644,optical neural network system,"an optical system of an optical neural network model for parallel data processing is disclosed. taking advantage of the fact that an auto-correlation matrix is symmetric with respect to a main diagonal and the weights for modulating the values of diagonals of the auto-correlation matrix are equal to each other, the configuration of an optical modulation unit is simplified by a one-dimensional modulation array on the one hand, and both positive and negative weights are capable of being computed at the same time on the other hand. in particular, the optical system makes up a second-order neural network exhibiting invariant characteristics against the translation and scale.",1993-06-15,"The title of the patent is optical neural network system and its abstract is an optical system of an optical neural network model for parallel data processing is disclosed. taking advantage of the fact that an auto-correlation matrix is symmetric with respect to a main diagonal and the weights for modulating the values of diagonals of the auto-correlation matrix are equal to each other, the configuration of an optical modulation unit is simplified by a one-dimensional modulation array on the one hand, and both positive and negative weights are capable of being computed at the same time on the other hand. in particular, the optical system makes up a second-order neural network exhibiting invariant characteristics against the translation and scale. dated 1993-06-15"
5222194,neural network with modification of neuron weights and reaction coefficient,"a neural network computation apparatus having a plurality of layers, each of the plurality of layers has at least an input layer and an output layer, each layer having a plurality of units, a plurality of links, each of the plurality of links connecting units on the plurality of layers, and changing means for changing input and output characteristics of a particular unit of the plurality of units and/or the weight of a particular link of the plurality of links in accordance with an output of the output layer after learning an example and with a particular rule. after the neural network computation apparatus learns an example, the changing means changes input and output characteristics of units and weights of links in accordance with outputs of the output layer and a particular rule. thus, a mutual operation between a logical knowledge and a pattern recognizing performance can be accomplished and thereby a determination close to that of a specialist can be accomplished. in other words, a proper determination in accordance with an experience can be made so as to deal with unknown patterns with high flexibility.",1993-06-22,"The title of the patent is neural network with modification of neuron weights and reaction coefficient and its abstract is a neural network computation apparatus having a plurality of layers, each of the plurality of layers has at least an input layer and an output layer, each layer having a plurality of units, a plurality of links, each of the plurality of links connecting units on the plurality of layers, and changing means for changing input and output characteristics of a particular unit of the plurality of units and/or the weight of a particular link of the plurality of links in accordance with an output of the output layer after learning an example and with a particular rule. after the neural network computation apparatus learns an example, the changing means changes input and output characteristics of units and weights of links in accordance with outputs of the output layer and a particular rule. thus, a mutual operation between a logical knowledge and a pattern recognizing performance can be accomplished and thereby a determination close to that of a specialist can be accomplished. in other words, a proper determination in accordance with an experience can be made so as to deal with unknown patterns with high flexibility. dated 1993-06-22"
5222195,dynamically stable associative learning neural system with one fixed weight,"a dynamically stable associative learning neural network system include a plurality of synapses (122,22-28), a non-linear function circuit (30) and an adaptive weight circuit (150) for adjusting the weight of each synapse based upon the present signal and the prior history of signals applied to the input of the particular synapse and the present signal and the prior history of signals applied to the input of a predetermined set of other collateral synapses. a flow-through neuron circuit (1110) embodiment includes a flow-through synapse (122) having a predetermined fixed weight. a neural network is formed by a set of flow-through neuron circuits connected by flow-through synapses to form separate paths between each input (215) and a corresponding output (245). in one embodiment (200), the neuron network is initialized by setting the adjustable synapses at some value near the minimum weight and setting the flow-through neuron circuits at some arbitrarily high weight. the neural network embodiments are taught by successively application of sets of inputs signals to the input terminals until a dynamic equilibrium is reached.",1993-06-22,"The title of the patent is dynamically stable associative learning neural system with one fixed weight and its abstract is a dynamically stable associative learning neural network system include a plurality of synapses (122,22-28), a non-linear function circuit (30) and an adaptive weight circuit (150) for adjusting the weight of each synapse based upon the present signal and the prior history of signals applied to the input of the particular synapse and the present signal and the prior history of signals applied to the input of a predetermined set of other collateral synapses. a flow-through neuron circuit (1110) embodiment includes a flow-through synapse (122) having a predetermined fixed weight. a neural network is formed by a set of flow-through neuron circuits connected by flow-through synapses to form separate paths between each input (215) and a corresponding output (245). in one embodiment (200), the neuron network is initialized by setting the adjustable synapses at some value near the minimum weight and setting the flow-through neuron circuits at some arbitrarily high weight. the neural network embodiments are taught by successively application of sets of inputs signals to the input terminals until a dynamic equilibrium is reached. dated 1993-06-22"
5222196,neural network shell for application programs,"a neural network shell has a defined interface to an application program. by interfacing with the neural network shell, any application program becomes a neural network application program. the neural network shell contains a set of utility programs that transfers data into and out of a neural network data structure. this set of utility programs allows an application program to define a new neural network model, create a neural network data structure, train a neural network, and run a neural network. once trained, the neural network data structure can be transported to other computer systems or to application programs written in different computing languages running on similar or different computer systems.",1993-06-22,"The title of the patent is neural network shell for application programs and its abstract is a neural network shell has a defined interface to an application program. by interfacing with the neural network shell, any application program becomes a neural network application program. the neural network shell contains a set of utility programs that transfers data into and out of a neural network data structure. this set of utility programs allows an application program to define a new neural network model, create a neural network data structure, train a neural network, and run a neural network. once trained, the neural network data structure can be transported to other computer systems or to application programs written in different computing languages running on similar or different computer systems. dated 1993-06-22"
5222210,method of displaying the state of an artificial neural network,"a computer simulator is provided for displaying the state of an artificial neural network in a simplified yet meaningful manner on a computer display terminal. the user may enter commands to select one or more areas of interest within the neural network for further information regarding its state of learning and operation. one display mode illustrates the output activity of each neuron as representatively sized and shaded boxes within the border of the neuron, while another display mode shows the connectivity as weighted synapses between a user-selected neuron and the remaining neurons of the network in a similar manner. a third display mode provides a tuning curve wherein the synapses associated with each of the neurons are represented within the borders of the same. both grid block and line graph type characterization are supported. the methodology allows large neural networks on the order of thousands of neurons to be displayed in a meaningful manner.",1993-06-22,"The title of the patent is method of displaying the state of an artificial neural network and its abstract is a computer simulator is provided for displaying the state of an artificial neural network in a simplified yet meaningful manner on a computer display terminal. the user may enter commands to select one or more areas of interest within the neural network for further information regarding its state of learning and operation. one display mode illustrates the output activity of each neuron as representatively sized and shaded boxes within the border of the neuron, while another display mode shows the connectivity as weighted synapses between a user-selected neuron and the remaining neurons of the network in a similar manner. a third display mode provides a tuning curve wherein the synapses associated with each of the neurons are represented within the borders of the same. both grid block and line graph type characterization are supported. the methodology allows large neural networks on the order of thousands of neurons to be displayed in a meaningful manner. dated 1993-06-22"
5224203,on-line process control neural network using data pointers,"an on-line process control neural network using data pointers allows the neural network to be easily configured to use data in a process control environment. the inputs, outputs, training inputs and errors can be retrieved and/or stored from any available data source without programming. the user of the neural network specifies data pointers indicating the particular computer system in which the data resides or will be stored; the type of data to be retrieved and/or stored; and the specific data value or storage location to be used. the data pointers include maximum, minimum, and maximum change limits, which can also serve as scaling limits for the neural network. data pointers indicating time-dependent data, such as time averages, also include time boundary specifiers. the data pointers are entered by the user of the neural network using pop-up menus and by completing fields in a template. an historical database provides both a source of input data and a storage function for output and error data.",1993-06-29,"The title of the patent is on-line process control neural network using data pointers and its abstract is an on-line process control neural network using data pointers allows the neural network to be easily configured to use data in a process control environment. the inputs, outputs, training inputs and errors can be retrieved and/or stored from any available data source without programming. the user of the neural network specifies data pointers indicating the particular computer system in which the data resides or will be stored; the type of data to be retrieved and/or stored; and the specific data value or storage location to be used. the data pointers include maximum, minimum, and maximum change limits, which can also serve as scaling limits for the neural network. data pointers indicating time-dependent data, such as time averages, also include time boundary specifiers. the data pointers are entered by the user of the neural network using pop-up menus and by completing fields in a template. an historical database provides both a source of input data and a storage function for output and error data. dated 1993-06-29"
5226092,method and apparatus for learning in a neural network,""" a method and apparatus for speeding and enhancing the """"learning"""" function of a computer configured as a multilayered, feed format artificial neural network using logistic functions as an activation function. the enhanced learning method provides a linear probing method for determining local minima values computed first along the gradient of the weight space and then adjusting the slope and direction of a linear probe line after determining the likelihood that a """"ravine"""" has been encountered in the terrain of the weight space. """,1993-07-06,"The title of the patent is method and apparatus for learning in a neural network and its abstract is "" a method and apparatus for speeding and enhancing the """"learning"""" function of a computer configured as a multilayered, feed format artificial neural network using logistic functions as an activation function. the enhanced learning method provides a linear probing method for determining local minima values computed first along the gradient of the weight space and then adjusting the slope and direction of a linear probe line after determining the likelihood that a """"ravine"""" has been encountered in the terrain of the weight space. "" dated 1993-07-06"
5227830,automatic camera,"in a focus detection device for a camera, a plurality of distance sensors each for detecting a distance to an object in a plurality of areas of a photographing image plane are provided and distance data obtained by the distance sensors for each object in each area is supplied to a main object detection circuit and a normalizing circuit. the normalizing circuit normalizes the distance data into a real number ranging from 0 to 1 and then supplies the same to a neural network. the neural network formed of a single-layered neuron units of which the synapse connection weighting factors are previously obtained by the learning process, calculates a vector difference between the distance data and the synapse connection weighting factors of each neuron unit, detects the minimum vector difference, and outputs position data of a main object corresponding to a neuron unit which gives the minimum vector difference. the position data of the main object is input to a main object detection circuit. one of the outputs from the distance sensors corresponding to the main object is selected by the main object detection circuit and is supplied to a focus detection circuit for effecting the calculation to detect the focus. an output of the focus detection circuit is supplied to a lens driving mechanism so as to adjust the focus.",1993-07-13,"The title of the patent is automatic camera and its abstract is in a focus detection device for a camera, a plurality of distance sensors each for detecting a distance to an object in a plurality of areas of a photographing image plane are provided and distance data obtained by the distance sensors for each object in each area is supplied to a main object detection circuit and a normalizing circuit. the normalizing circuit normalizes the distance data into a real number ranging from 0 to 1 and then supplies the same to a neural network. the neural network formed of a single-layered neuron units of which the synapse connection weighting factors are previously obtained by the learning process, calculates a vector difference between the distance data and the synapse connection weighting factors of each neuron unit, detects the minimum vector difference, and outputs position data of a main object corresponding to a neuron unit which gives the minimum vector difference. the position data of the main object is input to a main object detection circuit. one of the outputs from the distance sensors corresponding to the main object is selected by the main object detection circuit and is supplied to a focus detection circuit for effecting the calculation to detect the focus. an output of the focus detection circuit is supplied to a lens driving mechanism so as to adjust the focus. dated 1993-07-13"
5227835,teachable camera,a teachable camera 8 which includes an alterable template matching neural network 40 positioned between a microprocessor 10 that performs camera picture taking algorithms and the units 24-32 such as the shutter which control the characteristics of the picture. the network 40 alters the output of the algorithms to match the picture characteristics desired by the photographer. the network 40 is altered by a rule based expert system executing in a personal computer 70 which determines how to alter the matching template of the network 40.,1993-07-13,The title of the patent is teachable camera and its abstract is a teachable camera 8 which includes an alterable template matching neural network 40 positioned between a microprocessor 10 that performs camera picture taking algorithms and the units 24-32 such as the shutter which control the characteristics of the picture. the network 40 alters the output of the algorithms to match the picture characteristics desired by the photographer. the network 40 is altered by a rule based expert system executing in a personal computer 70 which determines how to alter the matching template of the network 40. dated 1993-07-13
5228113,accelerated training apparatus for back propagation networks,a supervised procedure for obtaining weight values for back-propagation neural networks is described. the method according to the invention performs a sequence of partial optimizations in order to determine values for the network connection weights. the partial optimization depends on a constrained representation of hidden weights derived from a singular value decomposition of the input space as well as an iterative least squares optimization solution for the output weights.,1993-07-13,The title of the patent is accelerated training apparatus for back propagation networks and its abstract is a supervised procedure for obtaining weight values for back-propagation neural networks is described. the method according to the invention performs a sequence of partial optimizations in order to determine values for the network connection weights. the partial optimization depends on a constrained representation of hidden weights derived from a singular value decomposition of the input space as well as an iterative least squares optimization solution for the output weights. dated 1993-07-13
5229623,"electric circuit using multiple differential negative resistance elements, semiconductor device and neuro chip using the same","a semiconductor device is disclosed, which includes a multiple negative differential resistance element having negative differential resistance characteristics at at least two places in the current-voltage characteristics, and which is suitable for constructing a neural network having a high density integration and a high reliability.",1993-07-20,"The title of the patent is electric circuit using multiple differential negative resistance elements, semiconductor device and neuro chip using the same and its abstract is a semiconductor device is disclosed, which includes a multiple negative differential resistance element having negative differential resistance characteristics at at least two places in the current-voltage characteristics, and which is suitable for constructing a neural network having a high density integration and a high reliability. dated 1993-07-20"
5235339,radar target discrimination systems using artificial neural network topology,"a system for distinguishing between a target and clutter analyzes frequency components of returned wave energy by one or more networks each having inputs receiving successive samples of the returned energy and having outputs individually connected to the inputs through multiplier elements providing selectable factors. the multipliers corresponding to each output are connected to the output through a summing element and a selectable and generally sigmoidal activation function. the factors may be bandpass filter coefficients or discrete fourier transform coefficients so as to generate frequency components of the energy. predetermined frequency characteristics of the returned energy may be detected by providing the outputs of a network to a network in which the factors are selected as correlation or convolution coefficients, are selected to integrate fed back outputs, or are selected to sum several outputs within a predetermined range. the activation functions may be selected for thresholding, linearity, limiting, or generation of logarithms.",1993-08-10,"The title of the patent is radar target discrimination systems using artificial neural network topology and its abstract is a system for distinguishing between a target and clutter analyzes frequency components of returned wave energy by one or more networks each having inputs receiving successive samples of the returned energy and having outputs individually connected to the inputs through multiplier elements providing selectable factors. the multipliers corresponding to each output are connected to the output through a summing element and a selectable and generally sigmoidal activation function. the factors may be bandpass filter coefficients or discrete fourier transform coefficients so as to generate frequency components of the energy. predetermined frequency characteristics of the returned energy may be detected by providing the outputs of a network to a network in which the factors are selected as correlation or convolution coefficients, are selected to integrate fed back outputs, or are selected to sum several outputs within a predetermined range. the activation functions may be selected for thresholding, linearity, limiting, or generation of logarithms. dated 1993-08-10"
5235440,"optical interconnector and highly interconnected, learning neural network incorporating optical interconnector therein","a variable weight optical interconnector is disclosed to include a projecting device and an interconnection weighting device remote from the projecting device. the projecting device projects a distribution of interconnecting light beams when illuminated by a spatially-modulated light pattern. the weighting device includes a photosensitive screen provided in optical alignment with the projecting device to independently control the intensity of each projected interconnecting beam to thereby assign an interconnection weight to each such beam. further in accordance with the present invention, a highly-interconnected optical neural network having learning capability is disclosed as including a spatial light modulator, a detecting device, an interconnector according to the present invention, and a device responsive to detection signals generated by the detecting device to modify the interconnection weights assigned by the photosensitive screen of the interconnector.",1993-08-10,"The title of the patent is optical interconnector and highly interconnected, learning neural network incorporating optical interconnector therein and its abstract is a variable weight optical interconnector is disclosed to include a projecting device and an interconnection weighting device remote from the projecting device. the projecting device projects a distribution of interconnecting light beams when illuminated by a spatially-modulated light pattern. the weighting device includes a photosensitive screen provided in optical alignment with the projecting device to independently control the intensity of each projected interconnecting beam to thereby assign an interconnection weight to each such beam. further in accordance with the present invention, a highly-interconnected optical neural network having learning capability is disclosed as including a spatial light modulator, a detecting device, an interconnector according to the present invention, and a device responsive to detection signals generated by the detecting device to modify the interconnection weights assigned by the photosensitive screen of the interconnector. dated 1993-08-10"
5235650,pattern classifier for character recognition,a pattern classifier for character recognition is constructed in accordance with a neural network model. the pattern classifier comprises (2n+1).times.(2n+1) input buffer amplifiers and m output buffer amplifiers. the input buffer amplifiers have an inverted output line and a non-inverted output line which intersect input lines to the output buffers. synapses are selectively arranged at the intersections of the output and input lines in accordance with predetermined mask patterns used in character recognition. pmos and nmos transistors are employed for the synapses.,1993-08-10,The title of the patent is pattern classifier for character recognition and its abstract is a pattern classifier for character recognition is constructed in accordance with a neural network model. the pattern classifier comprises (2n+1).times.(2n+1) input buffer amplifiers and m output buffer amplifiers. the input buffer amplifiers have an inverted output line and a non-inverted output line which intersect input lines to the output buffers. synapses are selectively arranged at the intersections of the output and input lines in accordance with predetermined mask patterns used in character recognition. pmos and nmos transistors are employed for the synapses. dated 1993-08-10
5235672,hardware for electronic neural network,"this application discloses hardware suitable for use in a neural network system. it makes use of z-technology modules, each containing densely packaged electronic circuitry. the modules provide access planes which are electrically connected to circuitry located on planar surfaces interfacing with such access planes. one such planar surface comprises a resistive feedback network. by combining two z-technology modules, whose stacked chips are in planes perpendicular to one another, and using switching networks between the two modules, the system provides bidirectional accessibility of each individual electronic element in the neural network to most or all of the other individual electronic elements in the system.",1993-08-10,"The title of the patent is hardware for electronic neural network and its abstract is this application discloses hardware suitable for use in a neural network system. it makes use of z-technology modules, each containing densely packaged electronic circuitry. the modules provide access planes which are electrically connected to circuitry located on planar surfaces interfacing with such access planes. one such planar surface comprises a resistive feedback network. by combining two z-technology modules, whose stacked chips are in planes perpendicular to one another, and using switching networks between the two modules, the system provides bidirectional accessibility of each individual electronic element in the neural network to most or all of the other individual electronic elements in the system. dated 1993-08-10"
5235673,enhanced neural network shell for application programs,"an enhanced neural network shell for application programs is disclosed. the user is prompted to enter in non-technical information about the specific problem type that the user wants solved by a neural network. the user also is prompted to indicate the input data usage information to the neural network. based on this information, the neural network shell creates a neural network data structure by automatically selecting an appropriate neural network model and automatically generating an appropriate number of inputs, outputs, and/or other model-specific parameters for the selected neural network model. the user is no longer required to have expertise in neural network technology to create a neural network data structure.",1993-08-10,"The title of the patent is enhanced neural network shell for application programs and its abstract is an enhanced neural network shell for application programs is disclosed. the user is prompted to enter in non-technical information about the specific problem type that the user wants solved by a neural network. the user also is prompted to indicate the input data usage information to the neural network. based on this information, the neural network shell creates a neural network data structure by automatically selecting an appropriate neural network model and automatically generating an appropriate number of inputs, outputs, and/or other model-specific parameters for the selected neural network model. the user is no longer required to have expertise in neural network technology to create a neural network data structure. dated 1993-08-10"
5237210,neural network accomodating parallel synaptic weight adjustments for correlation learning algorithms,a neural network providing correlation learning in a synapse cell coupled to a circuit for parallel implementation of weight adjustment in a broad class of learning algorithms. the circuit provides the learning portion of the synaptic operation and includes a pair of floating gate devices sharing a common floating gate member that stores the connection weight of the cell. parallel weight adjustments are performed in a predetermined number of cycles utilizing a novel debiasing technique.,1993-08-17,The title of the patent is neural network accomodating parallel synaptic weight adjustments for correlation learning algorithms and its abstract is a neural network providing correlation learning in a synapse cell coupled to a circuit for parallel implementation of weight adjustment in a broad class of learning algorithms. the circuit provides the learning portion of the synaptic operation and includes a pair of floating gate devices sharing a common floating gate member that stores the connection weight of the cell. parallel weight adjustments are performed in a predetermined number of cycles utilizing a novel debiasing technique. dated 1993-08-17
5239593,optical pattern recognition using detector and locator neural networks,"a system for performing optical pattern recognition includes a first detector neural network for detecting the presence of a particular optical pattern in an input image and a second locator neural network for locating and/or removing the particular optical pattern from the input image. the detector network and the locator network both comprise nodes which can take on the -1, +1, or undefined states. the nodes are arranged in layers and each node in a layer has a location corresponding to a pixel in the input image. a particular application of this neural network is in finding the amount field on a check and removing the line which borders the amount field.",1993-08-24,"The title of the patent is optical pattern recognition using detector and locator neural networks and its abstract is a system for performing optical pattern recognition includes a first detector neural network for detecting the presence of a particular optical pattern in an input image and a second locator neural network for locating and/or removing the particular optical pattern from the input image. the detector network and the locator network both comprise nodes which can take on the -1, +1, or undefined states. the nodes are arranged in layers and each node in a layer has a location corresponding to a pixel in the input image. a particular application of this neural network is in finding the amount field on a check and removing the line which borders the amount field. dated 1993-08-24"
5239594,self-organizing pattern classification neural network system,"a self-organizing pattern classification neural network system includes means for receiving incoming pattern of signals that were processed by feature extractors that extract feature vectors from the incoming signal. these feature vectors correspond to information regarding certain features of the incoming signal. the extracted feature vectors then each pass to separate self-organizing neural network classifiers. the classifiers compare the feature vectors to templates corresponding to respective classes and output the results of their comparisons. the output from the classifier for each class enter a discriminator. the discriminator generates a classification response indicating the best class for the input signal. the classification response includes information indicative of whether the classification is possible and also includes the identified best class. lastly, the system includes a learning trigger for transferring a correct glass signal to the self-organizing classifiers so that they can determine the validity of their classification results.",1993-08-24,"The title of the patent is self-organizing pattern classification neural network system and its abstract is a self-organizing pattern classification neural network system includes means for receiving incoming pattern of signals that were processed by feature extractors that extract feature vectors from the incoming signal. these feature vectors correspond to information regarding certain features of the incoming signal. the extracted feature vectors then each pass to separate self-organizing neural network classifiers. the classifiers compare the feature vectors to templates corresponding to respective classes and output the results of their comparisons. the output from the classifier for each class enter a discriminator. the discriminator generates a classification response indicating the best class for the input signal. the classification response includes information indicative of whether the classification is possible and also includes the identified best class. lastly, the system includes a learning trigger for transferring a correct glass signal to the self-organizing classifiers so that they can determine the validity of their classification results. dated 1993-08-24"
5239597,nearest neighbor dither image processing circuit,"a conversion circuit of binary dither image to multilevel image comprises a counter utilizing concepts of a neural network, an 8 bit register and 8 or gates, resulting in high speed of operation. the counter uses a neural network based on the hopfield model and is made up of an input synapse group, a first bias synapse group, a feedback synapse group, a second bias synapse group, a neuron group and an invertor group.",1993-08-24,"The title of the patent is nearest neighbor dither image processing circuit and its abstract is a conversion circuit of binary dither image to multilevel image comprises a counter utilizing concepts of a neural network, an 8 bit register and 8 or gates, resulting in high speed of operation. the counter uses a neural network based on the hopfield model and is made up of an input synapse group, a first bias synapse group, a feedback synapse group, a second bias synapse group, a neuron group and an invertor group. dated 1993-08-24"
5239618,data processing device with network structure and its learning processing method,"an output layer in a layered neural network uses a linear function or a designated region (linear region) of a threshold function instead of the threshold function to convert an input signal to an analog output signal when the basic unit uses the linear function, a limiter for limiting the output to a region between 1.0 and 0. when the basic unit uses the designated linear region of the threshold function, a limiter limits the output to a region between 0.8 and 0.2. upon a learning operation, the error propagation coefficient is determined as a constant value such as 1/6 and when the majority of the desired values are 1 or near 1, an error value regarding the opposite desired value 0 is amplified, and when the output values become equal to or more than 1, it is deemed that there is no error with regard to the output of more than 1 in case of many outputs 1, thereby speeding up an operation of updating the weight.",1993-08-24,"The title of the patent is data processing device with network structure and its learning processing method and its abstract is an output layer in a layered neural network uses a linear function or a designated region (linear region) of a threshold function instead of the threshold function to convert an input signal to an analog output signal when the basic unit uses the linear function, a limiter for limiting the output to a region between 1.0 and 0. when the basic unit uses the designated linear region of the threshold function, a limiter limits the output to a region between 0.8 and 0.2. upon a learning operation, the error propagation coefficient is determined as a constant value such as 1/6 and when the majority of the desired values are 1 or near 1, an error value regarding the opposite desired value 0 is amplified, and when the output values become equal to or more than 1, it is deemed that there is no error with regard to the output of more than 1 in case of many outputs 1, thereby speeding up an operation of updating the weight. dated 1993-08-24"
5239619,learning method for a data processing system having a multi-layer neural network,""" a learning method for a neural network having at least an input neuron layer, an output neuron layer, and a middle neuron layer between the input and output layers. each of the layers include a plurality of neurons which are coupled to corresponding neurons in adjacent neural layers. the learning method performs a learning function on the neurons of the middle layer on the basis of the respective outputs, or """"ignition patterns"""", of the neurons in the neural layers adjacent to the middle layer. the ignition pattern of neurons in the input layer is decided artificially according to a preferable image pattern to be input. the ignition pattern of neurons in the output layer is decided artificially according to the ignition pattern of the input layer neurons, wherein the ignition pattern of the output layer neurons is predetermined to correspond to a code or pattern preferable for a user. the ignition pattern of the middle layer neurons, coupled to the associated neurons of the respective input and output layers, is then decided according to the ignition pattern of the input layer and the output layer. """,1993-08-24,"The title of the patent is learning method for a data processing system having a multi-layer neural network and its abstract is "" a learning method for a neural network having at least an input neuron layer, an output neuron layer, and a middle neuron layer between the input and output layers. each of the layers include a plurality of neurons which are coupled to corresponding neurons in adjacent neural layers. the learning method performs a learning function on the neurons of the middle layer on the basis of the respective outputs, or """"ignition patterns"""", of the neurons in the neural layers adjacent to the middle layer. the ignition pattern of neurons in the input layer is decided artificially according to a preferable image pattern to be input. the ignition pattern of neurons in the output layer is decided artificially according to the ignition pattern of the input layer neurons, wherein the ignition pattern of the output layer neurons is predetermined to correspond to a code or pattern preferable for a user. the ignition pattern of the middle layer neurons, coupled to the associated neurons of the respective input and output layers, is then decided according to the ignition pattern of the input layer and the output layer. "" dated 1993-08-24"
5241509,arrangement of data cells and neural network system utilizing such an arrangement,"an arrangement of data cells which stores at least one matrix of data words which are arranged in rows and columns, the matrix being distributed in the arrangement in order to deliver/receive, via a single bus, permuted data words which correspond either to a row or to a column of the matrix. each data cell is connected to the single bus via series-connected switches which are associated with a respective addressing mode, the switches which address a same word of a same mode being directly controlled by a same selection signal. circulation members enable the original order of the data on the bus to be restored. an arrangement of this kind is used in a layered neural network system for executing the error backpropagation algorithm.",1993-08-31,"The title of the patent is arrangement of data cells and neural network system utilizing such an arrangement and its abstract is an arrangement of data cells which stores at least one matrix of data words which are arranged in rows and columns, the matrix being distributed in the arrangement in order to deliver/receive, via a single bus, permuted data words which correspond either to a row or to a column of the matrix. each data cell is connected to the single bus via series-connected switches which are associated with a respective addressing mode, the switches which address a same word of a same mode being directly controlled by a same selection signal. circulation members enable the original order of the data on the bus to be restored. an arrangement of this kind is used in a layered neural network system for executing the error backpropagation algorithm. dated 1993-08-31"
5241620,embedding neural networks into spreadsheet applications,"the present invention relates to a method of embedding a neural network into an application program such as a spreadsheet program. the method comprises providing an application program in which information is stored in rows and columns or a database containing fields and records and embedding a neural network in the application program or database using the stored information. the embedding step includes allocating unused memory in the application program and creating both a neural network engine and an application interface structure from the unused memory. once the neural network engine and an application interface structure have been created, the neural network may be trained using variable numerical and symbolic data stored within the application program. once training is completed, the neural network is ready for use, merely by using a recall function built into the applications program.",1993-08-31,"The title of the patent is embedding neural networks into spreadsheet applications and its abstract is the present invention relates to a method of embedding a neural network into an application program such as a spreadsheet program. the method comprises providing an application program in which information is stored in rows and columns or a database containing fields and records and embedding a neural network in the application program or database using the stored information. the embedding step includes allocating unused memory in the application program and creating both a neural network engine and an application interface structure from the unused memory. once the neural network engine and an application interface structure have been created, the neural network may be trained using variable numerical and symbolic data stored within the application program. once training is completed, the neural network is ready for use, merely by using a recall function built into the applications program. dated 1993-08-31"
5241845,neurocontrol for washing machines,"a fully automatic washing machine includes a detector for detecting a cloth volume, cloth type, soil degree and soil type in regard to clothes contained in a wash tub. a control device calculates a wash water stream in a wash step and a period of the wash step in the washing operation by a neurocontrol in which data of the cloth volume, the cloth type, soil degree and soil type are supplied to a neural network as input data. the neurocontrol is compensated for in accordance with the turbidity of a wash liquid detected at the time of completion of the wash step.",1993-09-07,"The title of the patent is neurocontrol for washing machines and its abstract is a fully automatic washing machine includes a detector for detecting a cloth volume, cloth type, soil degree and soil type in regard to clothes contained in a wash tub. a control device calculates a wash water stream in a wash step and a period of the wash step in the washing operation by a neurocontrol in which data of the cloth volume, the cloth type, soil degree and soil type are supplied to a neural network as input data. the neurocontrol is compensated for in accordance with the turbidity of a wash liquid detected at the time of completion of the wash step. dated 1993-09-07"
5243688,virtual neurocomputer architectures for neural networks,"the architectures for a scalable neural processor (snap) and a triangular scalable neural array processor (t-snap) are expanded to handle network simulations where the number of neurons to be modeled exceeds the number of physical neurons implemented. this virtual neural processing is described for three general virtual architectural approaches for handling the virtual neurons, one for snap and one for tsnap, and a third approach applied to both snap and tsnap.",1993-09-07,"The title of the patent is virtual neurocomputer architectures for neural networks and its abstract is the architectures for a scalable neural processor (snap) and a triangular scalable neural array processor (t-snap) are expanded to handle network simulations where the number of neurons to be modeled exceeds the number of physical neurons implemented. this virtual neural processing is described for three general virtual architectural approaches for handling the virtual neurons, one for snap and one for tsnap, and a third approach applied to both snap and tsnap. dated 1993-09-07"
5245672,object/anti-object neural network segmentation,"the system of the present invention applies self-organizing and/or supervd learning network methods to the problem of segmentation. the segmenter receives a visual field, implemented as a sliding window and distinguishes occurrences of complete characters from occurrences of parts of neighboring characters. images of isolated whole characters are true objects and the opposite of true objects are anti-objects, centered on the space between two characters. the window is moved across a line of text producing a sequence of images and the segmentation system distinguishes true objects from anti-objects. frames classified as anti-objects demarcate character boundaries, and frames classified as true objects represent detected character images. the system of the present invention may be a feedforward adaption using a symmetric triggering network. inputs to the network are applied directly to the separate associative memories of the network. the associative memories produce a best match pattern output for each part of the input data. the associative memories provide two or more subnetworks which define data subsets, such as objects or anti-objects, according to previously learned examples. multi-layer perceptron architecture may also be used in the system of the present invention rather than the symmetrically triggered feedforward adaptation with tradeoffs in training time but advantages in speed.",1993-09-14,"The title of the patent is object/anti-object neural network segmentation and its abstract is the system of the present invention applies self-organizing and/or supervd learning network methods to the problem of segmentation. the segmenter receives a visual field, implemented as a sliding window and distinguishes occurrences of complete characters from occurrences of parts of neighboring characters. images of isolated whole characters are true objects and the opposite of true objects are anti-objects, centered on the space between two characters. the window is moved across a line of text producing a sequence of images and the segmentation system distinguishes true objects from anti-objects. frames classified as anti-objects demarcate character boundaries, and frames classified as true objects represent detected character images. the system of the present invention may be a feedforward adaption using a symmetric triggering network. inputs to the network are applied directly to the separate associative memories of the network. the associative memories produce a best match pattern output for each part of the input data. the associative memories provide two or more subnetworks which define data subsets, such as objects or anti-objects, according to previously learned examples. multi-layer perceptron architecture may also be used in the system of the present invention rather than the symmetrically triggered feedforward adaptation with tradeoffs in training time but advantages in speed. dated 1993-09-14"
5245697,neural network processing apparatus for identifying an unknown image pattern as one of a plurality of instruction image patterns,"a neural network processing apparatus calculates an average of the absolute values of differences between the output values of all neurons and a center value whenever the output value of all neurons change, and calculates the difference between the average and the previous average. if the average is larger than a threshold or the previous average, the gain of a function in the network is decreased. if the average is smaller than the threshold or the previous average, the gain of the function is increased. then the controlled function is set to each neuron and the neural network is activated repeatedly to correctly identify an unknown multivalued image pattern.",1993-09-14,"The title of the patent is neural network processing apparatus for identifying an unknown image pattern as one of a plurality of instruction image patterns and its abstract is a neural network processing apparatus calculates an average of the absolute values of differences between the output values of all neurons and a center value whenever the output value of all neurons change, and calculates the difference between the average and the previous average. if the average is larger than a threshold or the previous average, the gain of a function in the network is decreased. if the average is smaller than the threshold or the previous average, the gain of the function is increased. then the controlled function is set to each neuron and the neural network is activated repeatedly to correctly identify an unknown multivalued image pattern. dated 1993-09-14"
5247206,neural network accommodating parallel synaptic weight adjustments in a single cycle,a neural network providing correlation learning in a synapse cell coupled to a circuit for parallel implementation of weight adjustment provides the learning portion of the synaptic operation and includes a floating gate device having a corresponding floating gate member that stores the connection weight of the cell. parallel weight adjustments are performed in a single operational cycle utilizing floating gate technology and control signals that facilitate programming/erasing operations.,1993-09-21,The title of the patent is neural network accommodating parallel synaptic weight adjustments in a single cycle and its abstract is a neural network providing correlation learning in a synapse cell coupled to a circuit for parallel implementation of weight adjustment provides the learning portion of the synaptic operation and includes a floating gate device having a corresponding floating gate member that stores the connection weight of the cell. parallel weight adjustments are performed in a single operational cycle utilizing floating gate technology and control signals that facilitate programming/erasing operations. dated 1993-09-21
5247445,control unit of an internal combustion engine control unit utilizing a neural network to reduce deviations between exhaust gas constituents and predetermined values,a control unit for an internal combustion engine that compensates for variations in injection valve flow rate characteristics by detecting an operation status of the engine and then using this status information to calculate a supply air amount or supply fuel amount in accordance with the detected status. exhaust gas constituents are detected and then used to correct the calculated supply air or supply fuel amount. the control unit compares the exhaust gas constituents with predetermined values and then uses a neural network to control the supply air amount or supply fuel amount to make any deviation between the exhaust gas constituents and the predetermined value approach zero.,1993-09-21,The title of the patent is control unit of an internal combustion engine control unit utilizing a neural network to reduce deviations between exhaust gas constituents and predetermined values and its abstract is a control unit for an internal combustion engine that compensates for variations in injection valve flow rate characteristics by detecting an operation status of the engine and then using this status information to calculate a supply air amount or supply fuel amount in accordance with the detected status. exhaust gas constituents are detected and then used to correct the calculated supply air or supply fuel amount. the control unit compares the exhaust gas constituents with predetermined values and then uses a neural network to control the supply air amount or supply fuel amount to make any deviation between the exhaust gas constituents and the predetermined value approach zero. dated 1993-09-21
5247584,signal processing unit for classifying objects on the basis of signals from sensors,"in a signal processing arrangement for classifying objects on the basis of signals from a plurality of different sensors each of the signals from the sensors is applied to a pair of neural networks. one neural network of each pair processes predetermined characteristics of the object and the other neural network processes movement or special data of the object such that these networks provide detection, identification and movement information specific for the sensors. feature vectors formed from this information specific for the sensors are applied to a neural network for determining the associations of the identification and movement information. the information obtained by this network is applied together with the feature vectors to a network for identifying and classifying the object. the information from the association and identification networks, respectively, are supplied together with the information specific for the sensors to an expert system which, by using further knowledge about data and facts of the potential objects, makes final decisions and conclusions for identification.",1993-09-21,"The title of the patent is signal processing unit for classifying objects on the basis of signals from sensors and its abstract is in a signal processing arrangement for classifying objects on the basis of signals from a plurality of different sensors each of the signals from the sensors is applied to a pair of neural networks. one neural network of each pair processes predetermined characteristics of the object and the other neural network processes movement or special data of the object such that these networks provide detection, identification and movement information specific for the sensors. feature vectors formed from this information specific for the sensors are applied to a neural network for determining the associations of the identification and movement information. the information obtained by this network is applied together with the feature vectors to a network for identifying and classifying the object. the information from the association and identification networks, respectively, are supplied together with the information specific for the sensors to an expert system which, by using further knowledge about data and facts of the potential objects, makes final decisions and conclusions for identification. dated 1993-09-21"
5247606,adaptively setting analog weights in a neural network and the like,"a method for adaptively setting analog weights in analog cells of a neural network and the like. the process starts by addressing a synapse cell in the network. a target weight for said addressed synapse cell is selected, and the current weight present on the synapse cell is measured. the amplitude and duration of a voltage pulse to be applied to said synapse cell to adjust said synapse cell in the direction of said target weight is calculated using a set of coefficients representing the the physical characteristics of the synapse cell. the voltage pulse is applied to the addressed synapse cell and the new weight of the synapse cell is re-measured. if the synapse cell weight is within acceptable limits of the target weight, the values of the coefficients are saved and the next adjacent synapse cell is addressed until all synapse cells are set. if the synapse cell is not within acceptable limits, new values for the coefficients are calculated in relation to the re-measured weight. a new voltage pulse is generated and applied to the synapse cell. the process is repeated until the weight of the synapse cell is set within an acceptable limit of the target weight.",1993-09-21,"The title of the patent is adaptively setting analog weights in a neural network and the like and its abstract is a method for adaptively setting analog weights in analog cells of a neural network and the like. the process starts by addressing a synapse cell in the network. a target weight for said addressed synapse cell is selected, and the current weight present on the synapse cell is measured. the amplitude and duration of a voltage pulse to be applied to said synapse cell to adjust said synapse cell in the direction of said target weight is calculated using a set of coefficients representing the the physical characteristics of the synapse cell. the voltage pulse is applied to the addressed synapse cell and the new weight of the synapse cell is re-measured. if the synapse cell weight is within acceptable limits of the target weight, the values of the coefficients are saved and the next adjacent synapse cell is addressed until all synapse cells are set. if the synapse cell is not within acceptable limits, new values for the coefficients are calculated in relation to the re-measured weight. a new voltage pulse is generated and applied to the synapse cell. the process is repeated until the weight of the synapse cell is set within an acceptable limit of the target weight. dated 1993-09-21"
5248899,neural network using photoelectric substance,"a neural network, and a method of storing information and retrieving it by such network. the network comprises neurons, synapses and switches, and when required also rectifying means. the network is based on a substance which undergoes a reversible change from stable state a to stable state b, and this substance can also be changed from state a to another state c, which change is also reversible, where each change provides a measurable electrical pulse. the change of state is brought about by means of illumination for a predetermined period of time at a certain wavelength, it being possible to convert a desired part of the substance from one state to the other.",1993-09-28,"The title of the patent is neural network using photoelectric substance and its abstract is a neural network, and a method of storing information and retrieving it by such network. the network comprises neurons, synapses and switches, and when required also rectifying means. the network is based on a substance which undergoes a reversible change from stable state a to stable state b, and this substance can also be changed from state a to another state c, which change is also reversible, where each change provides a measurable electrical pulse. the change of state is brought about by means of illumination for a predetermined period of time at a certain wavelength, it being possible to convert a desired part of the substance from one state to the other. dated 1993-09-28"
5249259,genetic algorithm technique for designing neural networks,"a generic algorithm search is applied to determine an optimum set of values (e.g., interconnection weights in a neural network), each value being associated with a pair of elements drawn from a universe of n elements, n an integer greater than zero, where the utility of any possible set of said values may be measured. an initial possible set of values is assembled, the values being organized in a matrix whose rows and columns correspond to the elements. a genetic algorithm operator is applied to generate successor matrices from said matrix. matrix computations are performed on the successor matrices to generate measures of the relative utilities of the successor matrices. a surviving matrix is selected from the successor matrices on the basis of the metrics. the steps are repeated until the metric of the surviving matrix is satisfactory.",1993-09-28,"The title of the patent is genetic algorithm technique for designing neural networks and its abstract is a generic algorithm search is applied to determine an optimum set of values (e.g., interconnection weights in a neural network), each value being associated with a pair of elements drawn from a universe of n elements, n an integer greater than zero, where the utility of any possible set of said values may be measured. an initial possible set of values is assembled, the values being organized in a matrix whose rows and columns correspond to the elements. a genetic algorithm operator is applied to generate successor matrices from said matrix. matrix computations are performed on the successor matrices to generate measures of the relative utilities of the successor matrices. a surviving matrix is selected from the successor matrices on the basis of the metrics. the steps are repeated until the metric of the surviving matrix is satisfactory. dated 1993-09-28"
5249954,integrated imaging sensor/neural network controller for combustion systems,"disclosed is an integrated imaging sensor/neural network controller for combustion control systems. the controller uses electronic imaging sensing of chemiluminescence from a combustion system, combined with neural network image processing, to sensitively identify and control a complex combustion system. the imaging system used is not adversely affected by the normal emissions variations caused by changes in burner load and flame position. by incorporating neural networks to learn emission patterns associated with combustor performance, control using image technology is fast enough to be used in a real time, closed loop control system. this advance in sensing and control strategy allows use of the spatial distribution of important parameters in the combustion system in identifying the overall operation condition of a given combustor and in formulating a control response accorded to a pre-determined control model.",1993-10-05,"The title of the patent is integrated imaging sensor/neural network controller for combustion systems and its abstract is disclosed is an integrated imaging sensor/neural network controller for combustion control systems. the controller uses electronic imaging sensing of chemiluminescence from a combustion system, combined with neural network image processing, to sensitively identify and control a complex combustion system. the imaging system used is not adversely affected by the normal emissions variations caused by changes in burner load and flame position. by incorporating neural networks to learn emission patterns associated with combustor performance, control using image technology is fast enough to be used in a real time, closed loop control system. this advance in sensing and control strategy allows use of the spatial distribution of important parameters in the combustion system in identifying the overall operation condition of a given combustor and in formulating a control response accorded to a pre-determined control model. dated 1993-10-05"
5250766,elevator control apparatus using neural network to predict car direction reversal floor,"an elevator control apparatus capable of predicting reversion floors of elevator cages accurately. the control apparatus comprises a neural network, in which traffic state data are fetched into the neural network, so that predicted values of floors where the moving direction of each cage is reversed are calculated as predicted reversion floors. in the elevator control apparatus, reversion floors near true reversion floors can be predicted flexibly correspondingly to traffic state and traffic volume.",1993-10-05,"The title of the patent is elevator control apparatus using neural network to predict car direction reversal floor and its abstract is an elevator control apparatus capable of predicting reversion floors of elevator cages accurately. the control apparatus comprises a neural network, in which traffic state data are fetched into the neural network, so that predicted values of floors where the moving direction of each cage is reversed are calculated as predicted reversion floors. in the elevator control apparatus, reversion floors near true reversion floors can be predicted flexibly correspondingly to traffic state and traffic volume. dated 1993-10-05"
5251269,multi-layer neural network modelled after the striate cortex for recognizing visual patterns,"a pattern recognition system includes at least one pair of basic associative units each having at least first and second unit ports for receiving pattern signal groups, respectively and a third unit port for outputting a pattern signal group. the pattern recognition system has characteristics of the type of pattern recognition carried out by living organisms. each of the basic units operates to derive weighting values for respective signals of the pattern signal groups inputted to the first and second unit ports of the basic unit itself in accordance with the degree of consistency between a previously given weighting pattern and respective patterns specified by the pattern signal groups inputted to the first and second unit ports of the basic unit itself. each of the basic units also operates to modulate the respective signals of the pattern signal groups inputted to the first and second unit ports of the basic unit in accordance with the derived weighting values and to totalize the modulated signals so as to form an output signal outputted form the third unit port of the basic unit itself. the third unit port of one of the basic unit pair is coupled to the first unit port of the other basic unit, and the third unit port of the other basic unit is coupled to the second unit port of the one basic unit. thus, the third unit port of one of the basic unit pair gives an recognition output.",1993-10-05,"The title of the patent is multi-layer neural network modelled after the striate cortex for recognizing visual patterns and its abstract is a pattern recognition system includes at least one pair of basic associative units each having at least first and second unit ports for receiving pattern signal groups, respectively and a third unit port for outputting a pattern signal group. the pattern recognition system has characteristics of the type of pattern recognition carried out by living organisms. each of the basic units operates to derive weighting values for respective signals of the pattern signal groups inputted to the first and second unit ports of the basic unit itself in accordance with the degree of consistency between a previously given weighting pattern and respective patterns specified by the pattern signal groups inputted to the first and second unit ports of the basic unit itself. each of the basic units also operates to modulate the respective signals of the pattern signal groups inputted to the first and second unit ports of the basic unit in accordance with the derived weighting values and to totalize the modulated signals so as to form an output signal outputted form the third unit port of the basic unit itself. the third unit port of one of the basic unit pair is coupled to the first unit port of the other basic unit, and the third unit port of the other basic unit is coupled to the second unit port of the one basic unit. thus, the third unit port of one of the basic unit pair gives an recognition output. dated 1993-10-05"
5251287,apparatus and method for neural processing,"the neural computing paradigm is characterized as a dynamic and highly computationally intensive system typically consisting of input weight multiplications, product summation, neural state calculations, and complete connectivity among the neurons. herein is described neural network architecture for a scalable neural array process (snap) which uses a unique interconnection and communication scheme within an array structure that provides high performance for completely connected network models such as the hopfield model. snap's packaging and expansion capabilities are addressed, demonstrating snap's scalability to larger networks. the array processor is made up of multiple sets of orthogonal interconnections and activity generators. each activity generator is responsive to an output signal in order to generate a neuron value. the interconnection structure also uses special adder trees which respond in a first state to generate an output signal and in a second state to communicate a neuron value back to the input of the array processor.",1993-10-05,"The title of the patent is apparatus and method for neural processing and its abstract is the neural computing paradigm is characterized as a dynamic and highly computationally intensive system typically consisting of input weight multiplications, product summation, neural state calculations, and complete connectivity among the neurons. herein is described neural network architecture for a scalable neural array process (snap) which uses a unique interconnection and communication scheme within an array structure that provides high performance for completely connected network models such as the hopfield model. snap's packaging and expansion capabilities are addressed, demonstrating snap's scalability to larger networks. the array processor is made up of multiple sets of orthogonal interconnections and activity generators. each activity generator is responsive to an output signal in order to generate a neuron value. the interconnection structure also uses special adder trees which respond in a first state to generate an output signal and in a second state to communicate a neuron value back to the input of the array processor. dated 1993-10-05"
5251626,apparatus and method for the detection and treatment of arrhythmias using a neural network,"an apparatus and method for the detection and treatment of arrhythmias using a processor having a neural network with a hierarchical arrangement including a first lower level for classifying individual waveforms, a second higher level for diagnosing detected arrhythmias and a third higher level for the application of therapy in response to a diagnosed arrhythmia. the neural network may be a back propogation neural network or an associative memory type neural network. the arrhythmias detected may be at least one of bradycardia, tachycardia and fibrillation. the apparatus may include a cardioverting/defibrillating pacemaker. in general, the apparatus acquires physiological signals representative of heart activity in a patient. a neural network receives the physiological signals and determines if any arrhythmia is present, and if present, selects therapy to be applied to the heart. a therapy generator then applies the therapy selected by the neural network. the physiological signals may be processed or unprocessed ecg signal, signals indicative of the properties of the blood including the presence of gases, blood temperature, and blood flow signals or signals representative of ventricular wall impedance or ventricular volume.",1993-10-12,"The title of the patent is apparatus and method for the detection and treatment of arrhythmias using a neural network and its abstract is an apparatus and method for the detection and treatment of arrhythmias using a processor having a neural network with a hierarchical arrangement including a first lower level for classifying individual waveforms, a second higher level for diagnosing detected arrhythmias and a third higher level for the application of therapy in response to a diagnosed arrhythmia. the neural network may be a back propogation neural network or an associative memory type neural network. the arrhythmias detected may be at least one of bradycardia, tachycardia and fibrillation. the apparatus may include a cardioverting/defibrillating pacemaker. in general, the apparatus acquires physiological signals representative of heart activity in a patient. a neural network receives the physiological signals and determines if any arrhythmia is present, and if present, selects therapy to be applied to the heart. a therapy generator then applies the therapy selected by the neural network. the physiological signals may be processed or unprocessed ecg signal, signals indicative of the properties of the blood including the presence of gases, blood temperature, and blood flow signals or signals representative of ventricular wall impedance or ventricular volume. dated 1993-10-12"
5252829,method of determining urea in milk,"the concentration of urea in a concentration range of 0-0.1% in a milk sample containing at least 1% fat, at least 1% dissolved lactose, and at least 1% protein, is determined with an accuracy better than 0.007%, expressed as standard error of prediction, by an infrared attenuation measuring technique, by determining, on the sample, the attenuation in the region of infrared radiation from 1000 cm.sup.-1 (10.0 .mu.m) to 4000 cm.sup.-1 (2.50 .mu.m), at least one determination being made in a waveband in the region from 1000 cm.sup.-1 (10.0 .mu.m) to 1800 cm.sup.-1 (5.56 .mu.m) in which urea absorbs, at least one other determination being made in a waveband in which fat absorbs, at least one further determination being made in a waveband where lactose absorbs, and at least one further determination being made in a waveband where protein absorbs; determining, on the basis of the thus determined attenuations and predetermined parameters established by multivariate calibration, the contribution from fat, lactose, and protein in the waveband where urea absorbs, and quantitatively assessing the concentration of urea in the sample on the basis of the absorption in the waveband where urea absorbs, and on the basis of the determined contribution from fat, lactose and protein in said waveband. the multivariate calibration may be performed by a partial least squares algorithm, principal component regression, multiple linear regression, or artificial neural network learning. using the method according to the invention, compensation for the influence on the urea measurement may further be performed for one or several of the following components: citric acid, free fatty acids, antibiotics, phosphates, somatic cells, bacteria, preservatives and casein.",1993-10-12,"The title of the patent is method of determining urea in milk and its abstract is the concentration of urea in a concentration range of 0-0.1% in a milk sample containing at least 1% fat, at least 1% dissolved lactose, and at least 1% protein, is determined with an accuracy better than 0.007%, expressed as standard error of prediction, by an infrared attenuation measuring technique, by determining, on the sample, the attenuation in the region of infrared radiation from 1000 cm.sup.-1 (10.0 .mu.m) to 4000 cm.sup.-1 (2.50 .mu.m), at least one determination being made in a waveband in the region from 1000 cm.sup.-1 (10.0 .mu.m) to 1800 cm.sup.-1 (5.56 .mu.m) in which urea absorbs, at least one other determination being made in a waveband in which fat absorbs, at least one further determination being made in a waveband where lactose absorbs, and at least one further determination being made in a waveband where protein absorbs; determining, on the basis of the thus determined attenuations and predetermined parameters established by multivariate calibration, the contribution from fat, lactose, and protein in the waveband where urea absorbs, and quantitatively assessing the concentration of urea in the sample on the basis of the absorption in the waveband where urea absorbs, and on the basis of the determined contribution from fat, lactose and protein in said waveband. the multivariate calibration may be performed by a partial least squares algorithm, principal component regression, multiple linear regression, or artificial neural network learning. using the method according to the invention, compensation for the influence on the urea measurement may further be performed for one or several of the following components: citric acid, free fatty acids, antibiotics, phosphates, somatic cells, bacteria, preservatives and casein. dated 1993-10-12"
5253327,optimization apparatus,"an optimization apparatus using a layered neural network having an input layer formed of input units and supplied with input data and an output layer formed of output units connected to the individual input units with specified synaptic weights, which comprises a calculator circuit for calculating, for each output unit, the degree of similarity between the input data and the synaptic weight as well as the evaluation function value by causing the optimization problem to correspond to the fired units in the output layer, a detector for detecting the best matching optimum output unit on the basis of the output of the calculator circuit, and a self-organization circuit for changing the synaptic weights of a group of the output units associated with the optimum unit detected by the detector.",1993-10-12,"The title of the patent is optimization apparatus and its abstract is an optimization apparatus using a layered neural network having an input layer formed of input units and supplied with input data and an output layer formed of output units connected to the individual input units with specified synaptic weights, which comprises a calculator circuit for calculating, for each output unit, the degree of similarity between the input data and the synaptic weight as well as the evaluation function value by causing the optimization problem to correspond to the fired units in the output layer, a detector for detecting the best matching optimum output unit on the basis of the output of the calculator circuit, and a self-organization circuit for changing the synaptic weights of a group of the output units associated with the optimum unit detected by the detector. dated 1993-10-12"
5253328,neural-network content-addressable memory,"a neural network content-addressable error-correcting memory system is disclosed including a plurality of hidden and visible processing units interconnected via a linear interconnection matrix. the network is symmetric and all self-connections are not present. all connections between processing units are present, except those connecting hidden units to other hidden units. each visible unit is connected to each other visible unit and to each hidden unit. a mean field theory learning and retrieval algorithm is also provided. bit patterns or code words are stored in the network via the learning algorithm. the retrieval algorithm retrieves error-corrected bit patterns in response to noisy or error-containing input bit patterns.",1993-10-12,"The title of the patent is neural-network content-addressable memory and its abstract is a neural network content-addressable error-correcting memory system is disclosed including a plurality of hidden and visible processing units interconnected via a linear interconnection matrix. the network is symmetric and all self-connections are not present. all connections between processing units are present, except those connecting hidden units to other hidden units. each visible unit is connected to each other visible unit and to each hidden unit. a mean field theory learning and retrieval algorithm is also provided. bit patterns or code words are stored in the network via the learning algorithm. the retrieval algorithm retrieves error-corrected bit patterns in response to noisy or error-containing input bit patterns. dated 1993-10-12"
5253329,neural network for processing both spatial and temporal data with time based back-propagation,"neural network algorithms have impressively demonstrated the capability of modelling spatial information. on the other hand, the application of parallel distributed models to processing of temporal data has been severely restricted. the invention introduces a novel technique which adds the dimension of time to the well known back-propagatio pac origin of the invention the invention described herein was made by employees of the united states government and ma be manufactured and used by or for the government of the united states of america for governmental purposes without payment of any royalties thereon or therefor.",1993-10-12,"The title of the patent is neural network for processing both spatial and temporal data with time based back-propagation and its abstract is neural network algorithms have impressively demonstrated the capability of modelling spatial information. on the other hand, the application of parallel distributed models to processing of temporal data has been severely restricted. the invention introduces a novel technique which adds the dimension of time to the well known back-propagatio pac origin of the invention the invention described herein was made by employees of the united states government and ma be manufactured and used by or for the government of the united states of america for governmental purposes without payment of any royalties thereon or therefor. dated 1993-10-12"
5253330,network architecture for the programmable emulation of artificial neural networks having digital operation,"a network architecture for the programmable emulation of large artificial neural networks ann having digital operation employs a plurality l of neuron units of identical structure, each equipped with m neurons, the inputs (e) thereof being connected to network inputs (e.sub.n) multiplied or branching via individual input registers (reg.sub.e). the outputs (a) of the neuron units are connectable to network outputs (a.sub.n) at different points in time via individual multiplexers (mux) and individual output registers (reg.sub.a) and the neuron units have individual auxiliary inputs via which signals can be supplied to them that represent weighting values (w) for weighting the appertaining neural connections and represent thresholds (0) for weighting input signals.",1993-10-12,"The title of the patent is network architecture for the programmable emulation of artificial neural networks having digital operation and its abstract is a network architecture for the programmable emulation of large artificial neural networks ann having digital operation employs a plurality l of neuron units of identical structure, each equipped with m neurons, the inputs (e) thereof being connected to network inputs (e.sub.n) multiplied or branching via individual input registers (reg.sub.e). the outputs (a) of the neuron units are connectable to network outputs (a.sub.n) at different points in time via individual multiplexers (mux) and individual output registers (reg.sub.a) and the neuron units have individual auxiliary inputs via which signals can be supplied to them that represent weighting values (w) for weighting the appertaining neural connections and represent thresholds (0) for weighting input signals. dated 1993-10-12"
5255342,pattern recognition system and method using neural network,"an inner product computing unit computes inner products of an input pattern whose category is unknown, and orthogonalized dictionary sets of a plurality of reference patterns whose categories are known. a nonlinear converting unit nonlinearly converts the inner products in accordance with a positive-negative symmetrical nonlinear function. a neural network unit or a statistical discriminant function computing unit performs predetermined computations of the nonlinearly converted values on the basis of preset coefficients in units of categories using a neural network or a statistical discriminant function. a determining section compares values calculated in units of categories using the preset coefficients with each other to discriminate a category to which the input pattern belongs.",1993-10-19,"The title of the patent is pattern recognition system and method using neural network and its abstract is an inner product computing unit computes inner products of an input pattern whose category is unknown, and orthogonalized dictionary sets of a plurality of reference patterns whose categories are known. a nonlinear converting unit nonlinearly converts the inner products in accordance with a positive-negative symmetrical nonlinear function. a neural network unit or a statistical discriminant function computing unit performs predetermined computations of the nonlinearly converted values on the basis of preset coefficients in units of categories using a neural network or a statistical discriminant function. a determining section compares values calculated in units of categories using the preset coefficients with each other to discriminate a category to which the input pattern belongs. dated 1993-10-19"
5255344,inference rule determining method and inference device,""" an inference rule determining process according to the present invention sequentially determines, using a learning function of a neural network model, a membership function representing a degree which the conditions of the if part of each inference rule is satisfied when input data is received to thereby obtain an optimal inference result without using experience rules. the inventive inference device uses an inference rule of the type """"if . . . then . . ."""" and includes a membership value determiner (1) which includes all of if part and has a neural network; individual inference quantity determiners (21)-(2r) which correspond to the respective then parts of the inference rules and determine the corresponding inference quantities for the inference rules; and a final inference quantity determiner which determines these inference quantities synthetically to obtain the final results of the inference. if the individual inference quantity determiners (2) each has a neural network structure, the non-linearity of the neural network models is used to obtain the result of the inference with high inference accuracy even if an object to be inferred is non-linear. """,1993-10-19,"The title of the patent is inference rule determining method and inference device and its abstract is "" an inference rule determining process according to the present invention sequentially determines, using a learning function of a neural network model, a membership function representing a degree which the conditions of the if part of each inference rule is satisfied when input data is received to thereby obtain an optimal inference result without using experience rules. the inventive inference device uses an inference rule of the type """"if . . . then . . ."""" and includes a membership value determiner (1) which includes all of if part and has a neural network; individual inference quantity determiners (21)-(2r) which correspond to the respective then parts of the inference rules and determine the corresponding inference quantities for the inference rules; and a final inference quantity determiner which determines these inference quantities synthetically to obtain the final results of the inference. if the individual inference quantity determiners (2) each has a neural network structure, the non-linearity of the neural network models is used to obtain the result of the inference with high inference accuracy even if an object to be inferred is non-linear. "" dated 1993-10-19"
5255346,method and apparatus for design of a vector quantizer,"a method and apparatus for the design of a robust vector quantizer is disclosed. the initial output vector set is equal to the centroid of a training sequence of input vectors. a neural-network simulation and neighborhood functions are utilized for splitting and optimizing the output vectors. in this manner, the entire output vector set is sensitive to each input vector and therefore optimal output vector locations with respect to specified distortion criteria are obtained. the resulting vector quantizer is robust for the class of signals represented by the training sequence.",1993-10-19,"The title of the patent is method and apparatus for design of a vector quantizer and its abstract is a method and apparatus for the design of a robust vector quantizer is disclosed. the initial output vector set is equal to the centroid of a training sequence of input vectors. a neural-network simulation and neighborhood functions are utilized for splitting and optimizing the output vectors. in this manner, the entire output vector set is sensitive to each input vector and therefore optimal output vector locations with respect to specified distortion criteria are obtained. the resulting vector quantizer is robust for the class of signals represented by the training sequence. dated 1993-10-19"
5255347,neural network with learning function,"a neural network system capable of performing integrated processing of a plurality of information includes a feature extractor group for extracting a plurality of learning feature data from learning data in a learning mode and a plurality of object feature data from object data to be processed in an execution mode, and an information processing unit for learning features of the learning data, based on the plurality of learning feature data from the feature extractor group and corresponding teacher data in the learning mode, and determining final learning result data from the plurality of object feature data from the feature extractor group in accordance with the learning result, including a logic representing relation between the plurality of object feature data in the execution mode.",1993-10-19,"The title of the patent is neural network with learning function and its abstract is a neural network system capable of performing integrated processing of a plurality of information includes a feature extractor group for extracting a plurality of learning feature data from learning data in a learning mode and a plurality of object feature data from object data to be processed in an execution mode, and an information processing unit for learning features of the learning data, based on the plurality of learning feature data from the feature extractor group and corresponding teacher data in the learning mode, and determining final learning result data from the plurality of object feature data from the feature extractor group in accordance with the learning result, including a logic representing relation between the plurality of object feature data in the execution mode. dated 1993-10-19"
5255348,"neural network for learning, recognition and recall of pattern sequences","a sequence processor for rapidly learning, recognizing and recalling temporal sequences. the processor, called the katamic system, is a biologically inspired artificial neural network based on a model of the functions of the cerebellum in the brain. the katamic system utilizes three basic types of neuron-like elements with different functional characteristics called predictrons, recognitrons and bi-stable switches. the katamic system is clock operated, processing input sequences pattern by pattern to produce an output pattern which is a prediction of the next pattern in the input sequence. the katamic system learns rapidly, has a large memory capacity, exhibits sequence completion and sequence recognition capability, and is fault and noise tolerant. the system's modular construction permits straightforward scaleability.",1993-10-19,"The title of the patent is neural network for learning, recognition and recall of pattern sequences and its abstract is a sequence processor for rapidly learning, recognizing and recalling temporal sequences. the processor, called the katamic system, is a biologically inspired artificial neural network based on a model of the functions of the cerebellum in the brain. the katamic system utilizes three basic types of neuron-like elements with different functional characteristics called predictrons, recognitrons and bi-stable switches. the katamic system is clock operated, processing input sequences pattern by pattern to produce an output pattern which is a prediction of the next pattern in the input sequence. the katamic system learns rapidly, has a large memory capacity, exhibits sequence completion and sequence recognition capability, and is fault and noise tolerant. the system's modular construction permits straightforward scaleability. dated 1993-10-19"
5255349,"""electronic neural network for solving """"traveling salesman"""" and similar global optimization problems""",""" this invention is a novel high-speed neural network based processor for solving the """"traveling salesman"""" and other global optimization problems. it comprises a novel hybrid architecture employing a binary synaptic array whose embodiment incorporates the fixed rules of the problem, such as the number of cities to be visited. the array is prompted by analog voltages representing variables such as distances. the processor incorporates two interconnected feedback networks, each of which solves part of the problem independently and simultaneously, yet which exchange information dynamically. """,1993-10-19,"The title of the patent is ""electronic neural network for solving """"traveling salesman"""" and similar global optimization problems"" and its abstract is "" this invention is a novel high-speed neural network based processor for solving the """"traveling salesman"""" and other global optimization problems. it comprises a novel hybrid architecture employing a binary synaptic array whose embodiment incorporates the fixed rules of the problem, such as the number of cities to be visited. the array is prompted by analog voltages representing variables such as distances. the processor incorporates two interconnected feedback networks, each of which solves part of the problem independently and simultaneously, yet which exchange information dynamically. "" dated 1993-10-19"
5255362,photo stimulated and controlled imaging neural network,"a photo stimulated and controlled imaging neural network for providing self generating learning sets and associative memory and programmability. an image to be recognized or detected is transferred to an imaging plane, which can be as simple as a lens or as complicated as a cathode ray tube. the imaging plane whose contents forms the input for a photo receptor array transfers the stimulus from the object to the photoreceptor array. the photoreceptor array responds to the stimulus provided by the imaging plane with various couplings between an array of neuron amplifiers. the photo receptor array comprises a plurality of synaptic photo controlled resistors which respond to the stimulus provided by the imaging plane. the individual neuron amplifiers settle into a set of on or off binary states based on the couplings of the photo controlled resistors which comprise the receptor array. the output states are equally weighted and as a whole constitute a particular learning set which is then passed onto a gate array where it can be utilized to make various decisions.",1993-10-19,"The title of the patent is photo stimulated and controlled imaging neural network and its abstract is a photo stimulated and controlled imaging neural network for providing self generating learning sets and associative memory and programmability. an image to be recognized or detected is transferred to an imaging plane, which can be as simple as a lens or as complicated as a cathode ray tube. the imaging plane whose contents forms the input for a photo receptor array transfers the stimulus from the object to the photoreceptor array. the photoreceptor array responds to the stimulus provided by the imaging plane with various couplings between an array of neuron amplifiers. the photo receptor array comprises a plurality of synaptic photo controlled resistors which respond to the stimulus provided by the imaging plane. the individual neuron amplifiers settle into a set of on or off binary states based on the couplings of the photo controlled resistors which comprise the receptor array. the output states are equally weighted and as a whole constitute a particular learning set which is then passed onto a gate array where it can be utilized to make various decisions. dated 1993-10-19"
5256911,neural network with multiplexed snyaptic processing,"in an apparatus for multiplexed operation of multi-cell neural network, the reference vector component values are stored as differential values in pairs of floating gate transistors. a long-tail pair differential transconductance multiplier is synthesized by selectively using the floating gate transistor pairs as the current source. appropriate transistor pairs are multiplexed into the network for forming a differential output current representative of the product of the input vector component applied to the differential input and the stored reference vector component stored in the multiplexed transistor pair that is switched into the multiplier network to function as the differential current source. pipelining and output multiplexing is also described in other preferred embodiments for increasing the effective output bandwidth of the network.",1993-10-26,"The title of the patent is neural network with multiplexed snyaptic processing and its abstract is in an apparatus for multiplexed operation of multi-cell neural network, the reference vector component values are stored as differential values in pairs of floating gate transistors. a long-tail pair differential transconductance multiplier is synthesized by selectively using the floating gate transistor pairs as the current source. appropriate transistor pairs are multiplexed into the network for forming a differential output current representative of the product of the input vector component applied to the differential input and the stored reference vector component stored in the multiplexed transistor pair that is switched into the multiplier network to function as the differential current source. pipelining and output multiplexing is also described in other preferred embodiments for increasing the effective output bandwidth of the network. dated 1993-10-26"
5257342,learning method for a data processing system with neighborhoods,"a learning method for a neural network type data processing system determines activation patterns in an input layer and output layer arbitrarily, increases weights of synapses in a middle layer and the output layer so that neuron activate with more than a certain rate among those corresponding to neurons in the input layer and the output layer and repeats the same process for each neuron in the middle layer. the input layer and output layer possess a plurality of neurons which activate and output certain data according to a specific result and the middle layer is between the input layer and output layer. the middle layer also possesses a plurality of neurons which are connected to each neuron in the input layer and output layer.",1993-10-26,"The title of the patent is learning method for a data processing system with neighborhoods and its abstract is a learning method for a neural network type data processing system determines activation patterns in an input layer and output layer arbitrarily, increases weights of synapses in a middle layer and the output layer so that neuron activate with more than a certain rate among those corresponding to neurons in the input layer and the output layer and repeats the same process for each neuron in the middle layer. the input layer and output layer possess a plurality of neurons which activate and output certain data according to a specific result and the middle layer is between the input layer and output layer. the middle layer also possesses a plurality of neurons which are connected to each neuron in the input layer and output layer. dated 1993-10-26"
5257343,intelligence information processing system,"an intelligence information processing system is composed of an associative memory and a serial processing-type computer. input pattern information is associated with the associative memory, and pattern recognition based on the computer evaluates an associative output. in accordance with this evaluation, an associative and restrictive condition is repeatedly added to the energy function of a neural network constituting the associative memory, thereby converging the associative output on a stable state of the energy. the converged associative output is verified with intelligence information stored in a computer memory. the associative and restrictive condition is again repeatedly added to the energy function in accordance with the verification so as to produce an output from the system.",1993-10-26,"The title of the patent is intelligence information processing system and its abstract is an intelligence information processing system is composed of an associative memory and a serial processing-type computer. input pattern information is associated with the associative memory, and pattern recognition based on the computer evaluates an associative output. in accordance with this evaluation, an associative and restrictive condition is repeatedly added to the energy function of a neural network constituting the associative memory, thereby converging the associative output on a stable state of the energy. the converged associative output is verified with intelligence information stored in a computer memory. the associative and restrictive condition is again repeatedly added to the energy function in accordance with the verification so as to produce an output from the system. dated 1993-10-26"
5258903,control circuit and power supply for televisions,"an adaptive feed forward control circuit and power supply for a television comprises a circuit for supplying energy from a source to a load, the load having energy requirements which vary in response to an input signal, for example a video signal. a feedback circuit generates a first correction signal indicative of a difference between an operating voltage or current level and a reference level. a neural network generates a second correction signal indicative of anticipated energy requirement variation by processing information in present values of the input signal. a control circuit, for example a pulse width modulating circuit, is responsive to the correction signals for controlling operation of the energy supplying circuit. the first and second correction signals are combined by a summing circuit. the neural network comprises a first signal adaptive circuit for the input signal and a second signal adaptive circuit for a processed version of the input signal. the processed input signal is linearly independent of the input signal to avoid redundancy in the weight factors. the square root of the input signal, for example, is appropriate for a switched mode power supply. the combination of outputs from the first and second signal adaptive circuits defines the second correction signal. a microprocessor can embody the neural network and provide the processed version of the input signal. the microprocessor can also embody the feedback circuit. the predictive correction signal can be adjusted responsive to the size and polarity of the energy requirement variation.",1993-11-02,"The title of the patent is control circuit and power supply for televisions and its abstract is an adaptive feed forward control circuit and power supply for a television comprises a circuit for supplying energy from a source to a load, the load having energy requirements which vary in response to an input signal, for example a video signal. a feedback circuit generates a first correction signal indicative of a difference between an operating voltage or current level and a reference level. a neural network generates a second correction signal indicative of anticipated energy requirement variation by processing information in present values of the input signal. a control circuit, for example a pulse width modulating circuit, is responsive to the correction signals for controlling operation of the energy supplying circuit. the first and second correction signals are combined by a summing circuit. the neural network comprises a first signal adaptive circuit for the input signal and a second signal adaptive circuit for a processed version of the input signal. the processed input signal is linearly independent of the input signal to avoid redundancy in the weight factors. the square root of the input signal, for example, is appropriate for a switched mode power supply. the combination of outputs from the first and second signal adaptive circuits defines the second correction signal. a microprocessor can embody the neural network and provide the processed version of the input signal. the microprocessor can also embody the feedback circuit. the predictive correction signal can be adjusted responsive to the size and polarity of the energy requirement variation. dated 1993-11-02"
5258934,charge domain bit serial vector-matrix multiplier and method thereof,a charge domain bit serial vector matrix multiplier for real time signal processing of mixed digital/analog signals for implementing opto-electronic neural networks and other signal processing functions. a combination of ccd and dcsd arrays permits vector/matrix multiplication with better than 10.sup.11 multiply accumulates per second on a one square centimeter chip. the ccd array portion of the invention is used to load and move charge packets into the dcsd array for processing therein. the ccd array is also used to empty the matrix of unwanted charge. the dcsd array is designed to store a plurality of charge packets representing the respective matrix values such as the synaptic interaction matrix of a neural network. the vector multiplicand may be applied in bit serial format. the row or sensor lines of the dcsd array are used to accumulate the results of the multiply operation. each such row output line is provided with a divide-by-two/accumulate ccd circuit which automatically compensates for the increasing value of the input vector element's bits from least significant bit to most significant bit. in a similar fashion each row output line can be provided with a multiply-by-two/accumulate ccd circuit which automatically accounts for the decreasing value of the input vector element's bits from most significant bit to least significant bit. the accumulated charge packet output of the array may be preferably converted to a digital signal compatible with the input vector configuration by utilizing a plurality of analog-to-digital converters.,1993-11-02,The title of the patent is charge domain bit serial vector-matrix multiplier and method thereof and its abstract is a charge domain bit serial vector matrix multiplier for real time signal processing of mixed digital/analog signals for implementing opto-electronic neural networks and other signal processing functions. a combination of ccd and dcsd arrays permits vector/matrix multiplication with better than 10.sup.11 multiply accumulates per second on a one square centimeter chip. the ccd array portion of the invention is used to load and move charge packets into the dcsd array for processing therein. the ccd array is also used to empty the matrix of unwanted charge. the dcsd array is designed to store a plurality of charge packets representing the respective matrix values such as the synaptic interaction matrix of a neural network. the vector multiplicand may be applied in bit serial format. the row or sensor lines of the dcsd array are used to accumulate the results of the multiply operation. each such row output line is provided with a divide-by-two/accumulate ccd circuit which automatically compensates for the increasing value of the input vector element's bits from least significant bit to most significant bit. in a similar fashion each row output line can be provided with a multiply-by-two/accumulate ccd circuit which automatically accounts for the decreasing value of the input vector element's bits from most significant bit to least significant bit. the accumulated charge packet output of the array may be preferably converted to a digital signal compatible with the input vector configuration by utilizing a plurality of analog-to-digital converters. dated 1993-11-02
5259064,signal processing apparatus having at least one neural network having pulse density signals as inputs and outputs,"a signal processing apparatus for controlling an object includes an input unit, a neural network, an output unit, a teaching unit, and an error signal generator for generating a teaching signal that makes the neural network learn in real time. an error signal generator generates an error signal from the teaching signal and information contained in the network output signal. the error signal controls the neural network so that the control output signal has correct control information with respect to the output signal from the controlled object.",1993-11-02,"The title of the patent is signal processing apparatus having at least one neural network having pulse density signals as inputs and outputs and its abstract is a signal processing apparatus for controlling an object includes an input unit, a neural network, an output unit, a teaching unit, and an error signal generator for generating a teaching signal that makes the neural network learn in real time. an error signal generator generates an error signal from the teaching signal and information contained in the network output signal. the error signal controls the neural network so that the control output signal has correct control information with respect to the output signal from the controlled object. dated 1993-11-02"
5259065,data processing system,"a data processing system of the neural network type. the system recognizes a predetermined shape by providing some connections that are inhibitory between a plurality of neurons in a neural layer of the neural network. if data is found in the inhibitory area, it makes it harder for the neurons in the correct area to fire. only when the neurons in the correct area fire is the predetermined shape recognized.",1993-11-02,"The title of the patent is data processing system and its abstract is a data processing system of the neural network type. the system recognizes a predetermined shape by providing some connections that are inhibitory between a plurality of neurons in a neural layer of the neural network. if data is found in the inhibitory area, it makes it harder for the neurons in the correct area to fire. only when the neurons in the correct area fire is the predetermined shape recognized. dated 1993-11-02"
5259384,ultrasonic bone-assessment apparatus and method,"non-invasive, quantitative in-vivo ultrasonic evaluation of bone is performed by subjecting bone to an acoustic excitation pulse supplied to one of two transducers on opposite sides of the bone, and involving a composite sine-wave signal consisting of repetitions of plural discrete ultrasonic frequencies that are spaced at approximately 2 mhz. signal-processing of received signal output of the other transducer is operative to sequentially average the most recently received given number of successive signals to obtain an averaged per-pulse signal and to produce a fourier transform of this signal. in a separate operation, the same transducer responds to the transmission and reception of the same excitation signal via a medium of known acoustic properties and path length to establish a reference signal, which is processed to produce its fourier transform. the two fourier transforms are comparatively evaluated to produce a bone-transfer function, which is then processed to derive the frequency-dependent specific-attenuation and group-velocity functions .mu.(f) and vg(f) associated with the bone-transfer function. the function vg(f) is related to the derivative of the phase of the bone-transfer function, as a function of frequency. a neural network, configured to generate an estimate of one or more of the desired bone-related quantities, is connected for response to the functions .mu.(f) and vg(f), whereby to generate the indicated estimates of bone status, namely, bone-density, bone-strength and fracture risk.",1993-11-09,"The title of the patent is ultrasonic bone-assessment apparatus and method and its abstract is non-invasive, quantitative in-vivo ultrasonic evaluation of bone is performed by subjecting bone to an acoustic excitation pulse supplied to one of two transducers on opposite sides of the bone, and involving a composite sine-wave signal consisting of repetitions of plural discrete ultrasonic frequencies that are spaced at approximately 2 mhz. signal-processing of received signal output of the other transducer is operative to sequentially average the most recently received given number of successive signals to obtain an averaged per-pulse signal and to produce a fourier transform of this signal. in a separate operation, the same transducer responds to the transmission and reception of the same excitation signal via a medium of known acoustic properties and path length to establish a reference signal, which is processed to produce its fourier transform. the two fourier transforms are comparatively evaluated to produce a bone-transfer function, which is then processed to derive the frequency-dependent specific-attenuation and group-velocity functions .mu.(f) and vg(f) associated with the bone-transfer function. the function vg(f) is related to the derivative of the phase of the bone-transfer function, as a function of frequency. a neural network, configured to generate an estimate of one or more of the desired bone-related quantities, is connected for response to the functions .mu.(f) and vg(f), whereby to generate the indicated estimates of bone status, namely, bone-density, bone-strength and fracture risk. dated 1993-11-09"
5260706,priority encoder,"a priority encoder using a mos array and neural network concepts is composed of an input side neuron group, an output side neuron group, a synapse group, a bias group and inverters. the encoder is simple in its construction and fast in its operating speed compared with the conventional priority encoders utilizing simple boolean logic.",1993-11-09,"The title of the patent is priority encoder and its abstract is a priority encoder using a mos array and neural network concepts is composed of an input side neuron group, an output side neuron group, a synapse group, a bias group and inverters. the encoder is simple in its construction and fast in its operating speed compared with the conventional priority encoders utilizing simple boolean logic. dated 1993-11-09"
5260871,method and apparatus for diagnosis of breast tumors,"an apparatus for distinguishing benign from malignant tumors in ultrasonic images of candidate tissue taken from a patient. a region of interest is located and defined on the ultrasonic image, including substantially all of the candidate tissue and excluding substantially all the normal tissue. the region of interest is digitized, generating an array of pixels intensity values. a first features is generated from the arrays of pixels corresponding to the angular second moment of the pixel intensity values. a second feature is generated from the array of pixels corresponding to the inverse contrast of the pixel intensity values. a third feature is generated from the array of pixels corresponding to the short run emphasis of the pixel intensity values. the first, second and third feature values are provided to a neural network. a set of trained weights are applied to the feature values, which generates a network output between 0 and 1, whereby the output values tend toward 1 when the candidate tissue is malignant and the output values tend toward 0 when the candidate tissue is benign.",1993-11-09,"The title of the patent is method and apparatus for diagnosis of breast tumors and its abstract is an apparatus for distinguishing benign from malignant tumors in ultrasonic images of candidate tissue taken from a patient. a region of interest is located and defined on the ultrasonic image, including substantially all of the candidate tissue and excluding substantially all the normal tissue. the region of interest is digitized, generating an array of pixels intensity values. a first features is generated from the arrays of pixels corresponding to the angular second moment of the pixel intensity values. a second feature is generated from the array of pixels corresponding to the inverse contrast of the pixel intensity values. a third feature is generated from the array of pixels corresponding to the short run emphasis of the pixel intensity values. the first, second and third feature values are provided to a neural network. a set of trained weights are applied to the feature values, which generates a network output between 0 and 1, whereby the output values tend toward 1 when the candidate tissue is malignant and the output values tend toward 0 when the candidate tissue is benign. dated 1993-11-09"
5261035,neural network architecture based on summation of phase-coherent alternating current signals,""" a neural network architecture has phase-coherent alternating current neural input signals. each input v.sub.k.sup.in is a two-phase pair of signals 180 degrees out of phase. capacitive coupling of both signals of n input pairs to a summation line gives a non-dissipative realization of the weighted sum ##equ1## with general real neural weights w.sub.ik. an alternating current offset signal proportional to u.sub.i is also capacitively coupled to the summation line. the signal on the summation line is passed through a low input capacitance follower/amplifier, a rectifier and a filter, producing a direct current signal proportional to the magnitude ##equ2## this signal is compared with a direct current threshold proportional to t.sub.i, and the resultant is used to gate a two-phase alternating current output signal. the output is therefore functionally related to the inputs by ##equ3## with .theta. the heaviside step function. this generalized neuron can directly compute the """"exclusive or"""" (xor) logical operation. alternative forms of the alternating current neuron using phase-shifters permit complex number inputs, outputs and neural weightings. """,1993-11-09,"The title of the patent is neural network architecture based on summation of phase-coherent alternating current signals and its abstract is "" a neural network architecture has phase-coherent alternating current neural input signals. each input v.sub.k.sup.in is a two-phase pair of signals 180 degrees out of phase. capacitive coupling of both signals of n input pairs to a summation line gives a non-dissipative realization of the weighted sum ##equ1## with general real neural weights w.sub.ik. an alternating current offset signal proportional to u.sub.i is also capacitively coupled to the summation line. the signal on the summation line is passed through a low input capacitance follower/amplifier, a rectifier and a filter, producing a direct current signal proportional to the magnitude ##equ2## this signal is compared with a direct current threshold proportional to t.sub.i, and the resultant is used to gate a two-phase alternating current output signal. the output is therefore functionally related to the inputs by ##equ3## with .theta. the heaviside step function. this generalized neuron can directly compute the """"exclusive or"""" (xor) logical operation. alternative forms of the alternating current neuron using phase-shifters permit complex number inputs, outputs and neural weightings. "" dated 1993-11-09"
5262632,integrated circuit for achieving pattern recognition,"an apparatus for massive computation in integrated circuits provides the ability to calculate multiple dot products between an image focused on the integrated circuit surface and many reference patterns built into the integrated circuit, and then give an output indication for all those reference patterns where the dot product exceeds a threshold. the implementation, using current mirrors for multiplication with fixed constants, permits the integrated circuit to achieve large amounts of computation per unit area. this apparatus permits a large input data bandwidth, and by virtue of having enough computation capacity to complete a processing task on one chip, the output bandwidth is greatly reduced as well. the apparatus is employed, as an example, in a neural network. a set of connections between nodes that modify the value of the signal passed from one node to the next. often many connections impinge on a node, and the summation of values at the node is further modified by a nonlinear function such as a threshold and amplitude limiter. values at the input nodes represent the signals to be evaluated by the network, and values at the outputs represent an evaluation by the network of the input signals. for instance, the input could be image pixels and the outputs could represent possible patterns to which the image could be assigned. the connections between weights are often determined and modified by training data, but they can also be prespecified in total or in part based on other information about the task of the network.",1993-11-16,"The title of the patent is integrated circuit for achieving pattern recognition and its abstract is an apparatus for massive computation in integrated circuits provides the ability to calculate multiple dot products between an image focused on the integrated circuit surface and many reference patterns built into the integrated circuit, and then give an output indication for all those reference patterns where the dot product exceeds a threshold. the implementation, using current mirrors for multiplication with fixed constants, permits the integrated circuit to achieve large amounts of computation per unit area. this apparatus permits a large input data bandwidth, and by virtue of having enough computation capacity to complete a processing task on one chip, the output bandwidth is greatly reduced as well. the apparatus is employed, as an example, in a neural network. a set of connections between nodes that modify the value of the signal passed from one node to the next. often many connections impinge on a node, and the summation of values at the node is further modified by a nonlinear function such as a threshold and amplitude limiter. values at the input nodes represent the signals to be evaluated by the network, and values at the outputs represent an evaluation by the network of the input signals. for instance, the input could be image pixels and the outputs could represent possible patterns to which the image could be assigned. the connections between weights are often determined and modified by training data, but they can also be prespecified in total or in part based on other information about the task of the network. dated 1993-11-16"
5263107,receptive field neural network with shift-invariant pattern recognition,"a neural network system and method of operating same wherein input data are initialized, then mapped onto a predetermined array for learning or recognition. the mapped information is divided into sub-input data or receptive fields, which are used for comparison of the input information with prelearned information having similar features, thereby allowing for correct classification of the input information. the receptive fields are shifted before the classification process, in order to generate a closest match between features which may be shifted at the time of input, and weights of the input information are updated based upon the closest-matching shifted position of the input information.",1993-11-16,"The title of the patent is receptive field neural network with shift-invariant pattern recognition and its abstract is a neural network system and method of operating same wherein input data are initialized, then mapped onto a predetermined array for learning or recognition. the mapped information is divided into sub-input data or receptive fields, which are used for comparison of the input information with prelearned information having similar features, thereby allowing for correct classification of the input information. the receptive fields are shifted before the classification process, in order to generate a closest match between features which may be shifted at the time of input, and weights of the input information are updated based upon the closest-matching shifted position of the input information. dated 1993-11-16"
5263121,neural network solution for interconnection apparatus,a neural network solution for routing calls through a three stage interconnection network selects an open path through the interconnection network if one exists. the neural network solution uses a neural network with a binary threshold. the weights of the neural network are fixed for all time and therefore are independent of the current state of the interconnection network. preferential call placement strategies are implemented by selecting appropriate external inputs to the neural network. an interconnection network controller stores information reflecting the current usage of the interconnection network and interfaces between the interconnection network and the neural network.,1993-11-16,The title of the patent is neural network solution for interconnection apparatus and its abstract is a neural network solution for routing calls through a three stage interconnection network selects an open path through the interconnection network if one exists. the neural network solution uses a neural network with a binary threshold. the weights of the neural network are fixed for all time and therefore are independent of the current state of the interconnection network. preferential call placement strategies are implemented by selecting appropriate external inputs to the neural network. an interconnection network controller stores information reflecting the current usage of the interconnection network and interfaces between the interconnection network and the neural network. dated 1993-11-16
5263122,neural network architecture,a frequency-based neural network in which the state of a neuron is indicated by the frequency of an impulse stream emitted by the neuron uses an interconnectivity structure employing a frequency-modulation multiplexing scheme to weight and communicate the pulse stream from an emitting neuron to receiving neurons in another network level.,1993-11-16,The title of the patent is neural network architecture and its abstract is a frequency-based neural network in which the state of a neuron is indicated by the frequency of an impulse stream emitted by the neuron uses an interconnectivity structure employing a frequency-modulation multiplexing scheme to weight and communicate the pulse stream from an emitting neuron to receiving neurons in another network level. dated 1993-11-16
5264734,difference calculating neural network utilizing switched capacitors,a difference calculating neural network is disclosed having an array of synapse cells arranged in rows and columns along pairs of row and column lines. the cells include a pair of floating gate devices which have their control gates coupled to receive one of a pair of complementary input voltages. the floating gate devices also have complementary threshold voltages such that packets of charge are produced from the synapse cells that are proportional to the difference between the input and voltage threshold. the charge packets are accumulated by the pairs of column lines in the array.,1993-11-23,The title of the patent is difference calculating neural network utilizing switched capacitors and its abstract is a difference calculating neural network is disclosed having an array of synapse cells arranged in rows and columns along pairs of row and column lines. the cells include a pair of floating gate devices which have their control gates coupled to receive one of a pair of complementary input voltages. the floating gate devices also have complementary threshold voltages such that packets of charge are produced from the synapse cells that are proportional to the difference between the input and voltage threshold. the charge packets are accumulated by the pairs of column lines in the array. dated 1993-11-23
5265192,method for the automated editing of seismic traces using an adaptive network,"an adaptive, or neural, network and a method of operating the same is disclosed which is particularly adapted for performing seismic trace editing for seismic shot records. the adaptive network is first trained according to the generalized delta rule. the disclosed training method includes backpropagation is performed according to the worst case error trace, including adjustment of the learning and momentum factors to increase as the worst case error decreases. slow convergence regions are detected, and methods applied to escape such regions including restoration of previously trimmed dormant links, renormalization of the weighting factor values, and the addition of new network layers with links between nodes that skip the hidden layer. after the training of the network, data corresponding to a discrete fast fourier transform of each trace, and to certain other attributes of the trace and adjacent traces thereto, are presented to the network. the network classifies the trace as good or noisy according to the inputs thereto, and to the weighting factors therewithin, such classification useful for ignoring noisy traces in subsequent data analysis. the analysis may be repeated for all of the traces in the shot record, and in multiple shot records.",1993-11-23,"The title of the patent is method for the automated editing of seismic traces using an adaptive network and its abstract is an adaptive, or neural, network and a method of operating the same is disclosed which is particularly adapted for performing seismic trace editing for seismic shot records. the adaptive network is first trained according to the generalized delta rule. the disclosed training method includes backpropagation is performed according to the worst case error trace, including adjustment of the learning and momentum factors to increase as the worst case error decreases. slow convergence regions are detected, and methods applied to escape such regions including restoration of previously trimmed dormant links, renormalization of the weighting factor values, and the addition of new network layers with links between nodes that skip the hidden layer. after the training of the network, data corresponding to a discrete fast fourier transform of each trace, and to certain other attributes of the trace and adjacent traces thereto, are presented to the network. the network classifies the trace as good or noisy according to the inputs thereto, and to the weighting factors therewithin, such classification useful for ignoring noisy traces in subsequent data analysis. the analysis may be repeated for all of the traces in the shot record, and in multiple shot records. dated 1993-11-23"
5267151,method and apparatus for detecting and identifying a condition,"a method and apparatus for sensing and classifying a condition of interest in a system from background noise in which a parameter representative of the condition of interest is sensed and an electrical signal representative of the sensed parameter is produced. the electrical signal is converted into a digital signal, this digital signal containing a signal of interest representative of the condition of interest and background noise. the digital signal is received by an artificial neural network which filters out the background noise to produce a filtered signal from the digital signal, and classifies the signal of interest from the filtered signal to produce an output representative of the classified signal.",1993-11-30,"The title of the patent is method and apparatus for detecting and identifying a condition and its abstract is a method and apparatus for sensing and classifying a condition of interest in a system from background noise in which a parameter representative of the condition of interest is sensed and an electrical signal representative of the sensed parameter is produced. the electrical signal is converted into a digital signal, this digital signal containing a signal of interest representative of the condition of interest and background noise. the digital signal is received by an artificial neural network which filters out the background noise to produce a filtered signal from the digital signal, and classifies the signal of interest from the filtered signal to produce an output representative of the classified signal. dated 1993-11-30"
5267165,data processing device and method for selecting data words contained in a dictionary,"a data processing device for selecting data words which are contained in a dictionary and which are nearest to a data word to be processed according to a correspondence criterion. the device includes: first apparatus for segmenting the space enclosing the assembly of data words of the dictionary; second apparatus for generating, for each segment, sub-dictionaries by making an arbitrary segment correspond, in accordance with the correspondence criterion, to words of a sub-dictionary; third apparatus for utilising the sub-dictionaries by determining, for an arbitrary data word to be processed, the segment with which it is associated, followed by determination, in accordance with the correspondence criterion, of that word or words among the words of the sub-dictionary associated with the segment which corresponds (correspond) best to the arbitrary data word to be processed. segmentation can be realised by means of a layered or tree-like neural network. the device may be used for data compression or data classification.",1993-11-30,"The title of the patent is data processing device and method for selecting data words contained in a dictionary and its abstract is a data processing device for selecting data words which are contained in a dictionary and which are nearest to a data word to be processed according to a correspondence criterion. the device includes: first apparatus for segmenting the space enclosing the assembly of data words of the dictionary; second apparatus for generating, for each segment, sub-dictionaries by making an arbitrary segment correspond, in accordance with the correspondence criterion, to words of a sub-dictionary; third apparatus for utilising the sub-dictionaries by determining, for an arbitrary data word to be processed, the segment with which it is associated, followed by determination, in accordance with the correspondence criterion, of that word or words among the words of the sub-dictionary associated with the segment which corresponds (correspond) best to the arbitrary data word to be processed. segmentation can be realised by means of a layered or tree-like neural network. the device may be used for data compression or data classification. dated 1993-11-30"
5267347,information processing element,"an information processing element for processing information with a function of neural network includes a semiconductor integrated circuit element portion comprising a plurality of neuron circuit regions constituting a neuron function among the neural network function, a molecular film element having a light-electricity function, provided on the circuit element portion, and the combination between the plurality of neurons is realized by utilizing a photoconductivity property of the molecular film element.",1993-11-30,"The title of the patent is information processing element and its abstract is an information processing element for processing information with a function of neural network includes a semiconductor integrated circuit element portion comprising a plurality of neuron circuit regions constituting a neuron function among the neural network function, a molecular film element having a light-electricity function, provided on the circuit element portion, and the combination between the plurality of neurons is realized by utilizing a photoconductivity property of the molecular film element. dated 1993-11-30"
5267502,weapons systems future muzzle velocity neural network,"in a device and method for predicting a future muzzle velocity of an indirect fire weapon 3, 7 means 9, 11 responsive to a measurement of muzzle velocity are adapted to implement an adaptive empirical prediction method to predict the future muzzle velocity. the invention also relates to an aiming system and method for an indirect-fire weapon 3, 7. the system comprises a muzzle velocity measuring device 5, and predictor means 9, 11 responsive to an output of the muzzle velocity measuring device 5 for determining a new elevation setting from the weapon. preferably, the predictor means utilizes an adaptive empirical prediction method such as a kalman filter or neural network.",1993-12-07,"The title of the patent is weapons systems future muzzle velocity neural network and its abstract is in a device and method for predicting a future muzzle velocity of an indirect fire weapon 3, 7 means 9, 11 responsive to a measurement of muzzle velocity are adapted to implement an adaptive empirical prediction method to predict the future muzzle velocity. the invention also relates to an aiming system and method for an indirect-fire weapon 3, 7. the system comprises a muzzle velocity measuring device 5, and predictor means 9, 11 responsive to an output of the muzzle velocity measuring device 5 for determining a new elevation setting from the weapon. preferably, the predictor means utilizes an adaptive empirical prediction method such as a kalman filter or neural network. dated 1993-12-07"
5268320,method of increasing the accuracy of an analog circuit employing floating gate memory devices,"a method for increasing the accuracy of an analog neural network which computers a sum-of-products between an input vector and a stored weight pattern is described. in one embodiment of the present invention, the method comprises initially training the network by programming the synapses with a certain weight pattern. the training may be carried out using any standard learning algorithm. preferably, a back-propagation learning algorithm is employed. next, the network is baked at an elevated temperature to effectuate a change in the weight pattern previously programmed during initial training. this change results from a charge redistribution which occurs within each of the synapses of the network. after baking, the network is then retrained to compensate for the change resulting from the charge redistribution. the baking and retraining steps may be successively repeated to increase the accuracy of the neural network to any desired level.",1993-12-07,"The title of the patent is method of increasing the accuracy of an analog circuit employing floating gate memory devices and its abstract is a method for increasing the accuracy of an analog neural network which computers a sum-of-products between an input vector and a stored weight pattern is described. in one embodiment of the present invention, the method comprises initially training the network by programming the synapses with a certain weight pattern. the training may be carried out using any standard learning algorithm. preferably, a back-propagation learning algorithm is employed. next, the network is baked at an elevated temperature to effectuate a change in the weight pattern previously programmed during initial training. this change results from a charge redistribution which occurs within each of the synapses of the network. after baking, the network is then retrained to compensate for the change resulting from the charge redistribution. the baking and retraining steps may be successively repeated to increase the accuracy of the neural network to any desired level. dated 1993-12-07"
5268684,apparatus for a neural network one-out-of-n encoder/decoder,"an artificial network for encoding the binary on-state of one-out-of-n inputs, say j, when only one state is on at a time wherein the jth on-state is represented by a suitable output level of an n-input mp type neuron operating in the non-saturated region of the neuron output nonlinearity. a single line transmits the encoded amplitude level signal to a decoder having n single input neural networks. the n outputs of the decoder are in the off-state except for the output corresponding to the active input node of the encoder",1993-12-07,"The title of the patent is apparatus for a neural network one-out-of-n encoder/decoder and its abstract is an artificial network for encoding the binary on-state of one-out-of-n inputs, say j, when only one state is on at a time wherein the jth on-state is represented by a suitable output level of an n-input mp type neuron operating in the non-saturated region of the neuron output nonlinearity. a single line transmits the encoded amplitude level signal to a decoder having n single input neural networks. the n outputs of the decoder are in the off-state except for the output corresponding to the active input node of the encoder dated 1993-12-07"
5268834,stable adaptive neural network controller,"an adaptive control system uses a neural network to provide adaptive control when the plant is operating within a normal operating range, but shifts to other types of control as the plant operating conditions move outside of the normal operating range. the controller uses a structure which allows the neural network parameters to be determined from minimal information about plant structure and the neural network is trained on-line during normal plant operation. the resulting system can be proven to be stable over all possible conditions. further, with the inventive techniques, the tracking accuracy can be controlled by appropriate network design.",1993-12-07,"The title of the patent is stable adaptive neural network controller and its abstract is an adaptive control system uses a neural network to provide adaptive control when the plant is operating within a normal operating range, but shifts to other types of control as the plant operating conditions move outside of the normal operating range. the controller uses a structure which allows the neural network parameters to be determined from minimal information about plant structure and the neural network is trained on-line during normal plant operation. the resulting system can be proven to be stable over all possible conditions. further, with the inventive techniques, the tracking accuracy can be controlled by appropriate network design. dated 1993-12-07"
5270950,apparatus and a method for locating a source of acoustic emission in a material,"an apparatus for locating a source of acoustic emission in a material comprises four spaced transducers coupled to the material. each transducer produces an output signal corresponding to a detected acoustic emission activity, and each output signal is amplified, rectified and enveloped before being supplied to a processor. artificially induced acoustic emission events, of known location, are generated in the material. the processor measures the times taken for each output signal corresponding to artificially induced acoustic emission events, to exceed two predetermined amplitudes from a datum time. a neural network analyzes the measured times to exceed the predetermined amplitudes for the output signals corresponding to the artificially induced acoustic emission events and infers the mathematical relationship between values of time and location of acoustic emission event. the times taken for each output signal, corresponding to acoustic emission events of unknown source location, to exceed two predetermined amplitudes from the datum are measured and are used to calculate the location of the unknown source with the mathematical relationship deduced by the neural network.",1993-12-14,"The title of the patent is apparatus and a method for locating a source of acoustic emission in a material and its abstract is an apparatus for locating a source of acoustic emission in a material comprises four spaced transducers coupled to the material. each transducer produces an output signal corresponding to a detected acoustic emission activity, and each output signal is amplified, rectified and enveloped before being supplied to a processor. artificially induced acoustic emission events, of known location, are generated in the material. the processor measures the times taken for each output signal corresponding to artificially induced acoustic emission events, to exceed two predetermined amplitudes from a datum time. a neural network analyzes the measured times to exceed the predetermined amplitudes for the output signals corresponding to the artificially induced acoustic emission events and infers the mathematical relationship between values of time and location of acoustic emission event. the times taken for each output signal, corresponding to acoustic emission events of unknown source location, to exceed two predetermined amplitudes from the datum are measured and are used to calculate the location of the unknown source with the mathematical relationship deduced by the neural network. dated 1993-12-14"
5271090,operational speed improvement for neural network,"higher operational speed is obtained without sacrificing computational accuracy and reliability in a neural network by interchanging a computationally complex nonlinear function with a similar but less complex nonlinear function in each neuron or computational element after each neuron of the network has been trained by an appropriate training algorithm for the classifying problem addressed by the neural network. in one exemplary embodiment, a hyperbolic tangent function is replaced by a piecewise linear threshold logic function.",1993-12-14,"The title of the patent is operational speed improvement for neural network and its abstract is higher operational speed is obtained without sacrificing computational accuracy and reliability in a neural network by interchanging a computationally complex nonlinear function with a similar but less complex nonlinear function in each neuron or computational element after each neuron of the network has been trained by an appropriate training algorithm for the classifying problem addressed by the neural network. in one exemplary embodiment, a hyperbolic tangent function is replaced by a piecewise linear threshold logic function. dated 1993-12-14"
5272723,waveform equalizer using a neural network,"a waveform equalizer for equalizing a distorted signal, contains a sampling unit, a time series generating unit, and an equalization neural network unit. the sampling unit samples the level of a distorted signal at a predetermined rate. the time series generating unit serially receives the sampled level and outputs in parallel a predetermined number of the levels which have been last received. the equalization neural network unit receives the outputs of the time series generating unit, and generates an equalized signal of the distorted signal based on the outputs of the time series generating unit using a set of equalization network weights which are preset therein. the waveform equalizer may further contain a distortion characteristic detecting unit, an equalization network weight holding unit, and a selector unit. the distortion characteristic detecting unit detects a distortion characteristic of the distorted signal. the equalization network weight holding unit holds a plurality of sets of equalization network weights each for being set in the equalization neural network unit. the selector unit selects one of the plurality of sets of equalization network weights according to the distortion characteristic which is detected in the distortion characteristic detecting unit, and supplies the selected set in the equalization neural network unit to set the selected set therein.",1993-12-21,"The title of the patent is waveform equalizer using a neural network and its abstract is a waveform equalizer for equalizing a distorted signal, contains a sampling unit, a time series generating unit, and an equalization neural network unit. the sampling unit samples the level of a distorted signal at a predetermined rate. the time series generating unit serially receives the sampled level and outputs in parallel a predetermined number of the levels which have been last received. the equalization neural network unit receives the outputs of the time series generating unit, and generates an equalized signal of the distorted signal based on the outputs of the time series generating unit using a set of equalization network weights which are preset therein. the waveform equalizer may further contain a distortion characteristic detecting unit, an equalization network weight holding unit, and a selector unit. the distortion characteristic detecting unit detects a distortion characteristic of the distorted signal. the equalization network weight holding unit holds a plurality of sets of equalization network weights each for being set in the equalization neural network unit. the selector unit selects one of the plurality of sets of equalization network weights according to the distortion characteristic which is detected in the distortion characteristic detecting unit, and supplies the selected set in the equalization neural network unit to set the selected set therein. dated 1993-12-21"
5274714,method and apparatus for determining and organizing feature vectors for neural network recognition,"a pattern recognition method and apparatus utilizes a neural network to recognize input images which are sufficiently similar to a database of previously stored images. images are first processed and subjected to a fourier transform which yields a power spectrum. an in-class to out-of-class study is performed on a typical collection of images in order to determine the most discriminatory regions of the fourier transform. a feature vector consisting of the (most discriminatory) information from the power spectrum of the fourier transform of the image is formed. feature vectors are input to a neural network having preferably two hidden layers, input dimensionality of the number of elements in the feature vector and output dimensionality of the number of data elements stored in the database. unique identifier numbers are preferably stored along with the feature vector. application of a query feature vector to the neural network results in an output vector. the output vector is subjected to statistical analysis to determine if a sufficiently high confidence level exists to indicate a successful identification whereupon a unique identifier number may be displayed.",1993-12-28,"The title of the patent is method and apparatus for determining and organizing feature vectors for neural network recognition and its abstract is a pattern recognition method and apparatus utilizes a neural network to recognize input images which are sufficiently similar to a database of previously stored images. images are first processed and subjected to a fourier transform which yields a power spectrum. an in-class to out-of-class study is performed on a typical collection of images in order to determine the most discriminatory regions of the fourier transform. a feature vector consisting of the (most discriminatory) information from the power spectrum of the fourier transform of the image is formed. feature vectors are input to a neural network having preferably two hidden layers, input dimensionality of the number of elements in the feature vector and output dimensionality of the number of data elements stored in the database. unique identifier numbers are preferably stored along with the feature vector. application of a query feature vector to the neural network results in an output vector. the output vector is subjected to statistical analysis to determine if a sufficiently high confidence level exists to indicate a successful identification whereupon a unique identifier number may be displayed. dated 1993-12-28"
5274742,combination problem solving method and apparatus,"by using the state transition of a highly interconnected neural network, in order to solve a combination problem, an energy function is set by the following procedure: (i) the energy function is set in correspondence to the size of the combination problem; (ii) the energy function is set for a combination problem to be solved by using an energy function which solved another combination problem of a different size from the combination problem to be solved. also, in order to solve a problem involving the cutting out a specific image from a whole image, as a combination problem when obtaining pixels representing a contour of an object, the energy function is set by either (i) or (ii) above.",1993-12-28,"The title of the patent is combination problem solving method and apparatus and its abstract is by using the state transition of a highly interconnected neural network, in order to solve a combination problem, an energy function is set by the following procedure: (i) the energy function is set in correspondence to the size of the combination problem; (ii) the energy function is set for a combination problem to be solved by using an energy function which solved another combination problem of a different size from the combination problem to be solved. also, in order to solve a problem involving the cutting out a specific image from a whole image, as a combination problem when obtaining pixels representing a contour of an object, the energy function is set by either (i) or (ii) above. dated 1993-12-28"
5274744,neural network for performing a relaxation process,"in accordance with the present invention, a neural network comprising an array of neurons (i.e. processing nodes) interconnected by synapses (i.e. weighted transmission links) is utilized to carry out a probabilistic relaxation process. the inventive neural network is especially suited for carrying out a variety of image processing tasks such as thresholding.",1993-12-28,"The title of the patent is neural network for performing a relaxation process and its abstract is in accordance with the present invention, a neural network comprising an array of neurons (i.e. processing nodes) interconnected by synapses (i.e. weighted transmission links) is utilized to carry out a probabilistic relaxation process. the inventive neural network is especially suited for carrying out a variety of image processing tasks such as thresholding. dated 1993-12-28"
5274745,method of processing information in artificial neural networks,"a method of processing information in an artificial neural network including a plurality of artificial neurons and weighted links coupling the neurons. in the method, those of the artificial neurons whose output values change by a value greater than a threshold value are selected. the output values of the selected neurons are calculated, and the influence which the changes in the output values of the selected neurons impose on the input values of the other artificial neurons is computed. the threshold value is changed such that an appropriate number of neurons are selected. the information processing in the artificial neural network is stopped when the threshold value decreased below a predetermined small value and the values output by all artificial neurons change by a value equal to or less than the threshold value.",1993-12-28,"The title of the patent is method of processing information in artificial neural networks and its abstract is a method of processing information in an artificial neural network including a plurality of artificial neurons and weighted links coupling the neurons. in the method, those of the artificial neurons whose output values change by a value greater than a threshold value are selected. the output values of the selected neurons are calculated, and the influence which the changes in the output values of the selected neurons impose on the input values of the other artificial neurons is computed. the threshold value is changed such that an appropriate number of neurons are selected. the information processing in the artificial neural network is stopped when the threshold value decreased below a predetermined small value and the values output by all artificial neurons change by a value equal to or less than the threshold value. dated 1993-12-28"
5274746,coupling element for semiconductor neural network device,"a neural network device includes internal data input lines, internal data output lines, coupling elements provided at the connections of the internal data input lines and the internal data output lines, word lines each for selecting one row of coupling elements. the coupling elements couple, with specific programmable coupling strengths, the associated internal data input lines to the associated internal data output lines. in a program mode, the internal data output lines serve as signal lines for transmitting the coupling strength information. each of the coupling elements includes memories constituted of cross-coupled inverters for storing the coupling strength information, first switching transistors responsive to signal potentials on associated word lines for connecting the memories to associated internal data output lines, second switching elements responsive to signal potentials on associated internal data input lines for transmitting the storage information in the memories to the associated internal data output lines. each of the internal data output lines has a pair of first and second internal data output lines.",1993-12-28,"The title of the patent is coupling element for semiconductor neural network device and its abstract is a neural network device includes internal data input lines, internal data output lines, coupling elements provided at the connections of the internal data input lines and the internal data output lines, word lines each for selecting one row of coupling elements. the coupling elements couple, with specific programmable coupling strengths, the associated internal data input lines to the associated internal data output lines. in a program mode, the internal data output lines serve as signal lines for transmitting the coupling strength information. each of the coupling elements includes memories constituted of cross-coupled inverters for storing the coupling strength information, first switching transistors responsive to signal potentials on associated word lines for connecting the memories to associated internal data output lines, second switching elements responsive to signal potentials on associated internal data input lines for transmitting the storage information in the memories to the associated internal data output lines. each of the internal data output lines has a pair of first and second internal data output lines. dated 1993-12-28"
5274748,electronic synapse circuit for artificial neural network,"an electronic synapse circuit is disclosed for multiplying an analog weight signal value by a digital state signal value to achieve a signed product value as a current which is capable of being summed with other such synapse circuit outputs. the circuit employs a storage multiplying digital-to-analog converter which provides storage for the analog weight signal value. additional circuitry permits programming different analog weight signal values into the circuit, performing four-quadrant multiplication, generating a current summable output, and maintaining the stored analog weight signal value at a substantially constant value independent of the digital state signal values.",1993-12-28,"The title of the patent is electronic synapse circuit for artificial neural network and its abstract is an electronic synapse circuit is disclosed for multiplying an analog weight signal value by a digital state signal value to achieve a signed product value as a current which is capable of being summed with other such synapse circuit outputs. the circuit employs a storage multiplying digital-to-analog converter which provides storage for the analog weight signal value. additional circuitry permits programming different analog weight signal values into the circuit, performing four-quadrant multiplication, generating a current summable output, and maintaining the stored analog weight signal value at a substantially constant value independent of the digital state signal values. dated 1993-12-28"
5276769,neural network learning apparatus and method,"a learning apparatus for use in a neural network system which has a plurality of classes representing different meanings. the learning apparatus is provided for learning a number of different patterns, inputted by input vectors, and classified in different classes. the learning apparatus is constructed by a computer and it includes a section for producing a plurality of output vectors representing different classes in response to an input vector, a section for obtaining a first largest output vector of all the output vectors, a section for obtaining a second largest output vector of all the output vectors, and a section for setting predetermined weights to the first and second largest output vectors, respectively, such that the first largest output vector is made larger, and the second largest output vector is made smaller. furthermore, a section for determining a ratio of the weighted first and second largest output vectors, respectively, is included. if the determined ratio is smaller than a predetermined value, the weighted first and second largest output vectors are further weighted to be made further larger and smaller, respectively.",1994-01-04,"The title of the patent is neural network learning apparatus and method and its abstract is a learning apparatus for use in a neural network system which has a plurality of classes representing different meanings. the learning apparatus is provided for learning a number of different patterns, inputted by input vectors, and classified in different classes. the learning apparatus is constructed by a computer and it includes a section for producing a plurality of output vectors representing different classes in response to an input vector, a section for obtaining a first largest output vector of all the output vectors, a section for obtaining a second largest output vector of all the output vectors, and a section for setting predetermined weights to the first and second largest output vectors, respectively, such that the first largest output vector is made larger, and the second largest output vector is made smaller. furthermore, a section for determining a ratio of the weighted first and second largest output vectors, respectively, is included. if the determined ratio is smaller than a predetermined value, the weighted first and second largest output vectors are further weighted to be made further larger and smaller, respectively. dated 1994-01-04"
5276770,training of neural network for multi-source data fusion,"a method of training a multilayer perceptron type neural network to provide a processor for fusion of target angle data detected by a plurality of sensors. the neural network includes a layer of input neurons at least equal in number to the number of sensors plus the maximum number of targets, at least one layer of inner neurons, and a plurality of output neurons forming an output layer. each neuron is connected to every neuron in adjacent layers by adjustable weighted synaptic connections. the method of training comprises the steps of (a) for each sensor, designing a plurality of the input neurons for receiving any target angle data, the number of designated input neurons for each sensor being at least as large as the maximum number of targets to be detected by the sensor; (b) for a known set of targets having a known target angle for each sensor, applying a signal related to each known target angle to the designated input neurons for each of the sensors, wherein the output neurons will produce an initial output; (c) for a selected one of the sensors, designating a plurality of the output neurons to correspond to the input neurons designated for the selected sensor and applying the signal related to the known target angles for the selected sensor to the designated output neurons to provide a designated output signal wherein the difference between the initial output and the designated output signal is used to adapt the weights throughout the neural network to provide an adjusted output signal; and (d) repeating steps (a)-(c) until the adjusted output signal corresponds to a desired output signal.",1994-01-04,"The title of the patent is training of neural network for multi-source data fusion and its abstract is a method of training a multilayer perceptron type neural network to provide a processor for fusion of target angle data detected by a plurality of sensors. the neural network includes a layer of input neurons at least equal in number to the number of sensors plus the maximum number of targets, at least one layer of inner neurons, and a plurality of output neurons forming an output layer. each neuron is connected to every neuron in adjacent layers by adjustable weighted synaptic connections. the method of training comprises the steps of (a) for each sensor, designing a plurality of the input neurons for receiving any target angle data, the number of designated input neurons for each sensor being at least as large as the maximum number of targets to be detected by the sensor; (b) for a known set of targets having a known target angle for each sensor, applying a signal related to each known target angle to the designated input neurons for each of the sensors, wherein the output neurons will produce an initial output; (c) for a selected one of the sensors, designating a plurality of the output neurons to correspond to the input neurons designated for the selected sensor and applying the signal related to the known target angles for the selected sensor to the designated output neurons to provide a designated output signal wherein the difference between the initial output and the designated output signal is used to adapt the weights throughout the neural network to provide an adjusted output signal; and (d) repeating steps (a)-(c) until the adjusted output signal corresponds to a desired output signal. dated 1994-01-04"
5276771,rapidly converging projective neural network,"a data processing system and method for solving pattern classification problems and function-fitting problems includes a neural network in which n-dimensional input vectors are augmented with at least one element to form an n+j-dimensional projected input vector, whose magnitude is then preferably normalized to lie on the surface of a hypersphere. weight vectors of at least a lowest intermediate layer of network nodes are preferably also constrained to lie on the n+j-dimensional surface. to train the network, the system compares network output values with known goal vectors, and an error function (which depends on all weights and threshold values of the intermediate and output nodes) is then minimized. in order to decrease the network's learning time even further, the weight vectors for the intermediate nodes are initially preferably set equal to known prototypes for the various classes of input vectors. furthermore, the invention also allows separation of the network into sub-networks, which are then trained individually and later recombined. the network is able to use both hyperspheres and hyperplanes to form decision boundaries, and, indeed, can converge to the one even if it initially assumes the other.",1994-01-04,"The title of the patent is rapidly converging projective neural network and its abstract is a data processing system and method for solving pattern classification problems and function-fitting problems includes a neural network in which n-dimensional input vectors are augmented with at least one element to form an n+j-dimensional projected input vector, whose magnitude is then preferably normalized to lie on the surface of a hypersphere. weight vectors of at least a lowest intermediate layer of network nodes are preferably also constrained to lie on the n+j-dimensional surface. to train the network, the system compares network output values with known goal vectors, and an error function (which depends on all weights and threshold values of the intermediate and output nodes) is then minimized. in order to decrease the network's learning time even further, the weight vectors for the intermediate nodes are initially preferably set equal to known prototypes for the various classes of input vectors. furthermore, the invention also allows separation of the network into sub-networks, which are then trained individually and later recombined. the network is able to use both hyperspheres and hyperplanes to form decision boundaries, and, indeed, can converge to the one even if it initially assumes the other. dated 1994-01-04"
5276772,real time adaptive probabilistic neural network system and method for data sorting,"an adaptive probabilistic neural network (apnn) includes a cluster processor circuit which generates a signal which represents a probability density function estimation value which is used to sort input pulse parameter data signals based upon a probability of obtaining a correct match with a group of input pulse parameter data signals that have already been sorted. in the apnn system, a pulse buffer memory circuit is contained within the cluster processor circuit and temporarily stores the assigned input pulse parameter data signals. the pulse buffer memory circuit is initially empty. as the input pulse parameter data signals are presented to the apnn, the system sorts the incoming data signals based on the probability density function estimation value signal generated by each currently operating cluster processor circuit. the current input pulse parameter data signal is sorted and stored in the pulse buffer memory circuit of the cluster processor circuit. a small probability density function estimation value signal indicates the current unassigned input pulse parameter data signal is not recognized by the apnn system. a large probability density function estimation value signal indicates a match and the current input pulse parameter data signal will be included within a particular cluster processor circuit.",1994-01-04,"The title of the patent is real time adaptive probabilistic neural network system and method for data sorting and its abstract is an adaptive probabilistic neural network (apnn) includes a cluster processor circuit which generates a signal which represents a probability density function estimation value which is used to sort input pulse parameter data signals based upon a probability of obtaining a correct match with a group of input pulse parameter data signals that have already been sorted. in the apnn system, a pulse buffer memory circuit is contained within the cluster processor circuit and temporarily stores the assigned input pulse parameter data signals. the pulse buffer memory circuit is initially empty. as the input pulse parameter data signals are presented to the apnn, the system sorts the incoming data signals based on the probability density function estimation value signal generated by each currently operating cluster processor circuit. the current input pulse parameter data signal is sorted and stored in the pulse buffer memory circuit of the cluster processor circuit. a small probability density function estimation value signal indicates the current unassigned input pulse parameter data signal is not recognized by the apnn system. a large probability density function estimation value signal indicates a match and the current input pulse parameter data signal will be included within a particular cluster processor circuit. dated 1994-01-04"
5276773,digital neural network executed in integrated circuit technology,"a digital neural network has a plurality of neurons (nr) completely meshed with one another, each of which comprises an evaluation stage having a plurality of evaluators (b) that is equal in number to the plurality of neurons (nr) and each of which comprises a decision stage having a decision unit (e). an adjustment information (inf.sub.e) that effects a defined pre-adjustment of the decision unit (e) can be supplied to every decision unit (e) by a pre-processing means via an information input. a weighting information (inf.sub.g) can be supplied to every evaluator (b) by a pre-processing means via an individual information input. an output information (inf.sub.a) can be output by every decision unit (e) to a post-processing means via a respective individual information output. the information outputs of the decision units (e) are each connected to an individual processing input of all evaluators (b) allocated to the appertaining decision unit (e). individual processing outputs of the evaluators (b) are connected to individual processing inputs of the decision unit (e) in the appertaining neuron (n), so that every output information (inf.sub.a) can be indirectly fed back onto every neuron (nr).",1994-01-04,"The title of the patent is digital neural network executed in integrated circuit technology and its abstract is a digital neural network has a plurality of neurons (nr) completely meshed with one another, each of which comprises an evaluation stage having a plurality of evaluators (b) that is equal in number to the plurality of neurons (nr) and each of which comprises a decision stage having a decision unit (e). an adjustment information (inf.sub.e) that effects a defined pre-adjustment of the decision unit (e) can be supplied to every decision unit (e) by a pre-processing means via an information input. a weighting information (inf.sub.g) can be supplied to every evaluator (b) by a pre-processing means via an individual information input. an output information (inf.sub.a) can be output by every decision unit (e) to a post-processing means via a respective individual information output. the information outputs of the decision units (e) are each connected to an individual processing input of all evaluators (b) allocated to the appertaining decision unit (e). individual processing outputs of the evaluators (b) are connected to individual processing inputs of the decision unit (e) in the appertaining neuron (n), so that every output information (inf.sub.a) can be indirectly fed back onto every neuron (nr). dated 1994-01-04"
5278755,method for determining image points in object images using neural networks,"an image point located in the region inside of an object image is determined from an image signal made up of a series of image signal components representing respective picture elements in a radiation image, which includes the object image and which has been recorded on a recording medium in accordance with a predetermined image recording menu. a plurality of different neural networks are prepared for a plurality of different image recording menus. each of the neural networks receives an image signal and generates outputs which represent an image point. a neural network, which is optimum for the predetermined image recording menu, is selected from the plurality of the neural networks. outputs, which represent the image point located in the region inside of the object image, are then obtained from the selected neural network.",1994-01-11,"The title of the patent is method for determining image points in object images using neural networks and its abstract is an image point located in the region inside of an object image is determined from an image signal made up of a series of image signal components representing respective picture elements in a radiation image, which includes the object image and which has been recorded on a recording medium in accordance with a predetermined image recording menu. a plurality of different neural networks are prepared for a plurality of different image recording menus. each of the neural networks receives an image signal and generates outputs which represent an image point. a neural network, which is optimum for the predetermined image recording menu, is selected from the plurality of the neural networks. outputs, which represent the image point located in the region inside of the object image, are then obtained from the selected neural network. dated 1994-01-11"
5278945,neural processor apparatus,"a neural processor apparatus implements a neural network at a low cost and with high efficiency by simultaneously processing a plurality of neurons using the same synaptic inputs. weight data is sequentially accessed from an external weight ram memory to minimize space on the ic. the input data and weight data may be configured as either a single, high-resolution input or a plurality of inputs having a lower resolution, whereby the plurality of inputs are processed simultaneously. a dynamic approximation method is implemented using a minimal amount of circuitry to provide high-resolution transformations in accordance with the transfer function of a given neuron model. the neural processor apparatus may be used to implement an entire neural network, or may be implemented using a plurality of devices, each device implementing a predetermined number of neural layers.",1994-01-11,"The title of the patent is neural processor apparatus and its abstract is a neural processor apparatus implements a neural network at a low cost and with high efficiency by simultaneously processing a plurality of neurons using the same synaptic inputs. weight data is sequentially accessed from an external weight ram memory to minimize space on the ic. the input data and weight data may be configured as either a single, high-resolution input or a plurality of inputs having a lower resolution, whereby the plurality of inputs are processed simultaneously. a dynamic approximation method is implemented using a minimal amount of circuitry to provide high-resolution transformations in accordance with the transfer function of a given neuron model. the neural processor apparatus may be used to implement an entire neural network, or may be implemented using a plurality of devices, each device implementing a predetermined number of neural layers. dated 1994-01-11"
5280564,neural network having an optimized transfer function for each neuron,"the characteristic data for determining the characteristics of the transfer functions (for example, sigmoid functions) of the neurons of the hidden layer and the output layer (the gradients of the sigmoid functions) of a neural network are learned and corrected in a manner similar to the correction of weighting data and threshold values. since at least one characteristic data which determines the characteristics of the transfer function of each neuron is learned, the transfer function characteristics can be different for different neurons in the network independently of the problem and/or the number of neurons, and be optimum. accordingly, a learning with high precision can be performed in a short time.",1994-01-18,"The title of the patent is neural network having an optimized transfer function for each neuron and its abstract is the characteristic data for determining the characteristics of the transfer functions (for example, sigmoid functions) of the neurons of the hidden layer and the output layer (the gradients of the sigmoid functions) of a neural network are learned and corrected in a manner similar to the correction of weighting data and threshold values. since at least one characteristic data which determines the characteristics of the transfer function of each neuron is learned, the transfer function characteristics can be different for different neurons in the network independently of the problem and/or the number of neurons, and be optimum. accordingly, a learning with high precision can be performed in a short time. dated 1994-01-18"
5280792,method and system for automatically classifying intracardiac electrograms,"the application is directed to a method for automatically classifying intracardiac electrograms, and a system for performing the method. in a further aspect, it concerns an implantable cardioverter defibrillator which incorporates the system and uses the method to monitor cardiac activity and deliver appropriate treatment. the method uses a combination of timing analysis and pattern matching using a neural network in order to correctly classify the electrograms. this technique allows both changes in rate and morphology to be taken into account.",1994-01-25,"The title of the patent is method and system for automatically classifying intracardiac electrograms and its abstract is the application is directed to a method for automatically classifying intracardiac electrograms, and a system for performing the method. in a further aspect, it concerns an implantable cardioverter defibrillator which incorporates the system and uses the method to monitor cardiac activity and deliver appropriate treatment. the method uses a combination of timing analysis and pattern matching using a neural network in order to correctly classify the electrograms. this technique allows both changes in rate and morphology to be taken into account. dated 1994-01-25"
5282131,control system for controlling a pulp washing system using a neural network controller,a control system for a countercurrent pulp washing process in which the pulp is formed as a pulp mat on at least one moving filter surface and the mat is supplied with rinse water to replace water in the pulp mat thereby reducing the soda loss in the mat before it is removed from the filter surface. the process is characterized by at least one predictable process variable including dissolved solids retained in the pulp mat. the system comprises a trainable neural network having a plurality of input neurons having input values applied thereto and output neurons for providing output values and means for training the neural network to provide predicted values for the predictable process variables.,1994-01-25,The title of the patent is control system for controlling a pulp washing system using a neural network controller and its abstract is a control system for a countercurrent pulp washing process in which the pulp is formed as a pulp mat on at least one moving filter surface and the mat is supplied with rinse water to replace water in the pulp mat thereby reducing the soda loss in the mat before it is removed from the filter surface. the process is characterized by at least one predictable process variable including dissolved solids retained in the pulp mat. the system comprises a trainable neural network having a plurality of input neurons having input values applied thereto and output neurons for providing output values and means for training the neural network to provide predicted values for the predictable process variables. dated 1994-01-25
5282261,neural network process measurement and control,"a computer neural network process measurement and control system and method uses real-time output data from a neural network to replace a sensor or laboratory input to a controller. the neural network can use readily available, inexpensive and reliable measurements from sensors as inputs, and produce predicted values of product properties as output data for input to the controller. the system and method overcome process deadtime, measurement deadtime, infrequent measurements, and measurement variability in laboratory data, thus providing improved control. an historical database can be used to provide a history of sensor and laboratory measurements to the neural network. the neural network can detect the appearance of new laboratory measurements in the history and automatically initiate retraining, on-line and in real-time. the system and method can use either a regulatory controller or a supervisory control architecture. a modular software implementation simplifies the building of multiple neural networks, and also optionally provides other control functions, such as supervisory controllers, expert systems, and statistical data filtering, thus allowing powerful extensions of the system and method. template specification for the neural network, and data specification using data pointers allow the system and method to be more easily implemented.",1994-01-25,"The title of the patent is neural network process measurement and control and its abstract is a computer neural network process measurement and control system and method uses real-time output data from a neural network to replace a sensor or laboratory input to a controller. the neural network can use readily available, inexpensive and reliable measurements from sensors as inputs, and produce predicted values of product properties as output data for input to the controller. the system and method overcome process deadtime, measurement deadtime, infrequent measurements, and measurement variability in laboratory data, thus providing improved control. an historical database can be used to provide a history of sensor and laboratory measurements to the neural network. the neural network can detect the appearance of new laboratory measurements in the history and automatically initiate retraining, on-line and in real-time. the system and method can use either a regulatory controller or a supervisory control architecture. a modular software implementation simplifies the building of multiple neural networks, and also optionally provides other control functions, such as supervisory controllers, expert systems, and statistical data filtering, thus allowing powerful extensions of the system and method. template specification for the neural network, and data specification using data pointers allow the system and method to be more easily implemented. dated 1994-01-25"
5283418,automated rotor welding processes using neural networks,"methods and apparatus for monitoring an arc welding process are disclosed. in a preferred embodiment, the present invention creates a digital representation of the arc created during welding and, using a neural network computer, determines if the arc is representative of normal or abnormal welding conditions. the neural network disclosed is trained to identify abnormal conditions and normal conditions and may be adaptively retrained to classify images that are not in the initial set of normal and abnormal images. in certain embodiments, other data, such as current, weld wire emission spectra, or shielding gas flow rate are also collected and the neural network is trained to monitor these data. also, in certain embodiments, an audio signal is collected from the vicinity of the welding process and is used by the neural network computer to further classify the arc as normal or abnormal. the present invention is most preferably implemented in repetitive and continuous welding operations, such as those encountered in the manufacture and rebuilding of steam turbines.",1994-02-01,"The title of the patent is automated rotor welding processes using neural networks and its abstract is methods and apparatus for monitoring an arc welding process are disclosed. in a preferred embodiment, the present invention creates a digital representation of the arc created during welding and, using a neural network computer, determines if the arc is representative of normal or abnormal welding conditions. the neural network disclosed is trained to identify abnormal conditions and normal conditions and may be adaptively retrained to classify images that are not in the initial set of normal and abnormal images. in certain embodiments, other data, such as current, weld wire emission spectra, or shielding gas flow rate are also collected and the neural network is trained to monitor these data. also, in certain embodiments, an audio signal is collected from the vicinity of the welding process and is used by the neural network computer to further classify the arc as normal or abnormal. the present invention is most preferably implemented in repetitive and continuous welding operations, such as those encountered in the manufacture and rebuilding of steam turbines. dated 1994-02-01"
5283746,manufacturing adjustment during article fabrication,"the use of neural networks has been employed to adjust processing during the fabrication of articles. for example, in the production of photolithographic masks by electron beam irradiation of a mask blank in a desired pattern, electrons scattered from the mask substrate cause distortion of the pattern. adjustment for such scattering is possible during the manufacturing process by employing an adjustment function determined by a neural network whose parameters are established relative to a prototypical mask pattern.",1994-02-01,"The title of the patent is manufacturing adjustment during article fabrication and its abstract is the use of neural networks has been employed to adjust processing during the fabrication of articles. for example, in the production of photolithographic masks by electron beam irradiation of a mask blank in a desired pattern, electrons scattered from the mask substrate cause distortion of the pattern. adjustment for such scattering is possible during the manufacturing process by employing an adjustment function determined by a neural network whose parameters are established relative to a prototypical mask pattern. dated 1994-02-01"
5283838,neural network apparatus,"when performing learning for a neural network, a plurality of learning vectors which belong to an arbitrary category are used, and self-organization learning in the category is carried out. as a result, the plurality of learning vectors which belong to the category are automatically clustered, and the contents of weight vectors in the neural network are set to representative vectors which exhibit common features of the learning vectors of each cluster. then, teacher-supervised learning is carried out for the neural network, using the thus set contents of the weight vectors as initial values thereof. in the learning process, an initial value of each weight vector is set to the representative vector of each cluster obtained by clustering. therefore, the number of calculations required until the teacher-supervised learning is converged is greatly reduced.",1994-02-01,"The title of the patent is neural network apparatus and its abstract is when performing learning for a neural network, a plurality of learning vectors which belong to an arbitrary category are used, and self-organization learning in the category is carried out. as a result, the plurality of learning vectors which belong to the category are automatically clustered, and the contents of weight vectors in the neural network are set to representative vectors which exhibit common features of the learning vectors of each cluster. then, teacher-supervised learning is carried out for the neural network, using the thus set contents of the weight vectors as initial values thereof. in the learning process, an initial value of each weight vector is set to the representative vector of each cluster obtained by clustering. therefore, the number of calculations required until the teacher-supervised learning is converged is greatly reduced. dated 1994-02-01"
5283855,neural network and method for training the neural network,"a method and apparatus are disclosed that modify [ies] and generalize [s] the use in artificial neural networks of the error backpropagation algorithm. each neuron unit first divides a plurality of weighted inputs into more than one group, then sums up weighted inputs in each group to provide each group's intermediate outputs, and finally processes the intermediate outputs to produce an output of the neuron unit. since the method uses, when modifying each weight, a partial differential coefficient generated by partially-differentiating the output of the neuron unit by each weighted input, the weight can be properly modified even if the output of a neuron unit as a function of intermediate outputs has a plurality of variables corresponding to the number of groups. since the conventional method uses only one differential coefficient, that is, the differential coefficient of the output of a neuron unit differentiated by the sum of all weighted inputs in a neuron unit, for all weights in a neuron unit, it may be said that the method according to the present invention generalizes the conventional method. the present invention is especially useful for pulse density neural networks which express data as an on-bit density of a bit string.",1994-02-01,"The title of the patent is neural network and method for training the neural network and its abstract is a method and apparatus are disclosed that modify [ies] and generalize [s] the use in artificial neural networks of the error backpropagation algorithm. each neuron unit first divides a plurality of weighted inputs into more than one group, then sums up weighted inputs in each group to provide each group's intermediate outputs, and finally processes the intermediate outputs to produce an output of the neuron unit. since the method uses, when modifying each weight, a partial differential coefficient generated by partially-differentiating the output of the neuron unit by each weighted input, the weight can be properly modified even if the output of a neuron unit as a function of intermediate outputs has a plurality of variables corresponding to the number of groups. since the conventional method uses only one differential coefficient, that is, the differential coefficient of the output of a neuron unit differentiated by the sum of all weighted inputs in a neuron unit, for all weights in a neuron unit, it may be said that the method according to the present invention generalizes the conventional method. the present invention is especially useful for pulse density neural networks which express data as an on-bit density of a bit string. dated 1994-02-01"
5285297,apparatus and method for color calibration,""" a method and apparatus for constructing, training and utilizing an artificial neural network (also termed herein a """"neural network"""", an ann, or an nn) in order to transform a first color value in a first color coordinate system into a second color value in a second color coordinate system. """,1994-02-08,"The title of the patent is apparatus and method for color calibration and its abstract is "" a method and apparatus for constructing, training and utilizing an artificial neural network (also termed herein a """"neural network"""", an ann, or an nn) in order to transform a first color value in a first color coordinate system into a second color value in a second color coordinate system. "" dated 1994-02-08"
5285523,apparatus for recognizing driving environment of vehicle,"an apparatus for recognizing driving environments of a vehicle including a plurality of sensors for detecting various parameters relating to driving conditions of the vehicle such as throttle valve open angle, vehicle running speed, brake pedal depression amount and gear shift range of an automatic transmission, first and second neuron interfaces for converting parameter values detected by the sensors into a plurality of input patterns having predetermined configuration, first and second neural networks having input layers to which corresponding input patterns are applied, hidden layers and output layers for producing recognition results, and a multiplexer for selecting one of the recognition results produced on the output layers of the first and second neural networks. the first neural network has a superior separating or recognizing and learning faculty, while the second neural network has a superior associating faculty. a accelerating pedal depression amount is detected by a sensor and a variation of the thus detected amount is compared with a reference value. when the variation is larger than the reference value, the recognition result produced by the first neural network is selected and when the variation is smaller than the reference value, the recognition result from the second neural network is selected.",1994-02-08,"The title of the patent is apparatus for recognizing driving environment of vehicle and its abstract is an apparatus for recognizing driving environments of a vehicle including a plurality of sensors for detecting various parameters relating to driving conditions of the vehicle such as throttle valve open angle, vehicle running speed, brake pedal depression amount and gear shift range of an automatic transmission, first and second neuron interfaces for converting parameter values detected by the sensors into a plurality of input patterns having predetermined configuration, first and second neural networks having input layers to which corresponding input patterns are applied, hidden layers and output layers for producing recognition results, and a multiplexer for selecting one of the recognition results produced on the output layers of the first and second neural networks. the first neural network has a superior separating or recognizing and learning faculty, while the second neural network has a superior associating faculty. a accelerating pedal depression amount is detected by a sensor and a variation of the thus detected amount is compared with a reference value. when the variation is larger than the reference value, the recognition result produced by the first neural network is selected and when the variation is smaller than the reference value, the recognition result from the second neural network is selected. dated 1994-02-08"
5285524,neural network with daisy chain control,"the present invention is a direct digitally implemented network system in which neural nodes 24, 26 and 28 which output to the same destination node 22 in the network share the same channel 30. if a set of nodes does not output any data to any node to which a second set of nodes outputs data (the two sets of nodes to not overlap or intersect), the two sets of nodes are independent and do not share a channel and have separate channels 120 and 122. the network is configured as parallel operating non-intersecting segments or independent sets where each segment has a segment communication channel or bus 30. each node in the independent set or segment is sequentially activated to produce an output by a daisy chain control signal. the outputs are thereby time division multiplexed over the channel 30 to the destination node 22.",1994-02-08,"The title of the patent is neural network with daisy chain control and its abstract is the present invention is a direct digitally implemented network system in which neural nodes 24, 26 and 28 which output to the same destination node 22 in the network share the same channel 30. if a set of nodes does not output any data to any node to which a second set of nodes outputs data (the two sets of nodes to not overlap or intersect), the two sets of nodes are independent and do not share a channel and have separate channels 120 and 122. the network is configured as parallel operating non-intersecting segments or independent sets where each segment has a segment communication channel or bus 30. each node in the independent set or segment is sequentially activated to produce an output by a daisy chain control signal. the outputs are thereby time division multiplexed over the channel 30 to the destination node 22. dated 1994-02-08"
5286947,apparatus and method for monitoring material removal from a workpiece,"an apparatus and method for monitoring material removal from a workpiece by a beam of energy during a material processing operation are disclosed. a detector is positioned for sensing optical emissions from the workpiece caused by removal of material when an energy beam pulse is incident upon the surface of the workpiece. a computing circuit, algorithm or artificial neural network is provided for determining a quantity of material removed from the sensed optical emissions in real-time during the material processing operation. analysis of the optical emission pulses caused by the material removal provides an indication of the efficiency of the material processing system and provides feedback for manual or automatic adjustment of material processing parameters during the material processing operation.",1994-02-15,"The title of the patent is apparatus and method for monitoring material removal from a workpiece and its abstract is an apparatus and method for monitoring material removal from a workpiece by a beam of energy during a material processing operation are disclosed. a detector is positioned for sensing optical emissions from the workpiece caused by removal of material when an energy beam pulse is incident upon the surface of the workpiece. a computing circuit, algorithm or artificial neural network is provided for determining a quantity of material removed from the sensed optical emissions in real-time during the material processing operation. analysis of the optical emission pulses caused by the material removal provides an indication of the efficiency of the material processing system and provides feedback for manual or automatic adjustment of material processing parameters during the material processing operation. dated 1994-02-15"
5287272,automated cytological specimen classification system and method,an automated screening system and method for cytological specimen classification in which a neural network is utilized in performance of the classification function. also included is an automated microscope and associated image processing circuitry.,1994-02-15,The title of the patent is automated cytological specimen classification system and method and its abstract is an automated screening system and method for cytological specimen classification in which a neural network is utilized in performance of the classification function. also included is an automated microscope and associated image processing circuitry. dated 1994-02-15
5287430,signal discrimination device using neural network,"a signal discrimination device using a neural network for discriminating input signals such as radar reception signals includes an adaptive code generator means for generating codes for representing the discrimination categories. the distances between the codes for closely related categories are smaller than the distances between the codes for remotely related categories. during the learning stage, the neural network is trained to output the codes for respective inputs. the discrimination result judgment means determines the categories by comparing the outputs of the neural network and the codes for the respective categories.",1994-02-15,"The title of the patent is signal discrimination device using neural network and its abstract is a signal discrimination device using a neural network for discriminating input signals such as radar reception signals includes an adaptive code generator means for generating codes for representing the discrimination categories. the distances between the codes for closely related categories are smaller than the distances between the codes for remotely related categories. during the learning stage, the neural network is trained to output the codes for respective inputs. the discrimination result judgment means determines the categories by comparing the outputs of the neural network and the codes for the respective categories. dated 1994-02-15"
5287431,neural network using liquid crystal for threshold and amplifiers for weights,a neural network type data processing system in which an optical input is received and a normalized optical output is generated. a plurality of light receiving regions of a photovoltaic material generate signals which are fed into amplifiers and summed. the gain of the amplifiers represent the synaptic weights. the output the summed amplified signals is then sent to a portion of a liquid crystal light valve where that portion of the liquid crystal light valve is used to produce a normalized light output.,1994-02-15,The title of the patent is neural network using liquid crystal for threshold and amplifiers for weights and its abstract is a neural network type data processing system in which an optical input is received and a normalized optical output is generated. a plurality of light receiving regions of a photovoltaic material generate signals which are fed into amplifiers and summed. the gain of the amplifiers represent the synaptic weights. the output the summed amplified signals is then sent to a portion of a liquid crystal light valve where that portion of the liquid crystal light valve is used to produce a normalized light output. dated 1994-02-15
5287533,apparatus for changing individual weight value of corresponding synaptic connection for succeeding learning process when past weight values satisfying predetermined condition,"the past record of the synaptic weight values set in the learning of a neural network is stored in a weight record memory. the past stored in the weight record memory is supplied to a control unit. if there exists a synaptic connection representing a record of weight values which have been used in a predetermined number of learning processes just prior to the present learning process and which satisfy a predetermined condition, the synaptic weight value used in the succeeding learning processes for the synaptic connection is re-set to a predetermined value by a weight setting unit. that is, the past record of the synaptic weight values is monitored, and the synaptic weight value which has been set in a learning process can be re-set as required.",1994-02-15,"The title of the patent is apparatus for changing individual weight value of corresponding synaptic connection for succeeding learning process when past weight values satisfying predetermined condition and its abstract is the past record of the synaptic weight values set in the learning of a neural network is stored in a weight record memory. the past stored in the weight record memory is supplied to a control unit. if there exists a synaptic connection representing a record of weight values which have been used in a predetermined number of learning processes just prior to the present learning process and which satisfy a predetermined condition, the synaptic weight value used in the succeeding learning processes for the synaptic connection is re-set to a predetermined value by a weight setting unit. that is, the past record of the synaptic weight values is monitored, and the synaptic weight value which has been set in a learning process can be re-set as required. dated 1994-02-15"
5289401,analog storage device for artificial neural network system,"an analog storage device employs an electrically erasable programmable transistor as its memory cell. the memory cell transistor has a source and a drain which are disposed spaced apart from each other on a semiconductive substrate to define a channel region therebetween, an insulated floating gate electrode which at least overlaps the channel region, and an insulated control gate electrode disposed above the insulated floating gate electrode. minority carriers are allowed to tunnel between the channel region and the insulated floating gate. the amount of carriers to be stored on the floating gate electrode is controlled such that it is in proportion to analog data to be stored therein. a variation in the internal field of the transistor which may occur when its floating gate electrode is being charged with minority carriers is monitored. when a field variation is detected, a voltage for compensating for the detected field variation is applied to the control gate electrode, whereby the linearity of analog storage is ensured.",1994-02-22,"The title of the patent is analog storage device for artificial neural network system and its abstract is an analog storage device employs an electrically erasable programmable transistor as its memory cell. the memory cell transistor has a source and a drain which are disposed spaced apart from each other on a semiconductive substrate to define a channel region therebetween, an insulated floating gate electrode which at least overlaps the channel region, and an insulated control gate electrode disposed above the insulated floating gate electrode. minority carriers are allowed to tunnel between the channel region and the insulated floating gate. the amount of carriers to be stored on the floating gate electrode is controlled such that it is in proportion to analog data to be stored therein. a variation in the internal field of the transistor which may occur when its floating gate electrode is being charged with minority carriers is monitored. when a field variation is detected, a voltage for compensating for the detected field variation is applied to the control gate electrode, whereby the linearity of analog storage is ensured. dated 1994-02-22"
5293453,error control codeword generating system and method based on a neural network,"a communication system and method that translates a first plurality of information symbols into a plurality of code words, transmits the plurality of code words through a communication channel receives the plurality of code words transmitted through the communication channel, deciphers the plurality of code words transmitted through the communication channel into a second plurality of information symbols that correspond to the first set plurality of information symbols, wherein the plurality of code words are derived from a reverse dynamical flow within a first neural network.",1994-03-08,"The title of the patent is error control codeword generating system and method based on a neural network and its abstract is a communication system and method that translates a first plurality of information symbols into a plurality of code words, transmits the plurality of code words through a communication channel receives the plurality of code words transmitted through the communication channel, deciphers the plurality of code words transmitted through the communication channel into a second plurality of information symbols that correspond to the first set plurality of information symbols, wherein the plurality of code words are derived from a reverse dynamical flow within a first neural network. dated 1994-03-08"
5293454,learning method of neural network,"a learning method of a neural network, in which from a set of learning patterns belonging to one category, specific learning patterns located at a region close to learning patterns belonging to another category are selected and learning of the neural network is performed by using the specific learning patterns so as to discriminate the categories from each other.",1994-03-08,"The title of the patent is learning method of neural network and its abstract is a learning method of a neural network, in which from a set of learning patterns belonging to one category, specific learning patterns located at a region close to learning patterns belonging to another category are selected and learning of the neural network is performed by using the specific learning patterns so as to discriminate the categories from each other. dated 1994-03-08"
5293456,object recognition system employing a sparse comparison neural network,"a neural network for comparing a known input to an unknown input comprises a first layer for receiving a first known input tensor and a first unknown input tensor. a second layer receives the first known and unknown input tensors. the second layer has at least one first trainable weight tensor associated with the first known input tensor and at least one second trainable weight tensor associated with the first unknown input tensor. the second layer includes at least one first processing element for transforming the first known input tensor on the first trainable weight tensor to produce a first known output and at least one second processing element for transforming the first unknown input tensor on the second trainable weight tensor to produce a first unknown output. the first known output comprises a first known output tensor of at least rank zero and has a third trainable weight tensor associated therewith. the first unknown output comprises a first unknown output tensor of at least rank zero and has a fourth trainable weight tensor associated therewith. the first known output tensor and the first unknown tensor are combined to form a second input tensor. a third layer receives the second input tensor. the third layer has at least one fifth trainable weight tensor associated with the second input tensor. the third layer includes at least one third processing element for transforming the second input tensor on the fifth trainable weight tensor, thereby comparing the first known output with the first unknown output and producing a resultant output. the resultant output is indicative of the degree of similarity between the first known input tensor and the first unknown input tensor.",1994-03-08,"The title of the patent is object recognition system employing a sparse comparison neural network and its abstract is a neural network for comparing a known input to an unknown input comprises a first layer for receiving a first known input tensor and a first unknown input tensor. a second layer receives the first known and unknown input tensors. the second layer has at least one first trainable weight tensor associated with the first known input tensor and at least one second trainable weight tensor associated with the first unknown input tensor. the second layer includes at least one first processing element for transforming the first known input tensor on the first trainable weight tensor to produce a first known output and at least one second processing element for transforming the first unknown input tensor on the second trainable weight tensor to produce a first unknown output. the first known output comprises a first known output tensor of at least rank zero and has a third trainable weight tensor associated therewith. the first unknown output comprises a first unknown output tensor of at least rank zero and has a fourth trainable weight tensor associated therewith. the first known output tensor and the first unknown tensor are combined to form a second input tensor. a third layer receives the second input tensor. the third layer has at least one fifth trainable weight tensor associated with the second input tensor. the third layer includes at least one third processing element for transforming the second input tensor on the fifth trainable weight tensor, thereby comparing the first known output with the first unknown output and producing a resultant output. the resultant output is indicative of the degree of similarity between the first known input tensor and the first unknown input tensor. dated 1994-03-08"
5293457,neural network integrated circuit device having self-organizing function,"an extension directed integrated circuit device having a learning function on a boltzmann model, includes a plurality of synapse representing units arrayed in a matrix, a plurality of neuron representing units, a plurality of educator signal control circuits, and a plurality of buffer circuits. each synapse representing unit is connected to a pair of axon signal transfer lines and a pair of dendrite signal transfer lines. each synapse representing unit includes a learning control circuit which derives synapse load change value data in accordance with predetermined learning rules in response to a first axon signal si and a second axon signal sj, a synapse load representing circuit which corrects a synapse load in response to the synapse load change valued data and holds the corrected synapse load value wij, a first synapse coupling operating circuit which derives a current signal indicating a product wij.multidot.si from the synapse load wij and the first axon signal si and transfers the same to a first dendrite signal line, and a second product signal indicating a product wij.multidot.sj from the synapse load wij and the second axon signal sj and transfers the same onto a second dendrite signal line.",1994-03-08,"The title of the patent is neural network integrated circuit device having self-organizing function and its abstract is an extension directed integrated circuit device having a learning function on a boltzmann model, includes a plurality of synapse representing units arrayed in a matrix, a plurality of neuron representing units, a plurality of educator signal control circuits, and a plurality of buffer circuits. each synapse representing unit is connected to a pair of axon signal transfer lines and a pair of dendrite signal transfer lines. each synapse representing unit includes a learning control circuit which derives synapse load change value data in accordance with predetermined learning rules in response to a first axon signal si and a second axon signal sj, a synapse load representing circuit which corrects a synapse load in response to the synapse load change valued data and holds the corrected synapse load value wij, a first synapse coupling operating circuit which derives a current signal indicating a product wij.multidot.si from the synapse load wij and the first axon signal si and transfers the same to a first dendrite signal line, and a second product signal indicating a product wij.multidot.sj from the synapse load wij and the second axon signal sj and transfers the same onto a second dendrite signal line. dated 1994-03-08"
5293458,mos multi-layer neural network and its design method,"disclosed is a multi-layer neural network and circuit design method. the multi-layer neural network receiving an m-bit input and generating an n-bit output comprises a neuron having a cascaded pair of cmos inverters and having an output node of the preceding cmos inverter among the pair of cmos inverters as its inverted output node and an output node of the succeeding cmos inverter as its non-inverted output node, an input layer having m neurons to receive the m-bit input, an output layer having n neurons to generate the n-bit output, at least one hidden layer provided with n neurons to transfer the input received from the input layer to the directly upper hidden layer or the output layer, an input synapse group in a matrix having each predetermined weight value to connect each output of neurons on the input layer to each neuron of the output layer and at least one hidden layer, at least one transfer synapse group in a matrix having each predetermined weight value to connect each output of neurons of the hidden layer to each neuron of its directly upper hidden layer or of the output layer, and a bias synapse group for biasing each input node of neurons of the hidden layers and the output layer.",1994-03-08,"The title of the patent is mos multi-layer neural network and its design method and its abstract is disclosed is a multi-layer neural network and circuit design method. the multi-layer neural network receiving an m-bit input and generating an n-bit output comprises a neuron having a cascaded pair of cmos inverters and having an output node of the preceding cmos inverter among the pair of cmos inverters as its inverted output node and an output node of the succeeding cmos inverter as its non-inverted output node, an input layer having m neurons to receive the m-bit input, an output layer having n neurons to generate the n-bit output, at least one hidden layer provided with n neurons to transfer the input received from the input layer to the directly upper hidden layer or the output layer, an input synapse group in a matrix having each predetermined weight value to connect each output of neurons on the input layer to each neuron of the output layer and at least one hidden layer, at least one transfer synapse group in a matrix having each predetermined weight value to connect each output of neurons of the hidden layer to each neuron of its directly upper hidden layer or of the output layer, and a bias synapse group for biasing each input node of neurons of the hidden layers and the output layer. dated 1994-03-08"
5293459,neural integrated circuit comprising learning means,"a neural integrated circuit, comprising a synaptic coefficient memory, a neuron state memory, resolving means and learning means which simultaneously operate in parallel on each of the synaptic coefficients in order to determine new synaptic coefficients. the learning means comprise means for performing a learning function on the states vj of input neurons and on a correction element si which is associated with each output neuron, and also comprise incrementation/decrementation elements which determine the new synaptic coefficients in parallel. the learning functions may be formed by logic and-gates and exclusive-or gates. the integrated circuit is used in a neural network system comprising a processing device.",1994-03-08,"The title of the patent is neural integrated circuit comprising learning means and its abstract is a neural integrated circuit, comprising a synaptic coefficient memory, a neuron state memory, resolving means and learning means which simultaneously operate in parallel on each of the synaptic coefficients in order to determine new synaptic coefficients. the learning means comprise means for performing a learning function on the states vj of input neurons and on a correction element si which is associated with each output neuron, and also comprise incrementation/decrementation elements which determine the new synaptic coefficients in parallel. the learning functions may be formed by logic and-gates and exclusive-or gates. the integrated circuit is used in a neural network system comprising a processing device. dated 1994-03-08"
5295130,apparatus and method for signal reproduction,an apparatus and method for reproducing pit information precisely from a magnetic optical disk without being adversely affected by heat accumulation. the signal reproducing apparatus reproduces signals using a neural network constituting a decoder that decodes pits on the disk. the signal reproducing method provides learning using a sigmoid function and carries out signal reproduction using a step function.,1994-03-15,The title of the patent is apparatus and method for signal reproduction and its abstract is an apparatus and method for reproducing pit information precisely from a magnetic optical disk without being adversely affected by heat accumulation. the signal reproducing apparatus reproduces signals using a neural network constituting a decoder that decodes pits on the disk. the signal reproducing method provides learning using a sigmoid function and carries out signal reproduction using a step function. dated 1994-03-15
5295197,information processing system using neural network learning function,"an information processing apparatus using a neural network learning function has, in one embodiment, a computer system and a pattern recognition apparatus associated with each other via a communication cable. the computer system includes a learning section having a first neural network and serves to adjust the weights of connection therein as a result of learning with a learning data signal supplied thereto from the pattern recognition apparatus via the communication cable. the pattern recognition apparatus includes an associative output section having a second neural network and receives data on the adjusted weights from the learning section via the communication cable to reconstruct the second neural network with the data on the adjusted weights. the pattern recognition apparatus with the associative output section having the reconstructed second neural network performs pattern recognition independently of the computer system with the communication cable being brought into an electrical isolation mode.",1994-03-15,"The title of the patent is information processing system using neural network learning function and its abstract is an information processing apparatus using a neural network learning function has, in one embodiment, a computer system and a pattern recognition apparatus associated with each other via a communication cable. the computer system includes a learning section having a first neural network and serves to adjust the weights of connection therein as a result of learning with a learning data signal supplied thereto from the pattern recognition apparatus via the communication cable. the pattern recognition apparatus includes an associative output section having a second neural network and receives data on the adjusted weights from the learning section via the communication cable to reconstruct the second neural network with the data on the adjusted weights. the pattern recognition apparatus with the associative output section having the reconstructed second neural network performs pattern recognition independently of the computer system with the communication cable being brought into an electrical isolation mode. dated 1994-03-15"
5295227,neural network learning system,"a neural network learning system is applied to extensive use in applications such as pattern and character recognizing operations, various controls, etc. the neural network learning system operates on, for example, a plurality of neural networks each having a different number of intermediate layer units to efficiently perform a learning process at a high speed with a reduced amount of hardware. a neural network system having a plurality of hierarchical neural networks each having an input layer, one or more intermediate layers and output layers is formed from a common input layer shared among two or more neural networks, or the common input layer and one or more intermediate layers and a learning controller for controlling a learning process performed by a plurality of neural networks.",1994-03-15,"The title of the patent is neural network learning system and its abstract is a neural network learning system is applied to extensive use in applications such as pattern and character recognizing operations, various controls, etc. the neural network learning system operates on, for example, a plurality of neural networks each having a different number of intermediate layer units to efficiently perform a learning process at a high speed with a reduced amount of hardware. a neural network system having a plurality of hierarchical neural networks each having an input layer, one or more intermediate layers and output layers is formed from a common input layer shared among two or more neural networks, or the common input layer and one or more intermediate layers and a learning controller for controlling a learning process performed by a plurality of neural networks. dated 1994-03-15"
5297232,wireless neural network and a wireless neural processing element,"a neural network is disclosed in which communication between processing elements occurs by radio waves in a waveguide. radio wave communication using common carrier signals by transceivers in a waveguide allows processing elements to communicate wirelessly and simultaneously. each processing element includes a radio frequency transceiver and an accompanying antenna which performs the neuron summing operation because input signals simultaneously received from plural processing elements by the antenna add. the weights on each input are provided by different spatial relationships between the transmitting processing elements and the receiving processing element which causes signal strength loses through the waveguide to be different. each receiving processing element performs a neural threshold or sigmoid operation on the summed signal received from the transceiver and then a strength (amplitude scaling) can be applied to the output before the processing element transmits that output to the other processing elements in the system. processing elements are grouped, allowing one group to transmit while the other group is receiving. wafer scale electronics including transceivers and analog processing elements are combined with a comparably sized waveguide to produce a compact device.",1994-03-22,"The title of the patent is wireless neural network and a wireless neural processing element and its abstract is a neural network is disclosed in which communication between processing elements occurs by radio waves in a waveguide. radio wave communication using common carrier signals by transceivers in a waveguide allows processing elements to communicate wirelessly and simultaneously. each processing element includes a radio frequency transceiver and an accompanying antenna which performs the neuron summing operation because input signals simultaneously received from plural processing elements by the antenna add. the weights on each input are provided by different spatial relationships between the transmitting processing elements and the receiving processing element which causes signal strength loses through the waveguide to be different. each receiving processing element performs a neural threshold or sigmoid operation on the summed signal received from the transceiver and then a strength (amplitude scaling) can be applied to the output before the processing element transmits that output to the other processing elements in the system. processing elements are grouped, allowing one group to transmit while the other group is receiving. wafer scale electronics including transceivers and analog processing elements are combined with a comparably sized waveguide to produce a compact device. dated 1994-03-22"
5298796,nonvolatile programmable neural network synaptic array,"a floating-gate mos transistor is implemented for use as a nonvolatile analog storage element of a synaptic cell used to implement an array of processing synaptic cells based on a four-quadrant analog multiplier requiring both x and y differential inputs, where one y input is uv programmable. these nonvolatile synaptic cells are disclosed fully connected in a 32.times.32 synaptic cell array using standard vlsi cmos technology.",1994-03-29,"The title of the patent is nonvolatile programmable neural network synaptic array and its abstract is a floating-gate mos transistor is implemented for use as a nonvolatile analog storage element of a synaptic cell used to implement an array of processing synaptic cells based on a four-quadrant analog multiplier requiring both x and y differential inputs, where one y input is uv programmable. these nonvolatile synaptic cells are disclosed fully connected in a 32.times.32 synaptic cell array using standard vlsi cmos technology. dated 1994-03-29"
5299285,neural network with dynamically adaptable neurons,"this invention is an adaptive neuron for use in neural network processors. the adaptive neuron participates in the supervised learning phase of operation on a coequal basis with the synapse matrix elements by adaptively changing its gain in a similar manner to the change of weights in the synapse io elements. in this manner, training time is decreased by as much as three orders of magnitude.",1994-03-29,"The title of the patent is neural network with dynamically adaptable neurons and its abstract is this invention is an adaptive neuron for use in neural network processors. the adaptive neuron participates in the supervised learning phase of operation on a coequal basis with the synapse matrix elements by adaptively changing its gain in a similar manner to the change of weights in the synapse io elements. in this manner, training time is decreased by as much as three orders of magnitude. dated 1994-03-29"
5299286,data processing system for implementing architecture of neural network subject to learning process,"data processing system implementing architecture of a neural network which is subject to a learning process, wherein the data processing system includes n.times.n synapses arranged in an array of j rows and i columns. a plurality of operational amplifiers respectively corresponding to the rows of the array are provided, with each operational amplifier defining a neuron. the input terminals of all of the synapses arranged in a respective column of the array are connected together and define n inputs of the neural network. the output terminals of the synapses arranged in a respective row of the array are connected together and serve as the inputs to a corresponding one of the plurality of operational amplifiers. each synapse includes a capacitor connected between ground potential and the input terminal for weighting the synapse by storing a weighting voltage applied thereto. a random access memory has digitally stored voltage values for weighting all of the synapses. a plurality of digital-analog converters, one for each column of the array of synapses, are connected to the random access memory for converting the digital voltage values for weighting the synapses into analog voltage values. the digital-analog converters provide respective outputs to the weighting terminals of the synapses of a column via respective electronic switches for each synapse. each row of the array includes a bistable circuit for driving the respective electronic switches under the control of a control section which also provides function commands and data to the random access memory.",1994-03-29,"The title of the patent is data processing system for implementing architecture of neural network subject to learning process and its abstract is data processing system implementing architecture of a neural network which is subject to a learning process, wherein the data processing system includes n.times.n synapses arranged in an array of j rows and i columns. a plurality of operational amplifiers respectively corresponding to the rows of the array are provided, with each operational amplifier defining a neuron. the input terminals of all of the synapses arranged in a respective column of the array are connected together and define n inputs of the neural network. the output terminals of the synapses arranged in a respective row of the array are connected together and serve as the inputs to a corresponding one of the plurality of operational amplifiers. each synapse includes a capacitor connected between ground potential and the input terminal for weighting the synapse by storing a weighting voltage applied thereto. a random access memory has digitally stored voltage values for weighting all of the synapses. a plurality of digital-analog converters, one for each column of the array of synapses, are connected to the random access memory for converting the digital voltage values for weighting the synapses into analog voltage values. the digital-analog converters provide respective outputs to the weighting terminals of the synapses of a column via respective electronic switches for each synapse. each row of the array includes a bistable circuit for driving the respective electronic switches under the control of a control section which also provides function commands and data to the random access memory. dated 1994-03-29"
5300770,apparatus for producing a porosity log of a subsurface formation corrected for detector standoff,"a borehole logging tool is lowered into a borehole traversing a subsurface formation and a neutron detector measures the die-away of nuclear radiation in the formation. intensity signals are produced representing the die-away of nuclear radiation as the logging tool traverses the borehole a signal processor, employing at least one neural network, processes the intensity signals and produces a standoff-corrected epithermal neutron lifetime signal to correct for standoff from the borehole wall encountered by the detector as the logging tool traverses the borehole. the signal processor further generates a porosity signal from the standoff-corrected epithermal neutron lifetime signal derived from measurements in borehole models at known porosities and conditions of detector standoff. a log is generated of such porosity signal versus depth as the logging tool traverses the borehole.",1994-04-05,"The title of the patent is apparatus for producing a porosity log of a subsurface formation corrected for detector standoff and its abstract is a borehole logging tool is lowered into a borehole traversing a subsurface formation and a neutron detector measures the die-away of nuclear radiation in the formation. intensity signals are produced representing the die-away of nuclear radiation as the logging tool traverses the borehole a signal processor, employing at least one neural network, processes the intensity signals and produces a standoff-corrected epithermal neutron lifetime signal to correct for standoff from the borehole wall encountered by the detector as the logging tool traverses the borehole. the signal processor further generates a porosity signal from the standoff-corrected epithermal neutron lifetime signal derived from measurements in borehole models at known porosities and conditions of detector standoff. a log is generated of such porosity signal versus depth as the logging tool traverses the borehole. dated 1994-04-05"
5301257,neural network,"to enable the pattern matching between a shifted input pattern and the standard pattern, a plurality of standard patterns are stored in a standard pattern associative memory network 12. a pattern shifted relative to the standard pattern is inputted to the input pattern network 11 and a restriction condition of when the input pattern is shifted relative to the standard pattern is stored in a coordinate associated network 14. in an association network 13, weights and biases are determined so that the respective units of the network 13 are activated most intensely when the input pattern and the standard pattern match correctly each other in response to the signals from the respective networks 11, 12, and 14.",1994-04-05,"The title of the patent is neural network and its abstract is to enable the pattern matching between a shifted input pattern and the standard pattern, a plurality of standard patterns are stored in a standard pattern associative memory network 12. a pattern shifted relative to the standard pattern is inputted to the input pattern network 11 and a restriction condition of when the input pattern is shifted relative to the standard pattern is stored in a coordinate associated network 14. in an association network 13, weights and biases are determined so that the respective units of the network 13 are activated most intensely when the input pattern and the standard pattern match correctly each other in response to the signals from the respective networks 11, 12, and 14. dated 1994-04-05"
5301681,device for detecting cancerous and precancerous conditions in a breast,"the present invention relates to a device for detecting and monitoring physiological conditions in mammalian tissue, and method for using the same. the device includes sensors for sensing physiological conditions and generating signals in response thereto and processor operatively associated with the sensors for receiving and manipulating the signals to produce a generalization indicative of normal and abnormal physiological condition of mammalian tissue. the processor is characterized to include a neural network having a predetermined solution spaced memory, the solution space memory including regions indicative of two (2) or more physiological conditions, wherein the generalization is characterized by the signals projected into the regions.",1994-04-12,"The title of the patent is device for detecting cancerous and precancerous conditions in a breast and its abstract is the present invention relates to a device for detecting and monitoring physiological conditions in mammalian tissue, and method for using the same. the device includes sensors for sensing physiological conditions and generating signals in response thereto and processor operatively associated with the sensors for receiving and manipulating the signals to produce a generalization indicative of normal and abnormal physiological condition of mammalian tissue. the processor is characterized to include a neural network having a predetermined solution spaced memory, the solution space memory including regions indicative of two (2) or more physiological conditions, wherein the generalization is characterized by the signals projected into the regions. dated 1994-04-12"
5303269,optically maximum a posteriori demodulator,"a system and method for optimal maximum a posteriori (map) demodulation. the present invention incorporates neural network technology, i.e., a hopfield network, (1) to replace the function of the traditional, suboptimal phase-locked loop in an fm receiver and/or (2) to optimally estimate a discrete phase value using an expected value (obtained from the mean of the prior probability distribution of the phase) and statistical dependence between different phase values in a block of samples (described by the covariance matrix of the prior phase distribution). the definition of the hopfield network includes particular bias currents, feedback weights and a sigmoid function for solving the nonlinear integral equation associated with optimal demodulation. the present invention also includes a signal classifier having a plurality of angled modulators for modeling different phase modulation processes.",1994-04-12,"The title of the patent is optically maximum a posteriori demodulator and its abstract is a system and method for optimal maximum a posteriori (map) demodulation. the present invention incorporates neural network technology, i.e., a hopfield network, (1) to replace the function of the traditional, suboptimal phase-locked loop in an fm receiver and/or (2) to optimally estimate a discrete phase value using an expected value (obtained from the mean of the prior probability distribution of the phase) and statistical dependence between different phase values in a block of samples (described by the covariance matrix of the prior phase distribution). the definition of the hopfield network includes particular bias currents, feedback weights and a sigmoid function for solving the nonlinear integral equation associated with optimal demodulation. the present invention also includes a signal classifier having a plurality of angled modulators for modeling different phase modulation processes. dated 1994-04-12"
5303311,method and apparatus for recognizing characters,a character recognition system identifies characters including hand written characters with a high degree of accuracy by use of spiral view codes for pels in the scanned character image. the spiral view codes are developed by comparing stroke length or distance to a remote stroke from a first radial view from a character pel to stroke length or distance to a remote stroke from a counterclockwise adjacent view for the same character pel. these spiral view codes are collected into a spiral view pattern for each character pel. the spiral view patterns for a character are accumulated to form a character vector. the character vector is analyzed by a linear decision network or a neural network.,1994-04-12,The title of the patent is method and apparatus for recognizing characters and its abstract is a character recognition system identifies characters including hand written characters with a high degree of accuracy by use of spiral view codes for pels in the scanned character image. the spiral view codes are developed by comparing stroke length or distance to a remote stroke from a first radial view from a character pel to stroke length or distance to a remote stroke from a counterclockwise adjacent view for the same character pel. these spiral view codes are collected into a spiral view pattern for each character pel. the spiral view patterns for a character are accumulated to form a character vector. the character vector is analyzed by a linear decision network or a neural network. dated 1994-04-12
5303328,neural network system for determining optimal solution,"a neural network system includes an input unit, an operation control unit, a parameter setting unit, a neural network group unit, and a display unit. the network group unit includes first and second neural networks. the first neural network operates according to the mean field approximation method to which the annealing is added, whereas the second neural network operates in accordance with the simulated annealing. each of the first an second neural networks includes a plurality of neurons each connected via synapses to neurons so as to weighting outputs from the neurons based on synapse weights, thereby computing an output related to a total of weighted outputs from the neurons according to an output function. the parameter setting unit is responsive to a setting instruction to generate neuron parameters including synapse weights, threshold values, and output functions, which are set to the first neural network and which are selective set to the second neural network. the operation control unit responsive to an input of a problem analyzes the problem and then generates a setting instruction based on a result of the analysis to output the result to the parameter setting unit. after the neuron parameters are set thereto, in order for the first and second neural network to selectively or to iteratively operate, the operation control unit controls operations of computations in the network group unit in accordance with the analysis result and then presents results of the computations in the network group unit on the display unit.",1994-04-12,"The title of the patent is neural network system for determining optimal solution and its abstract is a neural network system includes an input unit, an operation control unit, a parameter setting unit, a neural network group unit, and a display unit. the network group unit includes first and second neural networks. the first neural network operates according to the mean field approximation method to which the annealing is added, whereas the second neural network operates in accordance with the simulated annealing. each of the first an second neural networks includes a plurality of neurons each connected via synapses to neurons so as to weighting outputs from the neurons based on synapse weights, thereby computing an output related to a total of weighted outputs from the neurons according to an output function. the parameter setting unit is responsive to a setting instruction to generate neuron parameters including synapse weights, threshold values, and output functions, which are set to the first neural network and which are selective set to the second neural network. the operation control unit responsive to an input of a problem analyzes the problem and then generates a setting instruction based on a result of the analysis to output the result to the parameter setting unit. after the neuron parameters are set thereto, in order for the first and second neural network to selectively or to iteratively operate, the operation control unit controls operations of computations in the network group unit in accordance with the analysis result and then presents results of the computations in the network group unit on the display unit. dated 1994-04-12"
5303330,hybrid multi-layer neural networks,"a hybrid network 100 which combines a neural network of the self-organized type 110 with a plurality of neural networks of the supervised learning type 150,160,170 to successfully retrieve building address information from a database using imperfect textual retrieval keys. generally, the self-organized type is a kohonen feature map network, whereas each supervised learning type is a back propagation network. a user query 105 produces an activation response 111,112,113 from the self-organized network 110 and this response, along with a new query 151,161,171 derived from the original query 105, activates a selected one of the learning networks r.sub.1,r.sub.2,r.sub.m to retrieve the requested information.",1994-04-12,"The title of the patent is hybrid multi-layer neural networks and its abstract is a hybrid network 100 which combines a neural network of the self-organized type 110 with a plurality of neural networks of the supervised learning type 150,160,170 to successfully retrieve building address information from a database using imperfect textual retrieval keys. generally, the self-organized type is a kohonen feature map network, whereas each supervised learning type is a back propagation network. a user query 105 produces an activation response 111,112,113 from the self-organized network 110 and this response, along with a new query 151,161,171 derived from the original query 105, activates a selected one of the learning networks r.sub.1,r.sub.2,r.sub.m to retrieve the requested information. dated 1994-04-12"
5305204,digital image display apparatus with automatic window level and window width adjustment,"a digital image display apparatus for converting a pixel value of medical digital image data such as mri image data or ct image data into brightness in accordance with a display window including a window level and a window width of a display unit, determines the optimum window level and width for each image as follows. the apparatus obtains a histogram of pixel values from the digital image data and calculates brightness data of a pixel value having a highest frequency, brightness data of a pixel value at a boundary between a background and an image, area data of a portion having middle brightness within a display brightness range, area data of a portion having maximum brightness, and data indicating a ratio between an area of a portion having higher brightness than the middle brightness and an area of a portion having lower brightness than that obtained, when the digital image is to be displayed by a given display window on the basis of the histogram. the apparatus obtains image quality indicating clarity of the image displayed by the given window on the basis of the above data by using arithmetic operations or by using a neural network, thereby determining the optimum display window which provides a maximum image quality.",1994-04-19,"The title of the patent is digital image display apparatus with automatic window level and window width adjustment and its abstract is a digital image display apparatus for converting a pixel value of medical digital image data such as mri image data or ct image data into brightness in accordance with a display window including a window level and a window width of a display unit, determines the optimum window level and width for each image as follows. the apparatus obtains a histogram of pixel values from the digital image data and calculates brightness data of a pixel value having a highest frequency, brightness data of a pixel value at a boundary between a background and an image, area data of a portion having middle brightness within a display brightness range, area data of a portion having maximum brightness, and data indicating a ratio between an area of a portion having higher brightness than the middle brightness and an area of a portion having lower brightness than that obtained, when the digital image is to be displayed by a given display window on the basis of the histogram. the apparatus obtains image quality indicating clarity of the image displayed by the given window on the basis of the above data by using arithmetic operations or by using a neural network, thereby determining the optimum display window which provides a maximum image quality. dated 1994-04-19"
5305230,process control system and power plant process control system,"a process control system controls a large scale plant such as thermal power plant. this process control system includes a target setting unit for setting an operation target, a control unit for receiving a signal indicating the operation target and for outputting a controlled variable to operate the process, an evaluation unit for quantitatively evaluating operation characteristics corresponding to the operation target of the process operated on the basis of a signal indicating the controlled variable supplied from the control unit, a modification unit for extracting an optimum operation process qualitatively squaring or conforming with the evaluated value derived by the evaluation unit out of a modification rule predetermining operation unit in qualitative relation between the operation characteristics and the operation target of the process and for determining the modification rate of the control unit, a storage unit having a model of a neural network for storing a relation between the operation target and the modification rate derived by the modification unit as a connection state within a circuit, and a learning unit for making the model of the neural network learn the relation between the operation target and the modification rate.",1994-04-19,"The title of the patent is process control system and power plant process control system and its abstract is a process control system controls a large scale plant such as thermal power plant. this process control system includes a target setting unit for setting an operation target, a control unit for receiving a signal indicating the operation target and for outputting a controlled variable to operate the process, an evaluation unit for quantitatively evaluating operation characteristics corresponding to the operation target of the process operated on the basis of a signal indicating the controlled variable supplied from the control unit, a modification unit for extracting an optimum operation process qualitatively squaring or conforming with the evaluated value derived by the evaluation unit out of a modification rule predetermining operation unit in qualitative relation between the operation characteristics and the operation target of the process and for determining the modification rate of the control unit, a storage unit having a model of a neural network for storing a relation between the operation target and the modification rate derived by the modification unit as a connection state within a circuit, and a learning unit for making the model of the neural network learn the relation between the operation target and the modification rate. dated 1994-04-19"
5305235,monitoring diagnosis device for electrical appliance,"a monitoring diagnostic device for an electrical appliance such as gas insulated switchgear includes a sensor, such as an acceleration sensor, and a neural network including an input layer, an intermediate layer, and an output layer, each consisting of a plurality of neural elements. the input, intermediate and output layers are coupled to each other via a plurality of connection weights. the output of the sensor is first processed and then is supplied to the neural elements of the input layer. the connection weights are adjusted by means of learning data such that the output from the neural elements of the output layer of the neural network correctly identifies the causes of abnormality of the electrical appliance.",1994-04-19,"The title of the patent is monitoring diagnosis device for electrical appliance and its abstract is a monitoring diagnostic device for an electrical appliance such as gas insulated switchgear includes a sensor, such as an acceleration sensor, and a neural network including an input layer, an intermediate layer, and an output layer, each consisting of a plurality of neural elements. the input, intermediate and output layers are coupled to each other via a plurality of connection weights. the output of the sensor is first processed and then is supplied to the neural elements of the input layer. the connection weights are adjusted by means of learning data such that the output from the neural elements of the output layer of the neural network correctly identifies the causes of abnormality of the electrical appliance. dated 1994-04-19"
5305250,analog continuous-time mos vector multiplier circuit and a programmable mos realization for feedback neural networks,"a neuron circuit and a neural network including a four quadrant analog multiplier/summer circuit constructed in field effect transistors. the neuron circuit includes the analog multiplier/summer formed of an operational amplifier, plural sets of four field effect transistors, an rc circuit and a double inverter. the multiplier/summer circuit includes a set of four identical field effect transistors for each product implemented. this produces a four quadrant multiplication if the four field effect transistors operate in the triode mode. the output of the multiplier/summer is the sum of these products. the neural network includes a plurality of these neuron circuits. each neuron circuit receives an input and a set of synaptic weight inputs. the output of each neuron circuit is supplied to the corresponding feedback input of each neuron circuit. the multiplier/summer of each neuron circuit produces the sum of the product of each neuron circuit output and its corresponding synaptic weight. the individual neuron circuits and the neural network can be constructed in mos vlsi.",1994-04-19,"The title of the patent is analog continuous-time mos vector multiplier circuit and a programmable mos realization for feedback neural networks and its abstract is a neuron circuit and a neural network including a four quadrant analog multiplier/summer circuit constructed in field effect transistors. the neuron circuit includes the analog multiplier/summer formed of an operational amplifier, plural sets of four field effect transistors, an rc circuit and a double inverter. the multiplier/summer circuit includes a set of four identical field effect transistors for each product implemented. this produces a four quadrant multiplication if the four field effect transistors operate in the triode mode. the output of the multiplier/summer is the sum of these products. the neural network includes a plurality of these neuron circuits. each neuron circuit receives an input and a set of synaptic weight inputs. the output of each neuron circuit is supplied to the corresponding feedback input of each neuron circuit. the multiplier/summer of each neuron circuit produces the sum of the product of each neuron circuit output and its corresponding synaptic weight. the individual neuron circuits and the neural network can be constructed in mos vlsi. dated 1994-04-19"
5306893,weld acoustic monitor,"a system for real-time analysis of weld quality in an arc welding process. the system includes a transducer which receives acoustic signals generated during the welding process. the acoustic signals are then sampled and digitized. a signal processor calculates the root mean square and peak amplitudes of the digitized signals and transforms the digitized signal into a frequency domain signal. a data processor divides the frequency domain signal into a plurality of frequency bands and calculates the average power for each band. the average power values, in addition to the peak and root mean square amplitude values, are input to an artificial neural network for analysis of weld quality. arc current and/or arc voltage signals may be input to the a/d converter alone or in combination with the acoustic signal data for subsequent signal processing and neural network analysis.",1994-04-26,"The title of the patent is weld acoustic monitor and its abstract is a system for real-time analysis of weld quality in an arc welding process. the system includes a transducer which receives acoustic signals generated during the welding process. the acoustic signals are then sampled and digitized. a signal processor calculates the root mean square and peak amplitudes of the digitized signals and transforms the digitized signal into a frequency domain signal. a data processor divides the frequency domain signal into a plurality of frequency bands and calculates the average power for each band. the average power values, in addition to the peak and root mean square amplitude values, are input to an artificial neural network for analysis of weld quality. arc current and/or arc voltage signals may be input to the a/d converter alone or in combination with the acoustic signal data for subsequent signal processing and neural network analysis. dated 1994-04-26"
5307260,order entry apparatus for automatic estimation and its method,"order entry apparatus for automatic estimation a transformation model comprising a pattern composed of a plurality of parameters representing custom product specifications, production line conditions and factors for composing the estimates for production cost and completion date. the parameters of this transformation model are specified by learning, leading to estimation for the requested product specifications, in broad consideration of conditions such as those of the production line. the use of a neural network model as this pattern transformation model makes pattern transformation more flexible and pattern learning more efficient. estimation accuracy is also increased by entering values of predicted charges in production line conditions such as loads or stock obtained from resource requirements planning, process design, capacity requirements planning, etc.",1994-04-26,"The title of the patent is order entry apparatus for automatic estimation and its method and its abstract is order entry apparatus for automatic estimation a transformation model comprising a pattern composed of a plurality of parameters representing custom product specifications, production line conditions and factors for composing the estimates for production cost and completion date. the parameters of this transformation model are specified by learning, leading to estimation for the requested product specifications, in broad consideration of conditions such as those of the production line. the use of a neural network model as this pattern transformation model makes pattern transformation more flexible and pattern learning more efficient. estimation accuracy is also increased by entering values of predicted charges in production line conditions such as loads or stock obtained from resource requirements planning, process design, capacity requirements planning, etc. dated 1994-04-26"
5307444,voice analyzing system using hidden markov model and having plural neural network predictors,"an analyzing system analyzes object signals, particularly voice signals, by estimating a generation likelihood of an observation vector sequence being a time series of feature vectors with use of a markov model having a plurality of states and given transition probabilities from state to state. a state designation section determines a state i at a time t stochastically using the markov model. plural predictors, each of which is composed of a neural network and is provided per each state of the markov model, are provided for generating a predictional vector of the feature vector x.sub.t in the state i at the time t based on values of the feature vectors other than the feature vector x.sub.t. a first calculation section calculates an error vector of the predictional vector to the feature vector x.sub.t. a second calculation section calculates a generation likelihood of the error vector using a predetermined probability distribution of the error vector according to which the error vector is generated.",1994-04-26,"The title of the patent is voice analyzing system using hidden markov model and having plural neural network predictors and its abstract is an analyzing system analyzes object signals, particularly voice signals, by estimating a generation likelihood of an observation vector sequence being a time series of feature vectors with use of a markov model having a plurality of states and given transition probabilities from state to state. a state designation section determines a state i at a time t stochastically using the markov model. plural predictors, each of which is composed of a neural network and is provided per each state of the markov model, are provided for generating a predictional vector of the feature vector x.sub.t in the state i at the time t based on values of the feature vectors other than the feature vector x.sub.t. a first calculation section calculates an error vector of the predictional vector to the feature vector x.sub.t. a second calculation section calculates a generation likelihood of the error vector using a predetermined probability distribution of the error vector according to which the error vector is generated. dated 1994-04-26"
5309525,image processing apparatus using neural network,"disclosed is an image processing apparatus having an input device for inputting binary image data comprising a plurality of pixels which include a pixel of interest that is to be subjected to multivalued conversion, the plurality of pixels being contained in an area that is asymmetrical with respect to the position of the pixel of interest, and an multivalued converting device for executing processing, by a neural network, to restore the input binary image data to multivalued image data for the pixel of interest, whereby multivalued image data is estimated from binarized image data. it is possible to reduce the number of pixels referred to in arithmetic operations performed in the neural network.",1994-05-03,"The title of the patent is image processing apparatus using neural network and its abstract is disclosed is an image processing apparatus having an input device for inputting binary image data comprising a plurality of pixels which include a pixel of interest that is to be subjected to multivalued conversion, the plurality of pixels being contained in an area that is asymmetrical with respect to the position of the pixel of interest, and an multivalued converting device for executing processing, by a neural network, to restore the input binary image data to multivalued image data for the pixel of interest, whereby multivalued image data is estimated from binarized image data. it is possible to reduce the number of pixels referred to in arithmetic operations performed in the neural network. dated 1994-05-03"
5311182,method and apparatus for regenerating a distorted binary signal stream,"apparatus is shown for regenerating a signal stream of binary digits which has been distorted by intersymbol interference during passage through a channel (10 and 12) having insufficient channel bandwidth such that the channel output waveform comprises substantially an analog signal. (fig. 2 at b and d.) after equalization (24) the channel output is converted to a digital sample signal stream at analog-to-digital converter (26). the converter (26) output is supplied to shift register (28) from which successive groups of digital sample signals produced over a plurality of bit intervals of channel output are shifted to decoder (22). initialization bits that immediately precede the first group of binary digits to be regenerated also are supplied to decoder (22) through sector header reader (20) for use in decoding the first group of digital sample signals supplied to the decoder. during decoding of subsequent groups of digital sample signals, end bits (3,4 and 5) from the preceding group of regenerated binary digits are supplied to the decoder (22). the decoder includes a plurality of trained networks (40-1 through 40-5 and 50-1 through 50-m) of either the neural network or binary tree type.",1994-05-10,"The title of the patent is method and apparatus for regenerating a distorted binary signal stream and its abstract is apparatus is shown for regenerating a signal stream of binary digits which has been distorted by intersymbol interference during passage through a channel (10 and 12) having insufficient channel bandwidth such that the channel output waveform comprises substantially an analog signal. (fig. 2 at b and d.) after equalization (24) the channel output is converted to a digital sample signal stream at analog-to-digital converter (26). the converter (26) output is supplied to shift register (28) from which successive groups of digital sample signals produced over a plurality of bit intervals of channel output are shifted to decoder (22). initialization bits that immediately precede the first group of binary digits to be regenerated also are supplied to decoder (22) through sector header reader (20) for use in decoding the first group of digital sample signals supplied to the decoder. during decoding of subsequent groups of digital sample signals, end bits (3,4 and 5) from the preceding group of regenerated binary digits are supplied to the decoder (22). the decoder includes a plurality of trained networks (40-1 through 40-5 and 50-1 through 50-m) of either the neural network or binary tree type. dated 1994-05-10"
5311421,process control method and system for performing control of a controlled system by use of a neural network,"a method for controlling a controlled system by a controller such that a controlled variable can be brought into conformity with a desired value. with respect to at least one of input/output variables for a combined controlling-controlled system, which includes in combination the controller and the controlled system, and input/output variables for the controlled system, information containing its characteristics is taken out from the combined controlling-controlled system. the information with the characteristics contained therein is inputted to a neural network which has been caused beforehand to learn a correlation between the information containing the characteristics and control parameters. from the neural network, one or more of the control parameters, said one or more control parameters corresponding to a corresponding number of inputs to the neural network, are outputted to the controller.",1994-05-10,"The title of the patent is process control method and system for performing control of a controlled system by use of a neural network and its abstract is a method for controlling a controlled system by a controller such that a controlled variable can be brought into conformity with a desired value. with respect to at least one of input/output variables for a combined controlling-controlled system, which includes in combination the controller and the controlled system, and input/output variables for the controlled system, information containing its characteristics is taken out from the combined controlling-controlled system. the information with the characteristics contained therein is inputted to a neural network which has been caused beforehand to learn a correlation between the information containing the characteristics and control parameters. from the neural network, one or more of the control parameters, said one or more control parameters corresponding to a corresponding number of inputs to the neural network, are outputted to the controller. dated 1994-05-10"
5311600,method of edge detection in optical images using neural network classifier,"an image processor employing a camera, frame grabber and a new algorithm for detecting straight edges in optical images is disclosed. the algorithm is based on using a self-organizing unsupervised neural network learning to classify pixels on a digitized image and then extract the corresponding line parameters. the image processor is demonstrated on the specific application of edge detection for linewidth measurement in semiconductor lithography. the results are compared to results obtained by a standard straight edge detector based on the radon transform; good consistency is observed; however, superior speed is achieved for the proposed image processor. the results obtained by the proposed approach are also shown to be in agreement with scanning electron microscope (sem) measurements, which is known to have excellent accuracy but is an invasive measurement instrument. the method can thus be used for on-line measurement and control of microlithography processes and for alignment tasks as well.",1994-05-10,"The title of the patent is method of edge detection in optical images using neural network classifier and its abstract is an image processor employing a camera, frame grabber and a new algorithm for detecting straight edges in optical images is disclosed. the algorithm is based on using a self-organizing unsupervised neural network learning to classify pixels on a digitized image and then extract the corresponding line parameters. the image processor is demonstrated on the specific application of edge detection for linewidth measurement in semiconductor lithography. the results are compared to results obtained by a standard straight edge detector based on the radon transform; good consistency is observed; however, superior speed is achieved for the proposed image processor. the results obtained by the proposed approach are also shown to be in agreement with scanning electron microscope (sem) measurements, which is known to have excellent accuracy but is an invasive measurement instrument. the method can thus be used for on-line measurement and control of microlithography processes and for alignment tasks as well. dated 1994-05-10"
5313407,integrated active vibration cancellation and machine diagnostic system,a machine analyzer is connected to an active vibration cancellation system in order to identify the operating status of the moving machinery while using a minimum of additional parts and taking advantage of signal processing already occurring in the active vibration cancellation system. a preferred embodiment employs a neural network pattern classifier in connection with detecting operating states such as cylinder misfires in an internal combustion engine.,1994-05-17,The title of the patent is integrated active vibration cancellation and machine diagnostic system and its abstract is a machine analyzer is connected to an active vibration cancellation system in order to identify the operating status of the moving machinery while using a minimum of additional parts and taking advantage of signal processing already occurring in the active vibration cancellation system. a preferred embodiment employs a neural network pattern classifier in connection with detecting operating states such as cylinder misfires in an internal combustion engine. dated 1994-05-17
5313558,system for spatial and temporal pattern learning and recognition,"a neural network simulator that comprises a sensory window memory capable of providing a system of neuron elements with an input consisting of data generated from sequentially sampled spatial and/or temporal patterns. each neuron element comprises multiple levels, each of which is independently connected to the sensory window and/or to other neuron elements for receiving information corresponding to spatial and/or temporal patterns to be learned or recognized. each neuron level comprises a multiplicity of pairs of synaptic connections that record ratios of input information so received and compare them to prerecorded ratios corresponding to learned patterns. the comparison is carried out for each synaptic pair according to empirical activation functions that produce maximum activation of a particular pair when the current ratio matches the learned ratio. when a sufficiently large number of synaptic pairs in a level registers a high activation, the corresponding neuron is taken to have recognized the learned pattern and produces a recognition signal.",1994-05-17,"The title of the patent is system for spatial and temporal pattern learning and recognition and its abstract is a neural network simulator that comprises a sensory window memory capable of providing a system of neuron elements with an input consisting of data generated from sequentially sampled spatial and/or temporal patterns. each neuron element comprises multiple levels, each of which is independently connected to the sensory window and/or to other neuron elements for receiving information corresponding to spatial and/or temporal patterns to be learned or recognized. each neuron level comprises a multiplicity of pairs of synaptic connections that record ratios of input information so received and compare them to prerecorded ratios corresponding to learned patterns. the comparison is carried out for each synaptic pair according to empirical activation functions that produce maximum activation of a particular pair when the current ratio matches the learned ratio. when a sufficiently large number of synaptic pairs in a level registers a high activation, the corresponding neuron is taken to have recognized the learned pattern and produces a recognition signal. dated 1994-05-17"
5313559,method of and system for controlling learning in neural network,"a learning control method reduces overall learning time by displaying data related to an appropriate determination of learning protraction and a proper restoring method. prior to initiating the learning, the user is inquired about the current problem and a problem data set representing items associated with the problem is obtained. evaluation data indicating a state of learning obtained during the learning on the current problem is sequentially stored and displayed. when there is a high possibility of learning protraction during the learning, a message informing the user is displayed. when the learning is stopped by the user in this case, the problem data set and evaluation data set are stored. then, a list of restoring methods is displayed and a particular restoring method is selected by the user once the learning is stopped. the learning is restarted on the current problem in accordance with the selected restoring method.",1994-05-17,"The title of the patent is method of and system for controlling learning in neural network and its abstract is a learning control method reduces overall learning time by displaying data related to an appropriate determination of learning protraction and a proper restoring method. prior to initiating the learning, the user is inquired about the current problem and a problem data set representing items associated with the problem is obtained. evaluation data indicating a state of learning obtained during the learning on the current problem is sequentially stored and displayed. when there is a high possibility of learning protraction during the learning, a message informing the user is displayed. when the learning is stopped by the user in this case, the problem data set and evaluation data set are stored. then, a list of restoring methods is displayed and a particular restoring method is selected by the user once the learning is stopped. the learning is restarted on the current problem in accordance with the selected restoring method. dated 1994-05-17"
5315162,electrochemical synapses for artificial neural networks,an electrochemical synapse adapted for use in a neural network which includes an input terminal and an output terminal located at a distance of less than 100 microns from the input terminal. a permanent interconnect having controllable conductivity is located between the two inputs. the conductivity of the permanent interconnect is controlled by either growing or eliminating metallic whiskers between the inputs. the growth and elimination of whiskers provides a rapid and controllable electrochemical synapse. partial neural network systems are disclosed utilizing the electrochemical synapse.,1994-05-24,The title of the patent is electrochemical synapses for artificial neural networks and its abstract is an electrochemical synapse adapted for use in a neural network which includes an input terminal and an output terminal located at a distance of less than 100 microns from the input terminal. a permanent interconnect having controllable conductivity is located between the two inputs. the conductivity of the permanent interconnect is controlled by either growing or eliminating metallic whiskers between the inputs. the growth and elimination of whiskers provides a rapid and controllable electrochemical synapse. partial neural network systems are disclosed utilizing the electrochemical synapse. dated 1994-05-24
5315704,speech/voiceband data discriminator,"input signals are processed to generate a plurality of signals having different features according to whether the input signals are speech signals or voiceband data signals, and these plural signals are entered into a neural network to be determined whether they have features closer to those of speech signals or of voiceband data signals. the classifying function of the neural network is achieved by inputting samples of speech signals and voiceband data signals and learning how to obtain correct classification results.",1994-05-24,"The title of the patent is speech/voiceband data discriminator and its abstract is input signals are processed to generate a plurality of signals having different features according to whether the input signals are speech signals or voiceband data signals, and these plural signals are entered into a neural network to be determined whether they have features closer to those of speech signals or of voiceband data signals. the classifying function of the neural network is achieved by inputting samples of speech signals and voiceband data signals and learning how to obtain correct classification results. dated 1994-05-24"
5317675,neural network pattern recognition learning method,"a neural network includes an input layer composed of a plurality of cells receiving respective components of an input vector, an output layer composed of a plurality of cells representing attribute of the input vector, and an intermediate layer composed of a plurality of cells connected to all the cells of the input and output layers for producing a mapping to map a given input vector to its correct attribute. a learning method utilizing such neural network is carried out by image projecting the input vector into the partial dimensional space by a projection image operating means preliminarily prepared and by storing a coupling vector on the image projection space as well as the threshold and attribute vector.",1994-05-31,"The title of the patent is neural network pattern recognition learning method and its abstract is a neural network includes an input layer composed of a plurality of cells receiving respective components of an input vector, an output layer composed of a plurality of cells representing attribute of the input vector, and an intermediate layer composed of a plurality of cells connected to all the cells of the input and output layers for producing a mapping to map a given input vector to its correct attribute. a learning method utilizing such neural network is carried out by image projecting the input vector into the partial dimensional space by a projection image operating means preliminarily prepared and by storing a coupling vector on the image projection space as well as the threshold and attribute vector. dated 1994-05-31"
5317676,apparatus and method for facilitating use of a neural network,"a neural network development utility assists a developer in generating one or more filters for data to be input to or output from a neural network. a filter is a device which translates data in accordance with a data transformation definition contained in a translate template. source data for the neural network may be expressed in any arbitrary combination of symbolic or numeric fields in a data base. the developer selects those fields to be used from an interactive menu. the utility scans the selected field entries in the source data base to identify the logical type of each field, and creates a default translate template based on this scan. numeric data is automatically scaled. the developer may use the default template, or edit it from an interactive editor. when editing the template, the developer may select from a menu of commonly used neural network data formats, and from a menu of commonly used primitive mathematical operations. the developer may interactively define additional filters to perform data transformations in series, thus achieving more complex mathematical operations on the data. templates may be edited at any time during the development process. if a network does not appear to be giving satisfactory results, the developer may easily alter the template to present inputs in some other format.",1994-05-31,"The title of the patent is apparatus and method for facilitating use of a neural network and its abstract is a neural network development utility assists a developer in generating one or more filters for data to be input to or output from a neural network. a filter is a device which translates data in accordance with a data transformation definition contained in a translate template. source data for the neural network may be expressed in any arbitrary combination of symbolic or numeric fields in a data base. the developer selects those fields to be used from an interactive menu. the utility scans the selected field entries in the source data base to identify the logical type of each field, and creates a default translate template based on this scan. numeric data is automatically scaled. the developer may use the default template, or edit it from an interactive editor. when editing the template, the developer may select from a menu of commonly used neural network data formats, and from a menu of commonly used primitive mathematical operations. the developer may interactively define additional filters to perform data transformations in series, thus achieving more complex mathematical operations on the data. templates may be edited at any time during the development process. if a network does not appear to be giving satisfactory results, the developer may easily alter the template to present inputs in some other format. dated 1994-05-31"
5319587,computing element for neural networks,"a computing element for use in an array in a neural network. each computing element has k (k>1) input signal terminals, k input backpropagated signal terminals, k output backpropagated signal terminals and at least one output terminal. the input terminals of the computing element located in row i, column j of the array of computing elements receive a sequence of concurrent input signals on k parallel input lines representing a parallel input signal s.sub.ij having vector elements (s.sub.ij1, s.sub.ij2, s.sub.ij3, . . . , s.sub.ijk).sup.t. the k input backpropagated signal terminals are coupled to receive an m-dimensional (m<k) backpropagated signal vector characterized to provide a measure of the performance error of the computing element. the computing element comprises a weighting function means responsive to the concurrent input signal s.sub.ij for computing a k-dimensional weighting coefficient and a scalar activation signal u.sub.ij by computing a k-dimensional weighting-coefficient vector, w.sub.ij =(w.sub.ij1, w.sub.ij2, . . . , w.sub.ijk).sup.t, and by forming the vector inner product of the input signal s.sub.ij vector elements and the weighting-coefficient vector, w.sub.ij. feedback signals x.sub.ijk =w.sub.ijk *p.sub.ij are provided from the output backpropagated signal terminals where p.sub.ij provides a performance error of all columns of computing elements subsequent to this computing element weighted by the gain of the computing element. a nonlinear processor maps successive values of u.sub.ij through a nonlinear mapping function, m.sub.ij to provide a single-valued output signal y.sub.ij. the nonlinear processor responds to an m-dimensional backpropagated signal vector, x.sub.ij+1 =(x.sub.1,j+1,p ; x.sub.2,j+1,p ; x.sub.3,j+1,p ; . . . x.sub.m,j+1,p).sup.t characterized to provide a measure of the performance error of all computing elements subsequent to the computing element. vector x.sub.i,j+1 comprises the pth member of all backpropagated error vectors from the computing elements of column j+1.",1994-06-07,"The title of the patent is computing element for neural networks and its abstract is a computing element for use in an array in a neural network. each computing element has k (k>1) input signal terminals, k input backpropagated signal terminals, k output backpropagated signal terminals and at least one output terminal. the input terminals of the computing element located in row i, column j of the array of computing elements receive a sequence of concurrent input signals on k parallel input lines representing a parallel input signal s.sub.ij having vector elements (s.sub.ij1, s.sub.ij2, s.sub.ij3, . . . , s.sub.ijk).sup.t. the k input backpropagated signal terminals are coupled to receive an m-dimensional (m<k) backpropagated signal vector characterized to provide a measure of the performance error of the computing element. the computing element comprises a weighting function means responsive to the concurrent input signal s.sub.ij for computing a k-dimensional weighting coefficient and a scalar activation signal u.sub.ij by computing a k-dimensional weighting-coefficient vector, w.sub.ij =(w.sub.ij1, w.sub.ij2, . . . , w.sub.ijk).sup.t, and by forming the vector inner product of the input signal s.sub.ij vector elements and the weighting-coefficient vector, w.sub.ij. feedback signals x.sub.ijk =w.sub.ijk *p.sub.ij are provided from the output backpropagated signal terminals where p.sub.ij provides a performance error of all columns of computing elements subsequent to this computing element weighted by the gain of the computing element. a nonlinear processor maps successive values of u.sub.ij through a nonlinear mapping function, m.sub.ij to provide a single-valued output signal y.sub.ij. the nonlinear processor responds to an m-dimensional backpropagated signal vector, x.sub.ij+1 =(x.sub.1,j+1,p ; x.sub.2,j+1,p ; x.sub.3,j+1,p ; . . . x.sub.m,j+1,p).sup.t characterized to provide a measure of the performance error of all computing elements subsequent to the computing element. vector x.sub.i,j+1 comprises the pth member of all backpropagated error vectors from the computing elements of column j+1. dated 1994-06-07"
5319722,neural network for character recognition of rotated characters,"a process for converting characters arranged circularly, as for example about the center hole of a compact disk, into a linear arrangement. points are assigned to locations on the circular arrangement. these points are mapped to a linear arrangement. the number of points is selected to be greater than those of the original image to enhance the resolution of the resulting image. the location of the points is stored in an address array. the values of the pixels in the original image are then copied to a target array. the pixel values are then converted to binary values serving as input to a recognition neural network and a verification neural network.",1994-06-07,"The title of the patent is neural network for character recognition of rotated characters and its abstract is a process for converting characters arranged circularly, as for example about the center hole of a compact disk, into a linear arrangement. points are assigned to locations on the circular arrangement. these points are mapped to a linear arrangement. the number of points is selected to be greater than those of the original image to enhance the resolution of the resulting image. the location of the points is stored in an address array. the values of the pixels in the original image are then copied to a target array. the pixel values are then converted to binary values serving as input to a recognition neural network and a verification neural network. dated 1994-06-07"
5319738,neural network device,"this invention has an object to provide a practical neural network device. the first neural network device of this invention comprises an input circuit for performing predetermined processing of external input information and generating an input signal, an arithmetic processing circuit for performing an arithmetic operation of the input signal in accordance with a plurality of control parameters and generating an output signal, and a control circuit for controlling the control parameters of the arithmetic processing circuit so that the output signal is set to satisfy a predetermined relationship with the input signal, the control circuit including a first cumulative adder for performing cumulative summation of updating amounts of the control parameters for a plurality of proposition patterns supplied as the input signal during learning, and a second cumulative adder for adding currently used control parameter values to values obtained by the first cumulative adder to obtain new control parameter values.",1994-06-07,"The title of the patent is neural network device and its abstract is this invention has an object to provide a practical neural network device. the first neural network device of this invention comprises an input circuit for performing predetermined processing of external input information and generating an input signal, an arithmetic processing circuit for performing an arithmetic operation of the input signal in accordance with a plurality of control parameters and generating an output signal, and a control circuit for controlling the control parameters of the arithmetic processing circuit so that the output signal is set to satisfy a predetermined relationship with the input signal, the control circuit including a first cumulative adder for performing cumulative summation of updating amounts of the control parameters for a plurality of proposition patterns supplied as the input signal during learning, and a second cumulative adder for adding currently used control parameter values to values obtained by the first cumulative adder to obtain new control parameter values. dated 1994-06-07"
5319762,associative memory capable of matching a variable indicator in one string of characters with a portion of another string,""" an associative memory that finds the location of at least one string of characters in the associative memory that matches a string of characters presented sequentially as an input to the associative memory. the string of characters in the associative memory, the input string of characters, or both may include a specially marked characters, or set of characters, that acts as a """"variable indicator."""" the specially marked character, or set of characters, will """"match"""" a portion of the other string. a flag is set in the associative memory at either the starting locations or the ending locations of the matching strings. flags are provided only at locations of stored matching strings of characters found within a selected addressable area or areas. each flag can be moved from a first byte to a second byte in the associative memory that has a predetermined location relative to the first byte. a selection circuit selects one of the matching stored strings of characters by enabling a test signal which selects one of the flags to propagate through the associative memory circuit in a daisy-chain manner. the daisy-chain path is segmented in order to decrease the propagation time of the test signal. a summation circuit, useful in neural network applications, adds a number presented as at least one input byte to the associative memory to a number stored as at least one byte in the associative memory at the location of a stored string of characters that matches the input string. """,1994-06-07,"The title of the patent is associative memory capable of matching a variable indicator in one string of characters with a portion of another string and its abstract is "" an associative memory that finds the location of at least one string of characters in the associative memory that matches a string of characters presented sequentially as an input to the associative memory. the string of characters in the associative memory, the input string of characters, or both may include a specially marked characters, or set of characters, that acts as a """"variable indicator."""" the specially marked character, or set of characters, will """"match"""" a portion of the other string. a flag is set in the associative memory at either the starting locations or the ending locations of the matching strings. flags are provided only at locations of stored matching strings of characters found within a selected addressable area or areas. each flag can be moved from a first byte to a second byte in the associative memory that has a predetermined location relative to the first byte. a selection circuit selects one of the matching stored strings of characters by enabling a test signal which selects one of the flags to propagate through the associative memory circuit in a daisy-chain manner. the daisy-chain path is segmented in order to decrease the propagation time of the test signal. a summation circuit, useful in neural network applications, adds a number presented as at least one input byte to the associative memory to a number stored as at least one byte in the associative memory at the location of a stored string of characters that matches the input string. "" dated 1994-06-07"
5325388,optoelectronic waveguide neural architecture,"gaas base optical waveguide-based structure for a neural network is discld which may form the basic functional building block of a neural architecture in which the waveguide architecture contains at least three electrically active components which are electrically isolated from each other. the waveguide parameters are such that the laser light propagating through the waveguide is a single-mode in both transverse and lateral directions. a superlattice structure is incorporated in the waveguide core and results in electroabsorption of the input laser light. an electric field is supplied to the active components in order to change the transmission properties of the core material, thus modulating the light passing through the core material.",1994-06-28,"The title of the patent is optoelectronic waveguide neural architecture and its abstract is gaas base optical waveguide-based structure for a neural network is discld which may form the basic functional building block of a neural architecture in which the waveguide architecture contains at least three electrically active components which are electrically isolated from each other. the waveguide parameters are such that the laser light propagating through the waveguide is a single-mode in both transverse and lateral directions. a superlattice structure is incorporated in the waveguide core and results in electroabsorption of the input laser light. an electric field is supplied to the active components in order to change the transmission properties of the core material, thus modulating the light passing through the core material. dated 1994-06-28"
5325464,pyramid learning architecture neurocomputer,"the pyramid learning architecture neurocomputer (plan) is a scalable stacked pyramid arrangement of processor arrays. there are six processing levels in plan consisting of the pyramid base, level 6, containing n.sup.2 synapse processors (syps), level 5 containing multiple folded communicating adder tree structures (scats), level 4 made up of n completely connected neuron execution processors (neps), level 3 made up of multiple programmable communicating alu tree (pcats) structures, similar to level 5 scats but with programmable function capabilities in each tree node, level 2 containing the neuron instruction processor (nip), and level 1 comprising the host and user interface. the simplest processors are in the base level with each layer of processors increasing in computational power up to a general purpose host computer acting as the user interface. plan is scalable in direct neural network emulation and in virtual processing capability. consequently, depending upon performance and cost trade-offs, the number of physical neurons n to be implemented is chosen. a neural network model is mapped onto level 3 pcats, level 4 neps, level 5 scats, and level 6 syps. the neuron instruction processor, level 2, controls the neural network model through the level 3 programmable interconnection interface. in addition, the nip level controls the high speed and high capacity plan i/o interface necessary for large n massively parallel systems. this discussion describes the plan processors attached to a host computer and the overall control of the pyramid which constitutes the neurocomputer system.",1994-06-28,"The title of the patent is pyramid learning architecture neurocomputer and its abstract is the pyramid learning architecture neurocomputer (plan) is a scalable stacked pyramid arrangement of processor arrays. there are six processing levels in plan consisting of the pyramid base, level 6, containing n.sup.2 synapse processors (syps), level 5 containing multiple folded communicating adder tree structures (scats), level 4 made up of n completely connected neuron execution processors (neps), level 3 made up of multiple programmable communicating alu tree (pcats) structures, similar to level 5 scats but with programmable function capabilities in each tree node, level 2 containing the neuron instruction processor (nip), and level 1 comprising the host and user interface. the simplest processors are in the base level with each layer of processors increasing in computational power up to a general purpose host computer acting as the user interface. plan is scalable in direct neural network emulation and in virtual processing capability. consequently, depending upon performance and cost trade-offs, the number of physical neurons n to be implemented is chosen. a neural network model is mapped onto level 3 pcats, level 4 neps, level 5 scats, and level 6 syps. the neuron instruction processor, level 2, controls the neural network model through the level 3 programmable interconnection interface. in addition, the nip level controls the high speed and high capacity plan i/o interface necessary for large n massively parallel systems. this discussion describes the plan processors attached to a host computer and the overall control of the pyramid which constitutes the neurocomputer system. dated 1994-06-28"
5327228,system for improving the quality of television pictures using rule based dynamic control,"a system and method for automatically improving the picture quality of television or other video based images, based upon picture content, using improved rule based picture analysis and compensation techniques. the distribution of facial and non-facial tones in a television picture are determined and used to control of picture color quality using neural network techniques. a preferred embodiment of the invention adjusts the brightness, contrast and color saturation of a displayed picture based upon selected portions of the received picture signal. these controls are adjusted every 1/60th of a second (i.e. at the end of each field) to provide an improved visual display. analysis of the video signal in one field of a frame provides the information for adjusting the picture quality of the subsequent field of that frame.",1994-07-05,"The title of the patent is system for improving the quality of television pictures using rule based dynamic control and its abstract is a system and method for automatically improving the picture quality of television or other video based images, based upon picture content, using improved rule based picture analysis and compensation techniques. the distribution of facial and non-facial tones in a television picture are determined and used to control of picture color quality using neural network techniques. a preferred embodiment of the invention adjusts the brightness, contrast and color saturation of a displayed picture based upon selected portions of the received picture signal. these controls are adjusted every 1/60th of a second (i.e. at the end of each field) to provide an improved visual display. analysis of the video signal in one field of a frame provides the information for adjusting the picture quality of the subsequent field of that frame. dated 1994-07-05"
5327357,method of decarburizing molten metal in the refining of steel using neural networks,a method of decarburizing molten metal in the refining of steel using neural networks with a first neural network trained to analyze data representative of many process periods of one or more decarburization operations for providing an oxygen count for a preselected gas ratio of oxygen to diluent gas to cause the temperature of the molten metal bath to be decarburized to rise to a specified aim temperature and with a second neural network trained to analyze data representative of many process periods of one or more decarburization operations for providing an output schedule of oxygen counts to be injected into the bath to reduce the carbon level to a predetermined aim level in one or more successive stages corresponding to a preselected schedule of ratios of oxygen to diluent gas.,1994-07-05,The title of the patent is method of decarburizing molten metal in the refining of steel using neural networks and its abstract is a method of decarburizing molten metal in the refining of steel using neural networks with a first neural network trained to analyze data representative of many process periods of one or more decarburization operations for providing an oxygen count for a preselected gas ratio of oxygen to diluent gas to cause the temperature of the molten metal bath to be decarburized to rise to a specified aim temperature and with a second neural network trained to analyze data representative of many process periods of one or more decarburization operations for providing an output schedule of oxygen counts to be injected into the bath to reduce the carbon level to a predetermined aim level in one or more successive stages corresponding to a preselected schedule of ratios of oxygen to diluent gas. dated 1994-07-05
5329610,neural network employing absolute value calculating synapse,"a neural network employing absolute difference calculating synapse cells comprising a pair of floating gate devices coupled in parallel between an internal cell node and column line of the network. the network further includes a switched-capacitor circuit for summing all of the charges generated by all of the synapse cells within a column of the network. the circuit operates in response to a sequence of applied voltage pulses such that each cell generates a charge representing either the input, the weight, or the minimum/maximum of either the weight or the input. the accumulation of these charges represents the sum of the absolute value difference between the input voltages and the stored weights for a single column of the array.",1994-07-12,"The title of the patent is neural network employing absolute value calculating synapse and its abstract is a neural network employing absolute difference calculating synapse cells comprising a pair of floating gate devices coupled in parallel between an internal cell node and column line of the network. the network further includes a switched-capacitor circuit for summing all of the charges generated by all of the synapse cells within a column of the network. the circuit operates in response to a sequence of applied voltage pulses such that each cell generates a charge representing either the input, the weight, or the minimum/maximum of either the weight or the input. the accumulation of these charges represents the sum of the absolute value difference between the input voltages and the stored weights for a single column of the array. dated 1994-07-12"
5331182,organic light emitting device and preparation and use thereof,"an organic light-emitting device having a light-emitting layer contains the polymer comprising a repeating unit of the formula: equ --z--(x--y).sub.n -- (i) wherein n is at least 2; x is o, s, se or te; y is an aromatic or substituted aromatic group; z is a group containing an imide circle and also having the carrier transport layer and/or the light receiving layer contain the polymer comprising a repeating unit of the following formula: equ --(x--y).sub.n -- (ii) wherein n is at least 2; x is o, s, se or te; y is an aromatic or substituted aromatic group, which is used for a display or a light spatial modulator or an optical neural network system.",1994-07-19,"The title of the patent is organic light emitting device and preparation and use thereof and its abstract is an organic light-emitting device having a light-emitting layer contains the polymer comprising a repeating unit of the formula: equ --z--(x--y).sub.n -- (i) wherein n is at least 2; x is o, s, se or te; y is an aromatic or substituted aromatic group; z is a group containing an imide circle and also having the carrier transport layer and/or the light receiving layer contain the polymer comprising a repeating unit of the following formula: equ --(x--y).sub.n -- (ii) wherein n is at least 2; x is o, s, se or te; y is an aromatic or substituted aromatic group, which is used for a display or a light spatial modulator or an optical neural network system. dated 1994-07-19"
5331215,electrically adaptable neural network with post-processing circuitry,"a synaptic array according to the present invention comprises a plurality of electrically-adaptable elements. electrons may be placed onto and removed from a floating node in each electrically adaptable element associated with at least one mos insulated gate field effect transistor, usually the gate of the transistor, in an analog manner, by application of first and second electrical control signals generated in response to an adapt signal. the inputs to all synaptic elements in a row are connected to a common row input line. adapt inputs to all synaptic elements in a column are connected together to a common column adapt line. the current supplied to all amplifiers in a column is commonly provided by a sense line. in order to adapt the synaptic elements in the m row by n column matrix of the present invention, the voltages to which a given column n of the matrix is to be adapted are placed onto the input voltage lines, and the synaptic elements in column n are then simultaneously adapted by assertion of an adapt signal on the adapt line for column n. the vectors of input voltages for adapting successive columns may be placed sequentially onto the row input voltage lines and used to adapt the columns of synaptic elements by assertion of the adapt signals on the appropriate column adapt lines until the entire array is electrically adapted. after each synaptic element has been adapted, the current flowing through it will be maximized when the voltage at the input of the synaptic element equals the voltage to which the synaptic element has been adapted. an electrically adaptable winner-take-all circuit has its inputs connected to the column-sense lines of the array.",1994-07-19,"The title of the patent is electrically adaptable neural network with post-processing circuitry and its abstract is a synaptic array according to the present invention comprises a plurality of electrically-adaptable elements. electrons may be placed onto and removed from a floating node in each electrically adaptable element associated with at least one mos insulated gate field effect transistor, usually the gate of the transistor, in an analog manner, by application of first and second electrical control signals generated in response to an adapt signal. the inputs to all synaptic elements in a row are connected to a common row input line. adapt inputs to all synaptic elements in a column are connected together to a common column adapt line. the current supplied to all amplifiers in a column is commonly provided by a sense line. in order to adapt the synaptic elements in the m row by n column matrix of the present invention, the voltages to which a given column n of the matrix is to be adapted are placed onto the input voltage lines, and the synaptic elements in column n are then simultaneously adapted by assertion of an adapt signal on the adapt line for column n. the vectors of input voltages for adapting successive columns may be placed sequentially onto the row input voltage lines and used to adapt the columns of synaptic elements by assertion of the adapt signals on the appropriate column adapt lines until the entire array is electrically adapted. after each synaptic element has been adapted, the current flowing through it will be maximized when the voltage at the input of the synaptic element equals the voltage to which the synaptic element has been adapted. an electrically adaptable winner-take-all circuit has its inputs connected to the column-sense lines of the array. dated 1994-07-19"
5331422,video camera having an adaptive automatic iris control circuit,"a video camera includes a lens, an iris and an iris driving circuit, an image sensor, a circuit for dividing the picture into a plurality of sub-areas for extracting the luminance of each sub-area according to the luminance signal provided from the image sensor as a luminance distribution signal, a circuit for generating a signal defining a target value of an iris driving signal, an adaptive circuit using an artificial neural network to which the luminance distribution signal is input for carrying out adaptive conversion so that the offset between a provided teacher signal and its own output is minimized, and a switch for selecting either the target value signal or the output of the adaptive circuit to provide the same as a teacher signal to the adaptive circuit.",1994-07-19,"The title of the patent is video camera having an adaptive automatic iris control circuit and its abstract is a video camera includes a lens, an iris and an iris driving circuit, an image sensor, a circuit for dividing the picture into a plurality of sub-areas for extracting the luminance of each sub-area according to the luminance signal provided from the image sensor as a luminance distribution signal, a circuit for generating a signal defining a target value of an iris driving signal, an adaptive circuit using an artificial neural network to which the luminance distribution signal is input for carrying out adaptive conversion so that the offset between a provided teacher signal and its own output is minimized, and a switch for selecting either the target value signal or the output of the adaptive circuit to provide the same as a teacher signal to the adaptive circuit. dated 1994-07-19"
5331550,application of neural networks as an aid in medical diagnosis and general anomaly detection,"a method for computer-aided detection of anomalies in an image comprise the steps of: (1) dividing the image into a plurality of m.times.n regions; (2) subtracting the background from each of the regions; (3) for each of the regions, selecting a smaller p.times.q subregion; (4) normalizing the p.times.q subregion; (5) feeding the p.times.q subregions into a neural network system, the neural network system having plural member neural networks, each trained to recognize a particular preselected anomaly type; (6) comparing each output value of the plurality of member neural networks to a first threshold; (7) selecting a maximum value from the output values which are greater than the first threshold; (8) comparing the maximum value to a second threshold above which the presence of an anomaly is indicated, and storing the result; (9) clustering a plurality of the stored results to form clusters; and (10) marking the location of the clusters.",1994-07-19,"The title of the patent is application of neural networks as an aid in medical diagnosis and general anomaly detection and its abstract is a method for computer-aided detection of anomalies in an image comprise the steps of: (1) dividing the image into a plurality of m.times.n regions; (2) subtracting the background from each of the regions; (3) for each of the regions, selecting a smaller p.times.q subregion; (4) normalizing the p.times.q subregion; (5) feeding the p.times.q subregions into a neural network system, the neural network system having plural member neural networks, each trained to recognize a particular preselected anomaly type; (6) comparing each output value of the plurality of member neural networks to a first threshold; (7) selecting a maximum value from the output values which are greater than the first threshold; (8) comparing the maximum value to a second threshold above which the presence of an anomaly is indicated, and storing the result; (9) clustering a plurality of the stored results to form clusters; and (10) marking the location of the clusters. dated 1994-07-19"
5333125,optical information processing apparatus having a neural network for inducing an error signal,"an optical information processing apparatus is provided for at least either recording or reproducing information by projecting a converging light beam on a recording medium. the optical information processing apparatus includes a light detector for receiving the light beam, a neural network for inducing a tracking or focusing error signal from an output signal of the light detector, and a servo device for performing a tracking or focusing servo operation in accordance with the tracking or focusing error signal induced by the neural network.",1994-07-26,"The title of the patent is optical information processing apparatus having a neural network for inducing an error signal and its abstract is an optical information processing apparatus is provided for at least either recording or reproducing information by projecting a converging light beam on a recording medium. the optical information processing apparatus includes a light detector for receiving the light beam, a neural network for inducing a tracking or focusing error signal from an output signal of the light detector, and a servo device for performing a tracking or focusing servo operation in accordance with the tracking or focusing error signal induced by the neural network. dated 1994-07-26"
5333210,method and system for pattern analysis using a coarse-coded neural network,"a method and system for performing pattern analysis with a neural network coarse-code a pattern to be analyzed so as to form a plurality of sub-patterns collectively defined by data. each of the sub-patterns comprises sets of sub-pattern data. the neural network includes a plurality of fields, each field being associated with one of the sub-patterns so as to receive the sub-pattern data therefrom. training and testing by the neural network then proceeds in the usual way, with one modification: the transfer function thresholds the value obtained from summing the weighted products of each field over all sub-patterns associated with each pattern being analyzed by the system.",1994-07-26,"The title of the patent is method and system for pattern analysis using a coarse-coded neural network and its abstract is a method and system for performing pattern analysis with a neural network coarse-code a pattern to be analyzed so as to form a plurality of sub-patterns collectively defined by data. each of the sub-patterns comprises sets of sub-pattern data. the neural network includes a plurality of fields, each field being associated with one of the sub-patterns so as to receive the sub-pattern data therefrom. training and testing by the neural network then proceeds in the usual way, with one modification: the transfer function thresholds the value obtained from summing the weighted products of each field over all sub-patterns associated with each pattern being analyzed by the system. dated 1994-07-26"
5333238,method and apparatus for checking input-output characteristic of neural network,"an apparatus for checking the input-output characteristic of a neural network which has an input layer, an intermediate layer and an output layer. plural nodes of the input layer are related to plural nodes of the intermediate layer with plural connection weights while the plural nodes of the intermediate layer are also related to plural nodes of the output layer with plural connection weights. one of the nodes of the input layer is selected as a variable input element while the rest of the nodes are regarded as fixed input elements. a variable data input to the variable input element is varied within a predetermined variation range, while data input to the fixed input elements are fixed, so as to detect how the components of output data from the output layer change. the detected changes of the components of the output data are displayed as the input-output characteristic of the neural network.",1994-07-26,"The title of the patent is method and apparatus for checking input-output characteristic of neural network and its abstract is an apparatus for checking the input-output characteristic of a neural network which has an input layer, an intermediate layer and an output layer. plural nodes of the input layer are related to plural nodes of the intermediate layer with plural connection weights while the plural nodes of the intermediate layer are also related to plural nodes of the output layer with plural connection weights. one of the nodes of the input layer is selected as a variable input element while the rest of the nodes are regarded as fixed input elements. a variable data input to the variable input element is varied within a predetermined variation range, while data input to the fixed input elements are fixed, so as to detect how the components of output data from the output layer change. the detected changes of the components of the output data are displayed as the input-output characteristic of the neural network. dated 1994-07-26"
5333239,learning process system for use with a neural network structure data processing apparatus,"a learning process system is provided for a neural network. the neural network is a layered network comprising an input layer, an intermediate layer and an output layer formed of basic units. in the basic units, a plurality of inputs is multiplied by a weight signal and the products are accumulated, thereby supplying the sum of products. an output signal is obtained using a threshold value function in response to the sum of products. an error signal is generated by an error circuit in response to a difference between the output signal obtained from the output layer and a teacher signal. a weight updating signal is determined in a weight learning circuit by obtaining a weight value in which the sum of the error values falls within an allowable range. thus, the learning is performed in the layered neural network through use of a back propagation method. through such learning in the layered neural network, an updating quantity to be obtained in the present weight updating cycle is determined in response to a once delayed weight updating quantity signal in a previous weight updating cycle and a twice delayed weight updating quantity obtained at a twice-previous weight updating cycle prior to the previous weight updating cycle.",1994-07-26,"The title of the patent is learning process system for use with a neural network structure data processing apparatus and its abstract is a learning process system is provided for a neural network. the neural network is a layered network comprising an input layer, an intermediate layer and an output layer formed of basic units. in the basic units, a plurality of inputs is multiplied by a weight signal and the products are accumulated, thereby supplying the sum of products. an output signal is obtained using a threshold value function in response to the sum of products. an error signal is generated by an error circuit in response to a difference between the output signal obtained from the output layer and a teacher signal. a weight updating signal is determined in a weight learning circuit by obtaining a weight value in which the sum of the error values falls within an allowable range. thus, the learning is performed in the layered neural network through use of a back propagation method. through such learning in the layered neural network, an updating quantity to be obtained in the present weight updating cycle is determined in response to a once delayed weight updating quantity signal in a previous weight updating cycle and a twice delayed weight updating quantity obtained at a twice-previous weight updating cycle prior to the previous weight updating cycle. dated 1994-07-26"
5333240,neural network state diagnostic system for equipment,"a state diagnostic system for equipment is disclosed. the system is constructed of a neural network model for learning in advance one or more samples of information on vibrations, which are produced in a specific operation state of the equipment, in association with the corresponding operation state and obtaining an output signal corresponding to results of the learning when information on vibrations produced upon operation of the equipment is inputted; an input unit for inputting to the neural network model the information on the vibrations produced upon operation of the equipment; and an output unit for outputting the output signal from the neural network model as diagnostic results to a user. state diagnostic methods, learning system, preview/predict system, diagnosis training system, service life estimation assisting system, service life estimation assisting system, and maintenance assisting system are also disclosed.",1994-07-26,"The title of the patent is neural network state diagnostic system for equipment and its abstract is a state diagnostic system for equipment is disclosed. the system is constructed of a neural network model for learning in advance one or more samples of information on vibrations, which are produced in a specific operation state of the equipment, in association with the corresponding operation state and obtaining an output signal corresponding to results of the learning when information on vibrations produced upon operation of the equipment is inputted; an input unit for inputting to the neural network model the information on the vibrations produced upon operation of the equipment; and an output unit for outputting the output signal from the neural network model as diagnostic results to a user. state diagnostic methods, learning system, preview/predict system, diagnosis training system, service life estimation assisting system, service life estimation assisting system, and maintenance assisting system are also disclosed. dated 1994-07-26"
5333241,"neuron unit, neural network and signal processing method","a neuron unit processes a plurality of input signals and outputs an output signal which is indicative of a result of the processing. the neuron unit includes input lines for receiving the input signals, a forward process part including a supplying part for supplying weight functions and an operation part for carrying out an operation on each of the input signals using one of the weight functions and for outputting the output signal, and a self-learning part including a generating part for generating new weight functions based on errors between the output signal of the forward process part and teaching signals and a varying part for varying the weight functions supplied by the supplying part of the forward process part to the new weight functions generated by the generating part.",1994-07-26,"The title of the patent is neuron unit, neural network and signal processing method and its abstract is a neuron unit processes a plurality of input signals and outputs an output signal which is indicative of a result of the processing. the neuron unit includes input lines for receiving the input signals, a forward process part including a supplying part for supplying weight functions and an operation part for carrying out an operation on each of the input signals using one of the weight functions and for outputting the output signal, and a self-learning part including a generating part for generating new weight functions based on errors between the output signal of the forward process part and teaching signals and a varying part for varying the weight functions supplied by the supplying part of the forward process part to the new weight functions generated by the generating part. dated 1994-07-26"
5336937,programmable analog synapse and neural networks incorporating same,"an analog synapse circuit for an artificial neural network requiring less circuitry and interconnections than prior synapses, while affording better weight programming means uses two complementary floating-gate mosfets with tunneling injection in an inverter configuration, with each mosfet storing a weight value. this weight value is set by storing a charge injected by fowler-nordheim tunneling, or other tunneling means, into the floating-gate, which shifts the threshold voltage of the device. a programming line applies a current pulse to the mosfet floating gate to write or erase this stored charge, thereby adjusting the weight of the mosfet. the two mosfets are connected with the gate electrodes connected together and the drain electrodes connected together to provide a common gate and common drain between the two mosfets. an input line is connected to the common gate, and an output line is connected to the common drain. the source electrodes of each mosfet are connected to reference voltages. the synapse circuit may be used in either a feedforword or feedback network, and may be expanded from two to four quadrant operation. the synapse provides a single output current line which represents a function of the input voltage and the stored weights. a plurality of such synapses may be configured in a network, wherein the output lines of each synapse are connected at a current summing node at the input of a neuron. an active load in the input of the neuron allows for both excitatory and inhibitory output current from the synapse circuit.",1994-08-09,"The title of the patent is programmable analog synapse and neural networks incorporating same and its abstract is an analog synapse circuit for an artificial neural network requiring less circuitry and interconnections than prior synapses, while affording better weight programming means uses two complementary floating-gate mosfets with tunneling injection in an inverter configuration, with each mosfet storing a weight value. this weight value is set by storing a charge injected by fowler-nordheim tunneling, or other tunneling means, into the floating-gate, which shifts the threshold voltage of the device. a programming line applies a current pulse to the mosfet floating gate to write or erase this stored charge, thereby adjusting the weight of the mosfet. the two mosfets are connected with the gate electrodes connected together and the drain electrodes connected together to provide a common gate and common drain between the two mosfets. an input line is connected to the common gate, and an output line is connected to the common drain. the source electrodes of each mosfet are connected to reference voltages. the synapse circuit may be used in either a feedforword or feedback network, and may be expanded from two to four quadrant operation. the synapse provides a single output current line which represents a function of the input voltage and the stored weights. a plurality of such synapses may be configured in a network, wherein the output lines of each synapse are connected at a current summing node at the input of a neuron. an active load in the input of the neuron allows for both excitatory and inhibitory output current from the synapse circuit. dated 1994-08-09"
5337370,character recognition method employing non-character recognizer,"a method for recognition of an unconstrained written character image employs at least one non-character recognizer in addition to character recognizers for each candidate character. the input image is rejected if any of the non-character recognizers produces a match. the input image is recognized only if a) a character recognizer produces a match, and b) none of the non-character recognizers produces a match. the non-character recognizers are preferably trained on a) images which are not recognized by the set of character recognizers, b) images which are misclassified by at least one of the set of character recognizers result, and c) incorrectly segmented images. in the preferred embodiment of this invention individual segmented digit images are recognized for automatic sorting of mail by zip code. in the preferred embodiment plural character recognizers and at least one non-character recognizer are embodied in a neural network using a place encoding output. the preferred embodiment includes window encoding of the normalized image and plural cavity feature images derived from the normalized image. a set of encoded values is computed from the ratio of the number of stroke pixels within the overlapping windows divided by the size of the window.",1994-08-09,"The title of the patent is character recognition method employing non-character recognizer and its abstract is a method for recognition of an unconstrained written character image employs at least one non-character recognizer in addition to character recognizers for each candidate character. the input image is rejected if any of the non-character recognizers produces a match. the input image is recognized only if a) a character recognizer produces a match, and b) none of the non-character recognizers produces a match. the non-character recognizers are preferably trained on a) images which are not recognized by the set of character recognizers, b) images which are misclassified by at least one of the set of character recognizers result, and c) incorrectly segmented images. in the preferred embodiment of this invention individual segmented digit images are recognized for automatic sorting of mail by zip code. in the preferred embodiment plural character recognizers and at least one non-character recognizer are embodied in a neural network using a place encoding output. the preferred embodiment includes window encoding of the normalized image and plural cavity feature images derived from the normalized image. a set of encoded values is computed from the ratio of the number of stroke pixels within the overlapping windows divided by the size of the window. dated 1994-08-09"
5337395,spin: a sequential pipeline neurocomputer,"a neural network architecture consisting of input weight multiplications, product summation, neural state calculations, and complete connectivity among the neuron processing elements. neural networks are modelled using a sequential pipelined neurocomputer producing high performance with minimum hardware by sequentially processing each neuron in the completely connected network model. an n neuron network is implemented using multipliers, a pipelined adder tree structure, and activation functions. the activation functions are provided by using one activation function module and sequentially passing the n input product summations sequentially through it. one bus provides n.times.n communications by sequentially providing n neuron values to the multiplier registers. the neuron values are ensured of reaching corresponding multipliers through a tag compare function. the neuron information includes a source tag and a valid signal. higher performance is provided by connecting a number of the neurocomputers in a parallel.",1994-08-09,"The title of the patent is spin: a sequential pipeline neurocomputer and its abstract is a neural network architecture consisting of input weight multiplications, product summation, neural state calculations, and complete connectivity among the neuron processing elements. neural networks are modelled using a sequential pipelined neurocomputer producing high performance with minimum hardware by sequentially processing each neuron in the completely connected network model. an n neuron network is implemented using multipliers, a pipelined adder tree structure, and activation functions. the activation functions are provided by using one activation function module and sequentially passing the n input product summations sequentially through it. one bus provides n.times.n communications by sequentially providing n neuron values to the multiplier registers. the neuron values are ensured of reaching corresponding multipliers through a tag compare function. the neuron information includes a source tag and a valid signal. higher performance is provided by connecting a number of the neurocomputers in a parallel. dated 1994-08-09"
5339818,method for determining blood pressure utilizing a neural network,"a method and device for indirect, quantitative estimation of blood pressure attributes and similar variable physiological parameters utilizing indirect techniques. the method of practice includes (i) generating a sequence of signals which are quantitative dependent upon the variable parameter, (ii) transmitting and processing the signals within a computer system and associated neural network capable of generating a single output signal for the combined input signals, (iii) directly determining an actual value for the parameter concurrent with the indirect generation of signals of the previous steps, (iv) applying weighting factors within the neural network at interconnecting nodes to force the output signal of the neural network to match the true value of the parameter as determined invasively, (v) recording the input signals, weighting factors and true value as training data within memory of the computer, and (vi) repeating the previous steps to develop sufficient training data to enable the neural network to accurately estimate parameter value upon future receipt of on-line input signals. procedures are also described for preclassification of signals and artifact rejection. following training of the neural network, further direct measurement is unnecessary and the system is ready for diagnostic application and noninvasive estimation of parameter values.",1994-08-23,"The title of the patent is method for determining blood pressure utilizing a neural network and its abstract is a method and device for indirect, quantitative estimation of blood pressure attributes and similar variable physiological parameters utilizing indirect techniques. the method of practice includes (i) generating a sequence of signals which are quantitative dependent upon the variable parameter, (ii) transmitting and processing the signals within a computer system and associated neural network capable of generating a single output signal for the combined input signals, (iii) directly determining an actual value for the parameter concurrent with the indirect generation of signals of the previous steps, (iv) applying weighting factors within the neural network at interconnecting nodes to force the output signal of the neural network to match the true value of the parameter as determined invasively, (v) recording the input signals, weighting factors and true value as training data within memory of the computer, and (vi) repeating the previous steps to develop sufficient training data to enable the neural network to accurately estimate parameter value upon future receipt of on-line input signals. procedures are also described for preclassification of signals and artifact rejection. following training of the neural network, further direct measurement is unnecessary and the system is ready for diagnostic application and noninvasive estimation of parameter values. dated 1994-08-23"
5343044,infrared attenuation measuring system,"an infrared attenuation measuring system, for the quantitative determination of the concentration of one or more components in an aqueous fat-containing sample, such as milk, by an infrared attenuation technique. the system comprising a set of waveband-related parameters containing information enabling the system to calculate the concentrations substantially independent of the degree of homogenization of the fat-containing sample, and/or to determine the degree of homogenization of the sample. the parameter set contains parameters which are related to wavebands containing a high degree of information about the homogenization degree of the sample. this may be wavebands containing little information about the chemical components of the sample, and/or wavebands containing a substantial amount of information about the chemical components of the sample. the system is calibrated with aqueous fat-containing samples with different degrees of homogenization. the calculating means may be a microprocessor calculating the concentration of the component or components in question, or a neural network defined by the wave band-related parameters. the wavebands in which the attenuation values are determined in the infrared attenuation measuring means may be defined by optical filters, stationary grating and movable and/or multiple detectors, or by movable grating and one or several stationary or movable detectors. the infrared attenuation measuring means may comprise an interferometer, the data obtained from the interferometer being processed using fourier transformation; and infrared attenuation measuring means of this type may be a ftir spectrophotometer.",1994-08-30,"The title of the patent is infrared attenuation measuring system and its abstract is an infrared attenuation measuring system, for the quantitative determination of the concentration of one or more components in an aqueous fat-containing sample, such as milk, by an infrared attenuation technique. the system comprising a set of waveband-related parameters containing information enabling the system to calculate the concentrations substantially independent of the degree of homogenization of the fat-containing sample, and/or to determine the degree of homogenization of the sample. the parameter set contains parameters which are related to wavebands containing a high degree of information about the homogenization degree of the sample. this may be wavebands containing little information about the chemical components of the sample, and/or wavebands containing a substantial amount of information about the chemical components of the sample. the system is calibrated with aqueous fat-containing samples with different degrees of homogenization. the calculating means may be a microprocessor calculating the concentration of the component or components in question, or a neural network defined by the wave band-related parameters. the wavebands in which the attenuation values are determined in the infrared attenuation measuring means may be defined by optical filters, stationary grating and movable and/or multiple detectors, or by movable grating and one or several stationary or movable detectors. the infrared attenuation measuring means may comprise an interferometer, the data obtained from the interferometer being processed using fourier transformation; and infrared attenuation measuring means of this type may be a ftir spectrophotometer. dated 1994-08-30"
5343251,method and apparatus for classifying patterns of television programs and commercials based on discerning of broadcast audio and video signals,"a method and apparatus for classifying patterns of television programs and commercials, based on learning and discerning of broadcast audio and video signals, wherein the signals might be incomplete and the number of discerning features is large. the apparatus uses a discerner device containing feature extraction, an artificial neural network and control mechanisms. the discerner device classifies signal patterns into classes and stores, records or displays them accordingly on different storage, recording or display devices using the appropriate control mechanisms. the method is operable for classifying many types of signal patterns, but most importantly, those patterns of programs and commercials of television broadcast audio and video signals.",1994-08-30,"The title of the patent is method and apparatus for classifying patterns of television programs and commercials based on discerning of broadcast audio and video signals and its abstract is a method and apparatus for classifying patterns of television programs and commercials, based on learning and discerning of broadcast audio and video signals, wherein the signals might be incomplete and the number of discerning features is large. the apparatus uses a discerner device containing feature extraction, an artificial neural network and control mechanisms. the discerner device classifies signal patterns into classes and stores, records or displays them accordingly on different storage, recording or display devices using the appropriate control mechanisms. the method is operable for classifying many types of signal patterns, but most importantly, those patterns of programs and commercials of television broadcast audio and video signals. dated 1994-08-30"
5345077,method and apparatus for producing a porosity log of a subsurface formation corrected for detector standoff,"a borehole logging tool is lowered into a borehole traversing a subsurface formation and a neutron detector measures the die-away of nuclear radiation in the formation. a model of the die-away of nuclear radiation is produced using exponential terms varying as the sum of borehole, formation and thermal neutron background components. exponentially weighted moments of both the die-away measurements and die-away model are determined and equated. the equated moments are solved for the ratio of the borehole to formation amplitude components of the measurements. the formation decay constant is determined from at least the formation and thermal neutron background terms of the weighted measurement and model moments. an epithermal neutron lifetime is determined from the formation decay constant. this epithermal neutron lifetime and the amplitude ratio are used by a trained neural network to determine a lifetime correction and an apparent standoff. a standoff corrected lifetime is determined from the epithermal neutron lifetime and the lifetime correction. a porosity log of the formation is produced which is corrected for detector standoff from the borehole wall as a function of the standoff corrected epithermal neutron lifetime calibrated in borehole models of known porosities.",1994-09-06,"The title of the patent is method and apparatus for producing a porosity log of a subsurface formation corrected for detector standoff and its abstract is a borehole logging tool is lowered into a borehole traversing a subsurface formation and a neutron detector measures the die-away of nuclear radiation in the formation. a model of the die-away of nuclear radiation is produced using exponential terms varying as the sum of borehole, formation and thermal neutron background components. exponentially weighted moments of both the die-away measurements and die-away model are determined and equated. the equated moments are solved for the ratio of the borehole to formation amplitude components of the measurements. the formation decay constant is determined from at least the formation and thermal neutron background terms of the weighted measurement and model moments. an epithermal neutron lifetime is determined from the formation decay constant. this epithermal neutron lifetime and the amplitude ratio are used by a trained neural network to determine a lifetime correction and an apparent standoff. a standoff corrected lifetime is determined from the epithermal neutron lifetime and the lifetime correction. a porosity log of the formation is produced which is corrected for detector standoff from the borehole wall as a function of the standoff corrected epithermal neutron lifetime calibrated in borehole models of known porosities. dated 1994-09-06"
5345539,radar apparatus using neural network for azimuth and elevation detection,"radar apparatus used for point-source location, where an adaptive feed forward artificial neural network is used to calculate a position vector from image information provided by radar receiving element outputs. where an object is sensed within a field of view of a multiple output radar, then the radar receiving sensor element outputs are processed inputs as image vectors for use in input nodes of an input node layer of the artificial neural network. typically the neural network has the same number of input nodes as the number of sensor element outputs an array within the radar receiver. increased accuracy of point-source location can be achieved by increasing the number of hidden layers used, and/or increasing the number of nodes within each hidden layer. training of the artificial neural network is described for (5.times.1), (1.times.5) and (4.times.4) radar receiving arrays, and also for idealised and noisy data.",1994-09-06,"The title of the patent is radar apparatus using neural network for azimuth and elevation detection and its abstract is radar apparatus used for point-source location, where an adaptive feed forward artificial neural network is used to calculate a position vector from image information provided by radar receiving element outputs. where an object is sensed within a field of view of a multiple output radar, then the radar receiving sensor element outputs are processed inputs as image vectors for use in input nodes of an input node layer of the artificial neural network. typically the neural network has the same number of input nodes as the number of sensor element outputs an array within the radar receiver. increased accuracy of point-source location can be achieved by increasing the number of hidden layers used, and/or increasing the number of nodes within each hidden layer. training of the artificial neural network is described for (5.times.1), (1.times.5) and (4.times.4) radar receiving arrays, and also for idealised and noisy data. dated 1994-09-06"
5347591,method of and device for determining positioning between a hole and a wiring pattern on a printed circuit board by utilizing a set of area values,"a printed board, on which a wiring pattern and a through hole to be inspected are provided, is scanned pixel by pixel and is read optically. the data obtained by scanning is converted into an electric signal to obtain image data. on the basis of the image data thus obtained, a pattern image representing the wiring pattern an a hole image representing the through hole are obtained. a center and a radius of the hole image are obtained from the image data. then, a plurality of ring-shaped masks are obtained by magnifying the hole image at a plurality of magnifications. the size of the pattern image is normalized by the size of the hole image, and the, respective areas of overlapped regions between a plurality of the ringshaped masks and the pattern image are detected. since these areas are obtained with an isotropic method, they do not depend on the directions of a line entering a land. by simulation, correspondence between the area of the overlapped regions and a relative positional relation between the wiring pattern and the through hole is obtained in advance. on the basis of this known correspondence, a relative positional relation between the wiring pattern and the through hole is evaluated by utilizing a neural network that has been taught such known correspondence.",1994-09-13,"The title of the patent is method of and device for determining positioning between a hole and a wiring pattern on a printed circuit board by utilizing a set of area values and its abstract is a printed board, on which a wiring pattern and a through hole to be inspected are provided, is scanned pixel by pixel and is read optically. the data obtained by scanning is converted into an electric signal to obtain image data. on the basis of the image data thus obtained, a pattern image representing the wiring pattern an a hole image representing the through hole are obtained. a center and a radius of the hole image are obtained from the image data. then, a plurality of ring-shaped masks are obtained by magnifying the hole image at a plurality of magnifications. the size of the pattern image is normalized by the size of the hole image, and the, respective areas of overlapped regions between a plurality of the ringshaped masks and the pattern image are detected. since these areas are obtained with an isotropic method, they do not depend on the directions of a line entering a land. by simulation, correspondence between the area of the overlapped regions and a relative positional relation between the wiring pattern and the through hole is obtained in advance. on the basis of this known correspondence, a relative positional relation between the wiring pattern and the through hole is evaluated by utilizing a neural network that has been taught such known correspondence. dated 1994-09-13"
5347613,mos multi-layer neural network including a plurality of hidden layers interposed between synapse groups for performing pattern recognition,"disclosed is a multi-layer neural network and circuit design method. the multi-layer neural network receiving an m-bit input and generating an n-bit output comprises a neuron having a cascaded pair of cmos inverters and having an output node of the preceding cmos inverter among the pair of cmos inverters as its inverted output node and an output node of the succeeding cmos inverter as its non-inverted output node, an input layer having m neurons to receive the m-bit input, an output layer having n neurons to generate the n-bit output, at least one hidden layer provided with n neurons to transfer the input received from the input layer to every upper hidden layer and the output layer, an input synapse group in a matrix having each predetermined weight value to connect each output of neurons on the input layer to each neuron of the output layer and at least one hidden layer, at least one transfer synapse group in a matrix having each predetermined weight value to connect each output of neurons of the hidden layer to each neuron of every upper hidden layer and the output layer, and a bias synapse group for biasing each input node of neurons of the hidden layers and the output layer.",1994-09-13,"The title of the patent is mos multi-layer neural network including a plurality of hidden layers interposed between synapse groups for performing pattern recognition and its abstract is disclosed is a multi-layer neural network and circuit design method. the multi-layer neural network receiving an m-bit input and generating an n-bit output comprises a neuron having a cascaded pair of cmos inverters and having an output node of the preceding cmos inverter among the pair of cmos inverters as its inverted output node and an output node of the succeeding cmos inverter as its non-inverted output node, an input layer having m neurons to receive the m-bit input, an output layer having n neurons to generate the n-bit output, at least one hidden layer provided with n neurons to transfer the input received from the input layer to every upper hidden layer and the output layer, an input synapse group in a matrix having each predetermined weight value to connect each output of neurons on the input layer to each neuron of the output layer and at least one hidden layer, at least one transfer synapse group in a matrix having each predetermined weight value to connect each output of neurons of the hidden layer to each neuron of every upper hidden layer and the output layer, and a bias synapse group for biasing each input node of neurons of the hidden layers and the output layer. dated 1994-09-13"
5349541,method and apparatus utilizing neural networks to predict a specified signal value within a multi-element system,"a method and apparatus for predicting a signal value for a target element within a multi-element system is disclosed. the method includes modeling the multi-element system by defining fundamental physical relationships between the target element and other elements within the system. the resultant system model is in the form of a set of coupled non-linear differential equations. these differential equations are then approximated into linearized models about an operating point or series of operating points corresponding to the system behavior. the linearized differential equations are then subjected to a coupling analysis. the coupling analysis is employed to determine dynamic coupling between instruments. the coupling analysis assesses the degree of observability of the system and associated elements. the coupling analysis may be based upon observability tests, gramian analyses, or modal analyses. based upon the coupling analysis, coupled elements are selected. the coupled elements correspond to system elements which are strongly coupled to the target element. a neural network is then trained using previous process values corresponding to the coupled elements. thereafter, present operating system values corresponding to the coupled elements are fed to the trained neural network. the trained neural network processes the present operating system values to render a predicted value for the target element. this predicted value is then compared to the present system value to determine whether the target element is operating correctly.",1994-09-20,"The title of the patent is method and apparatus utilizing neural networks to predict a specified signal value within a multi-element system and its abstract is a method and apparatus for predicting a signal value for a target element within a multi-element system is disclosed. the method includes modeling the multi-element system by defining fundamental physical relationships between the target element and other elements within the system. the resultant system model is in the form of a set of coupled non-linear differential equations. these differential equations are then approximated into linearized models about an operating point or series of operating points corresponding to the system behavior. the linearized differential equations are then subjected to a coupling analysis. the coupling analysis is employed to determine dynamic coupling between instruments. the coupling analysis assesses the degree of observability of the system and associated elements. the coupling analysis may be based upon observability tests, gramian analyses, or modal analyses. based upon the coupling analysis, coupled elements are selected. the coupled elements correspond to system elements which are strongly coupled to the target element. a neural network is then trained using previous process values corresponding to the coupled elements. thereafter, present operating system values corresponding to the coupled elements are fed to the trained neural network. the trained neural network processes the present operating system values to render a predicted value for the target element. this predicted value is then compared to the present system value to determine whether the target element is operating correctly. dated 1994-09-20"
5349646,signal processing apparatus having at least one neural network,"a signal processing apparatus for controlling an object includes an input unit, a neural network, an output unit, a teaching unit, and an error signal generator for generating a teaching signal that makes the neural network learn in real time. an error signal generator generates an error signal from the teaching signal and information contained in the network output signal. the error signal controls the neural network so that the control output signal has correct control information with respect to the output signal from the controlled object.",1994-09-20,"The title of the patent is signal processing apparatus having at least one neural network and its abstract is a signal processing apparatus for controlling an object includes an input unit, a neural network, an output unit, a teaching unit, and an error signal generator for generating a teaching signal that makes the neural network learn in real time. an error signal generator generates an error signal from the teaching signal and information contained in the network output signal. the error signal controls the neural network so that the control output signal has correct control information with respect to the output signal from the controlled object. dated 1994-09-20"
5350953,digitally weighted neuron for artificial neural network,"a neuron for an artificial neural network provides digital weighting of input signals at a common portion of the neuron rather than at each synapse. the neuron is adapted for use of differential signals, and the weighting may be provided by field effect transistors of different widths, by subtracting a plurality of differential signal components from an opposite most significant component, or by subtracting one half of a differential signal component from the opposite next most significant component. the neuron may provide binary sign selection and digit selection by switching input and reference signals at each synapse.",1994-09-27,"The title of the patent is digitally weighted neuron for artificial neural network and its abstract is a neuron for an artificial neural network provides digital weighting of input signals at a common portion of the neuron rather than at each synapse. the neuron is adapted for use of differential signals, and the weighting may be provided by field effect transistors of different widths, by subtracting a plurality of differential signal components from an opposite most significant component, or by subtracting one half of a differential signal component from the opposite next most significant component. the neuron may provide binary sign selection and digit selection by switching input and reference signals at each synapse. dated 1994-09-27"
5351079,color balance adjusting apparatus using a decorrelating neural network,"a color balance adjusting apparatus for adjusting color imbalance due to the difference in color temperature between a preferable standard illuminant and an illuminant under which a color image is obtained, thereby making the colors of the image substantially identical to those obtained under the preferable standard illuminant. this apparatus comprises a decorrelating neural network for receiving three color component signals correlating with one another and indicative of an image, and minimizing the correlation thereamong, the network having learned so as to minimize the correlation among signals indicative of an image obtained under an illuminant, and a converter for mapping the output of the decorrelating means, into a space of the input image signals, with the use of the inverse matrix of a transfer matrix of the neural network having learned under the preferable standard illuminant. the output of the converter is generated as a signal obtained after white balance adjusting.",1994-09-27,"The title of the patent is color balance adjusting apparatus using a decorrelating neural network and its abstract is a color balance adjusting apparatus for adjusting color imbalance due to the difference in color temperature between a preferable standard illuminant and an illuminant under which a color image is obtained, thereby making the colors of the image substantially identical to those obtained under the preferable standard illuminant. this apparatus comprises a decorrelating neural network for receiving three color component signals correlating with one another and indicative of an image, and minimizing the correlation thereamong, the network having learned so as to minimize the correlation among signals indicative of an image obtained under an illuminant, and a converter for mapping the output of the decorrelating means, into a space of the input image signals, with the use of the inverse matrix of a transfer matrix of the neural network having learned under the preferable standard illuminant. the output of the converter is generated as a signal obtained after white balance adjusting. dated 1994-09-27"
5351311,neural network for detection and correction of local boundary misalignments between images,"a network is provided for the detection and correction of local boundary misalignments in a two-dimensional pixel space between a reference and transformed image. an input layer has input layer sections, each of which contains a plurality of input nodes associated with a cell. the cell is centered on a pixel and divided along a straightline orientation into first and second cell sections. each of the input nodes outputs a digital signal indicative of the presence or absence of a contrast gradient as measured by the two cell sections. a second layer has a plurality of second layer sections, each of which is associated with one of the input layer sections and contains a plurality of second layer nodes. each second layer node is responsive to a combination of input nodes to indicate the presence or absence of a boundary misalignment between the reference and transformed images. presence of a contrast gradient at the combination of nodes defines a local boundary misalignment. a third layer has a plurality of third layer nodes, each of which is associated with one of the second layer sections. each third layer node weights and combines outputs of the second layer nodes to output a signal defining a direction to shift the transformed image perpendicular to the straightline orientation. the third layer are outputs a signal defining the local correction of the local boundary misalignment between the reference and transformed images for the centered pixel.",1994-09-27,"The title of the patent is neural network for detection and correction of local boundary misalignments between images and its abstract is a network is provided for the detection and correction of local boundary misalignments in a two-dimensional pixel space between a reference and transformed image. an input layer has input layer sections, each of which contains a plurality of input nodes associated with a cell. the cell is centered on a pixel and divided along a straightline orientation into first and second cell sections. each of the input nodes outputs a digital signal indicative of the presence or absence of a contrast gradient as measured by the two cell sections. a second layer has a plurality of second layer sections, each of which is associated with one of the input layer sections and contains a plurality of second layer nodes. each second layer node is responsive to a combination of input nodes to indicate the presence or absence of a boundary misalignment between the reference and transformed images. presence of a contrast gradient at the combination of nodes defines a local boundary misalignment. a third layer has a plurality of third layer nodes, each of which is associated with one of the second layer sections. each third layer node weights and combines outputs of the second layer nodes to output a signal defining a direction to shift the transformed image perpendicular to the straightline orientation. the third layer are outputs a signal defining the local correction of the local boundary misalignment between the reference and transformed images for the centered pixel. dated 1994-09-27"
5353207,residual activation neural network,"a plant (72) is operable to receive control inputs c(t) and provide an output y(t). the plant (72) has associated therewith state variables s(t) that are not variable. a control network (74) is provided that accurately models the plant (72). the output of the control network (74) provides a predicted output which is combined with a desired output to generate an error. this error is back propagated through an inverse control network (76), which is the inverse of the control network (74) to generate a control error signal that is input to a distributed control system (73) to vary the control inputs to the plant (72) in order to change the output y(t) to meet the desired output. the control network (74) is comprised of a first network net 1 that is operable to store a representation of the dependency of the control variables on the state variables. the predicted result is subtracted from the actual state variable input and stored as a residual in a residual layer (102). the output of the residual layer (102) is input to a hidden layer (108) which also receives the control inputs to generate a predicted output in an output layer (106). during back propagation of error, the residual values in the residual layer (102) are latched and only the control inputs allowed to vary.",1994-10-04,"The title of the patent is residual activation neural network and its abstract is a plant (72) is operable to receive control inputs c(t) and provide an output y(t). the plant (72) has associated therewith state variables s(t) that are not variable. a control network (74) is provided that accurately models the plant (72). the output of the control network (74) provides a predicted output which is combined with a desired output to generate an error. this error is back propagated through an inverse control network (76), which is the inverse of the control network (74) to generate a control error signal that is input to a distributed control system (73) to vary the control inputs to the plant (72) in order to change the output y(t) to meet the desired output. the control network (74) is comprised of a first network net 1 that is operable to store a representation of the dependency of the control variables on the state variables. the predicted result is subtracted from the actual state variable input and stored as a residual in a residual layer (102). the output of the residual layer (102) is input to a hidden layer (108) which also receives the control inputs to generate a predicted output in an output layer (106). during back propagation of error, the residual values in the residual layer (102) are latched and only the control inputs allowed to vary. dated 1994-10-04"
5353382,programmable synapse for neural network applications,"a synapse for neural network applications providing four quadrant feed-forward and feed-back modes in addition to an outer-product learning capability allowing learning in-situ. the invention, in its preferred embodiment, utilizes a novel two-transistor implementation which permits each synapse to be built in an integrated circuit chip surface area of only 20 by 20 micrometers. one of the two transistors at each synapse of the present invention comprises a floating gate structure composed of a floating gate electrode and a control electrode which permits learning upon application of incident ultraviolet light. during ultraviolet light application, a floating gate electrode voltage may be altered to modify the weight of each synapse in accordance with preselected criteria, based upon the input and output weight change vector elements corresponding to that particular matrix element. the second transistor corresponding to each synapse of the present invention provides a novel method for applying a voltage to the control electrode of the aforementioned floating gate structure of the first transistor. the voltage applied to the control electrode and thus the proportionate change in the floating gate electrode of the first transistor may be made proportional to the product of the corresponding input weight change vector element and the corresponding output weight change vector element, by using slope controllable ramp generators and phase controllable pulse generators, only one set of which must be provided for the entire matrix of synapses herein disclosed.",1994-10-04,"The title of the patent is programmable synapse for neural network applications and its abstract is a synapse for neural network applications providing four quadrant feed-forward and feed-back modes in addition to an outer-product learning capability allowing learning in-situ. the invention, in its preferred embodiment, utilizes a novel two-transistor implementation which permits each synapse to be built in an integrated circuit chip surface area of only 20 by 20 micrometers. one of the two transistors at each synapse of the present invention comprises a floating gate structure composed of a floating gate electrode and a control electrode which permits learning upon application of incident ultraviolet light. during ultraviolet light application, a floating gate electrode voltage may be altered to modify the weight of each synapse in accordance with preselected criteria, based upon the input and output weight change vector elements corresponding to that particular matrix element. the second transistor corresponding to each synapse of the present invention provides a novel method for applying a voltage to the control electrode of the aforementioned floating gate structure of the first transistor. the voltage applied to the control electrode and thus the proportionate change in the floating gate electrode of the first transistor may be made proportional to the product of the corresponding input weight change vector element and the corresponding output weight change vector element, by using slope controllable ramp generators and phase controllable pulse generators, only one set of which must be provided for the entire matrix of synapses herein disclosed. dated 1994-10-04"
5353383,neural network circuit,"a neural network circuit including a number n of weight coefficients (w1-wn) corresponding to a number n of inputs, subtraction circuits for determining the difference between inputs and the weight coefficients in each input terminal, the result thereof being inputted into absolute value circuits, all calculation results of the absolute value circuits corresponding to the inputs and the weight coefficients being inputted into an addition circuit and accumulated, and this accumulation result determining the output value. a threshold value circuit determines the final output value, according to a step function pattern, a polygonal line pattern, or a sigmoid function pattern, depending on the object. in the case in which a neural network circuit is realized by means of digital circuits, the absolute value circuits can include simply ex-or logic (exclusive or) gates. furthermore, in the case in which the input terminals have two input paths and two weight coefficients corresponding to each input path, the neuron circuits form a recognition area having a flexible shape which is controlled by the weight coefficients.",1994-10-04,"The title of the patent is neural network circuit and its abstract is a neural network circuit including a number n of weight coefficients (w1-wn) corresponding to a number n of inputs, subtraction circuits for determining the difference between inputs and the weight coefficients in each input terminal, the result thereof being inputted into absolute value circuits, all calculation results of the absolute value circuits corresponding to the inputs and the weight coefficients being inputted into an addition circuit and accumulated, and this accumulation result determining the output value. a threshold value circuit determines the final output value, according to a step function pattern, a polygonal line pattern, or a sigmoid function pattern, depending on the object. in the case in which a neural network circuit is realized by means of digital circuits, the absolute value circuits can include simply ex-or logic (exclusive or) gates. furthermore, in the case in which the input terminals have two input paths and two weight coefficients corresponding to each input path, the neuron circuits form a recognition area having a flexible shape which is controlled by the weight coefficients. dated 1994-10-04"
5354957,artificially intelligent traffic modeling and prediction system,"a system for allocating hall calls in a group of elevators includes a plurality of neural network modules to model, learn and predict passenger arrival rates and passenger destination probabilities. the models learn the traffic occurring in a building by inputting to the neural networks traffic data previously stored. the neural networks then adjust their internal structure to make historic predictions based on data of the previous day and real time predictions based on data of the last ten minutes. the predictions of arrival rates are combined to provide optimum predictions. from every set of historic car calls and the optimum arrival rates, a matrix is constructed which stores entries representing the number of passengers with the same intended destination for each hall call. the traffic predictions are used separately or in combination by a group control to improve operating cost computations and car allocation, thereby reducing the travelling and waiting times of current and future passengers.",1994-10-11,"The title of the patent is artificially intelligent traffic modeling and prediction system and its abstract is a system for allocating hall calls in a group of elevators includes a plurality of neural network modules to model, learn and predict passenger arrival rates and passenger destination probabilities. the models learn the traffic occurring in a building by inputting to the neural networks traffic data previously stored. the neural networks then adjust their internal structure to make historic predictions based on data of the previous day and real time predictions based on data of the last ten minutes. the predictions of arrival rates are combined to provide optimum predictions. from every set of historic car calls and the optimum arrival rates, a matrix is constructed which stores entries representing the number of passengers with the same intended destination for each hall call. the traffic predictions are used separately or in combination by a group control to improve operating cost computations and car allocation, thereby reducing the travelling and waiting times of current and future passengers. dated 1994-10-11"
5355313,neural network interpretation of aeromagnetic data,"a method for determining depth to basement from aeromagnetic data utilizes neural networks to automate the laborious process of profile interpretation. the neural networks provide consistency, accuracy and overall quality without bias of interpretation.",1994-10-11,"The title of the patent is neural network interpretation of aeromagnetic data and its abstract is a method for determining depth to basement from aeromagnetic data utilizes neural networks to automate the laborious process of profile interpretation. the neural networks provide consistency, accuracy and overall quality without bias of interpretation. dated 1994-10-11"
5355434,method and apparatus for performing learning in a neural network,"an apparatus for carrying out learning operation of a neural network which has an input layer, an intermediate layer and an output layer. plural nodes of the input layer are related to plural nodes of the intermediate layer with plural connection weights, while the plural nodes of the intermediate layer are also related to plural nodes of the output layer with plural connection weights. although input data composed of plural components and teaching data composed of plural components are used in learning operation, some of the components are ineffective for the purpose of learning operation. during error calculation between output data from the neural network and the teaching data, the apparatus judges whether each of the components of the teaching data is effective or ineffective, and output errors corresponding to the ineffective components are regarded as zero. the connection weights are thereafter corrected based upon the calculated errors. the apparatus further comprises means for adding new input data and teaching data into a data base, and means for calculating a degree of heterogeneousness of the new teaching data. the new input data and teaching data are added to the data base only when the degree of heterogeneousness is smaller than a predetermined value.",1994-10-11,"The title of the patent is method and apparatus for performing learning in a neural network and its abstract is an apparatus for carrying out learning operation of a neural network which has an input layer, an intermediate layer and an output layer. plural nodes of the input layer are related to plural nodes of the intermediate layer with plural connection weights, while the plural nodes of the intermediate layer are also related to plural nodes of the output layer with plural connection weights. although input data composed of plural components and teaching data composed of plural components are used in learning operation, some of the components are ineffective for the purpose of learning operation. during error calculation between output data from the neural network and the teaching data, the apparatus judges whether each of the components of the teaching data is effective or ineffective, and output errors corresponding to the ineffective components are regarded as zero. the connection weights are thereafter corrected based upon the calculated errors. the apparatus further comprises means for adding new input data and teaching data into a data base, and means for calculating a degree of heterogeneousness of the new teaching data. the new input data and teaching data are added to the data base only when the degree of heterogeneousness is smaller than a predetermined value. dated 1994-10-11"
5355436,single layer neural network circuit for performing linearly separable and non-linearly separable logical operations,"a neural network provides both linearly separable and non-linearly separable logic operations, including the exclusive-or operation, on input signals in a single layer of circuits. the circuit weights the input signals with complex weights by multiplication and addition, and provides weighted signals to a neuron circuit (a neuron body or soma) which provides an output corresponding to the desired logical operation.",1994-10-11,"The title of the patent is single layer neural network circuit for performing linearly separable and non-linearly separable logical operations and its abstract is a neural network provides both linearly separable and non-linearly separable logic operations, including the exclusive-or operation, on input signals in a single layer of circuits. the circuit weights the input signals with complex weights by multiplication and addition, and provides weighted signals to a neuron circuit (a neuron body or soma) which provides an output corresponding to the desired logical operation. dated 1994-10-11"
5355437,neural network architecture for pattern recognition,"a data processing system according to the present invention provides a plurality of neural layers with neuron groups, each neuron group having a fixed number of neurons. the neurons of the neuron group each have an output coupled to a neuron of an adjacent neuron layer. each neuron layer has a plurality of neuron groups, and each neuron group has at least one neuron which also belongs to another neuron group, resulting in an overlap in the neuron groups. the number of neurons in the nth neural layer is determined on the basis of the number of neurons in the (n-1)th layer, the size of the neuron groups, and the degree of overlap between the adjacent neuron groups. in a variation of the data processing system of the present invention, the data processing system comprises a plurality of mutually independent data processing portions, each of which comprises a plurality of neural layers.",1994-10-11,"The title of the patent is neural network architecture for pattern recognition and its abstract is a data processing system according to the present invention provides a plurality of neural layers with neuron groups, each neuron group having a fixed number of neurons. the neurons of the neuron group each have an output coupled to a neuron of an adjacent neuron layer. each neuron layer has a plurality of neuron groups, and each neuron group has at least one neuron which also belongs to another neuron group, resulting in an overlap in the neuron groups. the number of neurons in the nth neural layer is determined on the basis of the number of neurons in the (n-1)th layer, the size of the neuron groups, and the degree of overlap between the adjacent neuron groups. in a variation of the data processing system of the present invention, the data processing system comprises a plurality of mutually independent data processing portions, each of which comprises a plurality of neural layers. dated 1994-10-11"
5355438,weighting and thresholding circuit for a neural network,"an analog circuit which performs weighting and thresholding for a neural network. each neuron of the neural network includes an operational amplifier receiving an input signal, the output of which is connected to a transistor. the transistor conducts only when the output exceeds a predetermined value thereby providing a threshold function. the output of the transistor is connected by a variable resistance to other inputs of other neurons. the output of each operational amplifier thereby corresponds to a weighted version of the input signal, which is adjusted for threshold and is also dependent on other neurons of the network.",1994-10-11,"The title of the patent is weighting and thresholding circuit for a neural network and its abstract is an analog circuit which performs weighting and thresholding for a neural network. each neuron of the neural network includes an operational amplifier receiving an input signal, the output of which is connected to a transistor. the transistor conducts only when the output exceeds a predetermined value thereby providing a threshold function. the output of the transistor is connected by a variable resistance to other inputs of other neurons. the output of each operational amplifier thereby corresponds to a weighted version of the input signal, which is adjusted for threshold and is also dependent on other neurons of the network. dated 1994-10-11"
5359699,method for using a feed forward neural network to perform classification with highly biased data,""" an artificial neural network detects points in feature space outside of a boundary determined by a set of sample data. the network is trained using pseudo data which compensates for the lack of original data representing """"abnormal"""" or novel combinations of features. the training process is done iteratively using a net bias parameter to close the boundary around the sample data. when the neural net stabilizes, the training process is complete. pseudo data is chosen using several disclosed methods. """,1994-10-25,"The title of the patent is method for using a feed forward neural network to perform classification with highly biased data and its abstract is "" an artificial neural network detects points in feature space outside of a boundary determined by a set of sample data. the network is trained using pseudo data which compensates for the lack of original data representing """"abnormal"""" or novel combinations of features. the training process is done iteratively using a net bias parameter to close the boundary around the sample data. when the neural net stabilizes, the training process is complete. pseudo data is chosen using several disclosed methods. "" dated 1994-10-25"
5359700,neural network incorporating difference neurons,"an artificial neural network incorporating difference type, non-mp (mccullough-pitts) neuron cells and a method and apparatus for training this network. more specifically, the output of each neuron cell is a nonlinear mapping of a distance metric of a difference vector and an offset. the difference vector is the difference between an input and a reference vector; the offset is representative of the radius of a hyperspheroidal discriminant function.",1994-10-25,"The title of the patent is neural network incorporating difference neurons and its abstract is an artificial neural network incorporating difference type, non-mp (mccullough-pitts) neuron cells and a method and apparatus for training this network. more specifically, the output of each neuron cell is a nonlinear mapping of a distance metric of a difference vector and an offset. the difference vector is the difference between an input and a reference vector; the offset is representative of the radius of a hyperspheroidal discriminant function. dated 1994-10-25"
5361201,real estate appraisal using predictive modeling,"an automated real estate appraisal system (100) and method generates estimates of real estate value using a predictive model such as a neural network (908). the predictive model (908) generates these estimates based on learned relationships among variables describing individual property characteristics (905) as well as general neighborhood characteristics at various levels of geographic specificity (906). the system (100) may also output reason codes indicating relative contributions (1009) of various variables to a particular result, and may generate reports (701) describing property valuations, market trend analyses, property conformity information, and recommendations regarding loans based on risk related to a property.",1994-11-01,"The title of the patent is real estate appraisal using predictive modeling and its abstract is an automated real estate appraisal system (100) and method generates estimates of real estate value using a predictive model such as a neural network (908). the predictive model (908) generates these estimates based on learned relationships among variables describing individual property characteristics (905) as well as general neighborhood characteristics at various levels of geographic specificity (906). the system (100) may also output reason codes indicating relative contributions (1009) of various variables to a particular result, and may generate reports (701) describing property valuations, market trend analyses, property conformity information, and recommendations regarding loans based on risk related to a property. dated 1994-11-01"
5361311,automated recongition of characters using optical filtering with positive and negative functions encoding pattern and relevance information,"a method and apparatus is described for recognition of hand printed characters using pairs of positive and negative correlative functions (pncfs), the pncfs including both pattern and relevance information, implemented by optical elements. a set of optical elements having varying optical density corresponding to a set of two-dimensional pncfs is generated. a pattern of illumination responsive to the image of the character to be identified is simultaneously transmitted through each of the optical elements implementing the pncfs. the amount of light transmitted through each of the elements is measured, providing a transmission coefficient. the transmission coefficients are the inputs to a neural network, such that the inputs to the neural network are a set of transmission coefficients resulting from transmission of light corresponding to a character to be identified through a complete set of optical elements implementing a set of pncfs. the neural network calculates weighted sums of the transmission coefficients. the neural network may be implemented as a network of resistors connected between input nodes, intermediate nodes, and output nodes. the output node having the highest voltage identifies the character to be identified.",1994-11-01,"The title of the patent is automated recongition of characters using optical filtering with positive and negative functions encoding pattern and relevance information and its abstract is a method and apparatus is described for recognition of hand printed characters using pairs of positive and negative correlative functions (pncfs), the pncfs including both pattern and relevance information, implemented by optical elements. a set of optical elements having varying optical density corresponding to a set of two-dimensional pncfs is generated. a pattern of illumination responsive to the image of the character to be identified is simultaneously transmitted through each of the optical elements implementing the pncfs. the amount of light transmitted through each of the elements is measured, providing a transmission coefficient. the transmission coefficients are the inputs to a neural network, such that the inputs to the neural network are a set of transmission coefficients resulting from transmission of light corresponding to a character to be identified through a complete set of optical elements implementing a set of pncfs. the neural network calculates weighted sums of the transmission coefficients. the neural network may be implemented as a network of resistors connected between input nodes, intermediate nodes, and output nodes. the output node having the highest voltage identifies the character to be identified. dated 1994-11-01"
5361326,enhanced interface for a neural network engine,"an interface for a neural network includes a generalized data translator and a certainty filter in the data path including the neural network for rendering a decision on raw data, possibly from a data processing application. the data translator is controlled with user-definable parameters and procedures contained in a property list in order to manipulate translation, truncation, mapping (including weighting) and other transformations of the raw data. the neuron to which the output of the data translator is applied is controlled by a code index contained in an action list. an external certainty threshold is also provided, preferably by the action list to filter the output of the neural network. the core program used with the conexns neurons for system maintenance also includes further core operations and size maintenance operations responsive to commands from the user of an application to cause operations to be performed with in the neural network as well as to create and update the property and action lists.",1994-11-01,"The title of the patent is enhanced interface for a neural network engine and its abstract is an interface for a neural network includes a generalized data translator and a certainty filter in the data path including the neural network for rendering a decision on raw data, possibly from a data processing application. the data translator is controlled with user-definable parameters and procedures contained in a property list in order to manipulate translation, truncation, mapping (including weighting) and other transformations of the raw data. the neuron to which the output of the data translator is applied is controlled by a code index contained in an action list. an external certainty threshold is also provided, preferably by the action list to filter the output of the neural network. the core program used with the conexns neurons for system maintenance also includes further core operations and size maintenance operations responsive to commands from the user of an application to cause operations to be performed with in the neural network as well as to create and update the property and action lists. dated 1994-11-01"
5361327,"waveform equalizer apparatus formed of neural network, and method of designing same","a waveform equalizer for reducing distortion of a digital signal produced from a digital data recording and playback system or transmission system is formed of a neural network having fixed weighting coefficients. respective values for the coefficients are established by generating a corresponding simulated neuron network, by software implementation using a computer, and by executing a neuron network learning operation using input values obtained from a distorted digital signal and teaching values obtained from an original digital signal which resulted in the distorted digital signal.",1994-11-01,"The title of the patent is waveform equalizer apparatus formed of neural network, and method of designing same and its abstract is a waveform equalizer for reducing distortion of a digital signal produced from a digital data recording and playback system or transmission system is formed of a neural network having fixed weighting coefficients. respective values for the coefficients are established by generating a corresponding simulated neuron network, by software implementation using a computer, and by executing a neuron network learning operation using input values obtained from a distorted digital signal and teaching values obtained from an original digital signal which resulted in the distorted digital signal. dated 1994-11-01"
5361328,data processing system using a neural network,a data processing system comprising a plurality of neural layers characterized in that a part of neurons in one of the layers are connected to neurons in the following layer.,1994-11-01,The title of the patent is data processing system using a neural network and its abstract is a data processing system comprising a plurality of neural layers characterized in that a part of neurons in one of the layers are connected to neurons in the following layer. dated 1994-11-01
5361628,system and method for processing test measurements collected from an internal combustion engine for diagnostic purposes,""" a system and method for processing test measurements collected from an internal combustion engine that is cold-tested for diagnostic purposes. test measurements are collected from an engine. the test measurements are """"pre-processed"""" by filtering and subsampling techniques so that principle component analysis can be applied to condense the quantity of test measurements while still retaining a statistically accurate indication of a majority of the original measurements. the pre-processed test measurements are then passed through one or more classifiers including: a neural network classifier, a fuzzy logic classifier, a cluster-based classifier (or """"spherical"""" classifier) and a genetic program classifier. results from these classifiers can be used to obtain a verdict about an engine (i.e., whether the engine is """"normal"""" or """"faulty""""). """,1994-11-08,"The title of the patent is system and method for processing test measurements collected from an internal combustion engine for diagnostic purposes and its abstract is "" a system and method for processing test measurements collected from an internal combustion engine that is cold-tested for diagnostic purposes. test measurements are collected from an engine. the test measurements are """"pre-processed"""" by filtering and subsampling techniques so that principle component analysis can be applied to condense the quantity of test measurements while still retaining a statistically accurate indication of a majority of the original measurements. the pre-processed test measurements are then passed through one or more classifiers including: a neural network classifier, a fuzzy logic classifier, a cluster-based classifier (or """"spherical"""" classifier) and a genetic program classifier. results from these classifiers can be used to obtain a verdict about an engine (i.e., whether the engine is """"normal"""" or """"faulty""""). "" dated 1994-11-08"
5364668,spatial light modulator and a neural network circuit,"the present invention relates to a spatial light modulator used for an optical computing system or display, which comprises a photoconductive layer having a rectification function for receiving incident lights to generate charges, an electrode for accumulating the charges and a ferroelectric liquid crystal layer for modulating the incident light according to bias voltage change with the accumulated charges, wherein the ferroelectric liquid crystal layer is arranged between a pair of alignment layers made of polyimide represented by the general formula (i); ##str1## wherein n is 2 or more, x is s, se or te, y is an aromatic group or a substituted aromatic group, and z is a group containing an aromatic group. in the spatial light modulator, any charges are not accumulated on the alignment layer with the driving time and thus a bistable memory condition can be realized.",1994-11-15,"The title of the patent is spatial light modulator and a neural network circuit and its abstract is the present invention relates to a spatial light modulator used for an optical computing system or display, which comprises a photoconductive layer having a rectification function for receiving incident lights to generate charges, an electrode for accumulating the charges and a ferroelectric liquid crystal layer for modulating the incident light according to bias voltage change with the accumulated charges, wherein the ferroelectric liquid crystal layer is arranged between a pair of alignment layers made of polyimide represented by the general formula (i); ##str1## wherein n is 2 or more, x is s, se or te, y is an aromatic group or a substituted aromatic group, and z is a group containing an aromatic group. in the spatial light modulator, any charges are not accumulated on the alignment layer with the driving time and thus a bistable memory condition can be realized. dated 1994-11-15"
5365158,driving control apparatus for induction motor,"a slip frequency type driving control apparatus includes a detector for detecting an input current and an input voltage to an induction motor, a calculating unit for calculating a rotor flux and a torque current of the induction motor, a neural network for outputting an excitation current command value and a slip frequency command value, and a vector control unit for performing vector control for the induction motor. the neural network receives a rotor flux command value, a torque current command value, a calculated rotor flux value, and a calculated torque current value, performs learning on the basis of a back-propagation law using signals output in correspondence with these inputs, and outputs an excitation current command value and a slip frequency command value. the vector control unit detects the torque current and the excitation current on the basis of the output slip frequency command value from the neural network and controls the induction motor in accordance with a deviation between the detected excitation current value and the excitation current command value and a deviation between the detected torque current value and the torque current command value.",1994-11-15,"The title of the patent is driving control apparatus for induction motor and its abstract is a slip frequency type driving control apparatus includes a detector for detecting an input current and an input voltage to an induction motor, a calculating unit for calculating a rotor flux and a torque current of the induction motor, a neural network for outputting an excitation current command value and a slip frequency command value, and a vector control unit for performing vector control for the induction motor. the neural network receives a rotor flux command value, a torque current command value, a calculated rotor flux value, and a calculated torque current value, performs learning on the basis of a back-propagation law using signals output in correspondence with these inputs, and outputs an excitation current command value and a slip frequency command value. the vector control unit detects the torque current and the excitation current on the basis of the output slip frequency command value from the neural network and controls the induction motor in accordance with a deviation between the detected excitation current value and the excitation current command value and a deviation between the detected torque current value and the torque current command value. dated 1994-11-15"
5365455,method and apparatus for automatic nucleic acid sequence determination,"a method and system for automated nucleic acid sequence determination of a polynucleotide, wherein a nucleic acid sequencing ladder comprises signals corresponding to oligonucleotides formed from the polynucleotide, comprising the step of correlating, particularly in a trained neural network or a scatter plot, an intensity variable for each signal in the nucleic acid sequencing ladder with an informative variable for that signal, wherein the informative variable comprises information from at least two adjacent signals in the nucleic acid sequencing ladder, such that each signal in the nucleic acid sequencing ladder identified so as to determine the nucleic acid sequence corresponding to the polynucleotide. in particular, the relative separation between consecutive signals, the relative intensities between consecutive signals, and a pattern recognition factor, which incorporates a comparison of relative separations and intensities of at least two adjacent signals with pattern recognition templates, can be used as informative variables. furthermore, this invention relates to a method and system for the on-the-fly resolution and extraction of information of signals contained in a digitized data stream involving calculation of the smoothed second derivative of a data point from the smoothed first derivative of the data point.",1994-11-15,"The title of the patent is method and apparatus for automatic nucleic acid sequence determination and its abstract is a method and system for automated nucleic acid sequence determination of a polynucleotide, wherein a nucleic acid sequencing ladder comprises signals corresponding to oligonucleotides formed from the polynucleotide, comprising the step of correlating, particularly in a trained neural network or a scatter plot, an intensity variable for each signal in the nucleic acid sequencing ladder with an informative variable for that signal, wherein the informative variable comprises information from at least two adjacent signals in the nucleic acid sequencing ladder, such that each signal in the nucleic acid sequencing ladder identified so as to determine the nucleic acid sequence corresponding to the polynucleotide. in particular, the relative separation between consecutive signals, the relative intensities between consecutive signals, and a pattern recognition factor, which incorporates a comparison of relative separations and intensities of at least two adjacent signals with pattern recognition templates, can be used as informative variables. furthermore, this invention relates to a method and system for the on-the-fly resolution and extraction of information of signals contained in a digitized data stream involving calculation of the smoothed second derivative of a data point from the smoothed first derivative of the data point. dated 1994-11-15"
5365460,neural network signal processor,"an apparatus and method using a neural network processor for target detection is described. an array of injection mode infrared detectors, whose output signals convey intensity change information of detected objects in a pulse train output form is combined with a frequency division multiplexer to apply the information to a minimum number of multiplexed channels, and transmit the detector output signals to the processor in a continuous mode. a multi-layer neural network processor is used to localize global information and concentrate on areas of interest through matrix transformation applied by the various neural layers.",1994-11-15,"The title of the patent is neural network signal processor and its abstract is an apparatus and method using a neural network processor for target detection is described. an array of injection mode infrared detectors, whose output signals convey intensity change information of detected objects in a pulse train output form is combined with a frequency division multiplexer to apply the information to a minimum number of multiplexed channels, and transmit the detector output signals to the processor in a continuous mode. a multi-layer neural network processor is used to localize global information and concentrate on areas of interest through matrix transformation applied by the various neural layers. dated 1994-11-15"
5367612,neurocontrolled adaptive process control system,"an adaptive process control system selectively controls vibrations in a given medium in real time. unwanted vibrations present at a point being monitored in a given medium are sensed, and the system generates an appropriate offsetting vibration that is applied to the medium at a convenient location, which may be remote from the monitored point. the system includes a vibration sensor, such as one or more accelerometers, that sense both input and output vibrations present within the medium; at least one vibration generator, such as an electromagnetic shaker, that generates appropriate offsetting vibrations that are applied to the medium at one or more appropriate locations; and a neural network controller that controls the vibration generator(s) so as to force the sensed vibration at the monitored point(s) to a desired level. the adaptive vibration cancellation provided by the invention takes place in real time, and without the need to process time-consuming complex mathematical algorithms. a specific embodiment of the neural network controller includes a plurality of 4-layer neural networks configured in an adaptive filtered-x configuration.",1994-11-22,"The title of the patent is neurocontrolled adaptive process control system and its abstract is an adaptive process control system selectively controls vibrations in a given medium in real time. unwanted vibrations present at a point being monitored in a given medium are sensed, and the system generates an appropriate offsetting vibration that is applied to the medium at a convenient location, which may be remote from the monitored point. the system includes a vibration sensor, such as one or more accelerometers, that sense both input and output vibrations present within the medium; at least one vibration generator, such as an electromagnetic shaker, that generates appropriate offsetting vibrations that are applied to the medium at one or more appropriate locations; and a neural network controller that controls the vibration generator(s) so as to force the sensed vibration at the monitored point(s) to a desired level. the adaptive vibration cancellation provided by the invention takes place in real time, and without the need to process time-consuming complex mathematical algorithms. a specific embodiment of the neural network controller includes a plurality of 4-layer neural networks configured in an adaptive filtered-x configuration. dated 1994-11-22"
5369503,method of picture compression by auto-organisation of a neural network,"the method of picture compression consists: pa1 in processing a current displaced pixel block in the picture to be compressed by using a network of n neurons, 1 to n, disposed in a single plane, each neuron having a number of inputs equal to the dimension of the blocks extracted from the picture, and such that, after a learning phase during which the network has auto-organised so that the neurons represent most probable input values, each neuron receives on its inputs, at the same instant, the pixels of the current block and outputs a signal, or potential, d.sub.j, which is positive and is all the lower the nearer is the state of the corresponding neuron j to the state of the inputs e.sub.i, pa1 in coding the current pixel block through a code associated with the neuron j whose potential d.sub.j is minimal. the invention applies, in particular, to the compression of television pictures.",1994-11-29,"The title of the patent is method of picture compression by auto-organisation of a neural network and its abstract is the method of picture compression consists: pa1 in processing a current displaced pixel block in the picture to be compressed by using a network of n neurons, 1 to n, disposed in a single plane, each neuron having a number of inputs equal to the dimension of the blocks extracted from the picture, and such that, after a learning phase during which the network has auto-organised so that the neurons represent most probable input values, each neuron receives on its inputs, at the same instant, the pixels of the current block and outputs a signal, or potential, d.sub.j, which is positive and is all the lower the nearer is the state of the corresponding neuron j to the state of the inputs e.sub.i, pa1 in coding the current pixel block through a code associated with the neuron j whose potential d.sub.j is minimal. the invention applies, in particular, to the compression of television pictures. dated 1994-11-29"
5369731,asynchronous control system for a neuro computer,"an asynchronous control system for a neuro computer, includes an inter-connected type neural network composed of a plurality of neurons for multiplying a plurality of input signals with corresponding weights, calculating a total sum-of-products of the input signals and weight, thereby providing the sum-of product signals, and converting the sum-of-product signal using a non-linear function. a weight memory is provided for storing data of the weights for said neurons, and a controller is provided for generating a control pattern which controls the neural network. a selector randomly selects one of the neurons which performs signal processing during one processing cycle.",1994-11-29,"The title of the patent is asynchronous control system for a neuro computer and its abstract is an asynchronous control system for a neuro computer, includes an inter-connected type neural network composed of a plurality of neurons for multiplying a plurality of input signals with corresponding weights, calculating a total sum-of-products of the input signals and weight, thereby providing the sum-of product signals, and converting the sum-of-product signal using a non-linear function. a weight memory is provided for storing data of the weights for said neurons, and a controller is provided for generating a control pattern which controls the neural network. a selector randomly selects one of the neurons which performs signal processing during one processing cycle. dated 1994-11-29"
5369773,neural network using virtual-zero,"a virtual-zero architecture is intended for use in a single instruction stream, multiple data stream (simd) processor which includes an input bus, an input unit, manipulation units, an output unit and an output bus. the virtual-zero architecture includes a memory unit (40) for storing data, an arithmetic unit (42) for mathematically operating on the data, a memory address generation unit (32) and an adder for computing a next memory address. the memory address generation unit (32) includes an address register (34) in the memory unit for identifying the address of a particular data block, a counter (38) for counting the number of memory addresses in a particular data block, and a rotation register (36) for providing a data-void address in the memory unit if and only if all of the entries in the data block are zero. the memory (40) and the address (32) units provide zero-value data blocks to the arithmetic unit (44) to simulate the data block having the data-void address during processing. the architecture may also be used to selectively handle input to a system.",1994-11-29,"The title of the patent is neural network using virtual-zero and its abstract is a virtual-zero architecture is intended for use in a single instruction stream, multiple data stream (simd) processor which includes an input bus, an input unit, manipulation units, an output unit and an output bus. the virtual-zero architecture includes a memory unit (40) for storing data, an arithmetic unit (42) for mathematically operating on the data, a memory address generation unit (32) and an adder for computing a next memory address. the memory address generation unit (32) includes an address register (34) in the memory unit for identifying the address of a particular data block, a counter (38) for counting the number of memory addresses in a particular data block, and a rotation register (36) for providing a data-void address in the memory unit if and only if all of the entries in the data block are zero. the memory (40) and the address (32) units provide zero-value data blocks to the arithmetic unit (44) to simulate the data block having the data-void address during processing. the architecture may also be used to selectively handle input to a system. dated 1994-11-29"
5371808,automated recognition of characters using optical filtering with maximum uncertainty - minimum variance (mumv) functions,"a method and apparatus is described for recognition of hand printed characters using maximum uncertainty--minimum variance (mumv) functions, such as gabor functions, implemented by optical elements. a set of optical elements having varying optical density corresponding to a set of two-dimensional mumv functions is generated. a pattern of illumination responsive to the image of the character to be identified is simultaneously transmitted through each of the optical elements implementing the mumv functions. the amount of light transmitted through each of the elements is measured, providing a transmission coefficient. such transmission coefficients are used as a set of inputs to a neural network, such that the inputs to the neural network are a set of transmission coefficients resulting from transmission of light corresponding to a character to be identified through a complete set of optical elements implementing a set of two-dimensional mumv functions. the neural network calculates weighted sums of the transmission coefficients. the neural network may be implemented as a network of resistors connected between input nodes, intermediate nodes, and output nodes. the output node having the highest voltage identifies the character to be identified.",1994-12-06,"The title of the patent is automated recognition of characters using optical filtering with maximum uncertainty - minimum variance (mumv) functions and its abstract is a method and apparatus is described for recognition of hand printed characters using maximum uncertainty--minimum variance (mumv) functions, such as gabor functions, implemented by optical elements. a set of optical elements having varying optical density corresponding to a set of two-dimensional mumv functions is generated. a pattern of illumination responsive to the image of the character to be identified is simultaneously transmitted through each of the optical elements implementing the mumv functions. the amount of light transmitted through each of the elements is measured, providing a transmission coefficient. such transmission coefficients are used as a set of inputs to a neural network, such that the inputs to the neural network are a set of transmission coefficients resulting from transmission of light corresponding to a character to be identified through a complete set of optical elements implementing a set of two-dimensional mumv functions. the neural network calculates weighted sums of the transmission coefficients. the neural network may be implemented as a network of resistors connected between input nodes, intermediate nodes, and output nodes. the output node having the highest voltage identifies the character to be identified. dated 1994-12-06"
5371809,neural network for improved classification of patterns which adds a best performing trial branch node to the network,""" each processing element has a number of weights for each input connection. these weights are coefficients of a polynomial equation. the use of quadratic nodes permits discrimination between body pixel and edge pixels, in which an intermediate value is present, using a grey scale image. in the training method of the present invention, the middle layer is initially one leaf node which is connected to each output node. the contribution of each leaf node to the total output error is determined and the weights of the inputs to the leaf nodes are adjusted to minimize the error. the leaf node that has the best chance of improving the total output error is then """"converted"""" into a branch node with two leaves. a branch node selected from a pool of trial branch nodes is used to replace the chosen leaf node. the trial branch nodes are then trained by gradient training to optimize the branch error function. from the set of trial branch nodes, the best performing node is selected and is substituted for the previously-selected leaf node. two new leaf nodes are then created from the newly-substituted best-performing-branch node. a leaf node is accepted or rejected based upon the number of times it was activated related to the correctness of the classification. once a leaf node is rejected, it is eliminated from any further operation, thereby minimizing the size of the network. integer mathematics can be generated within the network so that a separate floating point coprocessor is not required. """,1994-12-06,"The title of the patent is neural network for improved classification of patterns which adds a best performing trial branch node to the network and its abstract is "" each processing element has a number of weights for each input connection. these weights are coefficients of a polynomial equation. the use of quadratic nodes permits discrimination between body pixel and edge pixels, in which an intermediate value is present, using a grey scale image. in the training method of the present invention, the middle layer is initially one leaf node which is connected to each output node. the contribution of each leaf node to the total output error is determined and the weights of the inputs to the leaf nodes are adjusted to minimize the error. the leaf node that has the best chance of improving the total output error is then """"converted"""" into a branch node with two leaves. a branch node selected from a pool of trial branch nodes is used to replace the chosen leaf node. the trial branch nodes are then trained by gradient training to optimize the branch error function. from the set of trial branch nodes, the best performing node is selected and is substituted for the previously-selected leaf node. two new leaf nodes are then created from the newly-substituted best-performing-branch node. a leaf node is accepted or rejected based upon the number of times it was activated related to the correctness of the classification. once a leaf node is rejected, it is eliminated from any further operation, thereby minimizing the size of the network. integer mathematics can be generated within the network so that a separate floating point coprocessor is not required. "" dated 1994-12-06"
5371834,adaptive neuron model--an architecture for the rapid learning of nonlinear topological transformations,"a method and an apparatus for the rapid learning of nonlinear mappings and topological transformations using a dynamically reconfigurable artificial neural network is presented. this fully-recurrent adaptive neuron model (anm) network has been applied to the highly degenerative inverse kinematics problem in robotics, and its performance evaluation is bench-marked. once trained, the resulting neuromorphic architecture was implemented in custom analog neural network hardware and the parameters capturing the functional transformation downloaded onto the system. this neuroprocessor, capable of 10.sup.9 ops/sec, was interfaced directly to a three degree of freedom heathkit robotic manipulator. calculation of the hardware feed-forward pass for this mapping was benchmarked at .apprxeq.10 .mu.sec.",1994-12-06,"The title of the patent is adaptive neuron model--an architecture for the rapid learning of nonlinear topological transformations and its abstract is a method and an apparatus for the rapid learning of nonlinear mappings and topological transformations using a dynamically reconfigurable artificial neural network is presented. this fully-recurrent adaptive neuron model (anm) network has been applied to the highly degenerative inverse kinematics problem in robotics, and its performance evaluation is bench-marked. once trained, the resulting neuromorphic architecture was implemented in custom analog neural network hardware and the parameters capturing the functional transformation downloaded onto the system. this neuroprocessor, capable of 10.sup.9 ops/sec, was interfaced directly to a three degree of freedom heathkit robotic manipulator. calculation of the hardware feed-forward pass for this mapping was benchmarked at .apprxeq.10 .mu.sec. dated 1994-12-06"
5371835,inductively coupled neural network,a neural coupling includes a pair of primary and secondary members for coupling a precedent node to a subsequent node to synaptically transmit a signal through a neural network. a primary member responds to a signal from a precedent node to generate an inductive field according to the received signal. a secondary member spatially coupled to the primary member by the inductive field supplies to a subsequent node a corresponding signal in response to the inductive field to effect synaptic transmission of the signal through the neural network.,1994-12-06,The title of the patent is inductively coupled neural network and its abstract is a neural coupling includes a pair of primary and secondary members for coupling a precedent node to a subsequent node to synaptically transmit a signal through a neural network. a primary member responds to a signal from a precedent node to generate an inductive field according to the received signal. a secondary member spatially coupled to the primary member by the inductive field supplies to a subsequent node a corresponding signal in response to the inductive field to effect synaptic transmission of the signal through the neural network. dated 1994-12-06
5372015,air conditioner controller,"a controller for controlling an air conditioner comprises a room temperature sensor for measuring a room temperature, a room temperature change computation circuit for computing a difference between the room temperature and a set temperature, an outdoor temperature sensor for measuring an outdoor temperature, a neural network for computing load on the air conditioner according to the temperature change, outdoor temperature, and room temperature, and a controller for controlling, according to the computed load, the operation frequency of a coolant compressor by controlling a compressor motor through a variable frequency circuit, thereby speedily bringing the room temperature to the set temperature and stably maintaining the room temperature at the set temperature. the neural network learns various operation characteristics of a refrigerating cycle of the air conditioner, and according to a result of the learning, controls the air conditioner.",1994-12-13,"The title of the patent is air conditioner controller and its abstract is a controller for controlling an air conditioner comprises a room temperature sensor for measuring a room temperature, a room temperature change computation circuit for computing a difference between the room temperature and a set temperature, an outdoor temperature sensor for measuring an outdoor temperature, a neural network for computing load on the air conditioner according to the temperature change, outdoor temperature, and room temperature, and a controller for controlling, according to the computed load, the operation frequency of a coolant compressor by controlling a compressor motor through a variable frequency circuit, thereby speedily bringing the room temperature to the set temperature and stably maintaining the room temperature at the set temperature. the neural network learns various operation characteristics of a refrigerating cycle of the air conditioner, and according to a result of the learning, controls the air conditioner. dated 1994-12-13"
5373452,intangible sensor and method for making same,"an intangible sensor for measuring intangible properties of a substance and a method for making the sensor is described. the intangible sensor may be embodied in a mapping neural network model. the intangible sensor herein is a device that quantitatively measures complex intangible properties of a sample of a substance. the term intangible implies a subjective connotation such as in the taste, creaminess or softness of a substance or product and therefore can only be subjectively defined. although an intangible property is known to be a function of certain measurable physical properties of a substance, there are no known definitions of this function. the intangible sensor herein can implement this function simply without having any detailed knowledge of or making any analysis of the function.",1994-12-13,"The title of the patent is intangible sensor and method for making same and its abstract is an intangible sensor for measuring intangible properties of a substance and a method for making the sensor is described. the intangible sensor may be embodied in a mapping neural network model. the intangible sensor herein is a device that quantitatively measures complex intangible properties of a sample of a substance. the term intangible implies a subjective connotation such as in the taste, creaminess or softness of a substance or product and therefore can only be subjectively defined. although an intangible property is known to be a function of certain measurable physical properties of a substance, there are no known definitions of this function. the intangible sensor herein can implement this function simply without having any detailed knowledge of or making any analysis of the function. dated 1994-12-13"
5373566,neural network-based diacritical marker recognition system and method,"a diacritical marker recognition system and method recognizes diacritical markers in a character image based upon an analysis by a neural network of the portion of the character image most likely to contain a diacritical marker. once the neural network determines that a diacritical marker most likely exists in the character image, the system determines by using heuristics whether a diacritical marker exists or whether the character image appears to contain a diacritical marker which is actually a regular character.",1994-12-13,"The title of the patent is neural network-based diacritical marker recognition system and method and its abstract is a diacritical marker recognition system and method recognizes diacritical markers in a character image based upon an analysis by a neural network of the portion of the character image most likely to contain a diacritical marker. once the neural network determines that a diacritical marker most likely exists in the character image, the system determines by using heuristics whether a diacritical marker exists or whether the character image appears to contain a diacritical marker which is actually a regular character. dated 1994-12-13"
5375250,method of intelligent computing and neural-like processing of time and space functions,"the contents of an organic memory at an address a are multiplied by a quality factor q and then added to a signal id from a sensor or an external source of data, selected randomly and in real time, multiplied by a weight factor w. the sum obtained in this way is stored back at the same address a. the process is repeated for different values of id, w, q and a in a sequence that represents an evolution program. the result is a neural network with individually programmable, exponential output activation functions. meanwhile the contents of the organic memory are simultaneously scanned by a cpu from a second port without the need for special access protocols or timing changes in either the organic memory or the cpu. protocol-free, multi-ported memories facilitate the interfacing of multiprocessors and an address bus expansion circuit further reduces the need to access slow peripheral memories.",1994-12-20,"The title of the patent is method of intelligent computing and neural-like processing of time and space functions and its abstract is the contents of an organic memory at an address a are multiplied by a quality factor q and then added to a signal id from a sensor or an external source of data, selected randomly and in real time, multiplied by a weight factor w. the sum obtained in this way is stored back at the same address a. the process is repeated for different values of id, w, q and a in a sequence that represents an evolution program. the result is a neural network with individually programmable, exponential output activation functions. meanwhile the contents of the organic memory are simultaneously scanned by a cpu from a second port without the need for special access protocols or timing changes in either the organic memory or the cpu. protocol-free, multi-ported memories facilitate the interfacing of multiprocessors and an address bus expansion circuit further reduces the need to access slow peripheral memories. dated 1994-12-20"
5376962,neural network video image processor,""" a signal processing system for a video camera uses a single neural network to implement multiple nonlinear signal processing functions. in one example, the neural network implements gamma correction, contrast compression, color correction, high pass filtering and aperture correction as a combined function which is emulated by the network. the network is trained off-line using back propagation to emulate the entire composite function for a set of parameters which results in multiple sets of weighting factors. then, using the stored multiple sets of weighting factors as initial values, the neural network is """"re-trained"""" on-line for each new parameter setting. the use of a single neural network in place of the multiple dedicated processing functions reduces engineering effort to develop the product and may reduce the cost of the total system. """,1994-12-27,"The title of the patent is neural network video image processor and its abstract is "" a signal processing system for a video camera uses a single neural network to implement multiple nonlinear signal processing functions. in one example, the neural network implements gamma correction, contrast compression, color correction, high pass filtering and aperture correction as a combined function which is emulated by the network. the network is trained off-line using back propagation to emulate the entire composite function for a set of parameters which results in multiple sets of weighting factors. then, using the stored multiple sets of weighting factors as initial values, the neural network is """"re-trained"""" on-line for each new parameter setting. the use of a single neural network in place of the multiple dedicated processing functions reduces engineering effort to develop the product and may reduce the cost of the total system. "" dated 1994-12-27"
5376963,neural network video image processor,"a signal processing system for a video camera uses a single neural network to implement multiple nonlinear signal processing functions. in one example, the neural network implements gamma correction and contrast compression, in another example, color correction and aperture correction are added to the combined function emulated by the network. the network is trained using back propagation to emulate one function then a combination of two functions, then a combination of three functions, and so on. the programmed neural network replaces multiple pipelined signal processors in the video camera. the use of a single neural network in place of the multiple dedicated processing functions reduces engineering effort to develop the product and may reduce the cost of the total system.",1994-12-27,"The title of the patent is neural network video image processor and its abstract is a signal processing system for a video camera uses a single neural network to implement multiple nonlinear signal processing functions. in one example, the neural network implements gamma correction and contrast compression, in another example, color correction and aperture correction are added to the combined function emulated by the network. the network is trained using back propagation to emulate one function then a combination of two functions, then a combination of three functions, and so on. the programmed neural network replaces multiple pipelined signal processors in the video camera. the use of a single neural network in place of the multiple dedicated processing functions reduces engineering effort to develop the product and may reduce the cost of the total system. dated 1994-12-27"
5377108,method for predicting impact and an impact prediction system for realizing the same by using neural networks,"a method for predicting impact by using neural networks comprising steps of supplying a predetermined crash curve to a first neural network having an intermediate layer to train said first neural network by means of learning calculation and supplying a predetermined air bag deployment limit curve to a second neural network to train said second neural network, supplying data indicative of crash curve obtained by an acceleration sensing device on collision to said first and second neural networks, predicting in said first neural network a time instance at which a threshold displacement is going to reach based on the basis of the training result in said first neural network, comparing in said second neural network data indicative of crash curve on said collision and said air bag deployment limit curve to produce a decision signal of deploying the air bag according to the comparison result, calculating said decision signal and said time instance, to deploy the air bag depending on the impact, and supplying an operation command signal to an air bag deployment operation device.",1994-12-27,"The title of the patent is method for predicting impact and an impact prediction system for realizing the same by using neural networks and its abstract is a method for predicting impact by using neural networks comprising steps of supplying a predetermined crash curve to a first neural network having an intermediate layer to train said first neural network by means of learning calculation and supplying a predetermined air bag deployment limit curve to a second neural network to train said second neural network, supplying data indicative of crash curve obtained by an acceleration sensing device on collision to said first and second neural networks, predicting in said first neural network a time instance at which a threshold displacement is going to reach based on the basis of the training result in said first neural network, comparing in said second neural network data indicative of crash curve on said collision and said air bag deployment limit curve to produce a decision signal of deploying the air bag according to the comparison result, calculating said decision signal and said time instance, to deploy the air bag depending on the impact, and supplying an operation command signal to an air bag deployment operation device. dated 1994-12-27"
5377302,system for recognizing speech,"a pattern recognition system particularly useful for recognizing speech or handwriting. an input signal is first filtered by a filter bank having two stages where the outputs of the first stage is fed forward to the second stage of a significant number of filters and the output of the second stage is fed back to the first stage of a significant number of the filters. such feedback enhances the signal-to-noise ratio and resembles the coupling between the different sections of the basilar membrane of the cochlear. the output of the filter bank is a two-dimensional frequency-time representation of the original signal. a second set of filters which takes as input two-dimensional signals, detects the presence of elementary tonotopic features such as the onset, rise, fall and frequency of any significant tones in a speech signal. a third set of filters detects any contrasts in the elementary features at various levels of resolution. after such filtering, a neural network is employed to learn patterns formed from the multi-resolution contrasts in the identified features so that the system recognizes symbols from an input signal that is continuous in time. in the case of speech, the system recognizes continuous speech in a speaker-independent manner, and is also tolerant of noise.",1994-12-27,"The title of the patent is system for recognizing speech and its abstract is a pattern recognition system particularly useful for recognizing speech or handwriting. an input signal is first filtered by a filter bank having two stages where the outputs of the first stage is fed forward to the second stage of a significant number of filters and the output of the second stage is fed back to the first stage of a significant number of the filters. such feedback enhances the signal-to-noise ratio and resembles the coupling between the different sections of the basilar membrane of the cochlear. the output of the filter bank is a two-dimensional frequency-time representation of the original signal. a second set of filters which takes as input two-dimensional signals, detects the presence of elementary tonotopic features such as the onset, rise, fall and frequency of any significant tones in a speech signal. a third set of filters detects any contrasts in the elementary features at various levels of resolution. after such filtering, a neural network is employed to learn patterns formed from the multi-resolution contrasts in the identified features so that the system recognizes symbols from an input signal that is continuous in time. in the case of speech, the system recognizes continuous speech in a speaker-independent manner, and is also tolerant of noise. dated 1994-12-27"
5377305,outer product neural network,an outer product neural network provides a predetermined number of processing elements for extracting principal components of an input vector. the residual vector of the network when cascaded into a similar outer product neural network provides additional principal components defining a subspace orthogonal to the subspace defined by the principal components of the first network.,1994-12-27,The title of the patent is outer product neural network and its abstract is an outer product neural network provides a predetermined number of processing elements for extracting principal components of an input vector. the residual vector of the network when cascaded into a similar outer product neural network provides additional principal components defining a subspace orthogonal to the subspace defined by the principal components of the first network. dated 1994-12-27
5377307,system and method of global optimization using artificial neural networks,"a method of global optimization of complex, highly nonlinear, multivariant systems is described. an artificial neural network (ann) is trained to create an approximate inverse model. the desired behavior for a particular system is then input to the inverse model to derive approximate model parameters for the particular system. optimization of the approximate model parameters yields optimal model parameters. the method is applied to the synthesis of mechanical linkages where examples of a type of linkage mechanism are used to train an ann and derive the approximate inverse model. inverse models for a number of linkage mechanism types are derived and stored. for a linkage mechanism with unknown linkage parameters, a power spectrum representation of the coupler curve is developed and the inverse model for the type of linkage mechanism retrieved. the representation of the desired coupler curve is input and the approximate linkage parameters derived. optimization further refines the linkage parameters.",1994-12-27,"The title of the patent is system and method of global optimization using artificial neural networks and its abstract is a method of global optimization of complex, highly nonlinear, multivariant systems is described. an artificial neural network (ann) is trained to create an approximate inverse model. the desired behavior for a particular system is then input to the inverse model to derive approximate model parameters for the particular system. optimization of the approximate model parameters yields optimal model parameters. the method is applied to the synthesis of mechanical linkages where examples of a type of linkage mechanism are used to train an ann and derive the approximate inverse model. inverse models for a number of linkage mechanism types are derived and stored. for a linkage mechanism with unknown linkage parameters, a power spectrum representation of the coupler curve is developed and the inverse model for the type of linkage mechanism retrieved. the representation of the desired coupler curve is input and the approximate linkage parameters derived. optimization further refines the linkage parameters. dated 1994-12-27"
5381513,time series signal analyzer including neural network having path groups corresponding to states of markov chains,"a neural network with a high recognition rate when applied to static patterns is made applicable to dynamic time series patterns such as voice signals. plural units with one or more inputs and outputs are interconnected, and a unique load coefficient is assigned to each connection to weight the signals flowing through that connection. the neural network includes an input unit group to which are input the components of plural vectors included in the input feature vector series {y(t)}; an output unit which outputs the converted vectors, which are produced by passing the input vectors through each unit and the associated connections; and j paths from input unit group to the output unit group. the units are connected to form a hidden markov model wherein each signal path identified as j=1, 2, . . . , j corresponds to the same state.",1995-01-10,"The title of the patent is time series signal analyzer including neural network having path groups corresponding to states of markov chains and its abstract is a neural network with a high recognition rate when applied to static patterns is made applicable to dynamic time series patterns such as voice signals. plural units with one or more inputs and outputs are interconnected, and a unique load coefficient is assigned to each connection to weight the signals flowing through that connection. the neural network includes an input unit group to which are input the components of plural vectors included in the input feature vector series {y(t)}; an output unit which outputs the converted vectors, which are produced by passing the input vectors through each unit and the associated connections; and j paths from input unit group to the output unit group. the units are connected to form a hidden markov model wherein each signal path identified as j=1, 2, . . . , j corresponds to the same state. dated 1995-01-10"
5381515,two layer neural network comprised of neurons with improved input range and input offset,"a two-layer network according to the present invention is comprised of a first-layer array of electrically-adaptable synaptic elements, inter-layer connection circuitry comprised of electrically adaptable elements, and a second-layer array of electrically-adaptable synaptic elements. electrons may be placed onto and removed from a floating node associated with at least one mos transistor in each electrically adaptable element, usually comprising the gate of the transistor, in an analog manner, by application of first and second electrical control signals. a first electrical control signal controls the injection of electrons onto the floating node from an electron injection structure and the second electrical control signal controls the removal of electrons from the floating node by an electron removal structure. each synaptic element in the synaptic array comprises an adaptable cmos inverter or other amplifier circuit. the inputs to all first-layer synaptic elements in a row are connected to a common row input line. adapt inputs to all synaptic elements in a column are connected together to a common column adapt line. the outputs of all first layer synaptic elements in a column are connected to a common sense amplifier on a sense line. the outputs of the sense amplifiers are connected to the inputs of the synaptic elements of the second layer of the array. the outputs of all synaptic elements in a given row in the second layer of the array are connected to a common row output line. in order to adapt the synaptic elements in the array, the voltages to which the synaptic elements in a given column of the first layer of the array is to be adapted are placed onto the input voltage lines, and the synaptic elements in column n are then simultaneously adapted by assertion of an adapt signal on the adapt line for the column. the voltages to which the synaptic elements of the second layer of the array are to be adapted are placed on the row outputs lines.",1995-01-10,"The title of the patent is two layer neural network comprised of neurons with improved input range and input offset and its abstract is a two-layer network according to the present invention is comprised of a first-layer array of electrically-adaptable synaptic elements, inter-layer connection circuitry comprised of electrically adaptable elements, and a second-layer array of electrically-adaptable synaptic elements. electrons may be placed onto and removed from a floating node associated with at least one mos transistor in each electrically adaptable element, usually comprising the gate of the transistor, in an analog manner, by application of first and second electrical control signals. a first electrical control signal controls the injection of electrons onto the floating node from an electron injection structure and the second electrical control signal controls the removal of electrons from the floating node by an electron removal structure. each synaptic element in the synaptic array comprises an adaptable cmos inverter or other amplifier circuit. the inputs to all first-layer synaptic elements in a row are connected to a common row input line. adapt inputs to all synaptic elements in a column are connected together to a common column adapt line. the outputs of all first layer synaptic elements in a column are connected to a common sense amplifier on a sense line. the outputs of the sense amplifiers are connected to the inputs of the synaptic elements of the second layer of the array. the outputs of all synaptic elements in a given row in the second layer of the array are connected to a common row output line. in order to adapt the synaptic elements in the array, the voltages to which the synaptic elements in a given column of the first layer of the array is to be adapted are placed onto the input voltage lines, and the synaptic elements in column n are then simultaneously adapted by assertion of an adapt signal on the adapt line for the column. the voltages to which the synaptic elements of the second layer of the array are to be adapted are placed on the row outputs lines. dated 1995-01-10"
5383042,3 layer liquid crystal neural network with output layer displaying error value for optical weight updating,"an optical information processor for use as a matrix vector multiplier comprises a vector input spatial light modulator (1) and an optically addressed weight matrix spatial light modulator (3). a read beam (10) passes through the input modulator (1) and the weight modulator and onto a combined output transducer and error spatial light modulator (5). the error modulator (5) is then controlled in accordance with the difference between a target output vector and the output vector from the transducer (5), and modulates an update beam (11) which then passes through the input modulator (1) and onto the weight modulator (3). the weight modulator (3) represents a two-dimensional array of optical attenuation values which are updated in accordance with the optical radiation incident thereon during updating.",1995-01-17,"The title of the patent is 3 layer liquid crystal neural network with output layer displaying error value for optical weight updating and its abstract is an optical information processor for use as a matrix vector multiplier comprises a vector input spatial light modulator (1) and an optically addressed weight matrix spatial light modulator (3). a read beam (10) passes through the input modulator (1) and the weight modulator and onto a combined output transducer and error spatial light modulator (5). the error modulator (5) is then controlled in accordance with the difference between a target output vector and the output vector from the transducer (5), and modulates an update beam (11) which then passes through the input modulator (1) and onto the weight modulator (3). the weight modulator (3) represents a two-dimensional array of optical attenuation values which are updated in accordance with the optical radiation incident thereon during updating. dated 1995-01-17"
5384895,self-organizing neural network for classifying pattern signatures with `a posteriori` conditional class probability,"a self-organizing neural network and method for classifying a pattern signature having n-features is provided. the network provides a posteriori conditional class probability that the pattern signature belongs to a selected class from a plurality of classes with which the neural network was trained. in its training mode, a plurality of training vectors is processed to generate an n-feature, n-dimensional space defined by a set of non-overlapping trained clusters. each training vector has n-feature coordinates and a class coordinate. each trained cluster has a center and a radius defined by a vigilance parameter. the center of each trained cluster is a reference vector that represents a recursive mean of the n-feature coordinates from training vectors bounded by a corresponding trained cluster. each reference vector defines a fractional probability associated with the selected class based upon a ratio of i) a count of training vectors from the selected class that are bounded by the corresponding trained cluster to ii) a total count of training vectors bounded by the corresponding trained cluster. in the exercise mode, an input vector defines the pattern signature to be classified. the input vector has n-feature coordinates associated with an unknown class. one of the reference vectors is selected so as to minimize differences with the n-feature coordinates of the input vector. the fractional probability of the selected one of the reference vectors is the a posteriori conditional class probability that the input vector belongs to the selected class.",1995-01-24,"The title of the patent is self-organizing neural network for classifying pattern signatures with `a posteriori` conditional class probability and its abstract is a self-organizing neural network and method for classifying a pattern signature having n-features is provided. the network provides a posteriori conditional class probability that the pattern signature belongs to a selected class from a plurality of classes with which the neural network was trained. in its training mode, a plurality of training vectors is processed to generate an n-feature, n-dimensional space defined by a set of non-overlapping trained clusters. each training vector has n-feature coordinates and a class coordinate. each trained cluster has a center and a radius defined by a vigilance parameter. the center of each trained cluster is a reference vector that represents a recursive mean of the n-feature coordinates from training vectors bounded by a corresponding trained cluster. each reference vector defines a fractional probability associated with the selected class based upon a ratio of i) a count of training vectors from the selected class that are bounded by the corresponding trained cluster to ii) a total count of training vectors bounded by the corresponding trained cluster. in the exercise mode, an input vector defines the pattern signature to be classified. the input vector has n-feature coordinates associated with an unknown class. one of the reference vectors is selected so as to minimize differences with the n-feature coordinates of the input vector. the fractional probability of the selected one of the reference vectors is the a posteriori conditional class probability that the input vector belongs to the selected class. dated 1995-01-24"
5386149,data synapse expressing unit capable of refreshing stored synapse load,"a synapse expressing unit includes a capacitor for storing a synapse load value information in a form of electric charges, and a refresh control circuit for remedying the change in the amount of the electric charges stored in the capacitor. the refresh control circuit includes a comparator for comparing a potential at an electrode of the capacitor and a reference potential, and a drive circuit responsive to the comparator for recovering the electric charges of the capacitor through charge pumping operation. the synapse load value information is refreshed, and therefore a neural network circuit device reliably operating for a long time duration is provided.",1995-01-31,"The title of the patent is data synapse expressing unit capable of refreshing stored synapse load and its abstract is a synapse expressing unit includes a capacitor for storing a synapse load value information in a form of electric charges, and a refresh control circuit for remedying the change in the amount of the electric charges stored in the capacitor. the refresh control circuit includes a comparator for comparing a potential at an electrode of the capacitor and a reference potential, and a drive circuit responsive to the comparator for recovering the electric charges of the capacitor through charge pumping operation. the synapse load value information is refreshed, and therefore a neural network circuit device reliably operating for a long time duration is provided. dated 1995-01-31"
5386496,method and device for nonlinear transformation of colour information by neural network,"color information such as a combination of fundamental colors in the subtractive or additive color mixing, hue, chroma and value in the psychological attributes of color, or coordinate values in a uniform color space are transformed nonlinearly and mutually by using a neural network, particularly a multi-layered feedforward neural network sufficiently trained of the transformation on samples of known data.",1995-01-31,"The title of the patent is method and device for nonlinear transformation of colour information by neural network and its abstract is color information such as a combination of fundamental colors in the subtractive or additive color mixing, hue, chroma and value in the psychological attributes of color, or coordinate values in a uniform color space are transformed nonlinearly and mutually by using a neural network, particularly a multi-layered feedforward neural network sufficiently trained of the transformation on samples of known data. dated 1995-01-31"
5388164,method for judging particle agglutination patterns using neural networks,"particle patterns formed on an inclined bottom surface of a reaction vessel are photoelectrically detected to produce a two-dimensional image signal. the signal is processed to judge or classify the particle patterns into an agglutinated pattern, a non-agglutinated pattern or an uncertain pattern with the aid of a neural network. an image signal representing a particle pattern is first extracted, then the image signal is decomposed into a series of light intensity areas due to different contours of the inclined bottom surface. the integrated light intensities of each area are presented to a neural network. the neural network operates in a training mode and a classification mode. in the training mode the neural network is presented with numerous samples of decomposed images as well as their respective classification. in the classification mode the neural network will judge a decomposed image based on a generalization made during the training mode.",1995-02-07,"The title of the patent is method for judging particle agglutination patterns using neural networks and its abstract is particle patterns formed on an inclined bottom surface of a reaction vessel are photoelectrically detected to produce a two-dimensional image signal. the signal is processed to judge or classify the particle patterns into an agglutinated pattern, a non-agglutinated pattern or an uncertain pattern with the aid of a neural network. an image signal representing a particle pattern is first extracted, then the image signal is decomposed into a series of light intensity areas due to different contours of the inclined bottom surface. the integrated light intensities of each area are presented to a neural network. the neural network operates in a training mode and a classification mode. in the training mode the neural network is presented with numerous samples of decomposed images as well as their respective classification. in the classification mode the neural network will judge a decomposed image based on a generalization made during the training mode. dated 1995-02-07"
5388187,information processing device capable of optically writing synapse strength matrix,"an information processing device having neural network functions for performing information processing comprises: a semiconductor integrated circuit section including a plurality of neuronic circuit regions having a neuronic function which is one of said neural network functions, and first and second molecular film provided on the integrated circuit section. the first molecular film has a photoelectric function and the second molecular film has a light-emitting function. coupling between the plurality of neurons is realized through a combination or the light-emitting and light receiving functions of the first and second molecular films.",1995-02-07,"The title of the patent is information processing device capable of optically writing synapse strength matrix and its abstract is an information processing device having neural network functions for performing information processing comprises: a semiconductor integrated circuit section including a plurality of neuronic circuit regions having a neuronic function which is one of said neural network functions, and first and second molecular film provided on the integrated circuit section. the first molecular film has a photoelectric function and the second molecular film has a light-emitting function. coupling between the plurality of neurons is realized through a combination or the light-emitting and light receiving functions of the first and second molecular films. dated 1995-02-07"
5389764,automatic cooking appliance employing a neural network for cooking control,"a cooking appliance controls a cooking device on the basis of temperature information of an object to be cooked that is estimated from changes in physical characteristics. a neural network is taught, for a number of categories of food that are classified according to the temperature of the cooked and completed food, the relationship between changes in the physical characteristic, such as the temperature and humidity, generated during heating of the object to be cooked during cooking, and changes of temperature of the object at the center of the object and the surface of the object in order to provide for an automatic cooking operation.",1995-02-14,"The title of the patent is automatic cooking appliance employing a neural network for cooking control and its abstract is a cooking appliance controls a cooking device on the basis of temperature information of an object to be cooked that is estimated from changes in physical characteristics. a neural network is taught, for a number of categories of food that are classified according to the temperature of the cooked and completed food, the relationship between changes in the physical characteristic, such as the temperature and humidity, generated during heating of the object to be cooked during cooking, and changes of temperature of the object at the center of the object and the surface of the object in order to provide for an automatic cooking operation. dated 1995-02-14"
5390261,method and apparatus for pattern classification using distributed adaptive fuzzy windows,"a method for pattern classification and, in particular, a method which distributes the classification criteria across a neural network. the classification criteria for a pattern class is stored distributively in the neural network in two aspects. first, it manifests itself as one or more levels of templates, each of which represents a fuzzily unique perspective of the pattern class. second, the template at each level is represented by a set of fuzzy windows, each of which defines a classification criterion of a corresponding feature of the pattern class.",1995-02-14,"The title of the patent is method and apparatus for pattern classification using distributed adaptive fuzzy windows and its abstract is a method for pattern classification and, in particular, a method which distributes the classification criteria across a neural network. the classification criteria for a pattern class is stored distributively in the neural network in two aspects. first, it manifests itself as one or more levels of templates, each of which represents a fuzzily unique perspective of the pattern class. second, the template at each level is represented by a set of fuzzy windows, each of which defines a classification criterion of a corresponding feature of the pattern class. dated 1995-02-14"
5390284,learning method and apparatus for neural networks and simulator with neural network,"a neural network (100) has an input layer, a hidden layer, and an output layer. the neural network stores weight values which operate on data input at the input layer to generate output data at the output layer. an error computing unit (87) receives the output data and compares it with desired output data from a learning data storage unit (105) to calculate error values representing the difference. an error gradient computing unit (81) calculates an error gradient, i.e. rate and direction of error change. a ratio computing unit (82) computes a ratio or percentage of a prior conjugate vector and combines the ratio with the error gradient. a conjugate vector computing unit (83) generates a present line search conjugate vector from the error gradient value and a previously calculated line search gradient vector. a line search computing unit (95) includes a weight computing unit (88) which calculates a weight correction value. the weight correction value is compared (18) with a preselected maximum or upper limit correction value (.kappa.). the line search computing unit (95) limits adjustment of the weight values stored in the neural network in accordance with the maximum weight correction value.",1995-02-14,"The title of the patent is learning method and apparatus for neural networks and simulator with neural network and its abstract is a neural network (100) has an input layer, a hidden layer, and an output layer. the neural network stores weight values which operate on data input at the input layer to generate output data at the output layer. an error computing unit (87) receives the output data and compares it with desired output data from a learning data storage unit (105) to calculate error values representing the difference. an error gradient computing unit (81) calculates an error gradient, i.e. rate and direction of error change. a ratio computing unit (82) computes a ratio or percentage of a prior conjugate vector and combines the ratio with the error gradient. a conjugate vector computing unit (83) generates a present line search conjugate vector from the error gradient value and a previously calculated line search gradient vector. a line search computing unit (95) includes a weight computing unit (88) which calculates a weight correction value. the weight correction value is compared (18) with a preselected maximum or upper limit correction value (.kappa.). the line search computing unit (95) limits adjustment of the weight values stored in the neural network in accordance with the maximum weight correction value. dated 1995-02-14"
5390285,method and apparatus for training a neural network depending on average mismatch,a method of training a neural network (2) having dynamically adjustable parameters controlled by a controller (10) which determine the response of the network (2). a set of input vectors (i.sub.l to i.sub.n) are input to network (2) at an input port (4). the corresponding set of output vectors (o'.sub.l to o'.sub.n) provided by the network (2) are compared to a target set of output vectors (o.sub.l to o.sub.n) by an error logger (12) which provides to the controller (10) a measure of similarity of the two sets. the controller (10) is arranged to alter the dynamic parameters independence on the average number of occasions the output vectors are different from the respective target output vectors. measuring the similarity of the whole of the output set and target set and adjusting the parameters on this global measure rather than on the similarity of pairs of individual vectors provides enhanced training rates for neural networks having a data throughput rate that can be higher than the rate at which the parameters can be adjusted.,1995-02-14,The title of the patent is method and apparatus for training a neural network depending on average mismatch and its abstract is a method of training a neural network (2) having dynamically adjustable parameters controlled by a controller (10) which determine the response of the network (2). a set of input vectors (i.sub.l to i.sub.n) are input to network (2) at an input port (4). the corresponding set of output vectors (o'.sub.l to o'.sub.n) provided by the network (2) are compared to a target set of output vectors (o.sub.l to o.sub.n) by an error logger (12) which provides to the controller (10) a measure of similarity of the two sets. the controller (10) is arranged to alter the dynamic parameters independence on the average number of occasions the output vectors are different from the respective target output vectors. measuring the similarity of the whole of the output set and target set and adjusting the parameters on this global measure rather than on the similarity of pairs of individual vectors provides enhanced training rates for neural networks having a data throughput rate that can be higher than the rate at which the parameters can be adjusted. dated 1995-02-14
5393994,optical semiconductor device for neural network,"an optical semiconductor device is disclosed which includes a semiconductor laser having at least an active layer, reflecting means formed on the semiconductor laser for reflecting internal feedback light generated from the semiconductor laser and at least two phototransistors formed on the reflecting means for detecting light having a wavelength substantially identical to that of laser light oscillated from the active layer.",1995-02-28,"The title of the patent is optical semiconductor device for neural network and its abstract is an optical semiconductor device is disclosed which includes a semiconductor laser having at least an active layer, reflecting means formed on the semiconductor laser for reflecting internal feedback light generated from the semiconductor laser and at least two phototransistors formed on the reflecting means for detecting light having a wavelength substantially identical to that of laser light oscillated from the active layer. dated 1995-02-28"
5394257,optical neural network system,"an optical system of an optical neural network model for parallel data processing is disclosed. taking advantage of the fact that an auto-correlation matrix is symmetric with respect to a main diagonal and the weights for modulating the values of diagonals of the auto-correlation matrix are equal to each other, the configuration of an optical modulation unit is simplified by a one-dimensional modulation array on the one hand, and both positive and negative weights are capable of being computed at the same time on the other hand. in particular, the optical system makes up a second-order neural network exhibiting invariant characteristics against the translation and scale.",1995-02-28,"The title of the patent is optical neural network system and its abstract is an optical system of an optical neural network model for parallel data processing is disclosed. taking advantage of the fact that an auto-correlation matrix is symmetric with respect to a main diagonal and the weights for modulating the values of diagonals of the auto-correlation matrix are equal to each other, the configuration of an optical modulation unit is simplified by a one-dimensional modulation array on the one hand, and both positive and negative weights are capable of being computed at the same time on the other hand. in particular, the optical system makes up a second-order neural network exhibiting invariant characteristics against the translation and scale. dated 1995-02-28"
5394510,neural network-type data processing system recognizing a predetermined shape using inhibitory connections,"a data processing system of the neural network type. the system recognizes a predetermined shape by providing some connections that are inhibitory between a plurality of neurons in a neural layer of the neural network. if data is found in the inhibitory area, it makes it harder for the neurons in the correct area to fire. only when the neurons in the correct area fire is the predetermined shape recognized.",1995-02-28,"The title of the patent is neural network-type data processing system recognizing a predetermined shape using inhibitory connections and its abstract is a data processing system of the neural network type. the system recognizes a predetermined shape by providing some connections that are inhibitory between a plurality of neurons in a neural layer of the neural network. if data is found in the inhibitory area, it makes it harder for the neurons in the correct area to fire. only when the neurons in the correct area fire is the predetermined shape recognized. dated 1995-02-28"
5394511,coupling element for semiconductor neural network device,"a neural network device includes internal data input lines, internal data output lines, coupling elements provided at the connections of the internal data input lines and the internal data output lines. the coupling elements couple, with specific programmable coupling strengths, the associated internal data input lines to the associated internal data output lines. in a program mode, the internal data output lines serve as signal lines for transmitting the coupling strength information. each of the coupling elements includes storage elements, circuitry for writing a signal potential on an associated internal data output line, and circuitry for supplying a stored signal for a storage element into an associated internal data output line.",1995-02-28,"The title of the patent is coupling element for semiconductor neural network device and its abstract is a neural network device includes internal data input lines, internal data output lines, coupling elements provided at the connections of the internal data input lines and the internal data output lines. the coupling elements couple, with specific programmable coupling strengths, the associated internal data input lines to the associated internal data output lines. in a program mode, the internal data output lines serve as signal lines for transmitting the coupling strength information. each of the coupling elements includes storage elements, circuitry for writing a signal potential on an associated internal data output line, and circuitry for supplying a stored signal for a storage element into an associated internal data output line. dated 1995-02-28"
5396415,neruo-pid controller,pid controllers form a large proportion of controllers in use in many controlled systems today. this application describes how to use a neural network which receives pid inputs to be a controller and operate as a pid controller to save on retraining and provide other efficiencies in control. also shown is the user selectability between pid conventional controllers and neural network controllers.,1995-03-07,The title of the patent is neruo-pid controller and its abstract is pid controllers form a large proportion of controllers in use in many controlled systems today. this application describes how to use a neural network which receives pid inputs to be a controller and operate as a pid controller to save on retraining and provide other efficiencies in control. also shown is the user selectability between pid conventional controllers and neural network controllers. dated 1995-03-07
5396565,pattern recognition neural net insensitive to disturbances in inputs,"in a neural network assembly for use in pattern recognition, a memory region memorizes reference patterns along with their categories. a pattern associator neural network is connected to the memory region and is trained in accordance with a back-propagation training algorithm in a training phase of operation to correctly recognize the reference patterns according to the categories. an adaptive input region is connected to the neural network, supplied with roughly segmented patterns as input patterns, and trained in a recognition phase of operation following the training phase. and correctly processes the input patterns into processed patterns which the neural network can correctly recognize according to the categories. preferably, the assembly is operable in the training phase as comprising the memory region, the neural network, and a controlling part which is connected to the memory region and the neural network to train the neural network. in the recognition phase the assembly is operable as comprising the input region, the neural network and the controlling part which is now connected to the input region and the neural network and adjusts the input region so that each roughly segmented pattern may approach a pertinent one of the reference patterns.",1995-03-07,"The title of the patent is pattern recognition neural net insensitive to disturbances in inputs and its abstract is in a neural network assembly for use in pattern recognition, a memory region memorizes reference patterns along with their categories. a pattern associator neural network is connected to the memory region and is trained in accordance with a back-propagation training algorithm in a training phase of operation to correctly recognize the reference patterns according to the categories. an adaptive input region is connected to the neural network, supplied with roughly segmented patterns as input patterns, and trained in a recognition phase of operation following the training phase. and correctly processes the input patterns into processed patterns which the neural network can correctly recognize according to the categories. preferably, the assembly is operable in the training phase as comprising the memory region, the neural network, and a controlling part which is connected to the memory region and the neural network to train the neural network. in the recognition phase the assembly is operable as comprising the input region, the neural network and the controlling part which is now connected to the input region and the neural network and adjusts the input region so that each roughly segmented pattern may approach a pertinent one of the reference patterns. dated 1995-03-07"
5396580,translation of a neural network into a rule-based expert system,"a rule-based expert system is generated from a neural network. the neural network is trained in such a way as to avoid redundancy and to select input weights to the various processing elements in such a way as to nullify the input weights which have smaller absolute values. the neural network is translated into a set of rules by a heuristic search technique. additionally, the translation distinguishes between positive and negative attributes for efficiency and can adequately explore rule size exponential with a given parameter. both explicit and implicit knowledge of adapted neural networks are decoded and represented as if--then rules.",1995-03-07,"The title of the patent is translation of a neural network into a rule-based expert system and its abstract is a rule-based expert system is generated from a neural network. the neural network is trained in such a way as to avoid redundancy and to select input weights to the various processing elements in such a way as to nullify the input weights which have smaller absolute values. the neural network is translated into a set of rules by a heuristic search technique. additionally, the translation distinguishes between positive and negative attributes for efficiency and can adequately explore rule size exponential with a given parameter. both explicit and implicit knowledge of adapted neural networks are decoded and represented as if--then rules. dated 1995-03-07"
5396581,semiconductor neural network and operating method thereof,"a semiconductor neural network includes a coupling matrix having coupling elements arranged in a matrix which couple with specific coupling strengths internal data input lines to internal data output lines. the internal data output lines are divided into groups. the neural network further comprises weighting addition circuits provided corresponding to the groups of the internal data cutput lines. a weighting addition circuit includes weighing elements for adding weights to signals on the internal data output lines in the corresponding group and outputting the weighted signals, and an addition circuit for outputting a total sum of the outputs of those weighting elements. the internal data output lines are arranged to form pairs and the addition circuit has a first input terminal for receiving one weighting element output of each of the pairs in common, a second input terminal for receiving the other weighting element output of each of the pairs in common, and a sense amplifier for differentially amplifying signals at the first and second input terminals. the neural network further includes a circuit for detecting a change time of an input signal, a circuit responsive to an input signal change for equalizing the first and second input terminals for a predetermined period, and a circuit for activating the sense amplifier after the equalization is completed. the information retention capability of each coupling element is set according to the weight of an associated weighting element. this neural network can provide multi-valued expression of coupling strength with less number of coupling elements.",1995-03-07,"The title of the patent is semiconductor neural network and operating method thereof and its abstract is a semiconductor neural network includes a coupling matrix having coupling elements arranged in a matrix which couple with specific coupling strengths internal data input lines to internal data output lines. the internal data output lines are divided into groups. the neural network further comprises weighting addition circuits provided corresponding to the groups of the internal data cutput lines. a weighting addition circuit includes weighing elements for adding weights to signals on the internal data output lines in the corresponding group and outputting the weighted signals, and an addition circuit for outputting a total sum of the outputs of those weighting elements. the internal data output lines are arranged to form pairs and the addition circuit has a first input terminal for receiving one weighting element output of each of the pairs in common, a second input terminal for receiving the other weighting element output of each of the pairs in common, and a sense amplifier for differentially amplifying signals at the first and second input terminals. the neural network further includes a circuit for detecting a change time of an input signal, a circuit responsive to an input signal change for equalizing the first and second input terminals for a predetermined period, and a circuit for activating the sense amplifier after the equalization is completed. the information retention capability of each coupling element is set according to the weight of an associated weighting element. this neural network can provide multi-valued expression of coupling strength with less number of coupling elements. dated 1995-03-07"
5396817,tire inflation and velocity sensor,a system that will indicate tire inflation and vehicle velocity. an array of strain gage sensors is used to determine the distribution of contact forces along the width of a pneumatic tire. a neural network may be employed to classify the patterns of force sensed in this manner. additional processing of the data permits vehicle velocity to also be determined.,1995-03-14,The title of the patent is tire inflation and velocity sensor and its abstract is a system that will indicate tire inflation and vehicle velocity. an array of strain gage sensors is used to determine the distribution of contact forces along the width of a pneumatic tire. a neural network may be employed to classify the patterns of force sensed in this manner. additional processing of the data permits vehicle velocity to also be determined. dated 1995-03-14
5396896,medical pumping apparatus,this invention relates to a medical pumping apparatus which utilizes a neural network. the medical pumping apparatus continuously and automatically monitors fill status of the venous plexus and flow rate from the venous plexus and continuously and automatically controls the pressure and cycle rate of a pump capable of cyclically applying pressure to a part of the human body for the purpose of maximizing blood transfer therein.,1995-03-14,The title of the patent is medical pumping apparatus and its abstract is this invention relates to a medical pumping apparatus which utilizes a neural network. the medical pumping apparatus continuously and automatically monitors fill status of the venous plexus and flow rate from the venous plexus and continuously and automatically controls the pressure and cycle rate of a pump capable of cyclically applying pressure to a part of the human body for the purpose of maximizing blood transfer therein. dated 1995-03-14
5398187,correlation detection method and connectivity-structure estimation method for neurons,"the present invention provides an interneuron crossrelation identification technique and an interneuron connection-structure estimation technique for inferring a connection-structure and the strengths of the connectivities among a plurality of neurons required for constructing a neural network model, by obtaining crossrelations among time-course data of neurons. the interneuron crossrelation detection technique may include steps of: calculating conditional probabilities by, among other things, normalizing crosscoincidence histograms calculated from time-course data of activities of the neurons representing a train of action potentials of the neurons representing a train of action potentials of the neurons, and comparing trains of symbols representing time-course states of the activities of the neurons; distinguishing an inhibitory connectivity form an excitatory connectivity by comparing the conditional probabilities to each other; and quantitatively estimating the magnitude of crossrelation among the time-course data. the interneuron connection-structure estimation technique may include steps of: computing conditional probabilities by normalizing cross-coincidence histograms calculated from time-course data of activities of the neurons representing a train of action potentials of the neurons; computing conditional mutual information and three-point mutual information from the computed conditional probabilities; and inferring a connection structure among the neurons.",1995-03-14,"The title of the patent is correlation detection method and connectivity-structure estimation method for neurons and its abstract is the present invention provides an interneuron crossrelation identification technique and an interneuron connection-structure estimation technique for inferring a connection-structure and the strengths of the connectivities among a plurality of neurons required for constructing a neural network model, by obtaining crossrelations among time-course data of neurons. the interneuron crossrelation detection technique may include steps of: calculating conditional probabilities by, among other things, normalizing crosscoincidence histograms calculated from time-course data of activities of the neurons representing a train of action potentials of the neurons representing a train of action potentials of the neurons, and comparing trains of symbols representing time-course states of the activities of the neurons; distinguishing an inhibitory connectivity form an excitatory connectivity by comparing the conditional probabilities to each other; and quantitatively estimating the magnitude of crossrelation among the time-course data. the interneuron connection-structure estimation technique may include steps of: computing conditional probabilities by normalizing cross-coincidence histograms calculated from time-course data of activities of the neurons representing a train of action potentials of the neurons; computing conditional mutual information and three-point mutual information from the computed conditional probabilities; and inferring a connection structure among the neurons. dated 1995-03-14"
5398300,neural network having expert system functionality,"a method for performing a variety of expert system functions on any continuous-state feedforward neural network. these functions include decision-making, explanation, computation of confidence measures, and intelligent direction of information acquisition. additionally, the method converts the knowledge implicit in such a network into a set of explicit if-then rules.",1995-03-14,"The title of the patent is neural network having expert system functionality and its abstract is a method for performing a variety of expert system functions on any continuous-state feedforward neural network. these functions include decision-making, explanation, computation of confidence measures, and intelligent direction of information acquisition. additionally, the method converts the knowledge implicit in such a network into a set of explicit if-then rules. dated 1995-03-14"
5398302,method and apparatus for adaptive learning in neural networks,"a method enabling a neural network, having weights organized into a weight vector, to learn, comprising the steps of: (a) assigning a first memory location for storing a first learning rate, a second memory location for storing a momentum factor, a memory block for storing the weight vector, and a third memory location for storing a second learning rate; (b) initializing the learning rate, momentum factor, and weight vector; (c) storing the first learning rate, the momentum factor, and the weight vector into their respective memory locations; (d) saving the first learning rate in the second learning rate by storing it into the third memory location; (e) using a search technique to adjust the first learning rate to adjust the weight vector, and updating the first memory location and the memory block; (f) adapting the momentum factor using the first learning rate and the second learning rate; and repeating steps (c) through (f) until a predetermined convergence criterion has been met.",1995-03-14,"The title of the patent is method and apparatus for adaptive learning in neural networks and its abstract is a method enabling a neural network, having weights organized into a weight vector, to learn, comprising the steps of: (a) assigning a first memory location for storing a first learning rate, a second memory location for storing a momentum factor, a memory block for storing the weight vector, and a third memory location for storing a second learning rate; (b) initializing the learning rate, momentum factor, and weight vector; (c) storing the first learning rate, the momentum factor, and the weight vector into their respective memory locations; (d) saving the first learning rate in the second learning rate by storing it into the third memory location; (e) using a search technique to adjust the first learning rate to adjust the weight vector, and updating the first memory location and the memory block; (f) adapting the momentum factor using the first learning rate and the second learning rate; and repeating steps (c) through (f) until a predetermined convergence criterion has been met. dated 1995-03-14"
5400436,information retrieval system,an information retrieval system retrieves stored information on the basis of incomplete or noisy retrieval key information within a realistic processing time. the information to be recalled is stored in neural network associative memory. genetic algorithms are adopted to avoid a thorough retrieval of data in a huge breadth and depth of high-order space which must otherwise be searched due to ambiguity of retrieval key information. the information containing the given retrieval key is effectively sought. a processor employs a neural network. a memory is correctly recalled from partial inputs or inputs in which noise is extremely pronounced by repeating a sum-of-product operation of a firing pattern within a synapse coupling matrix and also threshold value processing.,1995-03-21,The title of the patent is information retrieval system and its abstract is an information retrieval system retrieves stored information on the basis of incomplete or noisy retrieval key information within a realistic processing time. the information to be recalled is stored in neural network associative memory. genetic algorithms are adopted to avoid a thorough retrieval of data in a huge breadth and depth of high-order space which must otherwise be searched due to ambiguity of retrieval key information. the information containing the given retrieval key is effectively sought. a processor employs a neural network. a memory is correctly recalled from partial inputs or inputs in which noise is extremely pronounced by repeating a sum-of-product operation of a firing pattern within a synapse coupling matrix and also threshold value processing. dated 1995-03-21
5400641,transformer oil gas extractor,"a device is disclosed for extracting dissolved gases from oil in an electrical transformer and for analyzing those gases using a plurality of gas sensors, a signal processor, and a neural network.",1995-03-28,"The title of the patent is transformer oil gas extractor and its abstract is a device is disclosed for extracting dissolved gases from oil in an electrical transformer and for analyzing those gases using a plurality of gas sensors, a signal processor, and a neural network. dated 1995-03-28"
5402359,method and apparatus for determining wiring routes by utilizing artificial neural networks,"a method for determining routes of wiring nets by utilizing artificial neural networks includes the steps of dividing the wired area into smaller areas, representing each boundary through which one wiring net passes is capable of passing as an artificial neuron, changing an output value of the artificial neuron according to whether or not the wiring net actually passes through the boundary, composing an artificial neural network in which the interaction between the artificial neurons is taken into consideration according to one or more prescribed conditions restricting each route of the wiring nets while changing the output values of the artificial neurons, converging the output values of all artificial neurons, and determining the routes of all wiring nets by judging whether or not each wiring net passes through a boundary according to the converged output values of the artificial neurons.",1995-03-28,"The title of the patent is method and apparatus for determining wiring routes by utilizing artificial neural networks and its abstract is a method for determining routes of wiring nets by utilizing artificial neural networks includes the steps of dividing the wired area into smaller areas, representing each boundary through which one wiring net passes is capable of passing as an artificial neuron, changing an output value of the artificial neuron according to whether or not the wiring net actually passes through the boundary, composing an artificial neural network in which the interaction between the artificial neurons is taken into consideration according to one or more prescribed conditions restricting each route of the wiring nets while changing the output values of the artificial neurons, converging the output values of all artificial neurons, and determining the routes of all wiring nets by judging whether or not each wiring net passes through a boundary according to the converged output values of the artificial neurons. dated 1995-03-28"
5402519,neural network system adapted for non-linear processing,"a neural network system includes a qualitative evaluation section, a neural network section, a quantifying section and a display section. the qualitative evaluation section qualitatively analyzes an unknown data supplied thereto, and normalizes the result of analysis within a predetermined range. the neural network section having a neural network with plural neurons computes the network output data from the normalized unknown data produced by the qualitative evaluation section. each neuron is connected to plural other neurons through synapses, each of which is assigned an individual weight coefficient. each neuron is adapted to output an output function value assigned thereto associated with the total sum of the products of the output from the neurons connected thereto and the synapse weight coefficient. the quantifying section quantifies the network output data to produce desired data. the desired data thus produced is displayed on the display section.",1995-03-28,"The title of the patent is neural network system adapted for non-linear processing and its abstract is a neural network system includes a qualitative evaluation section, a neural network section, a quantifying section and a display section. the qualitative evaluation section qualitatively analyzes an unknown data supplied thereto, and normalizes the result of analysis within a predetermined range. the neural network section having a neural network with plural neurons computes the network output data from the normalized unknown data produced by the qualitative evaluation section. each neuron is connected to plural other neurons through synapses, each of which is assigned an individual weight coefficient. each neuron is adapted to output an output function value assigned thereto associated with the total sum of the products of the output from the neurons connected thereto and the synapse weight coefficient. the quantifying section quantifies the network output data to produce desired data. the desired data thus produced is displayed on the display section. dated 1995-03-28"
5402520,neural network method and apparatus for retrieving signals embedded in noise and analyzing the retrieved signals,"an apparatus for retrieving signals embedded in noise and analyzing the signals. the apparatus includes an input device for receiving input signals having noise. at least one noise filter retrieves data signals embedded in the input signals. at least one adaptive pattern recognition filter generates coefficients of a polynomial expansion representing the pattern of the filtered data signals. a storage device stores the coefficients generated. it is determined when an event has occurred, the event being located at any position within the data signals. an adaptive autoregressive moving average pattern recognition filter generates coefficients of a polynomial expansion representing an enhanced pattern of filtered data signals. at least one weighting filter compares the stored patterns with the enhanced pattern of data signals.",1995-03-28,"The title of the patent is neural network method and apparatus for retrieving signals embedded in noise and analyzing the retrieved signals and its abstract is an apparatus for retrieving signals embedded in noise and analyzing the signals. the apparatus includes an input device for receiving input signals having noise. at least one noise filter retrieves data signals embedded in the input signals. at least one adaptive pattern recognition filter generates coefficients of a polynomial expansion representing the pattern of the filtered data signals. a storage device stores the coefficients generated. it is determined when an event has occurred, the event being located at any position within the data signals. an adaptive autoregressive moving average pattern recognition filter generates coefficients of a polynomial expansion representing an enhanced pattern of filtered data signals. at least one weighting filter compares the stored patterns with the enhanced pattern of data signals. dated 1995-03-28"
5402521,method for recognition of abnormal conditions using neural networks,"according to the present invention, a method for recognition of normal and abnormal conditions can be performed with at least one neural network. first, trend data of an object system, before a recognition-step, are entered as input data to an input layer of each neural network and data of this system at the recognition-step are entered as objective output data to an output layer of the neural network. thus, multiple sets of trend data showing at least one normal condition of this system are formed in the neural network in order to obtained learned weights and biases. next, output data at every recognition-step are predicted by entering actual trend data as input data to the neural network, while the learned weights and biases are utilized. then, the predicted output data are compared with actual output data at every recognition-step. finally, the normal and abnormal conditions of this system can be recognized by real time interpretation of deviations between the predicted output data and the actual output data. the method of the present invention particularly can be applied to a control system requiring the recognition of abnormal conditions such as a control system for the operation of a plant, an automobile, a robot, an aircraft, a marine vessel, a medical apparatus, security apparatus and the like.",1995-03-28,"The title of the patent is method for recognition of abnormal conditions using neural networks and its abstract is according to the present invention, a method for recognition of normal and abnormal conditions can be performed with at least one neural network. first, trend data of an object system, before a recognition-step, are entered as input data to an input layer of each neural network and data of this system at the recognition-step are entered as objective output data to an output layer of the neural network. thus, multiple sets of trend data showing at least one normal condition of this system are formed in the neural network in order to obtained learned weights and biases. next, output data at every recognition-step are predicted by entering actual trend data as input data to the neural network, while the learned weights and biases are utilized. then, the predicted output data are compared with actual output data at every recognition-step. finally, the normal and abnormal conditions of this system can be recognized by real time interpretation of deviations between the predicted output data and the actual output data. the method of the present invention particularly can be applied to a control system requiring the recognition of abnormal conditions such as a control system for the operation of a plant, an automobile, a robot, an aircraft, a marine vessel, a medical apparatus, security apparatus and the like. dated 1995-03-28"
5402522,dynamically stable associative learning neural system,"a dynamically stable associative learning neural network system include a plurality of synapses (122,22-28), a non-linear function circuit (30) and an adaptive weight circuit (150) for adjusting the weight of each synapse based upon the present signal and the prior history of signals applied to the input of the particular synapse and the present signal and the prior history of signals applied to the input of a predetermined set of other synapses. an embodiment of a conditional-signal neuron circuit (100) receives input signals from conditional stimuli and an unconditional-signal neuron circuit (110) receives input signals from unconditional stimuli. a neural network (200) is formed by a set of conditional-signal and unconditional-signal neuron circuits connected by flow-through synapses to form separate paths between each input (215) and a corresponding output (245). in one embodiment, the neural network (200) is initialized by varying the weight of the input signals from conditional stimuli, until a dynamic equilibrium is reached.",1995-03-28,"The title of the patent is dynamically stable associative learning neural system and its abstract is a dynamically stable associative learning neural network system include a plurality of synapses (122,22-28), a non-linear function circuit (30) and an adaptive weight circuit (150) for adjusting the weight of each synapse based upon the present signal and the prior history of signals applied to the input of the particular synapse and the present signal and the prior history of signals applied to the input of a predetermined set of other synapses. an embodiment of a conditional-signal neuron circuit (100) receives input signals from conditional stimuli and an unconditional-signal neuron circuit (110) receives input signals from unconditional stimuli. a neural network (200) is formed by a set of conditional-signal and unconditional-signal neuron circuits connected by flow-through synapses to form separate paths between each input (215) and a corresponding output (245). in one embodiment, the neural network (200) is initialized by varying the weight of the input signals from conditional stimuli, until a dynamic equilibrium is reached. dated 1995-03-28"
5404377,simultaneous transmission of data and audio signals by means of perceptual coding,""" a communication system for simultaneously transmitting data and audio signals via a conventional audio communications channel using perceptual coding techniques is disclosed. in a preferred embodiment, a first artificial neural network (nn) at an encoder monitors an audio channel to detect """"opportunities"""" to insert the data signal such that the inserted signals are masked by the audio signal, as defined by the """"perceptual entropy envelope"""" of the audio signal. under the control of the first nn a data signal containing, for example, an id or serial number, is encoded as one or more whitened direct sequence spread spectrum signals and/or a narrowband fsk data signal and transmitted at the time, frequency and level determined by the first nn such that the data signal is masked by the audio signal. the audio signal is combined with the spread spectrum and/or the fsk data signal(s) to form a composite signal, which is transmitted to one or more receiving locations via the audio channel. a decoder at each of the receiving locations comprises preprocessing circuitry, receiver sync circuitry and fsk decoder circuitry, as well as a second nn, which nn uses pattern and signature recognition techniques to perform block decoding, bit deinterleaving and acquisition confirm functions to recover the encoded id or serial number. """,1995-04-04,"The title of the patent is simultaneous transmission of data and audio signals by means of perceptual coding and its abstract is "" a communication system for simultaneously transmitting data and audio signals via a conventional audio communications channel using perceptual coding techniques is disclosed. in a preferred embodiment, a first artificial neural network (nn) at an encoder monitors an audio channel to detect """"opportunities"""" to insert the data signal such that the inserted signals are masked by the audio signal, as defined by the """"perceptual entropy envelope"""" of the audio signal. under the control of the first nn a data signal containing, for example, an id or serial number, is encoded as one or more whitened direct sequence spread spectrum signals and/or a narrowband fsk data signal and transmitted at the time, frequency and level determined by the first nn such that the data signal is masked by the audio signal. the audio signal is combined with the spread spectrum and/or the fsk data signal(s) to form a composite signal, which is transmitted to one or more receiving locations via the audio channel. a decoder at each of the receiving locations comprises preprocessing circuitry, receiver sync circuitry and fsk decoder circuitry, as well as a second nn, which nn uses pattern and signature recognition techniques to perform block decoding, bit deinterleaving and acquisition confirm functions to recover the encoded id or serial number. "" dated 1995-04-04"
5404422,speech recognition system with neural network,"a voice recognition apparatus capable of recognizing any word utterance by using a neural network, the apparatus includes a unit for inputting an input utterance and for outputting compressed feature variables of the input utterance a unit using a neural network and connected to the input unit for receiving the compressed feature variables output from the input unit and for outputting a value corresponding to a similarity between the input utterance and words to be recognized. the neural network unit has a first unit for outputting a value which corresponds to a similarity in partial phoneme series of a specific word among vocabularies to be recognized with respect to the input utterance. the neural network also has a second unit connected to the first unit for receiving all of the values output from the first unit and for outputting a value corresponding to a similarity in the specific word with respect to the input utterance. and the neural network also has a third unit connected to the second unit for receiving all of the values output from the second unit and for outputting a value corresponding to a classification of voice recognition in which the input utterance belongs.",1995-04-04,"The title of the patent is speech recognition system with neural network and its abstract is a voice recognition apparatus capable of recognizing any word utterance by using a neural network, the apparatus includes a unit for inputting an input utterance and for outputting compressed feature variables of the input utterance a unit using a neural network and connected to the input unit for receiving the compressed feature variables output from the input unit and for outputting a value corresponding to a similarity between the input utterance and words to be recognized. the neural network unit has a first unit for outputting a value which corresponds to a similarity in partial phoneme series of a specific word among vocabularies to be recognized with respect to the input utterance. the neural network also has a second unit connected to the first unit for receiving all of the values output from the first unit and for outputting a value corresponding to a similarity in the specific word with respect to the input utterance. and the neural network also has a third unit connected to the second unit for receiving all of the values output from the second unit and for outputting a value corresponding to a classification of voice recognition in which the input utterance belongs. dated 1995-04-04"
5404423,"method and apparatus for indetification, forecast, and control of a non-linear flow on a physical system network using a neural network","a neural network system for identifying forecasting, and controlling a non-linear flow on a physical system network, in which each branch between nodes in the physical system network is divided by a plurality of division points; a flow at each of the division points and a terminal point of each branch is calculated according to neural network model parameters specifying connections among the division points and the terminal point in a neural network model; an actual flow is measured at the terminal point of said each branch; an error of the calculated flow at the terminal point with respect to the measured actual flow at the terminal point is calculated; the neural network model parameters are adjusted to minimize the calculated error; and system dynamics parameters specifying dynamics of the physical system are determined according to the adjusted neural network model parameters. in addition, a target function to be optimized is calculated in terms of flows at terminal points of branches as a function of a control parameter specifying connecting and disconnecting of connections among branches at each node; and connections among branches at each node are connected/disconnected to optimize the calculated target function.",1995-04-04,"The title of the patent is method and apparatus for indetification, forecast, and control of a non-linear flow on a physical system network using a neural network and its abstract is a neural network system for identifying forecasting, and controlling a non-linear flow on a physical system network, in which each branch between nodes in the physical system network is divided by a plurality of division points; a flow at each of the division points and a terminal point of each branch is calculated according to neural network model parameters specifying connections among the division points and the terminal point in a neural network model; an actual flow is measured at the terminal point of said each branch; an error of the calculated flow at the terminal point with respect to the measured actual flow at the terminal point is calculated; the neural network model parameters are adjusted to minimize the calculated error; and system dynamics parameters specifying dynamics of the physical system are determined according to the adjusted neural network model parameters. in addition, a target function to be optimized is calculated in terms of flows at terminal points of branches as a function of a control parameter specifying connecting and disconnecting of connections among branches at each node; and connections among branches at each node are connected/disconnected to optimize the calculated target function. dated 1995-04-04"
5406581,control system for electric arc furnace,"an improved arc furnace regulator employs neural circuits connected in a multi-layer network configuration with various weighted relationships between the successive layers which are automatically changed over time as a function of an error signal by means of the back-propagation method so that the regulator gradually improves its control algorithm as a result of accumulated experience. the network is implemented in software which can be developed and run on a pc with extra co-computing capability for greater execution speed. a second trainable neural network which emulates the arc furnace is used to develop the error signal, and is trained in mutually exclusive time periods with the training of the regular network.",1995-04-11,"The title of the patent is control system for electric arc furnace and its abstract is an improved arc furnace regulator employs neural circuits connected in a multi-layer network configuration with various weighted relationships between the successive layers which are automatically changed over time as a function of an error signal by means of the back-propagation method so that the regulator gradually improves its control algorithm as a result of accumulated experience. the network is implemented in software which can be developed and run on a pc with extra co-computing capability for greater execution speed. a second trainable neural network which emulates the arc furnace is used to develop the error signal, and is trained in mutually exclusive time periods with the training of the regular network. dated 1995-04-11"
5408406,neural net based disturbance predictor for model predictive control,a control loop for controlling a process or plant which controls the process or plant via an actuator. the control loop receives from the process or plant a signal representative of the process or plant output. the loop includes a nominal controller that generates a control signal for the actuator which is used only in the absence of a predicted disturbance to the process or plant signal from a disturbance mode controller unit having a neural network conditioned for predicting and indicating a disturbance.,1995-04-18,The title of the patent is neural net based disturbance predictor for model predictive control and its abstract is a control loop for controlling a process or plant which controls the process or plant via an actuator. the control loop receives from the process or plant a signal representative of the process or plant output. the loop includes a nominal controller that generates a control signal for the actuator which is used only in the absence of a predicted disturbance to the process or plant signal from a disturbance mode controller unit having a neural network conditioned for predicting and indicating a disturbance. dated 1995-04-18
5408414,target route predicting apparatus utilizing characteristic parameters of terrain information,"a route predicting apparatus for monitoring a flying target and predicting a route thereof. even if the target is hidden by mountains or other obstacles thus preventing observation for a certain time, a reliable route prediction is carried out by utilizing the fact that a route of the target is restricted by geographical features. the apparatus comprises an observation unit for observing a target and outputting an observed value thereof, a terrain information unit for outputting characteristic parameters of stored terrain information, a prediction unit coupled to receive the observed value and the characteristic parameters of terrain information for performing a fuzzy inference to output a predicted observed value of the target, and a learning adjustment unit coupled to receive an error signal between a predicted observed value at the current time predicted a unit time before and a real observed value at the current time for adjusting the prediction unit by using a neural network.",1995-04-18,"The title of the patent is target route predicting apparatus utilizing characteristic parameters of terrain information and its abstract is a route predicting apparatus for monitoring a flying target and predicting a route thereof. even if the target is hidden by mountains or other obstacles thus preventing observation for a certain time, a reliable route prediction is carried out by utilizing the fact that a route of the target is restricted by geographical features. the apparatus comprises an observation unit for observing a target and outputting an observed value thereof, a terrain information unit for outputting characteristic parameters of stored terrain information, a prediction unit coupled to receive the observed value and the characteristic parameters of terrain information for performing a fuzzy inference to output a predicted observed value of the target, and a learning adjustment unit coupled to receive an error signal between a predicted observed value at the current time predicted a unit time before and a real observed value at the current time for adjusting the prediction unit by using a neural network. dated 1995-04-18"
5408585,internal connection method for neural networks,a method of internal connection for neural network linking successive neuron layers and useful in image analysis and image and signal processing is provided. the neuron output states of a new neuron layer are represented by functions obtained by the method. a weighted summation of functions which represent the output state of a neuron from a preceding layer is carried out. a saturation function is applied to the result of the weighted summation. a function is outputted which is representative of the output state of the neuron layer.,1995-04-18,The title of the patent is internal connection method for neural networks and its abstract is a method of internal connection for neural network linking successive neuron layers and useful in image analysis and image and signal processing is provided. the neuron output states of a new neuron layer are represented by functions obtained by the method. a weighted summation of functions which represent the output state of a neuron from a preceding layer is carried out. a saturation function is applied to the result of the weighted summation. a function is outputted which is representative of the output state of the neuron layer. dated 1995-04-18
5408586,historical database training method for neural networks,"an on-line training neural network for process control system and method trains by retrieving training sets from the stream of process data. the neural network detects the availability of new training data, and constructs a training set by retrieving the corresponding input data. the neural network is trained using the training set. over time, many training sets are presented to the neural network. when multiple presentations are needed to effectively train, a buffer of training sets is filled and updated as new training data becomes available. the size of the buffer is selected in accordance with the training needs of the neural network. once the buffer is full, a new training set bumps the oldest training set off the top of the buffer stack. the training sets in the buffer stack can be presented one or more times each time a new training set is constructed. an historical database of timestamped data can be used to construct training sets when training input data has a time delay from sample time to availability for the neural network. the timestamps of the training input data are used to select the appropriate timestamp at which input data is retrieved for use in the training set. using the historical database, the neural network can be trained retrospectively by searching the historical database and constructing training sets based on past data.",1995-04-18,"The title of the patent is historical database training method for neural networks and its abstract is an on-line training neural network for process control system and method trains by retrieving training sets from the stream of process data. the neural network detects the availability of new training data, and constructs a training set by retrieving the corresponding input data. the neural network is trained using the training set. over time, many training sets are presented to the neural network. when multiple presentations are needed to effectively train, a buffer of training sets is filled and updated as new training data becomes available. the size of the buffer is selected in accordance with the training needs of the neural network. once the buffer is full, a new training set bumps the oldest training set off the top of the buffer stack. the training sets in the buffer stack can be presented one or more times each time a new training set is constructed. an historical database of timestamped data can be used to construct training sets when training input data has a time delay from sample time to availability for the neural network. the timestamps of the training input data are used to select the appropriate timestamp at which input data is retrieved for use in the training set. using the historical database, the neural network can be trained retrospectively by searching the historical database and constructing training sets based on past data. dated 1995-04-18"
5408588,artificial neural network method and architecture,"an architecture and data processing method for a neural network that can approximate any mapping function between the input and output vectors without the use of hidden layers. the data processing is done at the sibling nodes (second row). it is based on the orthogonal expansion of the functions that map the input vector to the output vector. because the nodes of the second row are simply data processing stations, they remain passive during training. as a result the system is basically a single-layer linear network with a filter at its entrance. because of this it is free from the problems of local minima. the invention also includes a method that reduces the sum of the square of errors over all the output nodes to zero (0.000000) in fewer than ten cycles. this is done by initialization of the synaptic links with the coefficients of the orthogonal expansion. this feature makes it possible to design a computer chip which can perform the training process in real time. similarly, the ability to train in real time allows the system to retrain itself and improve its performance while executing its normal testing functions.",1995-04-18,"The title of the patent is artificial neural network method and architecture and its abstract is an architecture and data processing method for a neural network that can approximate any mapping function between the input and output vectors without the use of hidden layers. the data processing is done at the sibling nodes (second row). it is based on the orthogonal expansion of the functions that map the input vector to the output vector. because the nodes of the second row are simply data processing stations, they remain passive during training. as a result the system is basically a single-layer linear network with a filter at its entrance. because of this it is free from the problems of local minima. the invention also includes a method that reduces the sum of the square of errors over all the output nodes to zero (0.000000) in fewer than ten cycles. this is done by initialization of the synaptic links with the coefficients of the orthogonal expansion. this feature makes it possible to design a computer chip which can perform the training process in real time. similarly, the ability to train in real time allows the system to retrain itself and improve its performance while executing its normal testing functions. dated 1995-04-18"
5410477,control system for an automotive vehicle having apparatus for predicting the driving environment of the vehicle,"a total control system for an automotive vehicle assures vehicular behavior precisely following to a driver's demand for variation of driving environment and provide smooth transition in variation of the driving environment. the system includes a driving environment index predicting section predicting vehicular driving environment on the basis of a driving operation indicative amount, such as an accelerator depression magnitude, a brake depression magnitude, a steering angular position and so forth and a vehicular condition indicative amount, such as an engine speed, a vehicle speed, a longitudinal acceleration and so forth. based on the driving environment index derived by the predicting section, local control channels of the automotive vehicle are controlled. the driving environment index predicting section predicts the driving environment index by neural network or so forth to transfer to the local control channels though a vehicular local area network or a common memory. accordingly, variable control corresponding to the driving environment in the local control channels can be realized.",1995-04-25,"The title of the patent is control system for an automotive vehicle having apparatus for predicting the driving environment of the vehicle and its abstract is a total control system for an automotive vehicle assures vehicular behavior precisely following to a driver's demand for variation of driving environment and provide smooth transition in variation of the driving environment. the system includes a driving environment index predicting section predicting vehicular driving environment on the basis of a driving operation indicative amount, such as an accelerator depression magnitude, a brake depression magnitude, a steering angular position and so forth and a vehicular condition indicative amount, such as an engine speed, a vehicle speed, a longitudinal acceleration and so forth. based on the driving environment index derived by the predicting section, local control channels of the automotive vehicle are controlled. the driving environment index predicting section predicts the driving environment index by neural network or so forth to transfer to the local control channels though a vehicular local area network or a common memory. accordingly, variable control corresponding to the driving environment in the local control channels can be realized. dated 1995-04-25"
5412163,elevator control apparatus,"an elevator control apparatus determines the time required for a call to reach a hall and controls an operation of the car using the obtained estimated travel time. the elevator control apparatus includes an input data conversion unit for converting traffic data, including car position, car direction data, and data regarding car calls and hall calls into data that can be used as input data to a neural network. an estimated travel time operation unit including an input layer is provided for taking in the input data. an output layer is provided for outputting the estimated travel time. an intermediate layer is provided between the input and output layers in which a weighting factor is set. the estimated travel time operation unit comprises a neural network and an output data conversion unit for converting the estimated travel time output from the output layer into data that can be used for a predetermined control operation.",1995-05-02,"The title of the patent is elevator control apparatus and its abstract is an elevator control apparatus determines the time required for a call to reach a hall and controls an operation of the car using the obtained estimated travel time. the elevator control apparatus includes an input data conversion unit for converting traffic data, including car position, car direction data, and data regarding car calls and hall calls into data that can be used as input data to a neural network. an estimated travel time operation unit including an input layer is provided for taking in the input data. an output layer is provided for outputting the estimated travel time. an intermediate layer is provided between the input and output layers in which a weighting factor is set. the estimated travel time operation unit comprises a neural network and an output data conversion unit for converting the estimated travel time output from the output layer into data that can be used for a predetermined control operation. dated 1995-05-02"
5412256,neuron for use in self-learning neural network,"a neuron for use in a self-learning neural network comprises a current input node at which a plurality of synaptic input currents are summed using kirchoff's current law. the summed input currents are normalized using a coarse gain current normalizer. the normalized summed inputs current is then converted to a voltage using a current to voltage converter. this voltage is then amplified by a gain controlled cascode output amplifier. gain control inputs are provided in the output amplifier so that the neuron can be settled by the mean field approximation. a noise input stage is also connected to the output amplifier so that the neuron can be settled using simulated annealing. the resulting neuron is a variable gain, bi-directional current transimpedance neuron with a controllable noise input.",1995-05-02,"The title of the patent is neuron for use in self-learning neural network and its abstract is a neuron for use in a self-learning neural network comprises a current input node at which a plurality of synaptic input currents are summed using kirchoff's current law. the summed input currents are normalized using a coarse gain current normalizer. the normalized summed inputs current is then converted to a voltage using a current to voltage converter. this voltage is then amplified by a gain controlled cascode output amplifier. gain control inputs are provided in the output amplifier so that the neuron can be settled by the mean field approximation. a noise input stage is also connected to the output amplifier so that the neuron can be settled using simulated annealing. the resulting neuron is a variable gain, bi-directional current transimpedance neuron with a controllable noise input. dated 1995-05-02"
5412565,electronic synapse circuit for artificial neural network,"an electronic synapse circuit is disclosed for multiplying an analog weight signal value by a digital state signal value to achieve a signed product value as a current which is capable of being summed with other such synapse circuit outputs. the circuit employs a storage multiplying digital-to-analog converter which provides storage for the analog weight signal value. additional circuitry permits programming different analog weight signal values into the circuit, performing four-quadrant multiplication, generating a current summable output, and maintaining the stored analog weight signal value at a substantially constant value independent of the digital state signal values.",1995-05-02,"The title of the patent is electronic synapse circuit for artificial neural network and its abstract is an electronic synapse circuit is disclosed for multiplying an analog weight signal value by a digital state signal value to achieve a signed product value as a current which is capable of being summed with other such synapse circuit outputs. the circuit employs a storage multiplying digital-to-analog converter which provides storage for the analog weight signal value. additional circuitry permits programming different analog weight signal values into the circuit, performing four-quadrant multiplication, generating a current summable output, and maintaining the stored analog weight signal value at a substantially constant value independent of the digital state signal values. dated 1995-05-02"
5412576,selective assembly of component kits,a computerized method for the automatic selection of component kits from an inventory of component parts. a first list of component parts is created by a rule-based expert system and a second list of component parts is created by a previously trained node-based neural network. the first and second lists are then reconciled into a final list.,1995-05-02,The title of the patent is selective assembly of component kits and its abstract is a computerized method for the automatic selection of component kits from an inventory of component parts. a first list of component parts is created by a rule-based expert system and a second list of component parts is created by a previously trained node-based neural network. the first and second lists are then reconciled into a final list. dated 1995-05-02
5412670,n-bit parity neural network encoder,"a three layer artificial neural network having an n terminal input, a two cell hidden and a single cell output layer generates an output parity signal indicating whether an even or an odd number of binary bits are asserted at the n terminal input. the two hidden layer neural cells have activation functions that cause deviations about a linear response characteristic that allow the classification of a signal representative of the number of asserted input bits into odd and even groups. this network represents a significant reduction in the number of hidden units previously required because of the particular form of activation transfer characteristic used in the two hidden layer neural cells.",1995-05-02,"The title of the patent is n-bit parity neural network encoder and its abstract is a three layer artificial neural network having an n terminal input, a two cell hidden and a single cell output layer generates an output parity signal indicating whether an even or an odd number of binary bits are asserted at the n terminal input. the two hidden layer neural cells have activation functions that cause deviations about a linear response characteristic that allow the classification of a signal representative of the number of asserted input bits into odd and even groups. this network represents a significant reduction in the number of hidden units previously required because of the particular form of activation transfer characteristic used in the two hidden layer neural cells. dated 1995-05-02"
5412754,reverse time delay neural network for pattern generation,"trajectories are generated in response to an input label by using a reverse time delay neural network. the reverse time delay neural network comprises an input layer, a plurality of hidden layers, and an output layer, all arranged in succession so that the number of frames per layer increases as the network is traversed from the input layer to the output layer. additionally, the number of features decreases as the network is traversed from the input layer to the output layer. features of the trajectory are created from the input label so that a time series of frames can be output by the network. frames generally relate to particular epochs of time or time units and a frame includes a plurality of features. interconnection between layers is accomplished using differential neuron units. in the differential neuron unit, standard neuron weighting, summing, and nonlinear squashing functions are performed on the inputs thereto. moreover, the output of the differential neuron unit includes a contribution from the value of the corresponding feature in the previous frame in the same layer as the differential neuron unit output.",1995-05-02,"The title of the patent is reverse time delay neural network for pattern generation and its abstract is trajectories are generated in response to an input label by using a reverse time delay neural network. the reverse time delay neural network comprises an input layer, a plurality of hidden layers, and an output layer, all arranged in succession so that the number of frames per layer increases as the network is traversed from the input layer to the output layer. additionally, the number of features decreases as the network is traversed from the input layer to the output layer. features of the trajectory are created from the input label so that a time series of frames can be output by the network. frames generally relate to particular epochs of time or time units and a frame includes a plurality of features. interconnection between layers is accomplished using differential neuron units. in the differential neuron unit, standard neuron weighting, summing, and nonlinear squashing functions are performed on the inputs thereto. moreover, the output of the differential neuron unit includes a contribution from the value of the corresponding feature in the previous frame in the same layer as the differential neuron unit output. dated 1995-05-02"
5413029,system and method for improved weapons systems using a kalman filter,"in a device and method for predicting a future muzzle velocity of an indirect fire weapon 3, 7 means 9, 11 responsive to a measurement of muzzle velocity are adapted to implement an adaptive empirical prediction method to predict the future muzzle velocity. the invention also relates to an aiming system and method for an indirect-fire weapon 3, 7. the system comprises a muzzle velocity measuring device 5, and predictor means 9, 11 responsive to an output of the muzzle velocity measuring device 5 for determining a new elevation setting from the weapon. preferably, the predictor means utilizes an adaptive empirical prediction method such as a kalman filter or neural network.",1995-05-09,"The title of the patent is system and method for improved weapons systems using a kalman filter and its abstract is in a device and method for predicting a future muzzle velocity of an indirect fire weapon 3, 7 means 9, 11 responsive to a measurement of muzzle velocity are adapted to implement an adaptive empirical prediction method to predict the future muzzle velocity. the invention also relates to an aiming system and method for an indirect-fire weapon 3, 7. the system comprises a muzzle velocity measuring device 5, and predictor means 9, 11 responsive to an output of the muzzle velocity measuring device 5 for determining a new elevation setting from the weapon. preferably, the predictor means utilizes an adaptive empirical prediction method such as a kalman filter or neural network. dated 1995-05-09"
5413116,method and apparatus for diagnosing joints,"a method and apparatus for diagnosing joints based on sensed joint vibrations. accelerometers disposed on the skin adjacent to the joint detect vibrational patterns during movement of the joint. these patterns are then processed by one processor to generate a predetermined set of data parameters descriptive of the vibration pattern. also, the position and velocity of the joint during the vibration is recorded. this information from numerous patients with known joint conditions is used to train a adaptive interpreter, such as a neural network, to produce an output in response to these inputs which is indicative of the known joint condition. once trained, the adaptive interpreter can then interpret this set of parameters for an unknown joint to generate a fast and reliable diagnosis. the result is a non-subjective joint disorder classification system that can be utilized by persons without particular expertise in analyzing joint vibrational patterns.",1995-05-09,"The title of the patent is method and apparatus for diagnosing joints and its abstract is a method and apparatus for diagnosing joints based on sensed joint vibrations. accelerometers disposed on the skin adjacent to the joint detect vibrational patterns during movement of the joint. these patterns are then processed by one processor to generate a predetermined set of data parameters descriptive of the vibration pattern. also, the position and velocity of the joint during the vibration is recorded. this information from numerous patients with known joint conditions is used to train a adaptive interpreter, such as a neural network, to produce an output in response to these inputs which is indicative of the known joint condition. once trained, the adaptive interpreter can then interpret this set of parameters for an unknown joint to generate a fast and reliable diagnosis. the result is a non-subjective joint disorder classification system that can be utilized by persons without particular expertise in analyzing joint vibrational patterns. dated 1995-05-09"
5416308,transaction document reader,"a transaction document reader reads image data on a transaction document. the image data includes marking areas employed by a user to record marks and includes characters. a sensor senses, one row at a time, rows of image data as the document is transported across the sensor. each row includes a number of pixel areas. the pixel areas of the rows being aligned in columns. the sensor generates output signals representative of the pixel areas. the output signals are transformed into bit signals representative of each pixel area. the bit signals are stored in image memory which is organized into rows and columns of bit signals corresponding to the rows and columns of the pixel areas of the image data. the marking areas are located in image memory. the marks are identified in the marking areas while additional rows of bit signals are being stored and while other marking areas are being located by probing the bit signals in image memory. a specific area within image memory representative of a single character is also located. a neural network is utilized to recognize the character within specific area.",1995-05-16,"The title of the patent is transaction document reader and its abstract is a transaction document reader reads image data on a transaction document. the image data includes marking areas employed by a user to record marks and includes characters. a sensor senses, one row at a time, rows of image data as the document is transported across the sensor. each row includes a number of pixel areas. the pixel areas of the rows being aligned in columns. the sensor generates output signals representative of the pixel areas. the output signals are transformed into bit signals representative of each pixel area. the bit signals are stored in image memory which is organized into rows and columns of bit signals corresponding to the rows and columns of the pixel areas of the image data. the marking areas are located in image memory. the marks are identified in the marking areas while additional rows of bit signals are being stored and while other marking areas are being located by probing the bit signals in image memory. a specific area within image memory representative of a single character is also located. a neural network is utilized to recognize the character within specific area. dated 1995-05-16"
5416696,method and apparatus for translating words in an artificial neural network,"an apparatus for translating original words in an original sentence into words in a translated sentence consists of a sentence structure analysis dictionary, a translation dictionary, an inflection dictionary, a knowledge dictionary, grammar dictionaries of original language and translated language, a translation unit, and a processing device with an artificial neural network in which an external input is provided one after another to each of the artificial neurons assigned the words of the translated language linguistically corresponding to the original words classified as polysemy and semantically corresponding to the original words classified as monosemy, the external input provided to each of the artificial neurons is stored therein, and the value of the external input previously stored in the artificial neurons is uniformly reduced to again provide to each of the artificial neurons as past records each time the external input is provided to each of the artificial neurons.",1995-05-16,"The title of the patent is method and apparatus for translating words in an artificial neural network and its abstract is an apparatus for translating original words in an original sentence into words in a translated sentence consists of a sentence structure analysis dictionary, a translation dictionary, an inflection dictionary, a knowledge dictionary, grammar dictionaries of original language and translated language, a translation unit, and a processing device with an artificial neural network in which an external input is provided one after another to each of the artificial neurons assigned the words of the translated language linguistically corresponding to the original words classified as polysemy and semantically corresponding to the original words classified as monosemy, the external input provided to each of the artificial neurons is stored therein, and the value of the external input previously stored in the artificial neurons is uniformly reduced to again provide to each of the artificial neurons as past records each time the external input is provided to each of the artificial neurons. dated 1995-05-16"
5416850,associative pattern conversion system and adaption method thereof,"an associative pattern conversion system is disclosed which may be used for image recognition. the system includes an image input portion, an image processing portion and a recognition portion. the image processing portion includes a process unit for extracting characteristics and a frame memory for holding image data. the recognition portion, which includes a component for the learning of data to be associated, obtains the extracted characteristics from the image processing portion and performs associative pattern conversion from the image input portion. the system of the present invention may be applied to any neural network, preferably a matrix calculation type neural network.",1995-05-16,"The title of the patent is associative pattern conversion system and adaption method thereof and its abstract is an associative pattern conversion system is disclosed which may be used for image recognition. the system includes an image input portion, an image processing portion and a recognition portion. the image processing portion includes a process unit for extracting characteristics and a frame memory for holding image data. the recognition portion, which includes a component for the learning of data to be associated, obtains the extracted characteristics from the image processing portion and performs associative pattern conversion from the image input portion. the system of the present invention may be applied to any neural network, preferably a matrix calculation type neural network. dated 1995-05-16"
5416888,neural network for fuzzy reasoning,"a multi-layered type neural network for a fuzzy reasoning in which an if-part of a fuzzy rule is expressed by a membership function and a then-part of the fuzzy rule is expressed by a linear expression, the network comprising an if-part neural network for receiving if-part variables of all the fuzzy rules and calculating if-part membership values of all the fuzzy rules, an intermediate neural network for calculating, as a truth value of the premise of each fuzzy rule, a product of the if-part membership values for all the if-part variables, and a then-part neural network for calculating a first sum of the truth values of the premise of all the fuzzy rules, a second sum of a product of the truth values of the premise of all the fuzzy rules and then-part outputs of all the fuzzy rules, and dividing the second sum by the first sum to obtain a quotient as an inferential result.",1995-05-16,"The title of the patent is neural network for fuzzy reasoning and its abstract is a multi-layered type neural network for a fuzzy reasoning in which an if-part of a fuzzy rule is expressed by a membership function and a then-part of the fuzzy rule is expressed by a linear expression, the network comprising an if-part neural network for receiving if-part variables of all the fuzzy rules and calculating if-part membership values of all the fuzzy rules, an intermediate neural network for calculating, as a truth value of the premise of each fuzzy rule, a product of the if-part membership values for all the if-part variables, and a then-part neural network for calculating a first sum of the truth values of the premise of all the fuzzy rules, a second sum of a product of the truth values of the premise of all the fuzzy rules and then-part outputs of all the fuzzy rules, and dividing the second sum by the first sum to obtain a quotient as an inferential result. dated 1995-05-16"
5416889,method of optimizing combination by neural network,"a neural network for solving the problem of the optimization of the arrangement of n (at least two) parts which are combined with each other through connecting wires. after the weights of the synapses of a neuron which is allotted to each part are initially set, a learning process is repeated a predetermined number of times while satisfying restricting conditions. in the learning process, the fittest neurons for all the coordinates of the positions at which the parts are disposed are selected in accordance with a predetermined standard while serially updating the weights of the synapses of the other neurons so as to satisfy the restricting conditions. after the fittest neurons for all the coordinates of the positions are selected, judgement is made as to whether or not the arrangement obtained in the current learning cycle is closer to an optimal arrangement than any other arrangement which has been obtained in previous learning cycles.",1995-05-16,"The title of the patent is method of optimizing combination by neural network and its abstract is a neural network for solving the problem of the optimization of the arrangement of n (at least two) parts which are combined with each other through connecting wires. after the weights of the synapses of a neuron which is allotted to each part are initially set, a learning process is repeated a predetermined number of times while satisfying restricting conditions. in the learning process, the fittest neurons for all the coordinates of the positions at which the parts are disposed are selected in accordance with a predetermined standard while serially updating the weights of the synapses of the other neurons so as to satisfy the restricting conditions. after the fittest neurons for all the coordinates of the positions are selected, judgement is made as to whether or not the arrangement obtained in the current learning cycle is closer to an optimal arrangement than any other arrangement which has been obtained in previous learning cycles. dated 1995-05-16"
5416891,visual information processing device,"a visual information processing device having a neural network function and capable of visual information processing comprises a semiconductor integrated circuit device section equipped with a plurality of neuronic circuit regions realizing a neuron function included in the neural network function, and first and second molecular film sections provided on the integrated circuit device section. the first molecular film section comprises a light-receiving molecular film section including tij input elements having a photoelectric function and to which coupling strength levels (tij) between the plurality of neuronic circuit regions are optically written to realize electric connection between the neuronic circuit regions and image input elements for sensing visual images, each neuronic circuit region corresponding to one pixel. the second molecular film section comprises a light-emitting molecular film section including tij signal output elements having a light-emitting function to output tij matrix signals as matrix light emission patterns.",1995-05-16,"The title of the patent is visual information processing device and its abstract is a visual information processing device having a neural network function and capable of visual information processing comprises a semiconductor integrated circuit device section equipped with a plurality of neuronic circuit regions realizing a neuron function included in the neural network function, and first and second molecular film sections provided on the integrated circuit device section. the first molecular film section comprises a light-receiving molecular film section including tij input elements having a photoelectric function and to which coupling strength levels (tij) between the plurality of neuronic circuit regions are optically written to realize electric connection between the neuronic circuit regions and image input elements for sensing visual images, each neuronic circuit region corresponding to one pixel. the second molecular film section comprises a light-emitting molecular film section including tij signal output elements having a light-emitting function to output tij matrix signals as matrix light emission patterns. dated 1995-05-16"
5418710,simulator using a neural network,"a simulator includes a modelling simulate section in which properties of an apparatus to be controlled are modelled. in addition, this simulator includes a neural network in which learning is performed depending on a real control quantity for the apparatus to be controlled. a correction value relative to a simulation control quantity for the modeling simulate section is calculated in the neural network based on a process quantity output from a controlling unit. thereafter, the simulation control quantity is corrected depending on the calculated correction value. in practice, learning is performed in the neural network by changing a synapse load based on a real quantity for the apparatus to be controlled. as a result of the foregoing correction, the simulator can simulate the real apparatus more accurately.",1995-05-23,"The title of the patent is simulator using a neural network and its abstract is a simulator includes a modelling simulate section in which properties of an apparatus to be controlled are modelled. in addition, this simulator includes a neural network in which learning is performed depending on a real control quantity for the apparatus to be controlled. a correction value relative to a simulation control quantity for the modeling simulate section is calculated in the neural network based on a process quantity output from a controlling unit. thereafter, the simulation control quantity is corrected depending on the calculated correction value. in practice, learning is performed in the neural network by changing a synapse load based on a real quantity for the apparatus to be controlled. as a result of the foregoing correction, the simulator can simulate the real apparatus more accurately. dated 1995-05-23"
5419197,monitoring diagnostic apparatus using neural network,"a monitoring diagnostic apparatus for detecting an abnormality occurring in an object being monitored such as electrical equipment and determining the cause of the abnormality is disclosed. first of all, vibration or partial discharge occurring in the monitored object is detected by using a sensor installed in close proximity to the monitored object. a detection signal output by the sensor then undergoes predetermined signal processing such as the fourier transform and normalization. after the predetermined signal processing is completed, a neural network identifies the abnormality occurring in the monitored object and determining the cause of the abnormality. the neural network carries out a learning process based on causes of abnormalities occurring in the monitored object, outputting signals corresponding to the causes of the abnormalities.",1995-05-30,"The title of the patent is monitoring diagnostic apparatus using neural network and its abstract is a monitoring diagnostic apparatus for detecting an abnormality occurring in an object being monitored such as electrical equipment and determining the cause of the abnormality is disclosed. first of all, vibration or partial discharge occurring in the monitored object is detected by using a sensor installed in close proximity to the monitored object. a detection signal output by the sensor then undergoes predetermined signal processing such as the fourier transform and normalization. after the predetermined signal processing is completed, a neural network identifies the abnormality occurring in the monitored object and determining the cause of the abnormality. the neural network carries out a learning process based on causes of abnormalities occurring in the monitored object, outputting signals corresponding to the causes of the abnormalities. dated 1995-05-30"
5420939,method and apparatus for a focal neuron system,"a focal neuron system includes an image input layer comprising a plurality of input neurons, an input layer comprising a plurality of output neurons and a plurality of focal neurons located between the image input layer and the input layer. the input layer is an input to a conventional artificial neural network (ann), and a subject image is input to the image input layer for recognition in the ann. for each dimension of the image array, a maximum and minimum boundary are determined. based on the maximum and minimum boundaries, a focal neuron is selected such that the focal neuron provides appropriate scaling and translation of the subject image from the image layer to the input layer for each dimension. each activated neuron on the image input layer is projected through the selected focal neuron for each dimension. the selected focal neuron projects the subject image, with the appropriate scaling and translation, onto the input layer. the subject image is processed in the conventional ann for pattern recognition. by scaling and translating the images, the focal neuron system allows a conventional ann to operate on a much larger input space without a significant increase in the number of required neurons.",1995-05-30,"The title of the patent is method and apparatus for a focal neuron system and its abstract is a focal neuron system includes an image input layer comprising a plurality of input neurons, an input layer comprising a plurality of output neurons and a plurality of focal neurons located between the image input layer and the input layer. the input layer is an input to a conventional artificial neural network (ann), and a subject image is input to the image input layer for recognition in the ann. for each dimension of the image array, a maximum and minimum boundary are determined. based on the maximum and minimum boundaries, a focal neuron is selected such that the focal neuron provides appropriate scaling and translation of the subject image from the image layer to the input layer for each dimension. each activated neuron on the image input layer is projected through the selected focal neuron for each dimension. the selected focal neuron projects the subject image, with the appropriate scaling and translation, onto the input layer. the subject image is processed in the conventional ann for pattern recognition. by scaling and translating the images, the focal neuron system allows a conventional ann to operate on a much larger input space without a significant increase in the number of required neurons. dated 1995-05-30"
5420963,"apparatus including a neural network used for signal processing, such as signal clustering, signal identification, and a/d conversion","a signal processing apparatus using a neural network according to this invention includes a reference signal generating section for generating a plurality of reference signals having different signal values, a complement signal generating section for receiving the reference signals and an unknown input signal as an object to be processed, and generating a plurality of complement signals, indicating complement values of the corresponding reference signals with respect to a signal value obtained by multiplying the unknown input signal with a natural number, a multiplication section for receiving the reference signals and the complement signals, and multiplying the reference signals with the corresponding complement signals, and a neural network, in which a plurality of neurons are reciprocal-inhibition-coupled, the neurons receive the products obtained by the multiplication section, and the neuron, which receives the product having a largest value, outputs a spark signal.",1995-05-30,"The title of the patent is apparatus including a neural network used for signal processing, such as signal clustering, signal identification, and a/d conversion and its abstract is a signal processing apparatus using a neural network according to this invention includes a reference signal generating section for generating a plurality of reference signals having different signal values, a complement signal generating section for receiving the reference signals and an unknown input signal as an object to be processed, and generating a plurality of complement signals, indicating complement values of the corresponding reference signals with respect to a signal value obtained by multiplying the unknown input signal with a natural number, a multiplication section for receiving the reference signals and the complement signals, and multiplying the reference signals with the corresponding complement signals, and a neural network, in which a plurality of neurons are reciprocal-inhibition-coupled, the neurons receive the products obtained by the multiplication section, and the neuron, which receives the product having a largest value, outputs a spark signal. dated 1995-05-30"
5420964,"system, for learning an external evaluation standard","this invention pertains to neural network system for learning an external evaluation standard and for learning the evaluation from the outside for the processing result, in a system capable of internal evaluation of the correspondence between external information and the processing result of its own system for the input information. it purports to learn the external evaluation as the internal evaluation standard of the internal evaluation time. the learning system comprises an internal evaluation unit for evaluating an evaluation input pattern including input information at a first point in time and input information inputted at a point in time for the processing result of its own system for the input information according to the internal evaluation standard at a system execution time; and an evaluation desired pattern memory unit for making the external evaluation correspond with the evaluation input pattern and for memorizing it as an evaluation desired pattern for having the internal evaluation unit learn the external evaluation standard. the system is configured to have an internal evaluation unit to learn the evaluation desired pattern at a learning time of the system. this system is applicable in a robot control system.",1995-05-30,"The title of the patent is system, for learning an external evaluation standard and its abstract is this invention pertains to neural network system for learning an external evaluation standard and for learning the evaluation from the outside for the processing result, in a system capable of internal evaluation of the correspondence between external information and the processing result of its own system for the input information. it purports to learn the external evaluation as the internal evaluation standard of the internal evaluation time. the learning system comprises an internal evaluation unit for evaluating an evaluation input pattern including input information at a first point in time and input information inputted at a point in time for the processing result of its own system for the input information according to the internal evaluation standard at a system execution time; and an evaluation desired pattern memory unit for making the external evaluation correspond with the evaluation input pattern and for memorizing it as an evaluation desired pattern for having the internal evaluation unit learn the external evaluation standard. the system is configured to have an internal evaluation unit to learn the evaluation desired pattern at a learning time of the system. this system is applicable in a robot control system. dated 1995-05-30"
5422981,pattern recognition method and apparatus using a neural network,"a method and apparatus for using a neural network to categorize patterns from pattern feature data derived from the patterns. the average of all the learning data of the neural network is subtracted from the pattern feature data, and the result is input to the input layer of the neural network. the neural network outputs a value for each category and the pattern is categorized based on these values. the neural network includes an intermediate layer whose bias is set to zero and which includes a sigmoid transfer function which is symmetric with respect to the origin.",1995-06-06,"The title of the patent is pattern recognition method and apparatus using a neural network and its abstract is a method and apparatus for using a neural network to categorize patterns from pattern feature data derived from the patterns. the average of all the learning data of the neural network is subtracted from the pattern feature data, and the result is input to the input layer of the neural network. the neural network outputs a value for each category and the pattern is categorized based on these values. the neural network includes an intermediate layer whose bias is set to zero and which includes a sigmoid transfer function which is symmetric with respect to the origin. dated 1995-06-06"
5422982,neural networks containing variable resistors as synapses,"a synthetic neural network having a plurality of neuronal elements arranged in an input layer, an output layer, and a hidden layer between the input layer and the output layer. the network has a first plurality of synaptic weighting elements interconnecting the neuronal elements of the input layer with the neuronal elements of the hidden layer, and a second plurality of synaptic weighting elements interconnecting the neuronal elements of the hidden layer with the neuronal elements of the output layer. the improvement involves the synaptic weighting elements in the synthetic neural network being in the form of a silicon dioxide film derived from a hydrogen silsesquioxane resin. such a silicon dioxide film is characterized by a jv curve which includes both linear and non-linear regions.",1995-06-06,"The title of the patent is neural networks containing variable resistors as synapses and its abstract is a synthetic neural network having a plurality of neuronal elements arranged in an input layer, an output layer, and a hidden layer between the input layer and the output layer. the network has a first plurality of synaptic weighting elements interconnecting the neuronal elements of the input layer with the neuronal elements of the hidden layer, and a second plurality of synaptic weighting elements interconnecting the neuronal elements of the hidden layer with the neuronal elements of the output layer. the improvement involves the synaptic weighting elements in the synthetic neural network being in the form of a silicon dioxide film derived from a hydrogen silsesquioxane resin. such a silicon dioxide film is characterized by a jv curve which includes both linear and non-linear regions. dated 1995-06-06"
5422983,neural engine for emulating a neural network,the neural engine (20) is a hardware implementation of a neural network for use in real-time systems. the neural engine (20) includes a control circuit (26) and one or more multiply/accumulate circuits (28). each multiply/accumulate circuit (28) includes a parallel/serial arrangement of multiple multiplier/accumulators (84) interconnected with weight storage elements (80) to yield multiple neural weightings and sums in a single clock cycle. a neural processing language is used to program the neural engine (20) through a conventional host personal computer (22). the parallel processing permits very high processing speeds to permit real-time pattern classification capability.,1995-06-06,The title of the patent is neural engine for emulating a neural network and its abstract is the neural engine (20) is a hardware implementation of a neural network for use in real-time systems. the neural engine (20) includes a control circuit (26) and one or more multiply/accumulate circuits (28). each multiply/accumulate circuit (28) includes a parallel/serial arrangement of multiple multiplier/accumulators (84) interconnected with weight storage elements (80) to yield multiple neural weightings and sums in a single clock cycle. a neural processing language is used to program the neural engine (20) through a conventional host personal computer (22). the parallel processing permits very high processing speeds to permit real-time pattern classification capability. dated 1995-06-06
5423001,"data transmission and apparatus, data processing apparatus and a neural network which utilize phase shifted, modulated, convolutable pseudo noise","a data transmission method includes a step which modulates maximal-sequence codes which are phase shifted by different phase shift quantities based upon plural data for transmission, a step which then convolutes the modulated maximal-sequence codes to obtain transmission data, and afterwards, a step which receives the transmission data and obtains a cross-correlation of the transmission data with a maximal-sequence code which has been phase shifted by the same as the maximal-sequence code which corresponds to the data for transmission. a data transmission method also includes a step which modulates maximal-sequence codes which are phase shifted by different phase shift quantities based on plural data for transmission, then a step which convolutes the modulated maximal-sequence codes to obtain transmission data. the method previously obtains a time sequence code based on a weighting factor for all data and maximal-sequence codes which are phase shifted with corresponding phase shifting quantities, then obtains a cross-correlation of the transmission data and the time sequence code.",1995-06-06,"The title of the patent is data transmission and apparatus, data processing apparatus and a neural network which utilize phase shifted, modulated, convolutable pseudo noise and its abstract is a data transmission method includes a step which modulates maximal-sequence codes which are phase shifted by different phase shift quantities based upon plural data for transmission, a step which then convolutes the modulated maximal-sequence codes to obtain transmission data, and afterwards, a step which receives the transmission data and obtains a cross-correlation of the transmission data with a maximal-sequence code which has been phase shifted by the same as the maximal-sequence code which corresponds to the data for transmission. a data transmission method also includes a step which modulates maximal-sequence codes which are phase shifted by different phase shift quantities based on plural data for transmission, then a step which convolutes the modulated maximal-sequence codes to obtain transmission data. the method previously obtains a time sequence code based on a weighting factor for all data and maximal-sequence codes which are phase shifted with corresponding phase shifting quantities, then obtains a cross-correlation of the transmission data and the time sequence code. dated 1995-06-06"
5424736,latched neural network a/d converter,"an improved a/d converter is disclosed which employs variable voltage sources which provide reference levels for a plurality of voltage comparators. the outputs from the comparators feed forward into a latch. the higher-order outputs from the latch feed forward to the voltage sources for lower-order comparators. the reference levels of the lower-order comparators change based on the results from the higher-order outputs thus forming a neural network for performing the conversion. the latched outputs allow selection of a desired number of conversion steps or alternatively, the converter may be capable of determining when conversion is complete.",1995-06-13,"The title of the patent is latched neural network a/d converter and its abstract is an improved a/d converter is disclosed which employs variable voltage sources which provide reference levels for a plurality of voltage comparators. the outputs from the comparators feed forward into a latch. the higher-order outputs from the latch feed forward to the voltage sources for lower-order comparators. the reference levels of the lower-order comparators change based on the results from the higher-order outputs thus forming a neural network for performing the conversion. the latched outputs allow selection of a desired number of conversion steps or alternatively, the converter may be capable of determining when conversion is complete. dated 1995-06-13"
5424773,apparatus and method for generating a pseudo camera position image from a plurality of video images from different camera positions using a neural network,"the present invention generates a plurality of video image data of different camera positions for an object shot by a camera and modifies and synthesizes the video image data of a pseudo camera position sandwiched between the related camera angles from a plurality of video image data of the aforesaid different camera positions. by this, the video image of a pseudo camera position not actually shot can be obtained.",1995-06-13,"The title of the patent is apparatus and method for generating a pseudo camera position image from a plurality of video images from different camera positions using a neural network and its abstract is the present invention generates a plurality of video image data of different camera positions for an object shot by a camera and modifies and synthesizes the video image data of a pseudo camera position sandwiched between the related camera angles from a plurality of video image data of the aforesaid different camera positions. by this, the video image of a pseudo camera position not actually shot can be obtained. dated 1995-06-13"
5424959,interpretation of fluorescence fingerprints of crude oils and other hydrocarbon mixtures using neural networks,"an artificial intelligence system is used with a conglomeration of fluorescence data to provide a method of improving recognition of an unknown from its spectral pattern. customized neural network systems allow the ultimate organization and resourceful use of assumption-free variables already existing in a total scanning fluorescence database for a much more comprehensive, discrete and accurate differentiation and matching of spectra than is possible with human memory. the invention provides increased speed of fingerprinting analysis, accuracy and reliability together with a decreased learning curve and heightened objectivity for the analysis.",1995-06-13,"The title of the patent is interpretation of fluorescence fingerprints of crude oils and other hydrocarbon mixtures using neural networks and its abstract is an artificial intelligence system is used with a conglomeration of fluorescence data to provide a method of improving recognition of an unknown from its spectral pattern. customized neural network systems allow the ultimate organization and resourceful use of assumption-free variables already existing in a total scanning fluorescence database for a much more comprehensive, discrete and accurate differentiation and matching of spectra than is possible with human memory. the invention provides increased speed of fingerprinting analysis, accuracy and reliability together with a decreased learning curve and heightened objectivity for the analysis. dated 1995-06-13"
5425108,mobile type of automatic identification system for a car plate,""" an automatic identification system for a car-plate which comprises a photographing apparatus and an image-processing cpu. the system disclosed in the present invention can be mounted in a car for automatically identifying a car-plate that is under a still or moving condition. the photographing apparatus is used for taking the image of a car-plate, and the image is then transmitted into the image-processing cpu, which accurately extracts the characters of a car-plate using a """"fuzzy inference"""" technique, and identifies the characters through a character structure analysis neural network. with the automatic identification system of the present invention, an incorrect identification can be avoided in the event the image of a car-plate is blurred, and the characters is deformed or smeared. """,1995-06-13,"The title of the patent is mobile type of automatic identification system for a car plate and its abstract is "" an automatic identification system for a car-plate which comprises a photographing apparatus and an image-processing cpu. the system disclosed in the present invention can be mounted in a car for automatically identifying a car-plate that is under a still or moving condition. the photographing apparatus is used for taking the image of a car-plate, and the image is then transmitted into the image-processing cpu, which accurately extracts the characters of a car-plate using a """"fuzzy inference"""" technique, and identifies the characters through a character structure analysis neural network. with the automatic identification system of the present invention, an incorrect identification can be avoided in the event the image of a car-plate is blurred, and the characters is deformed or smeared. "" dated 1995-06-13"
5425130,apparatus for transforming voice using neural networks,"an apparatus for transforming a voice signal of a talker into a voice signal having characteristics of a different person provides apparatus for separating the talker's voice signal into a plurality of voice parameters including frequency components, a neural network for transforming at least some of the separated frequency components into those characteristic of the different person, and apparatus for combining the voice parameters for reconstituting the talker's voice signal having characteristics of the different person.",1995-06-13,"The title of the patent is apparatus for transforming voice using neural networks and its abstract is an apparatus for transforming a voice signal of a talker into a voice signal having characteristics of a different person provides apparatus for separating the talker's voice signal into a plurality of voice parameters including frequency components, a neural network for transforming at least some of the separated frequency components into those characteristic of the different person, and apparatus for combining the voice parameters for reconstituting the talker's voice signal having characteristics of the different person. dated 1995-06-13"
5426684,technique for finding the histogram region of interest for improved tone scale reproduction of digital radiographic images,"an image processing technique especially useful in processing digital radiographic images. a method for finding a histogram region of interest for improved tone scale reproduction of digital radiographic images includes the following steps. a digital radiographic image is randomly sampled with a sample having an appropriate size to delineate an object of interest. each sample is processed using texture analysis techniques to extract a plurality of texture features. using the extracted texture features, each sample is classified with a previously trained neural network classifier to determine its class. last, the pixel values belonging to the same class are accumulated to form separate histograms for each class. each of the histograms are then used to optimize tone scale reproduction.",1995-06-20,"The title of the patent is technique for finding the histogram region of interest for improved tone scale reproduction of digital radiographic images and its abstract is an image processing technique especially useful in processing digital radiographic images. a method for finding a histogram region of interest for improved tone scale reproduction of digital radiographic images includes the following steps. a digital radiographic image is randomly sampled with a sample having an appropriate size to delineate an object of interest. each sample is processed using texture analysis techniques to extract a plurality of texture features. using the extracted texture features, each sample is classified with a previously trained neural network classifier to determine its class. last, the pixel values belonging to the same class are accumulated to form separate histograms for each class. each of the histograms are then used to optimize tone scale reproduction. dated 1995-06-20"
5426720,neurocontrolled adaptive process control system,"an adaptive process control system selectively controls vibrations in a given medium in real time. unwanted vibrations present at a point being monitored in a given medium are sensed, and the system generates an appropriate offsetting vibration that is applied to the medium at a convenient location, which may be remote from the monitored point. the system includes a vibration sensor, such as one or more accelerometers, that sense both input and output vibrations present within the medium; at least one vibration generator, such as an electromagnetic shaker, that generates appropriate offsetting vibrations that are applied to the medium at one or more appropriate locations; and a neural network controller that controls the vibration generator(s) so as to force the sensed vibration at the monitored point(s) to a desired level. the adaptive vibration cancellation provided by the invention takes place in real time, and without the need to process time-consuming complex mathematical algorithms. a specific embodiment of the neural network controller includes a plurality of 4-layer neural networks configured in an adaptive filtered-x configuration.",1995-06-20,"The title of the patent is neurocontrolled adaptive process control system and its abstract is an adaptive process control system selectively controls vibrations in a given medium in real time. unwanted vibrations present at a point being monitored in a given medium are sensed, and the system generates an appropriate offsetting vibration that is applied to the medium at a convenient location, which may be remote from the monitored point. the system includes a vibration sensor, such as one or more accelerometers, that sense both input and output vibrations present within the medium; at least one vibration generator, such as an electromagnetic shaker, that generates appropriate offsetting vibrations that are applied to the medium at one or more appropriate locations; and a neural network controller that controls the vibration generator(s) so as to force the sensed vibration at the monitored point(s) to a desired level. the adaptive vibration cancellation provided by the invention takes place in real time, and without the need to process time-consuming complex mathematical algorithms. a specific embodiment of the neural network controller includes a plurality of 4-layer neural networks configured in an adaptive filtered-x configuration. dated 1995-06-20"
5426721,neural networks and methods for training neural networks,"novel neural networks and novel methods for training those networks are disclosed. the novel networks are feedforward networks having at least three layers of neurons. the training methods are easy to implement, converge rapidly, and are guaranteed to converge to a solution. a novel network structure is used, in which each corner of the input vector hypercube may be considered separately. the problem of mapping may be reduced to a sum of corner classification sub-problems. four efficient, alternative classification methods for use with the novel neural networks are also disclosed.",1995-06-20,"The title of the patent is neural networks and methods for training neural networks and its abstract is novel neural networks and novel methods for training those networks are disclosed. the novel networks are feedforward networks having at least three layers of neurons. the training methods are easy to implement, converge rapidly, and are guaranteed to converge to a solution. a novel network structure is used, in which each corner of the input vector hypercube may be considered separately. the problem of mapping may be reduced to a sum of corner classification sub-problems. four efficient, alternative classification methods for use with the novel neural networks are also disclosed. dated 1995-06-20"
5426757,data processing circuits in a neural network for processing first data stored in local register simultaneous with second data from a memory,"herein disclosed is a data processing system having a memory packaged therein for realizing a large-scale and high-speed parallel distributed processing and, especially, a data processing system for the neural network processing. the neural network processing system according to the present invention comprises: a memory circuit for storing neuron output values, connection weights, the desired values of outputs, and data necessary for learning; an input/output circuit for writing or reading data in or out of said memory circuit; a processing circuit for performing a processing for determining the neuron outputs such as the product, sum and nonlinear conversion of the data stored in said memory circuit, a comparison of the output value and its desired value, and a processing necessary for learning; and a control circuit for controlling the operations of said memory circuit, said input/output circuit and said processing circuit. the processing circuit is constructed to include at least one of an adder, a multiplier, a nonlinear transfer function circuit and a comparator so that at least a portion of the processing necessary for determining the neutron output values such as the product or sum may be accomplished in parallel. moreover, these circuits are shared among a plurality of neutrons and are operated in a time sharing manner to determine the plural neuron output values. still moreover, the aforementioned comparator compares the neuron output value determined and the desired value of the output in parallel.",1995-06-20,"The title of the patent is data processing circuits in a neural network for processing first data stored in local register simultaneous with second data from a memory and its abstract is herein disclosed is a data processing system having a memory packaged therein for realizing a large-scale and high-speed parallel distributed processing and, especially, a data processing system for the neural network processing. the neural network processing system according to the present invention comprises: a memory circuit for storing neuron output values, connection weights, the desired values of outputs, and data necessary for learning; an input/output circuit for writing or reading data in or out of said memory circuit; a processing circuit for performing a processing for determining the neuron outputs such as the product, sum and nonlinear conversion of the data stored in said memory circuit, a comparison of the output value and its desired value, and a processing necessary for learning; and a control circuit for controlling the operations of said memory circuit, said input/output circuit and said processing circuit. the processing circuit is constructed to include at least one of an adder, a multiplier, a nonlinear transfer function circuit and a comparator so that at least a portion of the processing necessary for determining the neutron output values such as the product or sum may be accomplished in parallel. moreover, these circuits are shared among a plurality of neutrons and are operated in a time sharing manner to determine the plural neuron output values. still moreover, the aforementioned comparator compares the neuron output value determined and the desired value of the output in parallel. dated 1995-06-20"
5428466,neural networks,a neural network has inputs formed by square array of optical modulators m.sub.ij and outputs by optical detectors d.sub.ij coupled to threshold comparators. a holographic plate includes a spatial modulator whose elements are controlled by a controller to form an array of optical beams from a coherent optical source. each optical beam optically interconnects a modulator m.sub.ij with a respective detector d.sub.ij. the weight values of the neural network are provided by the intensities of the optical beams. this obviates the need for an optical weighting mask between an array of light emitting diodes and a detector array allowing a higher density of lower power consumption components and reprogrammability of the network.,1995-06-27,The title of the patent is neural networks and its abstract is a neural network has inputs formed by square array of optical modulators m.sub.ij and outputs by optical detectors d.sub.ij coupled to threshold comparators. a holographic plate includes a spatial modulator whose elements are controlled by a controller to form an array of optical beams from a coherent optical source. each optical beam optically interconnects a modulator m.sub.ij with a respective detector d.sub.ij. the weight values of the neural network are provided by the intensities of the optical beams. this obviates the need for an optical weighting mask between an array of light emitting diodes and a detector array allowing a higher density of lower power consumption components and reprogrammability of the network. dated 1995-06-27
5428559,predictive control method and apparatus,"this invention relates to a predictive control method and apparatus for controlling a nonlinearly controlled object, such as a manipulator. the method and apparatus can control an object having unknown dynamic characteristics and can compensate for disturbances. a multi-layered neural network is provided as an identification model for the object to the controlled. a controlled variable signal received from the object and a manipulated variable signal for controlling the object are fed into the neural network to predict a future value of the controlled variable signal using forward calculation. the predicted future value is compared with a desired control variable value to produce an error value. the error value is input to the neural network to compute a correction value for the manipulated variable signal using back-propagation. the correction value is used to modify the manipulated variable signal value to thereby control the object.",1995-06-27,"The title of the patent is predictive control method and apparatus and its abstract is this invention relates to a predictive control method and apparatus for controlling a nonlinearly controlled object, such as a manipulator. the method and apparatus can control an object having unknown dynamic characteristics and can compensate for disturbances. a multi-layered neural network is provided as an identification model for the object to the controlled. a controlled variable signal received from the object and a manipulated variable signal for controlling the object are fed into the neural network to predict a future value of the controlled variable signal using forward calculation. the predicted future value is compared with a desired control variable value to produce an error value. the error value is input to the neural network to compute a correction value for the manipulated variable signal using back-propagation. the correction value is used to modify the manipulated variable signal value to thereby control the object. dated 1995-06-27"
5428710,fast temporal neural learning using teacher forcing,"a neural network is trained to output a time dependent target vector defined over a predetermined time interval in response to a time dependent input vector defined over the same time interval by applying corresponding elements of the error vector, or difference between the target vector and the actual neuron output vector, to the inputs of corresponding output neurons of the network corrective feedback. this feedback decreases the error and quickens the learning process, so that a much smaller number of training cycles are required to complete the learning process. a conventional gradient descent algorithm is employed to update the neural network parameters at the end of the predetermined time interval. the foregoing process is repeated in repetitive cycles until the actual output vector corresponds to the target vector. in the preferred embodiment, as the overall error of the neutral network output decreases during successive training cycles, the portion of the error fed back to the output neurons is decreased accordingly, allowing the network to learn with greater freedom from teacher forcing as the network parameters converge to their optimum values. the invention may also be used to train a neural network with stationary training and target vectors.",1995-06-27,"The title of the patent is fast temporal neural learning using teacher forcing and its abstract is a neural network is trained to output a time dependent target vector defined over a predetermined time interval in response to a time dependent input vector defined over the same time interval by applying corresponding elements of the error vector, or difference between the target vector and the actual neuron output vector, to the inputs of corresponding output neurons of the network corrective feedback. this feedback decreases the error and quickens the learning process, so that a much smaller number of training cycles are required to complete the learning process. a conventional gradient descent algorithm is employed to update the neural network parameters at the end of the predetermined time interval. the foregoing process is repeated in repetitive cycles until the actual output vector corresponds to the target vector. in the preferred embodiment, as the overall error of the neutral network output decreases during successive training cycles, the portion of the error fed back to the output neurons is decreased accordingly, allowing the network to learn with greater freedom from teacher forcing as the network parameters converge to their optimum values. the invention may also be used to train a neural network with stationary training and target vectors. dated 1995-06-27"
5428711,spatial light modulator and neural network,"a neural network system comprising input, intermediate and output layers interconnected through synapses, respectively is disclosed. each layer is comprised of a plurality of spatial light modulator units each of which is comprised of a photoconductive layer sandwiched between electrodes and a light modulation layer electrically connected to the photoconductive layer of which the light transmittance varies according to a voltage applied thereto, wherein electric currents induced by light bundles incident to the photoconductive layer are summed to cause a change in the voltage to be applied to the light modulation layer according to which the light transmittance is varied dependently thereon.",1995-06-27,"The title of the patent is spatial light modulator and neural network and its abstract is a neural network system comprising input, intermediate and output layers interconnected through synapses, respectively is disclosed. each layer is comprised of a plurality of spatial light modulator units each of which is comprised of a photoconductive layer sandwiched between electrodes and a light modulation layer electrically connected to the photoconductive layer of which the light transmittance varies according to a voltage applied thereto, wherein electric currents induced by light bundles incident to the photoconductive layer are summed to cause a change in the voltage to be applied to the light modulation layer according to which the light transmittance is varied dependently thereon. dated 1995-06-27"
5430830,adaptive weight adjusting circuit for an neural network,"a processing element utilizing the learning algorithm equ w.sub.i =w.sub.i-1 +a.sub.i where a.sub.i is one of .eta.*(x.sub.i -w.sub.i-1), +1, -1, and 0, and where: pa1 w.sub.i =a weight, pa1 w.sub.i-1 =a previous weight, pa1 .eta.=a plasticity signal, and pa1 x.sub.i =an input signal. a method of generating adaptive weight adjustments, a.sub.i, including generating .eta.*(x.sub.i -w.sub.i-1) and additional least significant bits of data representative of the term .eta.*(x.sub.i -w.sub.i-1). comparing the additional least significant bits of data to a random number and providing >, = and <, comparison signals. adding one of .eta.*(x.sub.i -w.sub.i-1), +1, -1 and 0, respectively, to the term w.sub.i-1 when one of the following occurs: at least one bit of .eta.*(x.sub.i -w.sub.i-1) equals one; .eta.*(x.sub.i -w.sub.i-1) equals zero, the comparison signal is < and the sign bit is +; .eta.* (x.sub.i -w.sub.i-1) equals zero, the comparison signal is < and the sign bit is -; or .eta.*(x.sub.i -w.sub.i-1) equals zero and the comparison signal is > or =.",1995-07-04,"The title of the patent is adaptive weight adjusting circuit for an neural network and its abstract is a processing element utilizing the learning algorithm equ w.sub.i =w.sub.i-1 +a.sub.i where a.sub.i is one of .eta.*(x.sub.i -w.sub.i-1), +1, -1, and 0, and where: pa1 w.sub.i =a weight, pa1 w.sub.i-1 =a previous weight, pa1 .eta.=a plasticity signal, and pa1 x.sub.i =an input signal. a method of generating adaptive weight adjustments, a.sub.i, including generating .eta.*(x.sub.i -w.sub.i-1) and additional least significant bits of data representative of the term .eta.*(x.sub.i -w.sub.i-1). comparing the additional least significant bits of data to a random number and providing >, = and <, comparison signals. adding one of .eta.*(x.sub.i -w.sub.i-1), +1, -1 and 0, respectively, to the term w.sub.i-1 when one of the following occurs: at least one bit of .eta.*(x.sub.i -w.sub.i-1) equals one; .eta.*(x.sub.i -w.sub.i-1) equals zero, the comparison signal is < and the sign bit is +; .eta.* (x.sub.i -w.sub.i-1) equals zero, the comparison signal is < and the sign bit is -; or .eta.*(x.sub.i -w.sub.i-1) equals zero and the comparison signal is > or =. dated 1995-07-04"
5432887,neural network system and method for factory floor scheduling,"methods are developed on a digital computer for performing work order scheduling activity in a dynamic factory floor environment, in a manner which enables scheduling heuristic knowledge from a scheduler to be encoded through an adaptive learning process, thus eliminating the need to define these rules explicitly. a sequential assignment paradigm incrementally builds up a final schedule from a partial schedule, assigning each work order to appropriate resources in turns, taking advantage of the parallel processing capability of neural networks by selecting the most appropriate resource combination (i.e. schedule generation) for each work order under simultaneous interaction of multiple scheduling constraints.",1995-07-11,"The title of the patent is neural network system and method for factory floor scheduling and its abstract is methods are developed on a digital computer for performing work order scheduling activity in a dynamic factory floor environment, in a manner which enables scheduling heuristic knowledge from a scheduler to be encoded through an adaptive learning process, thus eliminating the need to define these rules explicitly. a sequential assignment paradigm incrementally builds up a final schedule from a partial schedule, assigning each work order to appropriate resources in turns, taking advantage of the parallel processing capability of neural networks by selecting the most appropriate resource combination (i.e. schedule generation) for each work order under simultaneous interaction of multiple scheduling constraints. dated 1995-07-11"
5432889,vector neural networks,"a vector neural network (vnn) of interconnected neurons is provided in transition mappings of potential targets wherein the threshold (energy) of a single frame does not provide adequate information (energy) to declare a target position. the vnn enhances the signal-to-noise ratio (snr) by integrating target energy over multiple frames including the steps of postulating massive numbers of target tracks (the hypotheses), propagating these target tracks over multiple frames, and accommodating different velocity targets by pixel quantization. the vnn then defers thresholding to subsequent target stages when higher snr's are prevalent so that the loss of target information is minimized, and the vnn can declare both target location and velocity. the vnn can further include target maneuver detection by a process of energy balancing hypotheses.",1995-07-11,"The title of the patent is vector neural networks and its abstract is a vector neural network (vnn) of interconnected neurons is provided in transition mappings of potential targets wherein the threshold (energy) of a single frame does not provide adequate information (energy) to declare a target position. the vnn enhances the signal-to-noise ratio (snr) by integrating target energy over multiple frames including the steps of postulating massive numbers of target tracks (the hypotheses), propagating these target tracks over multiple frames, and accommodating different velocity targets by pixel quantization. the vnn then defers thresholding to subsequent target stages when higher snr's are prevalent so that the loss of target information is minimized, and the vnn can declare both target location and velocity. the vnn can further include target maneuver detection by a process of energy balancing hypotheses. dated 1995-07-11"
5434927,method and apparatus for machine vision classification and tracking,"a method and apparatus for classification and tracking objects in three-dimensional space is described. a machine vision system acquires images from roadway scenes and processes the images by analyzing the intensities of edge elements within the image. the system then applies fuzzy set theory to the location and angles of each pixel after the pixel intensities have been characterized by vectors. a neural network interprets the data created by the fuzzy set operators and classifies objects within the roadway scene. the system can also track objects within the roadway scene, such as vehicle, by forecasting potential track regions and then calculating match scores for each potential track region based on how well the edge elements from the target track regions match those from the source region as weighted by the extent the edge elements have moved.",1995-07-18,"The title of the patent is method and apparatus for machine vision classification and tracking and its abstract is a method and apparatus for classification and tracking objects in three-dimensional space is described. a machine vision system acquires images from roadway scenes and processes the images by analyzing the intensities of edge elements within the image. the system then applies fuzzy set theory to the location and angles of each pixel after the pixel intensities have been characterized by vectors. a neural network interprets the data created by the fuzzy set operators and classifies objects within the roadway scene. the system can also track objects within the roadway scene, such as vehicle, by forecasting potential track regions and then calculating match scores for each potential track region based on how well the edge elements from the target track regions match those from the source region as weighted by the extent the edge elements have moved. dated 1995-07-18"
5434950,method for making handover decisions in a radio communications network,"the invention relates to a method for making handover decisions in a radio communication network comprising a number of fixed base stations and a number of mobile units. the method utilizes an artificial neural network which is an image of the real network of a respective base station and which exhibit a behavior pattern learned through the acquisition of information from the network. thereafter, simulation is carried out in the neural network through the generation of a list of eligible base stations to which handover can be effected, every one of the eligible stations being given points. thereafter, a decision is made whether, or not, a handover will be effected by the network.",1995-07-18,"The title of the patent is method for making handover decisions in a radio communications network and its abstract is the invention relates to a method for making handover decisions in a radio communication network comprising a number of fixed base stations and a number of mobile units. the method utilizes an artificial neural network which is an image of the real network of a respective base station and which exhibit a behavior pattern learned through the acquisition of information from the network. thereafter, simulation is carried out in the neural network through the generation of a list of eligible base stations to which handover can be effected, every one of the eligible stations being given points. thereafter, a decision is made whether, or not, a handover will be effected by the network. dated 1995-07-18"
5434951,neural network system having minimum energy function value,"n neural networks having different set-values are provided, where n is an integer greater than 2. each neural network has plurality of artificial neurons and processes information. an optimal output detecting circuit receives outputs of the n neural networks and determines the optimal one of the neural networks based on the outputs of the n neural networks. an output circuit receives and outputs the output of the neural network detected by the optimal output detecting circuit.",1995-07-18,"The title of the patent is neural network system having minimum energy function value and its abstract is n neural networks having different set-values are provided, where n is an integer greater than 2. each neural network has plurality of artificial neurons and processes information. an optimal output detecting circuit receives outputs of the n neural networks and determines the optimal one of the neural networks based on the outputs of the n neural networks. an output circuit receives and outputs the output of the neural network detected by the optimal output detecting circuit. dated 1995-07-18"
5438644,translation of a neural network into a rule-based expert system,"a rule-based expert system is generated from a neural network. the neural network is trained in such a way as to avoid redundancy and to select input weights to the various processing elements in such a way as to nullify the input weights which have smaller absolute values. the neural network is translated into a set of rules by a heuristic search technique. additionally, the translation distinguishes between positive and negative attributes for efficiency and can adequately explore rule size exponential with a given parameter. both explicit and implicit knowledge of adapted neural networks are decoded and represented as if--then rules.",1995-08-01,"The title of the patent is translation of a neural network into a rule-based expert system and its abstract is a rule-based expert system is generated from a neural network. the neural network is trained in such a way as to avoid redundancy and to select input weights to the various processing elements in such a way as to nullify the input weights which have smaller absolute values. the neural network is translated into a set of rules by a heuristic search technique. additionally, the translation distinguishes between positive and negative attributes for efficiency and can adequately explore rule size exponential with a given parameter. both explicit and implicit knowledge of adapted neural networks are decoded and represented as if--then rules. dated 1995-08-01"
5438645,neural network which uses a monitor,"a neural network includes a monitor circuit for monitoring an output value of each neuron and a weighting calculation circuit for calculating a weighting between neurons in accordance with designation from the monitor circuit in order to determine the weightings between the neurons in accordance with an energy function expressed in the form of a sum of a constrain condition and an evaluation function. the monitor circuit monitors the output values of the neurons and transmits, to the weighting calculation circuit, information representing an unsatisfactory state in which a distribution of the output values does not satisfy the constrain condition. the weighting calculation circuit changes the weightings between the neurons, e.g., doubles the weightings, every time it receives this information.",1995-08-01,"The title of the patent is neural network which uses a monitor and its abstract is a neural network includes a monitor circuit for monitoring an output value of each neuron and a weighting calculation circuit for calculating a weighting between neurons in accordance with designation from the monitor circuit in order to determine the weightings between the neurons in accordance with an energy function expressed in the form of a sum of a constrain condition and an evaluation function. the monitor circuit monitors the output values of the neurons and transmits, to the weighting calculation circuit, information representing an unsatisfactory state in which a distribution of the output values does not satisfy the constrain condition. the weighting calculation circuit changes the weightings between the neurons, e.g., doubles the weightings, every time it receives this information. dated 1995-08-01"
5438646,feed-forward neural network,""" a forward feed neural network is disclosed using data flow techniques on a data flow microprocessor. as a result of this invention, a neural network is provided that has the capacity of """"learning"""" to distinguish among patterns of data which may differ recognizably from idealized cases, and is able to perform pattern recognition faster while utilizing less memory and fewer clock cycles than neural networks implemented on sequential processors. this implementation is simpler and faster because of an inherent similarity between the flow of information in the brain and in data flow architecture. """,1995-08-01,"The title of the patent is feed-forward neural network and its abstract is "" a forward feed neural network is disclosed using data flow techniques on a data flow microprocessor. as a result of this invention, a neural network is provided that has the capacity of """"learning"""" to distinguish among patterns of data which may differ recognizably from idealized cases, and is able to perform pattern recognition faster while utilizing less memory and fewer clock cycles than neural networks implemented on sequential processors. this implementation is simpler and faster because of an inherent similarity between the flow of information in the brain and in data flow architecture. "" dated 1995-08-01"
5439160,method and apparatus for obtaining reflow oven settings for soldering a pcb,"an artificial neural network is trained to recognize inputted thermal and physical features of a printed circuit board, for providing settings for a reflow oven for obtaining acceptable soldering of the printed circuit board.",1995-08-08,"The title of the patent is method and apparatus for obtaining reflow oven settings for soldering a pcb and its abstract is an artificial neural network is trained to recognize inputted thermal and physical features of a printed circuit board, for providing settings for a reflow oven for obtaining acceptable soldering of the printed circuit board. dated 1995-08-08"
5440150,non-crystalline silicon active device for large-scale digital and analog networks,"a non-crystalline silicon--preferably a-si:h--active device for use in a large-scale hardware implementation of an artificial neural network system having an analog and digital mixed morphology. a plurality of a-si:h thin-film transistors (tfts) implement addition, multiplication and weighting functionality and are arranged in a highly-connected morphology with other active and passive semiconductor elements. electrical signals are selectively applied to metal plates and light-emitting devices in order to locally or globally alter the threshold characteristics of the tfts.",1995-08-08,"The title of the patent is non-crystalline silicon active device for large-scale digital and analog networks and its abstract is a non-crystalline silicon--preferably a-si:h--active device for use in a large-scale hardware implementation of an artificial neural network system having an analog and digital mixed morphology. a plurality of a-si:h thin-film transistors (tfts) implement addition, multiplication and weighting functionality and are arranged in a highly-connected morphology with other active and passive semiconductor elements. electrical signals are selectively applied to metal plates and light-emitting devices in order to locally or globally alter the threshold characteristics of the tfts. dated 1995-08-08"
5440566,fault detection and diagnosis for printed circuit boards,"a method for detecting an diagnosing faults in printed circuit boards. emissivity data and thermal image data of the various components of the board are obtained. the emissivity data is used to correct the thermal image data. the corrected data is converted to device space data, which is then input to a previously trained neural network. the output of the neural network indicates the location of the fault. the method includes an improved method for obtaining an emissivity map and a method for providing suitable neural network input.",1995-08-08,"The title of the patent is fault detection and diagnosis for printed circuit boards and its abstract is a method for detecting an diagnosing faults in printed circuit boards. emissivity data and thermal image data of the various components of the board are obtained. the emissivity data is used to correct the thermal image data. the corrected data is converted to device space data, which is then input to a previously trained neural network. the output of the neural network indicates the location of the fault. the method includes an improved method for obtaining an emissivity map and a method for providing suitable neural network input. dated 1995-08-08"
5440651,pattern recognition neural network,"a multi-layered pattern recognition neural network that comprises an input layer (28) that is operable to be mapped onto an input space comprising a scan window (12). two hidden layers (30) and (32) map the input space to an output layer (16). the hidden layers utilize a local receptor field architecture and store representations of objects within the scan window (12) for mapping into one of a plurality of output nodes. each of the plurality of output nodes and associated representations stored in the hidden layer define an object that is centered within the scan window (12). when centered, the object and its associated representation in the hidden layer result in activation of the associated output node. the output node is only activated when the character is centered in the scan window (12). as the scan window (12) scans a string of text, the output nodes are only activated when the associated character moves within the substantial center of the scan window. the network is trained by backpropagation through various letter string such that the letter by itself within the substantial center of the scan window (12) will be recognized, and also the letter with constraints of additional letters on either side thereof will also be recognized. in addition, the center between characters is recognized when it is disposed substantially in the center of scan window (12), and a space is recognized when it is disposed within the substantial center of the scan window (12).",1995-08-08,"The title of the patent is pattern recognition neural network and its abstract is a multi-layered pattern recognition neural network that comprises an input layer (28) that is operable to be mapped onto an input space comprising a scan window (12). two hidden layers (30) and (32) map the input space to an output layer (16). the hidden layers utilize a local receptor field architecture and store representations of objects within the scan window (12) for mapping into one of a plurality of output nodes. each of the plurality of output nodes and associated representations stored in the hidden layer define an object that is centered within the scan window (12). when centered, the object and its associated representation in the hidden layer result in activation of the associated output node. the output node is only activated when the character is centered in the scan window (12). as the scan window (12) scans a string of text, the output nodes are only activated when the associated character moves within the substantial center of the scan window. the network is trained by backpropagation through various letter string such that the letter by itself within the substantial center of the scan window (12) will be recognized, and also the letter with constraints of additional letters on either side thereof will also be recognized. in addition, the center between characters is recognized when it is disposed substantially in the center of scan window (12), and a space is recognized when it is disposed within the substantial center of the scan window (12). dated 1995-08-08"
5440661,time series association learning,"an acoustic input is recognized from inferred articulatory movements output by a learned relationship between training acoustic waveforms and articulatory movements. the inferred movements are compared with template patterns prepared from training movements when the relationship was learned to regenerate an acoustic recognition. in a preferred embodiment, the acoustic articulatory relationships are learned by a neural network. subsequent input acoustic patterns then generate the inferred articulatory movements for use with the templates. articulatory movement data may be supplemented with characteristic acoustic information, e.g. relative power and high frequency data, to improve template recognition.",1995-08-08,"The title of the patent is time series association learning and its abstract is an acoustic input is recognized from inferred articulatory movements output by a learned relationship between training acoustic waveforms and articulatory movements. the inferred movements are compared with template patterns prepared from training movements when the relationship was learned to regenerate an acoustic recognition. in a preferred embodiment, the acoustic articulatory relationships are learned by a neural network. subsequent input acoustic patterns then generate the inferred articulatory movements for use with the templates. articulatory movement data may be supplemented with characteristic acoustic information, e.g. relative power and high frequency data, to improve template recognition. dated 1995-08-08"
5440670,error control system and method,"code word generation by recursive reverse flows in a neural network, and transmission systems encoding using such code words. the neural network (30) may be an array of operational amplifiers (34) as neurons with inverted amplifier output feedback through resistors (32) as the interconnection strengths to the amplifier inputs. the inversion of the amplifier output implies the dynamical flow of the neuron states is away from stored vectors; this contrasts with hopfield networks which have a dynamical flow to stored states and thus an associative memory function. the method of generating code words recursively uses this reverse dynamical flow with previously generated code words as the stored vectors. that is, the already generated code words define the stored vectors in a neural network, then the reverse dynamical flow finds a new vector away from the stored vectors, and lastly this new vector defines the next code word and the cycle repeats with the augmented set of stored vectors.",1995-08-08,"The title of the patent is error control system and method and its abstract is code word generation by recursive reverse flows in a neural network, and transmission systems encoding using such code words. the neural network (30) may be an array of operational amplifiers (34) as neurons with inverted amplifier output feedback through resistors (32) as the interconnection strengths to the amplifier inputs. the inversion of the amplifier output implies the dynamical flow of the neuron states is away from stored vectors; this contrasts with hopfield networks which have a dynamical flow to stored states and thus an associative memory function. the method of generating code words recursively uses this reverse dynamical flow with previously generated code words as the stored vectors. that is, the already generated code words define the stored vectors in a neural network, then the reverse dynamical flow finds a new vector away from the stored vectors, and lastly this new vector defines the next code word and the cycle repeats with the augmented set of stored vectors. dated 1995-08-08"
5442209,synapse mos transistor,"a synapse mos transistor has gate electrodes of different lengths, different widths or different lengths and widths, between one source region and one drain region. thus, when using the synapse mos transistor to implement a neural network, the chip area can be greatly reduced.",1995-08-15,"The title of the patent is synapse mos transistor and its abstract is a synapse mos transistor has gate electrodes of different lengths, different widths or different lengths and widths, between one source region and one drain region. thus, when using the synapse mos transistor to implement a neural network, the chip area can be greatly reduced. dated 1995-08-15"
5442543,"neural filter architecture for overcoming noise interference in a non-linear, adaptive manner","the non-linear filter architecture according to the invention provides a neural network for modelling a non-linear transfer function, there being supplied to the neural network, on the input side, the filter input signals f(n), . . . , f(n-i), . . . , f(n-m), a time index signal i and the values p(n), . . . , p(n-i), . . . , p(n-m) of the parameter vector p. the neural network uses these values to calculate, at each time i, output values which are summed for the m+1 times i=0, . . . , m, as a result of which the filter output function g(n) is formed. the invention can be used for implementing a method for overcoming noise signals in digital signal processing, by using a circuit arrangement or a software system. specifically, the invention can be used in a method for suppressing cardio-interference in magneto-encephalography. the invention can furthermore be used for overcoming motor noise.",1995-08-15,"The title of the patent is neural filter architecture for overcoming noise interference in a non-linear, adaptive manner and its abstract is the non-linear filter architecture according to the invention provides a neural network for modelling a non-linear transfer function, there being supplied to the neural network, on the input side, the filter input signals f(n), . . . , f(n-i), . . . , f(n-m), a time index signal i and the values p(n), . . . , p(n-i), . . . , p(n-m) of the parameter vector p. the neural network uses these values to calculate, at each time i, output values which are summed for the m+1 times i=0, . . . , m, as a result of which the filter output function g(n) is formed. the invention can be used for implementing a method for overcoming noise signals in digital signal processing, by using a circuit arrangement or a software system. specifically, the invention can be used in a method for suppressing cardio-interference in magneto-encephalography. the invention can furthermore be used for overcoming motor noise. dated 1995-08-15"
5442715,method and apparatus for cursive script recognition,"a method and apparatus for performing cursive script recognition is disclosed, wherein a cursive word, in the form of digitized data or bitmap, is simultaneously segmented and individual characters of the word are recognized using a scanning window that moves across a word field or segment. the bitmap data is preprocessed before being presented to a moving window letter center finding neural network that determines the spatial location of the centers of individual letters. character center data generated by the center finding neural network is then used to define the left and right edges of a fixed size window. the fixed window contains width normalized word segments that are presented to a second neural network which is taught to recognize the central character of the window based on portions of adjacent characters contained within the window. a final character recognition decision is made using a third neural network that combines the results of second neural network at three positions, namely, when the character to be recognized was in a precharacter position, a central position and a post character position relative to the third neural network.",1995-08-15,"The title of the patent is method and apparatus for cursive script recognition and its abstract is a method and apparatus for performing cursive script recognition is disclosed, wherein a cursive word, in the form of digitized data or bitmap, is simultaneously segmented and individual characters of the word are recognized using a scanning window that moves across a word field or segment. the bitmap data is preprocessed before being presented to a moving window letter center finding neural network that determines the spatial location of the centers of individual letters. character center data generated by the center finding neural network is then used to define the left and right edges of a fixed size window. the fixed window contains width normalized word segments that are presented to a second neural network which is taught to recognize the central character of the window based on portions of adjacent characters contained within the window. a final character recognition decision is made using a third neural network that combines the results of second neural network at three positions, namely, when the character to be recognized was in a precharacter position, a central position and a post character position relative to the third neural network. dated 1995-08-15"
5442730,adaptive job scheduling using neural network priority functions,"a job scheduler makes decisions concerning the order and frequency of access to a resource according to a substantially optimum delay cost function. the delay cost function is a single value function of one or more inputs, where at least one of the inputs is a delay time which increases as a job waits for service. the job scheduler is preferably used by a multi-user computer operating system to schedule jobs of different classes. the delay cost functions are preferably implemented by neural networks. the user specifies desired performance objectives for each job class. the computer system runs for a specified period of time, collecting data on system performance. the differences between the actual and desired performance objectives are computed, and used to adaptively train the neural network. the process repeats until the delay cost functions stabilize near optimum value. however, if the system configuration, workload, or desired performance objectives change, the neural network will again start to adapt.",1995-08-15,"The title of the patent is adaptive job scheduling using neural network priority functions and its abstract is a job scheduler makes decisions concerning the order and frequency of access to a resource according to a substantially optimum delay cost function. the delay cost function is a single value function of one or more inputs, where at least one of the inputs is a delay time which increases as a job waits for service. the job scheduler is preferably used by a multi-user computer operating system to schedule jobs of different classes. the delay cost functions are preferably implemented by neural networks. the user specifies desired performance objectives for each job class. the computer system runs for a specified period of time, collecting data on system performance. the differences between the actual and desired performance objectives are computed, and used to adaptively train the neural network. the process repeats until the delay cost functions stabilize near optimum value. however, if the system configuration, workload, or desired performance objectives change, the neural network will again start to adapt. dated 1995-08-15"
5443864,spatial light modulator and a neural network circuit,"the present invention relates to a spatial light modulator used for an optical computing system or display, which comprises a photoconductive layer having a rectification function for receiving incident lights to generate charges, an electrode for accumulating the charges and a ferroelectric liquid crystal layer for modulating the incident light according to bias voltage change with the accumulated charges, wherein the ferroelectric liquid crystal layer is arranged between a pair of alignment layers made of polyimide represented by the general formula (i); ##str1## wherein n is 2 or more, x is o, s, se or te, y is an aromatic group or a substituted aromatic group, and z is a group containing an aromatic group. in the spatial light modulator, any charges are not accumulated on the alignment layer with the driving time and thus a bistable memory condition can be realized.",1995-08-22,"The title of the patent is spatial light modulator and a neural network circuit and its abstract is the present invention relates to a spatial light modulator used for an optical computing system or display, which comprises a photoconductive layer having a rectification function for receiving incident lights to generate charges, an electrode for accumulating the charges and a ferroelectric liquid crystal layer for modulating the incident light according to bias voltage change with the accumulated charges, wherein the ferroelectric liquid crystal layer is arranged between a pair of alignment layers made of polyimide represented by the general formula (i); ##str1## wherein n is 2 or more, x is o, s, se or te, y is an aromatic group or a substituted aromatic group, and z is a group containing an aromatic group. in the spatial light modulator, any charges are not accumulated on the alignment layer with the driving time and thus a bistable memory condition can be realized. dated 1995-08-22"
5444796,method for unsupervised neural network classification with back propagation,"an unsupervised back propagation method for training neural networks. for a set of inputs, target outputs are assigned 1's and 0's randomly or arbitrarily for a small number of outputs. the learning process is initiated and the convergence of outputs towards targets is monitored. at intervals, the learning is paused, and the values for those targets for the outputs which are converging at a less than average rate, are changed (e.g., 0.fwdarw.1, or 1.fwdarw.0), and the learning is then resumed with the new targets. the process is continuously iterated and the outputs converge on a stable classification, thereby providing unsupervised back propagation. in a further embodiment, samples classified with the trained network may serve as the training sets for additional subdivisions to grow additional layers of a hierarchical classification tree which converges to indivisible branch tips. after training is completed, such a tree may be used to classify new unlabelled samples with high efficiency. in yet another embodiment, the unsupervised back propagation method of the present invention may be adapted to classify fuzzy sets.",1995-08-22,"The title of the patent is method for unsupervised neural network classification with back propagation and its abstract is an unsupervised back propagation method for training neural networks. for a set of inputs, target outputs are assigned 1's and 0's randomly or arbitrarily for a small number of outputs. the learning process is initiated and the convergence of outputs towards targets is monitored. at intervals, the learning is paused, and the values for those targets for the outputs which are converging at a less than average rate, are changed (e.g., 0.fwdarw.1, or 1.fwdarw.0), and the learning is then resumed with the new targets. the process is continuously iterated and the outputs converge on a stable classification, thereby providing unsupervised back propagation. in a further embodiment, samples classified with the trained network may serve as the training sets for additional subdivisions to grow additional layers of a hierarchical classification tree which converges to indivisible branch tips. after training is completed, such a tree may be used to classify new unlabelled samples with high efficiency. in yet another embodiment, the unsupervised back propagation method of the present invention may be adapted to classify fuzzy sets. dated 1995-08-22"
5444819,economic phenomenon predicting and analyzing system using neural network,"an economic phenomenon predicting and/or analyzing system using a neural network. in the disclosed system, time series data indicating economic phenomena are input to preparation modules, and moving-average values and their differences are generated. one of the preparation modules performs a predetermined process over the time series data indicating an economic phenomenon, i.e. the change of topix, to remove trends. a pattern sorter sorts the trend-free data into a certain number of groups. average values of various time series data, their differences and the result of pattern sorting are input to input layer neurons of the network. the network is provided in advance with learning information of the change of topix in the past. the output of the output layer neurons will be a value of prediction of the change of topix. for the output of hidden layer neurons, principal components are obtained by principal analysis modules. a correlation analysis module obtains a distribution of frequency of principal component rankings and analyzes the correlation between the explanation variants and the output of the neural network based on the obtained distribution of frequency.",1995-08-22,"The title of the patent is economic phenomenon predicting and analyzing system using neural network and its abstract is an economic phenomenon predicting and/or analyzing system using a neural network. in the disclosed system, time series data indicating economic phenomena are input to preparation modules, and moving-average values and their differences are generated. one of the preparation modules performs a predetermined process over the time series data indicating an economic phenomenon, i.e. the change of topix, to remove trends. a pattern sorter sorts the trend-free data into a certain number of groups. average values of various time series data, their differences and the result of pattern sorting are input to input layer neurons of the network. the network is provided in advance with learning information of the change of topix in the past. the output of the output layer neurons will be a value of prediction of the change of topix. for the output of hidden layer neurons, principal components are obtained by principal analysis modules. a correlation analysis module obtains a distribution of frequency of principal component rankings and analyzes the correlation between the explanation variants and the output of the neural network based on the obtained distribution of frequency. dated 1995-08-22"
5444820,adaptive system and method for predicting response times in a service environment,"a hybrid fuzzy logic/neural network prediction system and method is disclosed for predicting response times to service requests to a service provider. data from a historical database of records including customer requests and weather information are input to the hybrid system. the data is filtered to reject faulty data entries and data not necessarily useful for predicting response times to service requests such as customer comments are eliminated. a backpropagation neural network operating in a supervised learning mode is employed to decrease the effects of the inherent system nonlinearities. the prediction error from the neural network is trained to make predictions within a predetermined error limit. the neural network generates a prediction configuration; i.e. a set of neural network characteristics, for every record per geographical area, time frame, and month. a fuzzy logic classifier is used for further data reliability. a fuzzy logic classifier relying upon the fuzzy cell space predictor (fcsp) method is employed to improve predicted response times from year to year. the fuzzy logic classifier supervises the overall identification scheme and, for every record, computes a prediction configuration for its corresponding month in the preceding year. the fuzzy logic classifier then computes a prediction estimate for its neighboring months in the preceding year and computes the prediction estimate for the next time frame (i.e. morning and evening). the center of gravity method is used to smooth the different prediction estimates to obtain a final predicted response time.",1995-08-22,"The title of the patent is adaptive system and method for predicting response times in a service environment and its abstract is a hybrid fuzzy logic/neural network prediction system and method is disclosed for predicting response times to service requests to a service provider. data from a historical database of records including customer requests and weather information are input to the hybrid system. the data is filtered to reject faulty data entries and data not necessarily useful for predicting response times to service requests such as customer comments are eliminated. a backpropagation neural network operating in a supervised learning mode is employed to decrease the effects of the inherent system nonlinearities. the prediction error from the neural network is trained to make predictions within a predetermined error limit. the neural network generates a prediction configuration; i.e. a set of neural network characteristics, for every record per geographical area, time frame, and month. a fuzzy logic classifier is used for further data reliability. a fuzzy logic classifier relying upon the fuzzy cell space predictor (fcsp) method is employed to improve predicted response times from year to year. the fuzzy logic classifier supervises the overall identification scheme and, for every record, computes a prediction configuration for its corresponding month in the preceding year. the fuzzy logic classifier then computes a prediction estimate for its neighboring months in the preceding year and computes the prediction estimate for the next time frame (i.e. morning and evening). the center of gravity method is used to smooth the different prediction estimates to obtain a final predicted response time. dated 1995-08-22"
5444821,artificial neuron element with electrically programmable synaptic weight for neural networks,"a neuron element with electrically programmable synaptic weight for an artificial neural network features an excitatory-connection floating-gate transistor and an inhibitory-connection floating-gate transistor. the control gate electrodes of the two transistors are connected together, and the drain electrode of the inhibitory-connection transistor is connected to the source electrode of the excitatory-connection transistor. both of the excitatory-connection and inhibitory-connection transistors have programming electrodes. the control gate electrodes and the programming electrodes can be utilized to program the threshold voltages of the transistors and thus the synaptic weight of the neuron element.",1995-08-22,"The title of the patent is artificial neuron element with electrically programmable synaptic weight for neural networks and its abstract is a neuron element with electrically programmable synaptic weight for an artificial neural network features an excitatory-connection floating-gate transistor and an inhibitory-connection floating-gate transistor. the control gate electrodes of the two transistors are connected together, and the drain electrode of the inhibitory-connection transistor is connected to the source electrode of the excitatory-connection transistor. both of the excitatory-connection and inhibitory-connection transistors have programming electrodes. the control gate electrodes and the programming electrodes can be utilized to program the threshold voltages of the transistors and thus the synaptic weight of the neuron element. dated 1995-08-22"
5444822,semiconductor integrated circuit device carrying out parallel operational processing with electronically implemented neural network,"a semiconductor integrated circuit device electrically simulating a vital neural network includes neuron units. each neuron unit includes a plurality of laterally connected synapse units, an accumulator for accumulatively adding the outputs of the final synapse unit in the lateral connection, and a nonlinear processor for carrying out a predetermined nonlinear operational processing on the output of the accumulator. the number of the neuron units and the number of synapse units per neuron unit satisfy a relation of an integer multiple. the number of regularly operating neuron units can be made equal to that of the synapse units per neuron unit, whereby it is possible to prevent the neuron units from performing meaningless operations and an efficient neural network can be obtained.",1995-08-22,"The title of the patent is semiconductor integrated circuit device carrying out parallel operational processing with electronically implemented neural network and its abstract is a semiconductor integrated circuit device electrically simulating a vital neural network includes neuron units. each neuron unit includes a plurality of laterally connected synapse units, an accumulator for accumulatively adding the outputs of the final synapse unit in the lateral connection, and a nonlinear processor for carrying out a predetermined nonlinear operational processing on the output of the accumulator. the number of the neuron units and the number of synapse units per neuron unit satisfy a relation of an integer multiple. the number of regularly operating neuron units can be made equal to that of the synapse units per neuron unit, whereby it is possible to prevent the neuron units from performing meaningless operations and an efficient neural network can be obtained. dated 1995-08-22"
5444824,enhanced neural network shell for application programs,"an enhanced neural network shell for application programs is disclosed. the user is prompted to enter in non-technical information about the specific problem type that the user wants solved by a neural network. the user also is prompted to indicate the input data usage information to the neural network. based on this information, the neural network shell creates a neural network data structure by automatically selecting an appropriate neural network model and automatically generating an appropriate number of inputs, outputs, and/or other model-specific parameters for the selected neural network model. the user is no longer required to have expertise in neural network technology to create a neural network data structure.",1995-08-22,"The title of the patent is enhanced neural network shell for application programs and its abstract is an enhanced neural network shell for application programs is disclosed. the user is prompted to enter in non-technical information about the specific problem type that the user wants solved by a neural network. the user also is prompted to indicate the input data usage information to the neural network. based on this information, the neural network shell creates a neural network data structure by automatically selecting an appropriate neural network model and automatically generating an appropriate number of inputs, outputs, and/or other model-specific parameters for the selected neural network model. the user is no longer required to have expertise in neural network technology to create a neural network data structure. dated 1995-08-22"
5445020,tire inflation sensor,"a system that will indicate tire inflation. a two-dimensional array of sensors is used to determine the distribution of contact forces over the footprint of a pneumatic tire. a neural network may be employed to classify the patterns of force sensed in this manner, and to closely estimate the actual tire inflation pressure.",1995-08-29,"The title of the patent is tire inflation sensor and its abstract is a system that will indicate tire inflation. a two-dimensional array of sensors is used to determine the distribution of contact forces over the footprint of a pneumatic tire. a neural network may be employed to classify the patterns of force sensed in this manner, and to closely estimate the actual tire inflation pressure. dated 1995-08-29"
5446543,method and apparatus for extracting a pattern of color from an object using a neural network,"a pattern extracting apparatus extracts a pattern of a desired color from an object containing a plurality of color patterns. in the apparatus, the object is irradiated by light to provide data relating to color components and the data relating to color components are converted into munsell color data h (hue), v (value) and c (chroma). after that, a pattern of a specific color is extracted from the object based on the munsell color data.",1995-08-29,"The title of the patent is method and apparatus for extracting a pattern of color from an object using a neural network and its abstract is a pattern extracting apparatus extracts a pattern of a desired color from an object containing a plurality of color patterns. in the apparatus, the object is irradiated by light to provide data relating to color components and the data relating to color components are converted into munsell color data h (hue), v (value) and c (chroma). after that, a pattern of a specific color is extracted from the object based on the munsell color data. dated 1995-08-29"
5446828,nonlinear neural network oscillator,"a nonlinear oscillator (10) includes a neural network (12) having at least one output (12a) for outputting a one dimensional vector. the neural network includes a plurality of layers, including an input layer, an output layer, and at least one hidden layer. each of the layers includes at least one processing element (pe) that is interconnected to processing elements of adjacent layers. the input layer has an input coupled to the at least one output and includes an analog delay line (14) having a plurality of taps each of which outputs a time-delayed sample of the one dimensional output vector. each of the taps is connected to each one of the processing elements of the at least one hidden layer for providing a time-delayed sample of the one dimensional output vector thereto. the nonlinear oscillator further includes a feedback network (16) that is interposed between the output of the neural network and the input of the input layer for modifying a magnitude and/or a polarity of the one dimensional output vector prior to the sample of the one dimensional output vector being applied to the input of the analog delay line. the analog delay line is capable of being shifted in either a first or a second direction. connection weights of the neural network are trained on a deterministic sequence of data from a chaotic source or may be a representation of a stochastic process, wherein each of the weights is randomly selected.",1995-08-29,"The title of the patent is nonlinear neural network oscillator and its abstract is a nonlinear oscillator (10) includes a neural network (12) having at least one output (12a) for outputting a one dimensional vector. the neural network includes a plurality of layers, including an input layer, an output layer, and at least one hidden layer. each of the layers includes at least one processing element (pe) that is interconnected to processing elements of adjacent layers. the input layer has an input coupled to the at least one output and includes an analog delay line (14) having a plurality of taps each of which outputs a time-delayed sample of the one dimensional output vector. each of the taps is connected to each one of the processing elements of the at least one hidden layer for providing a time-delayed sample of the one dimensional output vector thereto. the nonlinear oscillator further includes a feedback network (16) that is interposed between the output of the neural network and the input of the input layer for modifying a magnitude and/or a polarity of the one dimensional output vector prior to the sample of the one dimensional output vector being applied to the input of the analog delay line. the analog delay line is capable of being shifted in either a first or a second direction. connection weights of the neural network are trained on a deterministic sequence of data from a chaotic source or may be a representation of a stochastic process, wherein each of the weights is randomly selected. dated 1995-08-29"
5446829,artificial network for temporal sequence processing,"a computer-based artificial network is presented that is capable of learning, recognizing, and generating temporal-spatial sequences. the network system includes time-delays and artificial neural subnetworks. the system generally includes three parts: (1) comparator units, (2) a parallel array of subnetworks and (3) delayed feedback lines from the output of the system to the neural subnetwork layer.",1995-08-29,"The title of the patent is artificial network for temporal sequence processing and its abstract is a computer-based artificial network is presented that is capable of learning, recognizing, and generating temporal-spatial sequences. the network system includes time-delays and artificial neural subnetworks. the system generally includes three parts: (1) comparator units, (2) a parallel array of subnetworks and (3) delayed feedback lines from the output of the system to the neural subnetwork layer. dated 1995-08-29"
5447166,neurocognitive adaptive computer interface method and system based on on-line measurement of the user's mental effort,"a human-computer interface uses neuroelectric signals recorded from the user's scalp i.e. electroencephalograms (eegs) to alter the program being run by the computer, for example to present less or more difficult material to the user, depending on the user's neurocognitive on-line workload score. each user is tested with a standard battery of tasks, while wearing an eeg hat, to calibrate a neurocognitive workload function. the calibrated function is user-specific and is obtained by modifying a neural network pattern analyzer which has been previously trained to index neurocognitive workload using data from a group of subjects performing the same battery of tasks.",1995-09-05,"The title of the patent is neurocognitive adaptive computer interface method and system based on on-line measurement of the user's mental effort and its abstract is a human-computer interface uses neuroelectric signals recorded from the user's scalp i.e. electroencephalograms (eegs) to alter the program being run by the computer, for example to present less or more difficult material to the user, depending on the user's neurocognitive on-line workload score. each user is tested with a standard battery of tasks, while wearing an eeg hat, to calibrate a neurocognitive workload function. the calibrated function is user-specific and is obtained by modifying a neural network pattern analyzer which has been previously trained to index neurocognitive workload using data from a group of subjects performing the same battery of tasks. dated 1995-09-05"
5448476,distributed water flow predicting device,"the distributed water flow predicting device comprises seasonal prediction model learning means for learning weight coefficients by back propagation over a prediction model a neural network model for predicting daily distributed water flows and specific characters of an hourly distributed water flow pattern for a season, based on processed actual seasonal weather data and seasonal distributed water flows, so as to identify the prediction model. seasonal distributed water flow predicting means pridicts daily distributed water flows and specific characters of hourly distributed water flow patterns for a required season by using the prediction model, based on inputted information about the weather, temperature, a weekday or a holiday. the seasonal distributed water flow predicting means compares the specific characters of an hourly distributed water flow pattern given by the prediction model with specific characters of past actual distributed water flow patterns so as to retrieve a most similar hourly distributed flow pattern from the past actual distributed water flow patterns, giving the retrieved pattern as a predicted hourly distributed water flow pattern, and multiplys the predicted daily distributed water flow by the predicted hourly distributed water flow pattern to predict an hourly distributed water flow.",1995-09-05,"The title of the patent is distributed water flow predicting device and its abstract is the distributed water flow predicting device comprises seasonal prediction model learning means for learning weight coefficients by back propagation over a prediction model a neural network model for predicting daily distributed water flows and specific characters of an hourly distributed water flow pattern for a season, based on processed actual seasonal weather data and seasonal distributed water flows, so as to identify the prediction model. seasonal distributed water flow predicting means pridicts daily distributed water flows and specific characters of hourly distributed water flow patterns for a required season by using the prediction model, based on inputted information about the weather, temperature, a weekday or a holiday. the seasonal distributed water flow predicting means compares the specific characters of an hourly distributed water flow pattern given by the prediction model with specific characters of past actual distributed water flow patterns so as to retrieve a most similar hourly distributed flow pattern from the past actual distributed water flow patterns, giving the retrieved pattern as a predicted hourly distributed water flow pattern, and multiplys the predicted daily distributed water flow by the predicted hourly distributed water flow pattern to predict an hourly distributed water flow. dated 1995-09-05"
5448484,neural network-based vehicle detection system and method,"the present invention is directed to a neural network-based system for detecting the presence of a vehicle within a traffic scene. the vehicle detection system comprises an apparatus for producing an image signal representative of an image of the traffic scene and a trainable neural network for identifying the presence of a vehicle within the traffic scene. the present invention is also directed to a method for detecting the presence of a vehicle within a traffic scene. the vehicle detection method includes the steps of producing an image signal representative of an image of the traffic scene, collecting a training set of these image signals, training a neural network from this training set of image signals to correctly identify the presence of a vehicle within the traffic scene and performing surveillance of the traffic scene with the trained neural network to detect the presence of a vehicle.",1995-09-05,"The title of the patent is neural network-based vehicle detection system and method and its abstract is the present invention is directed to a neural network-based system for detecting the presence of a vehicle within a traffic scene. the vehicle detection system comprises an apparatus for producing an image signal representative of an image of the traffic scene and a trainable neural network for identifying the presence of a vehicle within the traffic scene. the present invention is also directed to a method for detecting the presence of a vehicle within a traffic scene. the vehicle detection method includes the steps of producing an image signal representative of an image of the traffic scene, collecting a training set of these image signals, training a neural network from this training set of image signals to correctly identify the presence of a vehicle within the traffic scene and performing surveillance of the traffic scene with the trained neural network to detect the presence of a vehicle. dated 1995-09-05"
5448503,acoustic monitor,"a system for real-time analysis of weld quality in an arc welding process. he system includes a transducer which receives acoustic signals generated during the welding process. the acoustic signals are then sampled and digitized. a signal processor calculates the root mean square and peak amplitudes of the digitized signals and transforms the digitized signal into a frequency domain signal. a data processor divides the frequency domain signal into a plurality of frequency bands and calculates the average power for each band. the average power values, in addition to the peak and root mean square amplitude values, are input to an artificial neural network for analysis of weld quality. arc current and/or arc voltage signals may be input to the a/d converter alone or in combination with the acoustic signal data for subsequent signal processing and neural network analysis.",1995-09-05,"The title of the patent is acoustic monitor and its abstract is a system for real-time analysis of weld quality in an arc welding process. he system includes a transducer which receives acoustic signals generated during the welding process. the acoustic signals are then sampled and digitized. a signal processor calculates the root mean square and peak amplitudes of the digitized signals and transforms the digitized signal into a frequency domain signal. a data processor divides the frequency domain signal into a plurality of frequency bands and calculates the average power for each band. the average power values, in addition to the peak and root mean square amplitude values, are input to an artificial neural network for analysis of weld quality. arc current and/or arc voltage signals may be input to the a/d converter alone or in combination with the acoustic signal data for subsequent signal processing and neural network analysis. dated 1995-09-05"
5448681,intelligent controller with neural network and reinforcement learning,"a plant controller using reinforcement learning for controlling a plant includes action and critic networks with enhanced learning for generating a plant control signal. learning is enhanced within the action network by using a neural network configured to operate according to unsupervised learning techniques based upon a kohonen feature map. learning is enhanced within the critic network by using a distance parameter which represents the difference between the actual and desired states of the quantitative performance, or output, of the plant when generating the reinforcement signal for the action network.",1995-09-05,"The title of the patent is intelligent controller with neural network and reinforcement learning and its abstract is a plant controller using reinforcement learning for controlling a plant includes action and critic networks with enhanced learning for generating a plant control signal. learning is enhanced within the action network by using a neural network configured to operate according to unsupervised learning techniques based upon a kohonen feature map. learning is enhanced within the critic network by using a distance parameter which represents the difference between the actual and desired states of the quantitative performance, or output, of the plant when generating the reinforcement signal for the action network. dated 1995-09-05"
5448682,programmable multilayer neural network,"a programmable multilayer neural network includes a weight storing circuit for storing the weight of each synapse to perform an intended function, an interfacing circuit for transmitting the weight value stored in the storing circuit to each synapse, and a multilayer neural network circuit programmed to have the weight from the weight storing circuit and for outputting an intended output.",1995-09-05,"The title of the patent is programmable multilayer neural network and its abstract is a programmable multilayer neural network includes a weight storing circuit for storing the weight of each synapse to perform an intended function, an interfacing circuit for transmitting the weight value stored in the storing circuit to each synapse, and a multilayer neural network circuit programmed to have the weight from the weight storing circuit and for outputting an intended output. dated 1995-09-05"
5448684,"neural network, neuron, and method for recognizing a missing input valve","a neuron (100) has a null-inhibiting function so that null inputs do not affect the output of the neuron (100) or updating of its weights. the neuron (100) provides a net value based on a sum of products of each of several inputs, and corresponding weight and null values, and provides an output in response to the net value. a neural network (40) which uses such a neuron (100) has a first segmented layer (41) in which each segment (50-52) corresponds to a manufacturing process step (60-62). each segment of the first layer (41) receives as inputs measured values associated with the process step (60-62). a second layer (42) connected to the first layer (4l), is non-segmented to model the entire manufacturing process (80). the first (41) and second (42) layers are both unsupervised and competitive. a third layer (43) connected to the second layer (42) then estimates parameters of the manufacturing process (80) and is unsupervised and noncompetitive.",1995-09-05,"The title of the patent is neural network, neuron, and method for recognizing a missing input valve and its abstract is a neuron (100) has a null-inhibiting function so that null inputs do not affect the output of the neuron (100) or updating of its weights. the neuron (100) provides a net value based on a sum of products of each of several inputs, and corresponding weight and null values, and provides an output in response to the net value. a neural network (40) which uses such a neuron (100) has a first segmented layer (41) in which each segment (50-52) corresponds to a manufacturing process step (60-62). each segment of the first layer (41) receives as inputs measured values associated with the process step (60-62). a second layer (42) connected to the first layer (4l), is non-segmented to model the entire manufacturing process (80). the first (41) and second (42) layers are both unsupervised and competitive. a third layer (43) connected to the second layer (42) then estimates parameters of the manufacturing process (80) and is unsupervised and noncompetitive. dated 1995-09-05"
5450315,apparatus using a neural network for power factor calculation,a weld controller utilizes a neural network to compute power factor of a secondary circuit of a welding transformer supply power to a workpiece through a pair of welding electrodes. phase controlled switches supply power to the transformer and the neural network uses the phase angle at the time of energization of the switches and the length of time that the switches conduct to compute the power factor. the computed power factor is compared with previous computations of power factor online and any changes are interpreted as changes in resistance of the secondary circuit. this provides a measure of contact wear and the weld controller can compensate for these changes by increasing weld power. the neural network is adaptable for use with other types of control systems.,1995-09-12,The title of the patent is apparatus using a neural network for power factor calculation and its abstract is a weld controller utilizes a neural network to compute power factor of a secondary circuit of a welding transformer supply power to a workpiece through a pair of welding electrodes. phase controlled switches supply power to the transformer and the neural network uses the phase angle at the time of energization of the switches and the length of time that the switches conduct to compute the power factor. the computed power factor is compared with previous computations of power factor online and any changes are interpreted as changes in resistance of the secondary circuit. this provides a measure of contact wear and the weld controller can compensate for these changes by increasing weld power. the neural network is adaptable for use with other types of control systems. dated 1995-09-12
5450527,method for converting an existing expert system into one utilizing one or more neural networks,"a technique for converting an existing expert system into one incorporating one or more neural networks includes the steps of separating the knowledge base and inference engine of the existing expert system, identifying the external and internal inputs and outputs, identifying subsystems from the inputs and outputs, using a neural network for each subsystem, training each neural network to learn the production rules of its associated subsystem, and computing exact or interpolated outputs from a given set of inputs. each neural network utilizes a training algorithm which does not require repetitive training and which yields a global minimum to each given set of inputs.",1995-09-12,"The title of the patent is method for converting an existing expert system into one utilizing one or more neural networks and its abstract is a technique for converting an existing expert system into one incorporating one or more neural networks includes the steps of separating the knowledge base and inference engine of the existing expert system, identifying the external and internal inputs and outputs, identifying subsystems from the inputs and outputs, using a neural network for each subsystem, training each neural network to learn the production rules of its associated subsystem, and computing exact or interpolated outputs from a given set of inputs. each neural network utilizes a training algorithm which does not require repetitive training and which yields a global minimum to each given set of inputs. dated 1995-09-12"
5450528,self-learning neural multi-layer network and learning method thereof,"a self-learning multi layer neural network and the learning method thereof are characterized in that n-bit input data and m-bit desired output data are received, a weight value of each synapse is adjusted so as to produce output data corresponding to the input data, and self-learning is performed while proceeding to a next layer. thus, it is not necessary for the user to input and adjust all the weight values of the respective synapse while the network performs self-learning and a desired function.",1995-09-12,"The title of the patent is self-learning neural multi-layer network and learning method thereof and its abstract is a self-learning multi layer neural network and the learning method thereof are characterized in that n-bit input data and m-bit desired output data are received, a weight value of each synapse is adjusted so as to produce output data corresponding to the input data, and self-learning is performed while proceeding to a next layer. thus, it is not necessary for the user to input and adjust all the weight values of the respective synapse while the network performs self-learning and a desired function. dated 1995-09-12"
5450529,neural network shell for application programs,"a neural network shell has a defined interface to an application program. by interfacing with the neural network shell, any application program becomes a neural network application program. the neural network shell contains a set of utility programs that transfers data into and out of a neural network data structure. this set of utility programs allows an application program to define a new neural network model, create a neural network data structure, train a neural network, and run a neural network. once trained, the neural network data structure can be transported to other computer systems or to application programs written in different computing languages running on similar or different computer systems.",1995-09-12,"The title of the patent is neural network shell for application programs and its abstract is a neural network shell has a defined interface to an application program. by interfacing with the neural network shell, any application program becomes a neural network application program. the neural network shell contains a set of utility programs that transfers data into and out of a neural network data structure. this set of utility programs allows an application program to define a new neural network model, create a neural network data structure, train a neural network, and run a neural network. once trained, the neural network data structure can be transported to other computer systems or to application programs written in different computing languages running on similar or different computer systems. dated 1995-09-12"
5452400,method of optimizing a combination using a neural network,"a neural network for solving the problem of the optimization of the arrangement of at least two parts which are combined with each other through connecting wires. one neuron is assigned to each part. each neuron has a synapse with a weight. weights of the synapses of a neuron are initially set, and a learning process is repeated a while satisfying restricting conditions. in the learning process, the fittest neurons for all the coordinates of the positions at which the parts are disposed are selected in accordance with a predetermined standard while serially updating the weights of the synapses of the neurons so as to satisfy the restricting conditions. after the fittest neurons for all the coordinates of the positions are selected, judgment is made as to whether or not the arrangement obtained in the current learning cycle is closer to an optimal arrangement than any other arrangement which has been obtained in previous learning cycles. the optimal arrangement obtained from a set of learning cycles is provided as an indication of a quasi-optimal arrangement.",1995-09-19,"The title of the patent is method of optimizing a combination using a neural network and its abstract is a neural network for solving the problem of the optimization of the arrangement of at least two parts which are combined with each other through connecting wires. one neuron is assigned to each part. each neuron has a synapse with a weight. weights of the synapses of a neuron are initially set, and a learning process is repeated a while satisfying restricting conditions. in the learning process, the fittest neurons for all the coordinates of the positions at which the parts are disposed are selected in accordance with a predetermined standard while serially updating the weights of the synapses of the neurons so as to satisfy the restricting conditions. after the fittest neurons for all the coordinates of the positions are selected, judgment is made as to whether or not the arrangement obtained in the current learning cycle is closer to an optimal arrangement than any other arrangement which has been obtained in previous learning cycles. the optimal arrangement obtained from a set of learning cycles is provided as an indication of a quasi-optimal arrangement. dated 1995-09-19"
5452402,neural network circuit for adaptively controlling the coupling of neurons,"in a multi-layered neural network circuit provided with an input layer having input vectors, an intermediate layer having networks in tree-like structure whose outputs are necessarily determined by the values of the input vectors and whose number corresponds to the number of the input vectors of the input layer, and an output layer having plural output units for integrating all outputs of the intermediate layer, provided are learning-time memories for memorizing the numbers of times at learning in paths between the intermediate layer and the respective output units, threshold processing circuits for threshold-processing the outputs of the leaning-time memories, and connection control circuits to be controlled by the outputs of the threshold processing circuits for controlling connection of paths between the intermediate layer and the output units. the outputs of the intermediate layer connected by the connection control circuits are summed in each output unit. thus, the neural network circuit for recognizing an image or the like can execute recognition and learning of data to be recognized at high speed with small circuit size, and the recognition accuracy for unlearned data is high.",1995-09-19,"The title of the patent is neural network circuit for adaptively controlling the coupling of neurons and its abstract is in a multi-layered neural network circuit provided with an input layer having input vectors, an intermediate layer having networks in tree-like structure whose outputs are necessarily determined by the values of the input vectors and whose number corresponds to the number of the input vectors of the input layer, and an output layer having plural output units for integrating all outputs of the intermediate layer, provided are learning-time memories for memorizing the numbers of times at learning in paths between the intermediate layer and the respective output units, threshold processing circuits for threshold-processing the outputs of the leaning-time memories, and connection control circuits to be controlled by the outputs of the threshold processing circuits for controlling connection of paths between the intermediate layer and the output units. the outputs of the intermediate layer connected by the connection control circuits are summed in each output unit. thus, the neural network circuit for recognizing an image or the like can execute recognition and learning of data to be recognized at high speed with small circuit size, and the recognition accuracy for unlearned data is high. dated 1995-09-19"
5453676,trainable drive system for a windshield wiper,"a neural network provides automatic control of a windshield wiper. a rain sensor generates sensing signals indicating a rain pattern, and the neural network generates wiping demand signals indicating a wiping action desired by the driver. a training unit uses manually generated wiping supervision signals to create weight factors for the neural network.",1995-09-26,"The title of the patent is trainable drive system for a windshield wiper and its abstract is a neural network provides automatic control of a windshield wiper. a rain sensor generates sensing signals indicating a rain pattern, and the neural network generates wiping demand signals indicating a wiping action desired by the driver. a training unit uses manually generated wiping supervision signals to create weight factors for the neural network. dated 1995-09-26"
5454358,driving power control apparatus for internal combustion engine,"a neural network of a neuro computer has learned a prediction pattern based on the vehicle speed and time series data of the acceleration stroke received before the shipment of a vehicle. based on the vehicle speed and time series data of the acceleration stroke, a request on the acceleration is predicted by a neural network computer during driving. at this time, an acceleration sensor or the like is not used, and the acceleration request is predicted more directly. based on the prediction result, the throttle sensitivity is computed and set. a throttle computer refers to the throttle sensitivity to open or close the throttle valve to control the output of the engine.",1995-10-03,"The title of the patent is driving power control apparatus for internal combustion engine and its abstract is a neural network of a neuro computer has learned a prediction pattern based on the vehicle speed and time series data of the acceleration stroke received before the shipment of a vehicle. based on the vehicle speed and time series data of the acceleration stroke, a request on the acceleration is predicted by a neural network computer during driving. at this time, an acceleration sensor or the like is not used, and the acceleration request is predicted more directly. based on the prediction result, the throttle sensitivity is computed and set. a throttle computer refers to the throttle sensitivity to open or close the throttle valve to control the output of the engine. dated 1995-10-03"
5455583,combined conventional/neural network analog to digital converter,an analog-to-digital converter employs both flash and neural converters for converting an analog input voltage. the flash converter converts the higher-order bits in a single clock cycle. values of the lower-order bits are determined by outputs from comparators with reference voltages provided by digital-to-analog converters. the d/a converters receive inputs from the flash converter as well as from those comparators which provide output results for higher-order bits. this interconnection of d/a converters and comparators thus forms a neural network for determining the value of the lower-order bits.,1995-10-03,The title of the patent is combined conventional/neural network analog to digital converter and its abstract is an analog-to-digital converter employs both flash and neural converters for converting an analog input voltage. the flash converter converts the higher-order bits in a single clock cycle. values of the lower-order bits are determined by outputs from comparators with reference voltages provided by digital-to-analog converters. the d/a converters receive inputs from the flash converter as well as from those comparators which provide output results for higher-order bits. this interconnection of d/a converters and comparators thus forms a neural network for determining the value of the lower-order bits. dated 1995-10-03
5455890,method for structuring an expert system utilizing one or more neural networks,"neural networks learn expert system rules, for either business or real-time applications, to improve the robustness and speed of execution of the expert system. one or more neural networks are constructed which incorporate the production rules of one or more expert systems. each neural network is constructed of neurons or neuron circuits each having only one significant processing element in the form of a multiplier. each neural network utilizes a training algorithm which does not require repetitive training and which yields a global minimum to each given set of input vectors.",1995-10-03,"The title of the patent is method for structuring an expert system utilizing one or more neural networks and its abstract is neural networks learn expert system rules, for either business or real-time applications, to improve the robustness and speed of execution of the expert system. one or more neural networks are constructed which incorporate the production rules of one or more expert systems. each neural network is constructed of neurons or neuron circuits each having only one significant processing element in the form of a multiplier. each neural network utilizes a training algorithm which does not require repetitive training and which yields a global minimum to each given set of input vectors. dated 1995-10-03"
5455891,system and method for a learning neural network for generating random directions for weight changes,"a learning neural network (30) implements a random weight change learning algorithm within a weight adjustment mechanism (28) for manipulating the weights applied to inputs of the network (30) in order to achieve a desired functionality for the network (30). weights are changed randomly from an initial state with a small incremental weight change of either +.delta. or -.delta.. if the overall network output decreases by the weight change, the same weight change is iterated until the error increases. if, however, the overall network error increases, the weights are changed randomly again. after iterating the foregoing methodology, the network error gradually decreases and finally reaches approximately zero. furthermore, a shift mechanism (36) and a multiplier (38) are employed as a weight application mechanism (16). the shift mechanisms (36) are connected in series with a random line (35) and are connected in parallel with a shift line (44). a random direction is successively channelled through the shift mechanisms (36) via the random line (35) under the control of the shift line ( 44) so that only a single random number need be generated for all of the shift mechanisms (36) within the neural network (30).",1995-10-03,"The title of the patent is system and method for a learning neural network for generating random directions for weight changes and its abstract is a learning neural network (30) implements a random weight change learning algorithm within a weight adjustment mechanism (28) for manipulating the weights applied to inputs of the network (30) in order to achieve a desired functionality for the network (30). weights are changed randomly from an initial state with a small incremental weight change of either +.delta. or -.delta.. if the overall network output decreases by the weight change, the same weight change is iterated until the error increases. if, however, the overall network error increases, the weights are changed randomly again. after iterating the foregoing methodology, the network error gradually decreases and finally reaches approximately zero. furthermore, a shift mechanism (36) and a multiplier (38) are employed as a weight application mechanism (16). the shift mechanisms (36) are connected in series with a random line (35) and are connected in parallel with a shift line (44). a random direction is successively channelled through the shift mechanisms (36) via the random line (35) under the control of the shift line ( 44) so that only a single random number need be generated for all of the shift mechanisms (36) within the neural network (30). dated 1995-10-03"
5455892,method for training a neural network for classifying an unknown signal with respect to known signals,"the device includes a neural network with an input layer 3, an internal layer 4, and an output layer 5. this network is designed to classify data vectors to classes, the synaptic weights in the network being determined through programming on the basis of specimens whose classes are known. each class is defined during programming as corresponding to a set of neurons of which each represents a domain which contains a fixed number of specimens. the network includes a number of neurons and synaptic weights which have been determined as a function of the classes thus defined.",1995-10-03,"The title of the patent is method for training a neural network for classifying an unknown signal with respect to known signals and its abstract is the device includes a neural network with an input layer 3, an internal layer 4, and an output layer 5. this network is designed to classify data vectors to classes, the synaptic weights in the network being determined through programming on the basis of specimens whose classes are known. each class is defined during programming as corresponding to a set of neurons of which each represents a domain which contains a fixed number of specimens. the network includes a number of neurons and synaptic weights which have been determined as a function of the classes thus defined. dated 1995-10-03"
5456548,control apparatus for line marking machines,"a control system is provided to enable pavement line marking apparatus to refurbish old lines on a pavement surface. the control system has at least one old line detector adapted to scan a predetermined width of the pavement surface while being carried forwardly along the pavement. the old line detector is capable of recognizing old line pattern transition points as taught by a neural network and in response to such recognition, control activation of new line material applicators to repeat accurately said line pattern changes.",1995-10-10,"The title of the patent is control apparatus for line marking machines and its abstract is a control system is provided to enable pavement line marking apparatus to refurbish old lines on a pavement surface. the control system has at least one old line detector adapted to scan a predetermined width of the pavement surface while being carried forwardly along the pavement. the old line detector is capable of recognizing old line pattern transition points as taught by a neural network and in response to such recognition, control activation of new line material applicators to repeat accurately said line pattern changes. dated 1995-10-10"
5457770,speaker independent speech recognition system and method using neural network and/or dp matching technique,"a system and method for recognizing an utterance of a speech in which each reference pattern stored in a dictionary is constituted by a series of phonemes of a word to be recognized, each phoneme having a predetermined length of continued time and having a series of frames and a lattice point (i, j) of an i-th number phoneme at an j-th number frame having a discriminating score derived from neural networks for the corresponding phoneme. when the series of phonemes recognized by a phoneme recognition block is compared with each reference pattern, one i of the input series of phonemes recognized by the phoneme recognition block being calculated as a matching score as gk(i, j); ##equ1## wherein ak(i, j) denotes an output score value of the neural networks of the j-th number phoneme at the j-th number frame of the reference pattern and p denoted a penalty constant to avoid an extreme shrinkage of the phonemes, a total matching score is calculated as gk (i, j), i denoting the number of frames of the input series of phonemes and j denoting the number of phonemes of the reference pattern k, and one of the reference patterns which gives a maximum matching score is output as the word recognition.",1995-10-10,"The title of the patent is speaker independent speech recognition system and method using neural network and/or dp matching technique and its abstract is a system and method for recognizing an utterance of a speech in which each reference pattern stored in a dictionary is constituted by a series of phonemes of a word to be recognized, each phoneme having a predetermined length of continued time and having a series of frames and a lattice point (i, j) of an i-th number phoneme at an j-th number frame having a discriminating score derived from neural networks for the corresponding phoneme. when the series of phonemes recognized by a phoneme recognition block is compared with each reference pattern, one i of the input series of phonemes recognized by the phoneme recognition block being calculated as a matching score as gk(i, j); ##equ1## wherein ak(i, j) denotes an output score value of the neural networks of the j-th number phoneme at the j-th number frame of the reference pattern and p denoted a penalty constant to avoid an extreme shrinkage of the phonemes, a total matching score is calculated as gk (i, j), i denoting the number of frames of the input series of phonemes and j denoting the number of phonemes of the reference pattern k, and one of the reference patterns which gives a maximum matching score is output as the word recognition. dated 1995-10-10"
5459636,position and orientation estimation neural network system and method,"disclosed are a system and method for determining the pose (translation, rotation, and scale), or position and orientation, of a model object that best matches a target object located in image data. through an iterative process small adjustments are made to the original position and orientation of the model object until it converges to a state that best matches the target object contained in the image data. edge data representative of edges of the target object and edge data representative of the model object are processed for each data point in the model object relative to each point in the target object to produce a set of minimum distance vectors between the model object and the target object. a neural network estimates translation, rotation, and scaling adjustments that are to be made to the model object. pose of the model object is adjusted relative to the target object based upon the estimated translation, rotation, and scaling adjustments provided by the neural network. iterative calculation of the minimum distance vectors, estimation of the translation, rotation, and scaling adjustments, and adjustment of the position and orientation of the model object is adapted to reposition the model object until it substantially overlays the target object. final position of the model object provides an estimate of the position and orientation of the target object in the digitized image.",1995-10-17,"The title of the patent is position and orientation estimation neural network system and method and its abstract is disclosed are a system and method for determining the pose (translation, rotation, and scale), or position and orientation, of a model object that best matches a target object located in image data. through an iterative process small adjustments are made to the original position and orientation of the model object until it converges to a state that best matches the target object contained in the image data. edge data representative of edges of the target object and edge data representative of the model object are processed for each data point in the model object relative to each point in the target object to produce a set of minimum distance vectors between the model object and the target object. a neural network estimates translation, rotation, and scaling adjustments that are to be made to the model object. pose of the model object is adjusted relative to the target object based upon the estimated translation, rotation, and scaling adjustments provided by the neural network. iterative calculation of the minimum distance vectors, estimation of the translation, rotation, and scaling adjustments, and adjustment of the position and orientation of the model object is adapted to reposition the model object until it substantially overlays the target object. final position of the model object provides an estimate of the position and orientation of the target object in the digitized image. dated 1995-10-17"
5459665,transportation system traffic controlling system using a neural network,"a traffic volume estimating apparatus 1a estimates the traffic volumes of traffic apparatus, and a traffic flow presuming apparatus 1b presumes the traffic flows generating the estimated traffic volumes. a presumption function constructing apparatus 1c corrects the presumption functions of the traffic flow presuming apparatus 1b on actually measured traffic volumes, traffic flow presumption results and control results. a control result detecting apparatus 1g detects the control results and the drive results of the traffic apparatus. further, a control parameter setting apparatus 1d sets control parameters on traffic flow presumption results, and corrects the control parameters according to the control results and the drive results.",1995-10-17,"The title of the patent is transportation system traffic controlling system using a neural network and its abstract is a traffic volume estimating apparatus 1a estimates the traffic volumes of traffic apparatus, and a traffic flow presuming apparatus 1b presumes the traffic flows generating the estimated traffic volumes. a presumption function constructing apparatus 1c corrects the presumption functions of the traffic flow presuming apparatus 1b on actually measured traffic volumes, traffic flow presumption results and control results. a control result detecting apparatus 1g detects the control results and the drive results of the traffic apparatus. further, a control parameter setting apparatus 1d sets control parameters on traffic flow presumption results, and corrects the control parameters according to the control results and the drive results. dated 1995-10-17"
5459817,neural network with learning function based on simulated annealing and monte-carlo method,"a neural network with a learning function which does not require the backward propagation of the signals for the learning, which is applicable for a case involving the feedback of the synapses or the loop formed by the synapses, and which enables the construction of a large scale neural network by using compact and inexpensive circuit elements. an evaluation value is calculated according to a difference between each output signal of the network and a corresponding teacher signal; a manner of updating the synapse weight factor of each synapse is determined according to an evaluation value change between a present value and a previous value of evaluation value on a basis of the simulated annealing; a randomly changing update control signal is generated according to a random number; and a synapse weight factor of each synapse is updated according to the generated update control signal and the determined manner of updating on a basis of the monte-carlo method.",1995-10-17,"The title of the patent is neural network with learning function based on simulated annealing and monte-carlo method and its abstract is a neural network with a learning function which does not require the backward propagation of the signals for the learning, which is applicable for a case involving the feedback of the synapses or the loop formed by the synapses, and which enables the construction of a large scale neural network by using compact and inexpensive circuit elements. an evaluation value is calculated according to a difference between each output signal of the network and a corresponding teacher signal; a manner of updating the synapse weight factor of each synapse is determined according to an evaluation value change between a present value and a previous value of evaluation value on a basis of the simulated annealing; a randomly changing update control signal is generated according to a random number; and a synapse weight factor of each synapse is updated according to the generated update control signal and the determined manner of updating on a basis of the monte-carlo method. dated 1995-10-17"
5461322,autonulling dc bridge for real-time sensor transduction and voltage measurements,a feedback operated dc bridge circuit for monitoring the voltage variations in a voltage divider circuit using a voltage controlled resistance component to reach a null balance across the bridge. amplification is provided at higher accuracy near the null point when the voltage difference across the bridge is zero. the feedback bridge circuit includes an integrator which directly drives the controlling component to the value of the resistance in an unknown branch to force the null condition. the voltage controlled component (configured as a discrete metal oxide semiconductor device or bipolar junction transistor) and the balancing scheme are suitable for microfabrication and provides noise-rejection enhancement. the interconnected integral feedback of the autonulling dc bridge enables both a neural network for pre-processing sensor input in a spatial domain as well as general analog computation that mimics a first order differential equation in the form of the system state equation.,1995-10-24,The title of the patent is autonulling dc bridge for real-time sensor transduction and voltage measurements and its abstract is a feedback operated dc bridge circuit for monitoring the voltage variations in a voltage divider circuit using a voltage controlled resistance component to reach a null balance across the bridge. amplification is provided at higher accuracy near the null point when the voltage difference across the bridge is zero. the feedback bridge circuit includes an integrator which directly drives the controlling component to the value of the resistance in an unknown branch to force the null condition. the voltage controlled component (configured as a discrete metal oxide semiconductor device or bipolar junction transistor) and the balancing scheme are suitable for microfabrication and provides noise-rejection enhancement. the interconnected integral feedback of the autonulling dc bridge enables both a neural network for pre-processing sensor input in a spatial domain as well as general analog computation that mimics a first order differential equation in the form of the system state equation. dated 1995-10-24
5461559,hierarchical control system for molecular beam epitaxy,"a multi-featured control system which improves the manufacturing capability of the thin-film semiconductor growth process. this system improves repeatability and accuracy of the process, reduces the manpower requirements to operate mbe, and improves the mbe environment for scientific investigation. this system has three levels of feedback control. the first level improves the precision and tracking of the process variables, flux, and substrate temperature. the second level comprises an expert system that uses sensors to monitor the status of the product in order to tailor the process plan in real time so that the exact qualities desired are achieved. the third level features a continuously evolving neural network model of the process which is used to recommend the recipe and command inputs to achieve a desired goal. the third level is particularly useful during the development process for new materials. all three levels require models of the process which are updated during automatic process identification experiments.",1995-10-24,"The title of the patent is hierarchical control system for molecular beam epitaxy and its abstract is a multi-featured control system which improves the manufacturing capability of the thin-film semiconductor growth process. this system improves repeatability and accuracy of the process, reduces the manpower requirements to operate mbe, and improves the mbe environment for scientific investigation. this system has three levels of feedback control. the first level improves the precision and tracking of the process variables, flux, and substrate temperature. the second level comprises an expert system that uses sensors to monitor the status of the product in order to tailor the process plan in real time so that the exact qualities desired are achieved. the third level features a continuously evolving neural network model of the process which is used to recommend the recipe and command inputs to achieve a desired goal. the third level is particularly useful during the development process for new materials. all three levels require models of the process which are updated during automatic process identification experiments. dated 1995-10-24"
5461696,decision directed adaptive neural network,a method for adapting a decision directed adaptive neural network (10). the method finds the best matches between a plurality of input data vectors (16) and an associated plurality of input portion of weight vectors. the input portion of the weight vectors are adapted. the identification codes (12) which represent the sequence of best matched weight vectors are stored in a memory (12) and the associated output portion of weight vectors (22) are output. a sequence of output portion of weight vectors (22) is matched with predetermined models (21). a sequence of labels (24) associated with the best matched model is stored which identifies the categories of match data. the labels (24) are sequentially combined with the identification codes (12) to build adaptation vectors. the adaptation vectors are then used to sequentially adapt the output portion of weight vectors (22).,1995-10-24,The title of the patent is decision directed adaptive neural network and its abstract is a method for adapting a decision directed adaptive neural network (10). the method finds the best matches between a plurality of input data vectors (16) and an associated plurality of input portion of weight vectors. the input portion of the weight vectors are adapted. the identification codes (12) which represent the sequence of best matched weight vectors are stored in a memory (12) and the associated output portion of weight vectors (22) are output. a sequence of output portion of weight vectors (22) is matched with predetermined models (21). a sequence of labels (24) associated with the best matched model is stored which identifies the categories of match data. the labels (24) are sequentially combined with the identification codes (12) to build adaptation vectors. the adaptation vectors are then used to sequentially adapt the output portion of weight vectors (22). dated 1995-10-24
5461697,speaker recognition system using neural network,"a speaker recognition system for recognizing a speaker from an input voice using a neural network, in which a feature quantity extracted from the input voice is timewise averaged to create an input pattern to the neural network. the averaging technique is such that the input voice is equally divided timewise into a plurality of blocks in a simple manner and that such feature quantity is averaged every block. the feature quantity includes a frequency characteristic, pitch frequency, linear prediction coefficient, and partial self-correlation (parcor) coefficient of the voice.",1995-10-24,"The title of the patent is speaker recognition system using neural network and its abstract is a speaker recognition system for recognizing a speaker from an input voice using a neural network, in which a feature quantity extracted from the input voice is timewise averaged to create an input pattern to the neural network. the averaging technique is such that the input voice is equally divided timewise into a plurality of blocks in a simple manner and that such feature quantity is averaged every block. the feature quantity includes a frequency characteristic, pitch frequency, linear prediction coefficient, and partial self-correlation (parcor) coefficient of the voice. dated 1995-10-24"
5461698,method for modelling similarity function using neural network,"given a set of objects (a, b, c, . . . ), each described by a set of attribute values, and given a classification of these objects into categories, a similarity function accounts well for this classification when only a small number of objects are not correctly classified. a method for modelling a similarity function using a neural network comprises the steps of: (a) inputting feature vectors to a raw input stage of a neural network respectively for object s in the given category, for other objects g in the same category being compared the object s, and for object b outside the given category; (b) coupling the raw inputs of feature vectors for s, g, and b to an input layer of the neural network performing respective set operations required for the similarity function so as to have a property of monotonicity; (c) coupling the input elements of the input layer to respective processing elements of an hidden layer of the neural network for computing similarity function results adaptively with different values of a coefficient w of the similarity function; (d) coupling the processing elements of the hidden layer to respective output elements of an output layer of the neural network for providing respective outputs of an error function measuring the extent to which object s is more similar to object g than to object b; and (e) obtaining an optimal coefficient w by back propagation through the neural network which minimizes the error outputs of the error function.",1995-10-24,"The title of the patent is method for modelling similarity function using neural network and its abstract is given a set of objects (a, b, c, . . . ), each described by a set of attribute values, and given a classification of these objects into categories, a similarity function accounts well for this classification when only a small number of objects are not correctly classified. a method for modelling a similarity function using a neural network comprises the steps of: (a) inputting feature vectors to a raw input stage of a neural network respectively for object s in the given category, for other objects g in the same category being compared the object s, and for object b outside the given category; (b) coupling the raw inputs of feature vectors for s, g, and b to an input layer of the neural network performing respective set operations required for the similarity function so as to have a property of monotonicity; (c) coupling the input elements of the input layer to respective processing elements of an hidden layer of the neural network for computing similarity function results adaptively with different values of a coefficient w of the similarity function; (d) coupling the processing elements of the hidden layer to respective output elements of an output layer of the neural network for providing respective outputs of an error function measuring the extent to which object s is more similar to object g than to object b; and (e) obtaining an optimal coefficient w by back propagation through the neural network which minimizes the error outputs of the error function. dated 1995-10-24"
5461699,forecasting using a neural network and a statistical forecast,"a system and method for forecasting that combines a neural network with a statistical forecast is presented. a neural network having an input layer, a hidden layer, and an output layer with each layer having one or more nodes is presented. each node in the input layer is connected to each node in the hidden layer and each node in the hidden layer is connected to each node in the output layer. each connection between nodes has an associated weight. one node in the input layer is connected to a statistical forecast that is produced by a statistical model. all other nodes in the input layer are connected to a different historical datum from the set of historical data. the neural network being operative by outputting a forecast, the output of the output layer nodes, when presented with input data. the weights associated with the connections of the neural network are first adjusted by a training device. the training device applies a plurality of training sets to the neural network, each training set consisting of historical data, an associated statistical output and a desired forecast, with each set of training data the training device determines a difference between the forecast produced by the neural network given the training data and the desired forecast, the training device then adjusts the weights of the neural network based on the difference.",1995-10-24,"The title of the patent is forecasting using a neural network and a statistical forecast and its abstract is a system and method for forecasting that combines a neural network with a statistical forecast is presented. a neural network having an input layer, a hidden layer, and an output layer with each layer having one or more nodes is presented. each node in the input layer is connected to each node in the hidden layer and each node in the hidden layer is connected to each node in the output layer. each connection between nodes has an associated weight. one node in the input layer is connected to a statistical forecast that is produced by a statistical model. all other nodes in the input layer are connected to a different historical datum from the set of historical data. the neural network being operative by outputting a forecast, the output of the output layer nodes, when presented with input data. the weights associated with the connections of the neural network are first adjusted by a training device. the training device applies a plurality of training sets to the neural network, each training set consisting of historical data, an associated statistical output and a desired forecast, with each set of training data the training device determines a difference between the forecast produced by the neural network given the training data and the desired forecast, the training device then adjusts the weights of the neural network based on the difference. dated 1995-10-24"
5463548,method and system for differential diagnosis based on clinical and radiological information using artificial neural networks,"a method and system for computer-aided differential diagnosis of diseases, and in particular, computer-aided differential diagnosis using neural networks. a first embodiment of the neural network distinguishes between a plurality of interstitial lung diseases on the basis of inputted clinical parameters and radiographic information. a second embodiment distinguishes between malignant and benign mammographic cases based upon similar inputted clinical and radiographic information. the neural networks were first trained using a hypothetical data base made up of hypothetical cases for each of the interstitial lung diseases and for malignant and benign cases. the performance of the neural network was evaluated using receiver operating characteristics (roc) analysis. the decision performance of the neural network was compared to experienced radiologists and achieved a high performance comparable to that of the experienced radiologists. the neural network according to the invention can be made up of a single network or a plurality of successive or parallel networks. the neural network according to the invention can also be interfaced to a computer which provides computerized automated lung texture analysis to supply radiographic input data in an objective and automated manner.",1995-10-31,"The title of the patent is method and system for differential diagnosis based on clinical and radiological information using artificial neural networks and its abstract is a method and system for computer-aided differential diagnosis of diseases, and in particular, computer-aided differential diagnosis using neural networks. a first embodiment of the neural network distinguishes between a plurality of interstitial lung diseases on the basis of inputted clinical parameters and radiographic information. a second embodiment distinguishes between malignant and benign mammographic cases based upon similar inputted clinical and radiographic information. the neural networks were first trained using a hypothetical data base made up of hypothetical cases for each of the interstitial lung diseases and for malignant and benign cases. the performance of the neural network was evaluated using receiver operating characteristics (roc) analysis. the decision performance of the neural network was compared to experienced radiologists and achieved a high performance comparable to that of the experienced radiologists. the neural network according to the invention can be made up of a single network or a plurality of successive or parallel networks. the neural network according to the invention can also be interfaced to a computer which provides computerized automated lung texture analysis to supply radiographic input data in an objective and automated manner. dated 1995-10-31"
5463717,inductively coupled neural network,"a data processing system based on the concept of a neural network includes a normalizing circuit and driving elements. each driving element has an output inductor magnetically coupled to an input inductor of the normalizing circuit. in the normalizing circuit, the input inductor is coupled to receive an input signal. the circuit also has a switching circuit responsive to the input signal and a switched inductor energized in response to the switched signal. the switched inductor comprising either a hooked or a spiral inductor.",1995-10-31,"The title of the patent is inductively coupled neural network and its abstract is a data processing system based on the concept of a neural network includes a normalizing circuit and driving elements. each driving element has an output inductor magnetically coupled to an input inductor of the normalizing circuit. in the normalizing circuit, the input inductor is coupled to receive an input signal. the circuit also has a switching circuit responsive to the input signal and a switched inductor energized in response to the switched signal. the switched inductor comprising either a hooked or a spiral inductor. dated 1995-10-31"
5465204,"heuristic control system employing expert system, neural network and training pattern generating and controlling system","a heuristic control system and method for use with a computer-aided design, capable of learning complicated control and achieving an optimizing task with a fewer number of iteration. the heuristic control system includes: rule-based system for choosing and evaluating a rule among a plurality of given rules; training system for choosing strongly a rule whose evaluation result is favorable based on a predetermined value which evaluates an evaluation result of the rule-based system, and for generating a learning pattern; and neural network for designing an optimized circuit based on a signal fed from the training system and for sending a resultant signal to the rule-based system for another iteration of heuristic control. the learning method includes the steps of: choosing and evaluating a rule among a plurality of given rules; choosing strongly a rule whose evaluation result is favorable based on a predetermined value; generating a learning pattern which brings a desirable result based on the evaluating step; designing an optimized circuit based on the learning pattern generated; and choosing and evaluating iteratively a rule among a plurality of given rules.",1995-11-07,"The title of the patent is heuristic control system employing expert system, neural network and training pattern generating and controlling system and its abstract is a heuristic control system and method for use with a computer-aided design, capable of learning complicated control and achieving an optimizing task with a fewer number of iteration. the heuristic control system includes: rule-based system for choosing and evaluating a rule among a plurality of given rules; training system for choosing strongly a rule whose evaluation result is favorable based on a predetermined value which evaluates an evaluation result of the rule-based system, and for generating a learning pattern; and neural network for designing an optimized circuit based on a signal fed from the training system and for sending a resultant signal to the rule-based system for another iteration of heuristic control. the learning method includes the steps of: choosing and evaluating a rule among a plurality of given rules; choosing strongly a rule whose evaluation result is favorable based on a predetermined value; generating a learning pattern which brings a desirable result based on the evaluating step; designing an optimized circuit based on the learning pattern generated; and choosing and evaluating iteratively a rule among a plurality of given rules. dated 1995-11-07"
5465308,pattern recognition system,"a method and apparatus under software control for pattern recognition utilizes a neural network implementation to recognize two dimensional input images which are sufficiently similar to a database of previously stored two dimensional images. images are first image processed and subjected to a fourier transform which yields a power spectrum. an in-class to out-of-class study is performed on a typical collection of images in order to determine the most discriminatory regions of the fourier transform. a feature vector consisting of the highest order (most discriminatory) magnitude information from the power spectrum of the fourier transform of the image is formed. feature vectors are input to a neural network having preferably two hidden layers, input dimensionality of the number of elements in the feature vector and output dimensionality of the number of data elements stored in the database. unique identifier numbers are preferably stored along with the feature vector. application of a query feature vector to the neural network will result in an output vector. the output vector is subjected to statistical analysis to determine if a sufficiently high confidence level exists to indicate that a successful identification has been made. where a successful identification has occurred, the unique identifier number may be displayed.",1995-11-07,"The title of the patent is pattern recognition system and its abstract is a method and apparatus under software control for pattern recognition utilizes a neural network implementation to recognize two dimensional input images which are sufficiently similar to a database of previously stored two dimensional images. images are first image processed and subjected to a fourier transform which yields a power spectrum. an in-class to out-of-class study is performed on a typical collection of images in order to determine the most discriminatory regions of the fourier transform. a feature vector consisting of the highest order (most discriminatory) magnitude information from the power spectrum of the fourier transform of the image is formed. feature vectors are input to a neural network having preferably two hidden layers, input dimensionality of the number of elements in the feature vector and output dimensionality of the number of data elements stored in the database. unique identifier numbers are preferably stored along with the feature vector. application of a query feature vector to the neural network will result in an output vector. the output vector is subjected to statistical analysis to determine if a sufficiently high confidence level exists to indicate that a successful identification has been made. where a successful identification has occurred, the unique identifier number may be displayed. dated 1995-11-07"
5465320,"method of automated learning, an apparatus therefor, and a system incorporating such an apparatus","in order to speed up and simplify automated learning of rules by a neural network making use of fuzzy logic, data from a system is analyzed by a teaching data creation means which groups the data into clusters and then selects a representative data item from each group for subsequent analysis. the selected data items are passed to a rule extraction means which investigates relationships between the data items, to derive rules, but eliminates rules which have only an insignificant effect on the system. the results are candidate rules which are stored in a first rule base. the candidate rules are then compared with rules in a second rule base to check for duplication and/or contradiction. only those rules which are not duplicated and not contradictory are stored in the second rule base. hence, when fuzzy inference is used to control the system on the basis of rules in the second rule base, only valid rules which provide a significant effect on the system are used.",1995-11-07,"The title of the patent is method of automated learning, an apparatus therefor, and a system incorporating such an apparatus and its abstract is in order to speed up and simplify automated learning of rules by a neural network making use of fuzzy logic, data from a system is analyzed by a teaching data creation means which groups the data into clusters and then selects a representative data item from each group for subsequent analysis. the selected data items are passed to a rule extraction means which investigates relationships between the data items, to derive rules, but eliminates rules which have only an insignificant effect on the system. the results are candidate rules which are stored in a first rule base. the candidate rules are then compared with rules in a second rule base to check for duplication and/or contradiction. only those rules which are not duplicated and not contradictory are stored in the second rule base. hence, when fuzzy inference is used to control the system on the basis of rules in the second rule base, only valid rules which provide a significant effect on the system are used. dated 1995-11-07"
5465620,micromechanical vibratory gyroscope sensor array,a micromechanical gyroscopic sensor array for detecting rotational movement includes a plurality of vibrational microgyroscopic sensing elements fixedly oriented in a common plane relative to one another such that an in-plane direction and an out-of-plane direction orthogonal thereto are defined. a mechanism for driving each microgyroscopic sensing element at a predetermined in-plane drive frequency is provided. each sensing element has a predefined out-of-plane resonant frequency selected such that a range of in-plane frequencies and/or out-of-plane resonant frequency differences exist among the sensing elements. a resonant gate transistor is associated with each sensing element for sensing out-of-plane motion and producing a sense signal representative thereof. signal processing then converts the sense signal to a signal representative of rotational movement experienced by the micromechanical gyroscopic sensor array. a learning function can be added to the sensor array by associating adjustable weights with each sensing element and employing a neural network having feedback control of the adjustable weights.,1995-11-14,The title of the patent is micromechanical vibratory gyroscope sensor array and its abstract is a micromechanical gyroscopic sensor array for detecting rotational movement includes a plurality of vibrational microgyroscopic sensing elements fixedly oriented in a common plane relative to one another such that an in-plane direction and an out-of-plane direction orthogonal thereto are defined. a mechanism for driving each microgyroscopic sensing element at a predetermined in-plane drive frequency is provided. each sensing element has a predefined out-of-plane resonant frequency selected such that a range of in-plane frequencies and/or out-of-plane resonant frequency differences exist among the sensing elements. a resonant gate transistor is associated with each sensing element for sensing out-of-plane motion and producing a sense signal representative thereof. signal processing then converts the sense signal to a signal representative of rotational movement experienced by the micromechanical gyroscopic sensor array. a learning function can be added to the sensor array by associating adjustable weights with each sensing element and employing a neural network having feedback control of the adjustable weights. dated 1995-11-14
5467427,memory capacity neural network,hopfield and bam neural network training or learning rules allowing memorization of a greater number of patterns. successive over-relaxation is used in the learning rules based on the training patterns and the output vectors. neural networks trained in this manner can better serve as the neural networks in a variety of pattern recognition and element correlation systems.,1995-11-14,The title of the patent is memory capacity neural network and its abstract is hopfield and bam neural network training or learning rules allowing memorization of a greater number of patterns. successive over-relaxation is used in the learning rules based on the training patterns and the output vectors. neural networks trained in this manner can better serve as the neural networks in a variety of pattern recognition and element correlation systems. dated 1995-11-14
5467428,artificial neural network method and architecture adaptive signal filtering,"an architecture and data processing method for a neural network that can approximate any mapping function between the input and output vectors without the use of hidden layers. the data processing is done at the sibling nodes (second row). it is based on the orthogonal expansion of the functions that map the input vector to the output vector. because the nodes of the second row are simply data processing stations, they remain passive during training. as a result the system is basically a single-layer linear network with a filter at its entrance. because of this it is free from the problems of local minima. the invention also includes a method that reduces the sum of the square of errors over all the output nodes to zero (0.000000) in fewer than ten cycles. this is done by initialization of the synaptic links with the coefficients of the orthogonal expansion. this feature makes it possible to design a computer chip which can perform the training process in real time. similarly, the ability to train in real time allows the system to retrain itself and improve its performance while executing its normal testing functions. because the second synaptic link values represent the frequency spectrum of the signal appearing on a given output node, by training the onn with all n sibling nodes and using only some of them in testing, we can create a low pass, a high pass or a band pass filter.",1995-11-14,"The title of the patent is artificial neural network method and architecture adaptive signal filtering and its abstract is an architecture and data processing method for a neural network that can approximate any mapping function between the input and output vectors without the use of hidden layers. the data processing is done at the sibling nodes (second row). it is based on the orthogonal expansion of the functions that map the input vector to the output vector. because the nodes of the second row are simply data processing stations, they remain passive during training. as a result the system is basically a single-layer linear network with a filter at its entrance. because of this it is free from the problems of local minima. the invention also includes a method that reduces the sum of the square of errors over all the output nodes to zero (0.000000) in fewer than ten cycles. this is done by initialization of the synaptic links with the coefficients of the orthogonal expansion. this feature makes it possible to design a computer chip which can perform the training process in real time. similarly, the ability to train in real time allows the system to retrain itself and improve its performance while executing its normal testing functions. because the second synaptic link values represent the frequency spectrum of the signal appearing on a given output node, by training the onn with all n sibling nodes and using only some of them in testing, we can create a low pass, a high pass or a band pass filter. dated 1995-11-14"
5467429,neural network circuit,"a neural network circuit including a number n of weight coefficients (w1-wn) corresponding to a number n of inputs, subtraction circuits for determining the difference between inputs and the weight coefficients in each input terminal, the result thereof being inputted into absolute value circuits, all calculation results of the absolute value circuits corresponding to the inputs and the weight coefficients being inputted into an addition circuit and accumulated, and this accumulation result determining the output valve. a threshold valve circuit determines the final output value, according to a step function pattern, a polygonal line pattern, or a sigmoid function pattern, depending on the object. in the case in which a neural network circuit is realized by means of digital circuits, the absolute value circuits can include simply ex-or logic (exclusive or) gates. furthermore, in the case in which the input terminals have two input paths and two weight coefficients corresponding to each input path, the neuron circuits form a recognition area having a flexible shape which is controlled by the weight coefficients.",1995-11-14,"The title of the patent is neural network circuit and its abstract is a neural network circuit including a number n of weight coefficients (w1-wn) corresponding to a number n of inputs, subtraction circuits for determining the difference between inputs and the weight coefficients in each input terminal, the result thereof being inputted into absolute value circuits, all calculation results of the absolute value circuits corresponding to the inputs and the weight coefficients being inputted into an addition circuit and accumulated, and this accumulation result determining the output valve. a threshold valve circuit determines the final output value, according to a step function pattern, a polygonal line pattern, or a sigmoid function pattern, depending on the object. in the case in which a neural network circuit is realized by means of digital circuits, the absolute value circuits can include simply ex-or logic (exclusive or) gates. furthermore, in the case in which the input terminals have two input paths and two weight coefficients corresponding to each input path, the neuron circuits form a recognition area having a flexible shape which is controlled by the weight coefficients. dated 1995-11-14"
5467883,active neural network control of wafer attributes in a plasma etch process,"the present invention is predicated upon the fact that an emission trace from a plasma glow used in fabricating integrated circuits contains information about phenoma which cause variations in the fabrication process such as age of the plasma reactor, densities of the wafers exposed to the plasma, chemistry of the plasma, and concentration of the remaining material. in accordance with the present invention, a method for using neural networks to determine plasma etch end-point times in an integrated circuit fabrication process is disclosed. the end-point time is based on in-situ monitoring of the optical emission trace. the back-propagation method is used to train the network. more generally, a neural network can be used to regulate control variables and materials in a manufacturing process to yield an output product with desired quality attributes. an identified process signature which reflects the relation between the quality attribute and the process may be used to train the neural network.",1995-11-21,"The title of the patent is active neural network control of wafer attributes in a plasma etch process and its abstract is the present invention is predicated upon the fact that an emission trace from a plasma glow used in fabricating integrated circuits contains information about phenoma which cause variations in the fabrication process such as age of the plasma reactor, densities of the wafers exposed to the plasma, chemistry of the plasma, and concentration of the remaining material. in accordance with the present invention, a method for using neural networks to determine plasma etch end-point times in an integrated circuit fabrication process is disclosed. the end-point time is based on in-situ monitoring of the optical emission trace. the back-propagation method is used to train the network. more generally, a neural network can be used to regulate control variables and materials in a manufacturing process to yield an output product with desired quality attributes. an identified process signature which reflects the relation between the quality attribute and the process may be used to train the neural network. dated 1995-11-21"
5471381,intelligent servomechanism controller,"a controller for a servomechanism (such as a computer disk drive) includes both a conventional dumb controller and a neural network controller working in conjunction with one another. in one embodiment, the neural network controller and dumb controller operate in a quasi-series configuration, with the neural network controller receiving and processing the output of the dumb controller to produce a servomechanism control signal. in another embodiment, the neural network controller and dumb controller operate in a quasi-parallel configuration, with the outputs of the neural network controller and dumb controller being combined to produce an intelligent servomechanism control signal. in yet another embodiment, the neural network controller and dumb controller operate in a quasi-series configuration during the recall phase, following the learning phase in which the neural network controller is trained to develop an indirect performance model of the serial combination of the dumb controller and servomechanism. in still another embodiment, the neural network controller and dumb controller operate in a quasi-series configuration during the recall phase, while a model neural network is used to develop a direct performance model during the learning phase of the serial combination of the dumb controller and servomechanism which is then used to train the neural network controller for its operation during the recall phase.",1995-11-28,"The title of the patent is intelligent servomechanism controller and its abstract is a controller for a servomechanism (such as a computer disk drive) includes both a conventional dumb controller and a neural network controller working in conjunction with one another. in one embodiment, the neural network controller and dumb controller operate in a quasi-series configuration, with the neural network controller receiving and processing the output of the dumb controller to produce a servomechanism control signal. in another embodiment, the neural network controller and dumb controller operate in a quasi-parallel configuration, with the outputs of the neural network controller and dumb controller being combined to produce an intelligent servomechanism control signal. in yet another embodiment, the neural network controller and dumb controller operate in a quasi-series configuration during the recall phase, following the learning phase in which the neural network controller is trained to develop an indirect performance model of the serial combination of the dumb controller and servomechanism. in still another embodiment, the neural network controller and dumb controller operate in a quasi-series configuration during the recall phase, while a model neural network is used to develop a direct performance model during the learning phase of the serial combination of the dumb controller and servomechanism which is then used to train the neural network controller for its operation during the recall phase. dated 1995-11-28"
5471557,speech recognition system utilizing a neural network,"a speech recognition system for recognizing the remote-controlling vocal commands of tv sets and vcrs comprises a microphone for receiving the speech pronounced by a user; a speech analyzer for analyzing the speech input via the microphone; circuitry for detecting a vocal section of the speech from the speech analyzer and performing a time-axis normalization and a binarization for the detected vocal section; and a multilayer neural network for receiving the binarization data from the aforementioned circuitry and then performing the learning, to thereby output the speech recognition result. accordingly, the present invention can enhance the recognition ratio of speech.",1995-11-28,"The title of the patent is speech recognition system utilizing a neural network and its abstract is a speech recognition system for recognizing the remote-controlling vocal commands of tv sets and vcrs comprises a microphone for receiving the speech pronounced by a user; a speech analyzer for analyzing the speech input via the microphone; circuitry for detecting a vocal section of the speech from the speech analyzer and performing a time-axis normalization and a binarization for the detected vocal section; and a multilayer neural network for receiving the binarization data from the aforementioned circuitry and then performing the learning, to thereby output the speech recognition result. accordingly, the present invention can enhance the recognition ratio of speech. dated 1995-11-28"
5471627,systolic array image processing system and method,"a systolic array of processing elements is connected to receive weight inputs and multiplexed data inputs for operation in two dimension convolution mode, or fully-connected neural network mode, or in cooperative, competitive neural network mode. feature vector or two-dimensional image data is retrieved from external data memory and is transformed via input look-up table to input data for the systolic array. the convoluted image or outputs from the systolic array are scaled and transformed via output look-up table for storage in the external data memory. the architecture of the system allows it to calculate convolutions of any size within the same physical systolic array, merely by adjusting the programs that control the data flow.",1995-11-28,"The title of the patent is systolic array image processing system and method and its abstract is a systolic array of processing elements is connected to receive weight inputs and multiplexed data inputs for operation in two dimension convolution mode, or fully-connected neural network mode, or in cooperative, competitive neural network mode. feature vector or two-dimensional image data is retrieved from external data memory and is transformed via input look-up table to input data for the systolic array. the convoluted image or outputs from the systolic array are scaled and transformed via output look-up table for storage in the external data memory. the architecture of the system allows it to calculate convolutions of any size within the same physical systolic array, merely by adjusting the programs that control the data flow. dated 1995-11-28"
5473532,intelligent machining system,"an intelligent machining system employs a neural network for calculating machining conditions on the basis of attribute data or a workpiece and a grinding machine. the system comprises a reference machining condition calculating unit, a neural network which receives the attribute data and provides corrections, and a correcting unit for correcting the reference machining conditions by using the corrections. corrections which cannot be determined by means of empirical expressions or theoretical expressions are determined appropriately by the neural network previously learned. the system determines corrections for the machining conditions on the basis of machining errors decided by a neural network. the system also detects time-series machining phenomena by sensors, processes the output detection signals of the sensors by a neural network to obtain machining circumstance data. the feed rate of the tool is controlled by a fuzzy inference on the basis of the machining circumstance data.",1995-12-05,"The title of the patent is intelligent machining system and its abstract is an intelligent machining system employs a neural network for calculating machining conditions on the basis of attribute data or a workpiece and a grinding machine. the system comprises a reference machining condition calculating unit, a neural network which receives the attribute data and provides corrections, and a correcting unit for correcting the reference machining conditions by using the corrections. corrections which cannot be determined by means of empirical expressions or theoretical expressions are determined appropriately by the neural network previously learned. the system determines corrections for the machining conditions on the basis of machining errors decided by a neural network. the system also detects time-series machining phenomena by sensors, processes the output detection signals of the sensors by a neural network to obtain machining circumstance data. the feed rate of the tool is controlled by a fuzzy inference on the basis of the machining circumstance data. dated 1995-12-05"
5473631,simultaneous transmission of data and audio signals by means of perceptual coding,""" a communication system for simultaneously transmitting data and audio signals via a conventional audio communications channel using perceptual coding techniques is disclosed. in a preferred embodiment, a first artificial neural network (nn) at an encoder monitors an audio channel to detect """"opportunities"""" to insert the data signal such that the inserted signals are masked by the audio signal, as defined by the """"perceptual entropy envelope"""" of the audio signal. under the control of the first nn a data signal containing, for example, an id or serial number, is encoded as one or more whitened direct sequence spread spectrum signals and/or a narrowband fsk data signal and transmitted at the time, frequency and level determined by the first nn such that the data signal is masked by the audio signal. the audio signal is combined with the spread spectrum and/or the fsk data signal(s) to form a composite signal, which is transmitted to one or more receiving locations via the audio channel. a decoder at each of the receiving locations comprises preprocessing circuitry, receiver sync circuitry and fsk decoder circuitry, as well as a second nn, which nn uses pattern and signature recognition techniques to perform block decoding, bit deinterleaving and acquisition confirm functions to recover the encoded id or serial number. """,1995-12-05,"The title of the patent is simultaneous transmission of data and audio signals by means of perceptual coding and its abstract is "" a communication system for simultaneously transmitting data and audio signals via a conventional audio communications channel using perceptual coding techniques is disclosed. in a preferred embodiment, a first artificial neural network (nn) at an encoder monitors an audio channel to detect """"opportunities"""" to insert the data signal such that the inserted signals are masked by the audio signal, as defined by the """"perceptual entropy envelope"""" of the audio signal. under the control of the first nn a data signal containing, for example, an id or serial number, is encoded as one or more whitened direct sequence spread spectrum signals and/or a narrowband fsk data signal and transmitted at the time, frequency and level determined by the first nn such that the data signal is masked by the audio signal. the audio signal is combined with the spread spectrum and/or the fsk data signal(s) to form a composite signal, which is transmitted to one or more receiving locations via the audio channel. a decoder at each of the receiving locations comprises preprocessing circuitry, receiver sync circuitry and fsk decoder circuitry, as well as a second nn, which nn uses pattern and signature recognition techniques to perform block decoding, bit deinterleaving and acquisition confirm functions to recover the encoded id or serial number. "" dated 1995-12-05"
5473730,high efficiency learning network,"nodal outputs are discretized to values of s2.sup.n where n is an integer and s is equal to +1 or -1. during forward propagation, this offers the advantage of forming a product of a nodal output and a weight using a simple shift operation rather than a multiply operation. replacing multiply operations with shift operations through out a neural network improves response times and permits building larger networks that have broader applicability. training is also improved by increasing the efficiency of backward propagation. the multiplications involved in backward propagation are reduced to shift operations by discretizing the errors associated with each node so that they are represented as 2.sup.n where n is an integer and s is equal to +1 or -1.",1995-12-05,"The title of the patent is high efficiency learning network and its abstract is nodal outputs are discretized to values of s2.sup.n where n is an integer and s is equal to +1 or -1. during forward propagation, this offers the advantage of forming a product of a nodal output and a weight using a simple shift operation rather than a multiply operation. replacing multiply operations with shift operations through out a neural network improves response times and permits building larger networks that have broader applicability. training is also improved by increasing the efficiency of backward propagation. the multiplications involved in backward propagation are reduced to shift operations by discretizing the errors associated with each node so that they are represented as 2.sup.n where n is an integer and s is equal to +1 or -1. dated 1995-12-05"
5475768,high accuracy optical character recognition using neural networks with centroid dithering,"pattern recognition, for instance optical character recognition, is achieved by training a neural network, scanning an image, segmenting the image to detect a pattern, preprocessing the detected pattern, and applying the preprocessed detected pattern to the trained neural network. the preprocessing includes determining a centroid of the pattern and centrally positioning the centroid in a frame containing the pattern. the training of the neural network includes randomly displacing template patterns within frames before applying the template patterns to the neural network.",1995-12-12,"The title of the patent is high accuracy optical character recognition using neural networks with centroid dithering and its abstract is pattern recognition, for instance optical character recognition, is achieved by training a neural network, scanning an image, segmenting the image to detect a pattern, preprocessing the detected pattern, and applying the preprocessed detected pattern to the trained neural network. the preprocessing includes determining a centroid of the pattern and centrally positioning the centroid in a frame containing the pattern. the training of the neural network includes randomly displacing template patterns within frames before applying the template patterns to the neural network. dated 1995-12-12"
5475794,semiconductor neural network and operating method thereof,"a semiconductor neural network includes a coupling matrix having coupling elements arranged in a matrix which couple with specific coupling strengths internal data input lines to internal data output lines. the internal data output lines are divided into groups. the neural network further comprises weighting addition circuits provided corresponding to the groups of the internal data output lines. a weighting addition circuit includes weighing elements for adding weights to signals on the internal data output lines in the corresponding group and outputting the weighted signals, and an addition circuit for outputting a total sum of the outputs of those weighting elements. the internal data output lines are arranged to form pairs and the addition circuit has a first input terminal for receiving one weighting element output of each of the pairs in common, a second input terminal for receiving the other weighting element output of each of the pairs in common, and a sense amplifier for differentially amplifying signals at the first and second input terminals. the neural network further includes a circuit for detecting a change time of an input signal, a circuit responsive to an input signal change for equalizing the first and second input terminals for a predetermined period, and a circuit for activating the sense amplifier after the equalization is completed. the information retention capability of each coupling element is set according to the weight of an associated weighting element. this neural network can provide multi-valued expression of coupling strength with less number of coupling elements.",1995-12-12,"The title of the patent is semiconductor neural network and operating method thereof and its abstract is a semiconductor neural network includes a coupling matrix having coupling elements arranged in a matrix which couple with specific coupling strengths internal data input lines to internal data output lines. the internal data output lines are divided into groups. the neural network further comprises weighting addition circuits provided corresponding to the groups of the internal data output lines. a weighting addition circuit includes weighing elements for adding weights to signals on the internal data output lines in the corresponding group and outputting the weighted signals, and an addition circuit for outputting a total sum of the outputs of those weighting elements. the internal data output lines are arranged to form pairs and the addition circuit has a first input terminal for receiving one weighting element output of each of the pairs in common, a second input terminal for receiving the other weighting element output of each of the pairs in common, and a sense amplifier for differentially amplifying signals at the first and second input terminals. the neural network further includes a circuit for detecting a change time of an input signal, a circuit responsive to an input signal change for equalizing the first and second input terminals for a predetermined period, and a circuit for activating the sense amplifier after the equalization is completed. the information retention capability of each coupling element is set according to the weight of an associated weighting element. this neural network can provide multi-valued expression of coupling strength with less number of coupling elements. dated 1995-12-12"
5475795,neural processing devices for handling real-valued inputs,"a device for use in a neural processing network. the network includes a memory having a plurality of storage locations each having stored a number representing a probability. each storage location is selectively addressible to cause the contents of the location to be read to an input of a comparator. a noise generator inputs to the comparator a random number representing noise. at an output of the comparator, an output signal appears having a first or second value depending on the values of the numbers received from the addressed storage location and the noise generator. the probability of the output signal has a given one of the first and second values determined by the number at the addressed location. the address inputs for the memory are derived from a real-to-spike frequency translator which has real values as its input vector.",1995-12-12,"The title of the patent is neural processing devices for handling real-valued inputs and its abstract is a device for use in a neural processing network. the network includes a memory having a plurality of storage locations each having stored a number representing a probability. each storage location is selectively addressible to cause the contents of the location to be read to an input of a comparator. a noise generator inputs to the comparator a random number representing noise. at an output of the comparator, an output signal appears having a first or second value depending on the values of the numbers received from the addressed storage location and the noise generator. the probability of the output signal has a given one of the first and second values determined by the number at the addressed location. the address inputs for the memory are derived from a real-to-spike frequency translator which has real values as its input vector. dated 1995-12-12"
5477444,"control system using an adaptive neural network for target and path optimization for a multivariable, nonlinear process","a control system having four major components: a target optimizer, a path optimizer, a neural network adaptation controller and a neural network. in the target optimizer, the controlled variables are optimized to provide the most economically desirable outputs, subject to operating constraints. various manipulated variable and disturbance values are provided for modeling purposes. the neural network receives as inputs a plurality of settings for each manipulated and disturbance variable. for target optimization all the neural network input values are set equal to produce a steady state controlled variable value. the entire process is repeated with differing manipulated variable values until an optimal solution develops. the resulting target controlled and manipulated variable values are provided to the path optimizer to allow the manipulated variables to be adjusted to obtain the target output. various manipulated variable values are developed to model moves from current to desired values. in this case trend indicating values of the manipulated and disturbance variables are provided to produce time varying values of the controlled variables. the process is repeated until an optimal path is obtained, at which time the manipulated variable values are applied to the actual process. on a periodic basis all of the disturbance, manipulated and controlled variables are sampled to find areas where the training of the neural network is sparse or where high dynamic conditions are indicated. these values are added to the set of values used to train the neural network.",1995-12-19,"The title of the patent is control system using an adaptive neural network for target and path optimization for a multivariable, nonlinear process and its abstract is a control system having four major components: a target optimizer, a path optimizer, a neural network adaptation controller and a neural network. in the target optimizer, the controlled variables are optimized to provide the most economically desirable outputs, subject to operating constraints. various manipulated variable and disturbance values are provided for modeling purposes. the neural network receives as inputs a plurality of settings for each manipulated and disturbance variable. for target optimization all the neural network input values are set equal to produce a steady state controlled variable value. the entire process is repeated with differing manipulated variable values until an optimal solution develops. the resulting target controlled and manipulated variable values are provided to the path optimizer to allow the manipulated variables to be adjusted to obtain the target output. various manipulated variable values are developed to model moves from current to desired values. in this case trend indicating values of the manipulated and disturbance variables are provided to produce time varying values of the controlled variables. the process is repeated until an optimal path is obtained, at which time the manipulated variable values are applied to the actual process. on a periodic basis all of the disturbance, manipulated and controlled variables are sampled to find areas where the training of the neural network is sparse or where high dynamic conditions are indicated. these values are added to the set of values used to train the neural network. dated 1995-12-19"
5479169,multiple neural network analog to digital converter for simultaneously processing multiple samples,an improved analog-to digital converter employs multiple sample and hold circuits to simultaneously supply multiple neural network a/d converters with samples of an analog input voltage so that the neural networks may simultaneously perform conversion of the different samples into a lower-order portion of the digital signals. a single fast a/d converter converts each sample into a higher-order portion of each digital signal.,1995-12-26,The title of the patent is multiple neural network analog to digital converter for simultaneously processing multiple samples and its abstract is an improved analog-to digital converter employs multiple sample and hold circuits to simultaneously supply multiple neural network a/d converters with samples of an analog input voltage so that the neural networks may simultaneously perform conversion of the different samples into a lower-order portion of the digital signals. a single fast a/d converter converts each sample into a higher-order portion of each digital signal. dated 1995-12-26
5479563,boundary extracting system from a sentence,"the present invention extracts boundaries from a sentence with no need for linguistic knowledge or complicated grammatical rules. upon extracting a clause/phrase boundary, words are classified according to part-of-speech numbers of words which form inputted sentence information. then, an input pattern representing part-of-speech numbers of a target word is checked to determine whether a clause/phrase boundary exists before or after the target word; a plurality of words before and after the target words is then applied to a neural network. among units in the output layer of the neural network, a unit having the output larger than a threshold is determined to refer to a clause/phrase boundary of the target word. upon extracting a subject-predicate boundary, words are classified in word number, and an input pattern corresponding to a plurality of words are applied to the neural network. the neural network comprises output units for a subject and a predicate, and a boundary is extracted by an inputted pattern which changes the output of these units.",1995-12-26,"The title of the patent is boundary extracting system from a sentence and its abstract is the present invention extracts boundaries from a sentence with no need for linguistic knowledge or complicated grammatical rules. upon extracting a clause/phrase boundary, words are classified according to part-of-speech numbers of words which form inputted sentence information. then, an input pattern representing part-of-speech numbers of a target word is checked to determine whether a clause/phrase boundary exists before or after the target word; a plurality of words before and after the target words is then applied to a neural network. among units in the output layer of the neural network, a unit having the output larger than a threshold is determined to refer to a clause/phrase boundary of the target word. upon extracting a subject-predicate boundary, words are classified in word number, and an input pattern corresponding to a plurality of words are applied to the neural network. the neural network comprises output units for a subject and a predicate, and a boundary is extracted by an inputted pattern which changes the output of these units. dated 1995-12-26"
5479569,intelligence information processing method,"an intelligence information processing system is composed of an associative memory and a serial processing-type computer. input pattern information is associated with the associative memory, and pattern recognition based on the computer evaluates an associative output. in accordance with this evaluation, an associative and restrictive condition is repeatedly added to the energy function of a neural network constituting the associative memory, thereby converging the associative output on a stable state of the energy. the converged associative output is verified with intelligence information stored in a computer memory. the associative and restrictive condition is again repeatedly added to the energy function in accordance with the verification so as to produce an output from the system.",1995-12-26,"The title of the patent is intelligence information processing method and its abstract is an intelligence information processing system is composed of an associative memory and a serial processing-type computer. input pattern information is associated with the associative memory, and pattern recognition based on the computer evaluates an associative output. in accordance with this evaluation, an associative and restrictive condition is repeatedly added to the energy function of a neural network constituting the associative memory, thereby converging the associative output on a stable state of the energy. the converged associative output is verified with intelligence information stored in a computer memory. the associative and restrictive condition is again repeatedly added to the energy function in accordance with the verification so as to produce an output from the system. dated 1995-12-26"
5479571,"neural node network and model, and method of teaching same","the present invention is a fully connected feed forward network that includes at least one hidden layer 16. the hidden layer 16 includes nodes 20 in which the output of the node is fed back to that node as an input with a unit delay produced by a delay device 24 occurring in the feedback path 22 (local feedback). each node within each layer also receives a delayed output (crosstalk) produced by a delay unit 36 from all the other nodes within the same layer 16. the node performs a transfer function operation based on the inputs from the previous layer and the delayed outputs. the network can be implemented as analog or digital or within a general purpose processor. two teaching methods can be used: (1) back propagation of weight calculation that includes the local feedback and the crosstalk or (2) more preferably a feed forward gradient decent which immediately follows the output computations and which also includes the local feedback and the crosstalk. subsequent to the gradient propagation, the weights can be normalized, thereby preventing convergence to a local optimum. education of the network can be incremental both on and off-line. an educated network is suitable for modeling and controlling dynamic nonlinear systems and time series systems and predicting the outputs as well as hidden states and parameters. the educated network can also be further educated during on-line processing.",1995-12-26,"The title of the patent is neural node network and model, and method of teaching same and its abstract is the present invention is a fully connected feed forward network that includes at least one hidden layer 16. the hidden layer 16 includes nodes 20 in which the output of the node is fed back to that node as an input with a unit delay produced by a delay device 24 occurring in the feedback path 22 (local feedback). each node within each layer also receives a delayed output (crosstalk) produced by a delay unit 36 from all the other nodes within the same layer 16. the node performs a transfer function operation based on the inputs from the previous layer and the delayed outputs. the network can be implemented as analog or digital or within a general purpose processor. two teaching methods can be used: (1) back propagation of weight calculation that includes the local feedback and the crosstalk or (2) more preferably a feed forward gradient decent which immediately follows the output computations and which also includes the local feedback and the crosstalk. subsequent to the gradient propagation, the weights can be normalized, thereby preventing convergence to a local optimum. education of the network can be incremental both on and off-line. an educated network is suitable for modeling and controlling dynamic nonlinear systems and time series systems and predicting the outputs as well as hidden states and parameters. the educated network can also be further educated during on-line processing. dated 1995-12-26"
5479572,artificial neural network (ann) classifier apparatus for selecting related computer routines and methods,"an artificial neural network (ann) classifier provides a series of outputs indicative of a series of classes to which input feature vectors are classified. the ann provides only one output for each input feature vector to partition said input into one class. the one output of the classifier is coupled to the interrupt input of an associated digital computer or cpu. upon receipt of the output, the cpu immediately interrupts a main program and executes an interrupt service routine which is triggered by the output of the classifier. in this manner, the cpu is immediately accessed in the interrupt mode by the transition of one of the output class processing elements when activated.",1995-12-26,"The title of the patent is artificial neural network (ann) classifier apparatus for selecting related computer routines and methods and its abstract is an artificial neural network (ann) classifier provides a series of outputs indicative of a series of classes to which input feature vectors are classified. the ann provides only one output for each input feature vector to partition said input into one class. the one output of the classifier is coupled to the interrupt input of an associated digital computer or cpu. upon receipt of the output, the cpu immediately interrupts a main program and executes an interrupt service routine which is triggered by the output of the classifier. in this manner, the cpu is immediately accessed in the interrupt mode by the transition of one of the output class processing elements when activated. dated 1995-12-26"
5479574,method and apparatus for adaptive classification,"a neural network and method for pipeline operation within a neural network which permits rapid classification of input vectors provided thereto is disclosed. in a training mode, a plurality of training input features are presented to the neural network and distances between the plurality of training features and a plurality of prototype weight values are concurrently computed. in response to an indication of a last training epoch count values for each of the prototype weight values are stored in a memory to thereby allow the neural network to operate in a probabilistic classification mode.",1995-12-26,"The title of the patent is method and apparatus for adaptive classification and its abstract is a neural network and method for pipeline operation within a neural network which permits rapid classification of input vectors provided thereto is disclosed. in a training mode, a plurality of training input features are presented to the neural network and distances between the plurality of training features and a plurality of prototype weight values are concurrently computed. in response to an indication of a last training epoch count values for each of the prototype weight values are stored in a memory to thereby allow the neural network to operate in a probabilistic classification mode. dated 1995-12-26"
5479575,self-organizing neural network for pattern classification,"a neural network includes a plurality of input nodes for receiving the respective elements of the input vector. a copy of all of the elements of the input vector is sent to the next level of nodes in the neural network denoted as intermediate nodes. the intermediate nodes each encode a separate template pattern. they compare the actual input pattern with the template and generate a signal indicative of the difference between the input pattern and the template pattern. each of the templates encoded in the intermediate nodes has a class associated with it. the difference calculated by the intermediate nodes is passed to an output node for each of the intermediate nodes at a given class. the output node then selects the minimum difference amongst the values sent from the intermediate nodes. this lowest difference for the class represented by the output node is then forwarded to a selector. the selector receives such values from each of the output nodes of all of the classes and then selects that to output value which is a minimum difference. the selector in turn, generates a signal indicative of the class of the intermediate node that sent the smallest difference value.",1995-12-26,"The title of the patent is self-organizing neural network for pattern classification and its abstract is a neural network includes a plurality of input nodes for receiving the respective elements of the input vector. a copy of all of the elements of the input vector is sent to the next level of nodes in the neural network denoted as intermediate nodes. the intermediate nodes each encode a separate template pattern. they compare the actual input pattern with the template and generate a signal indicative of the difference between the input pattern and the template pattern. each of the templates encoded in the intermediate nodes has a class associated with it. the difference calculated by the intermediate nodes is passed to an output node for each of the intermediate nodes at a given class. the output node then selects the minimum difference amongst the values sent from the intermediate nodes. this lowest difference for the class represented by the output node is then forwarded to a selector. the selector receives such values from each of the output nodes of all of the classes and then selects that to output value which is a minimum difference. the selector in turn, generates a signal indicative of the class of the intermediate node that sent the smallest difference value. dated 1995-12-26"
5479576,neural network learning system inferring an input-output relationship from a set of given input and output samples,"a neural network learning system in which an input-output relationship is inferred. the system includes a probability density part for determining a probability density on a sum space of an input space and an output space from a set of given input and output samples by learning, the probability density on the sum space being defined to have a parameter, and an inference part for inferring a probability density function based on the probability density from the probability density part, so that an input-output relationship of the samples is inferred from the probability density function having a parameter value determined by learning, the learning of the parameter being repeated until the value of a predefined parameter differential function using a prescribed maximum likelihood method is smaller than a prescribed reference value.",1995-12-26,"The title of the patent is neural network learning system inferring an input-output relationship from a set of given input and output samples and its abstract is a neural network learning system in which an input-output relationship is inferred. the system includes a probability density part for determining a probability density on a sum space of an input space and an output space from a set of given input and output samples by learning, the probability density on the sum space being defined to have a parameter, and an inference part for inferring a probability density function based on the probability density from the probability density part, so that an input-output relationship of the samples is inferred from the probability density function having a parameter value determined by learning, the learning of the parameter being repeated until the value of a predefined parameter differential function using a prescribed maximum likelihood method is smaller than a prescribed reference value. dated 1995-12-26"
5479579,cascaded vlsi neural network architecture for on-line learning,"high-speed, analog, fully-parallel and asynchronous building blocks are cascaded for larger sizes and enhanced resolution. a hardware-compatible algorithm permits hardware-in-the-loop learning despite limited weight resolution. a computation-intensive feature classification application has been demonstrated with this flexible hardware and new algorithm at high speed. this result indicates that these building block chips can be embedded as application-specific-coprocessors for solving real-world problems at extremely high data rates.",1995-12-26,"The title of the patent is cascaded vlsi neural network architecture for on-line learning and its abstract is high-speed, analog, fully-parallel and asynchronous building blocks are cascaded for larger sizes and enhanced resolution. a hardware-compatible algorithm permits hardware-in-the-loop learning despite limited weight resolution. a computation-intensive feature classification application has been demonstrated with this flexible hardware and new algorithm at high speed. this result indicates that these building block chips can be embedded as application-specific-coprocessors for solving real-world problems at extremely high data rates. dated 1995-12-26"
5481644,neural network speech recognition apparatus recognizing the frequency of successively input identical speech data sequences,"the speech recognition apparatus recognizes a frequency of successively input identical speech data sequences. the speech recognition apparatus includes a speech recognition non-layered neural network unit. speech data sequence is inputted as feature vectors from a feature extracting unit. the neural network performs speech recognition and determines whether the input speech data sequence matches at least one predetermined speech data sequence. the neural network generates a speech recognition signal when the input speech data sequence matches the at least one predetermined speech data sequence. a recognition signal detecting unit outputs a reset instruction signal each time the neural network generates the speech recognition signal. an internal state value setting unit resets the neural network unit to an initial state each time the recognition signal detecting unit outputs the reset instruction signal. since the neural network unit is reset each time the speech recognition signal is outputted, accurate detection can be achieved even when speech data sequence to be recognized is inputted successively.",1996-01-02,"The title of the patent is neural network speech recognition apparatus recognizing the frequency of successively input identical speech data sequences and its abstract is the speech recognition apparatus recognizes a frequency of successively input identical speech data sequences. the speech recognition apparatus includes a speech recognition non-layered neural network unit. speech data sequence is inputted as feature vectors from a feature extracting unit. the neural network performs speech recognition and determines whether the input speech data sequence matches at least one predetermined speech data sequence. the neural network generates a speech recognition signal when the input speech data sequence matches the at least one predetermined speech data sequence. a recognition signal detecting unit outputs a reset instruction signal each time the neural network generates the speech recognition signal. an internal state value setting unit resets the neural network unit to an initial state each time the recognition signal detecting unit outputs the reset instruction signal. since the neural network unit is reset each time the speech recognition signal is outputted, accurate detection can be achieved even when speech data sequence to be recognized is inputted successively. dated 1996-01-02"
5483170,integrated circuit fault testing implementing voltage supply rail pulsing and corresponding instantaneous current response analysis,"a method and apparatus for detecting faults in digital, analog, and hybrid integrated circuits is disclosed. a single test vector employing bias voltage on input used in conjunction with pulsing the power supply rails is used to allow detection of the various faults which may be present. the instantaneous rail current (i.sub.dd) is then employed for analysis of the circuit, preferably by neural network.",1996-01-09,"The title of the patent is integrated circuit fault testing implementing voltage supply rail pulsing and corresponding instantaneous current response analysis and its abstract is a method and apparatus for detecting faults in digital, analog, and hybrid integrated circuits is disclosed. a single test vector employing bias voltage on input used in conjunction with pulsing the power supply rails is used to allow detection of the various faults which may be present. the instantaneous rail current (i.sub.dd) is then employed for analysis of the circuit, preferably by neural network. dated 1996-01-09"
5483446,method and apparatus for estimating a vehicle maneuvering state and method and apparatus for controlling a vehicle running characteristic,"an apparatus for estimating a vehicle maneuvering state and controlling a vehicle running characteristic includes a controller having a fuzzy estimating function and a neural network function. the controller carries out frequency analysis on vehicle driving parameters such as vehicle speed, steering angle, opening degree of an accelerator, and longitudinal acceleration and lateral acceleration of a vehicle, to thereby determine a mean value and variance of each parameter. it implements fuzzy inference based on a traveling time ratio, an average speed, and an average lateral acceleration, which are obtained from vehicle speed and/or steering angle, to thereby calculate road traffic condition parameters, including a city area degree, a jammed road degree, and a mountainous road degree. according to the neural network function, the controller further determines an output parameter, indicative of the vehicle maneuvering state, by subjecting the mean value and variance of the vehicle driving parameters and the weighted total sum of the road traffic condition parameters to nonlinear conversion. then, it variably controls the operating characteristic of a vehicle-mounted apparatus such as a rear-wheel steering controlling unit in accordance with the output parameter, thereby variably controlling the vehicle running characteristic.",1996-01-09,"The title of the patent is method and apparatus for estimating a vehicle maneuvering state and method and apparatus for controlling a vehicle running characteristic and its abstract is an apparatus for estimating a vehicle maneuvering state and controlling a vehicle running characteristic includes a controller having a fuzzy estimating function and a neural network function. the controller carries out frequency analysis on vehicle driving parameters such as vehicle speed, steering angle, opening degree of an accelerator, and longitudinal acceleration and lateral acceleration of a vehicle, to thereby determine a mean value and variance of each parameter. it implements fuzzy inference based on a traveling time ratio, an average speed, and an average lateral acceleration, which are obtained from vehicle speed and/or steering angle, to thereby calculate road traffic condition parameters, including a city area degree, a jammed road degree, and a mountainous road degree. according to the neural network function, the controller further determines an output parameter, indicative of the vehicle maneuvering state, by subjecting the mean value and variance of the vehicle driving parameters and the weighted total sum of the road traffic condition parameters to nonlinear conversion. then, it variably controls the operating characteristic of a vehicle-mounted apparatus such as a rear-wheel steering controlling unit in accordance with the output parameter, thereby variably controlling the vehicle running characteristic. dated 1996-01-09"
5483620,learning machine synapse processor system apparatus,"a neural synapse processor apparatus having a neuron architecture for the synapse processing elements of the apparatus. the apparatus which we prefer will have a n neuron structure having synapse processing units that contain instruction and data storage units, receive instructions and data, and execute instructions. the n neuron structure should contain communicating adder trees, neuron activation function units, and an arrangement for communicating both instructions, data, and the outputs of neuron activation function units back to the input synapse processing units by means of the communicating adder trees. the apparatus can be structured as a bit-serial or word parallel system. the preferred structure contains n.sup.2 synapse processing units, each associated with a connection weight in the n neural network to be emulated, placed in the form of a n by n matrix that has been folded along the diagonal and made up of diagonal cells and general cells. diagonal cells, each utilizing a single synapse processing unit, are associated with the diagonal connection weights of the folded n by n connection weight matrix and general cells, each of which has two synapse processing units merged together, and which are associated with the symmetric connection weights of the folded n by n connection weight matrix. the back-propagation learning algorithm is first discussed followed by a presentation of the learning machine synapse processor architecture. an example implementation of the back-propagation learning algorithm is then presented. this is followed by a boltzmann like machine example and data parallel examples mapped onto the architecture",1996-01-09,"The title of the patent is learning machine synapse processor system apparatus and its abstract is a neural synapse processor apparatus having a neuron architecture for the synapse processing elements of the apparatus. the apparatus which we prefer will have a n neuron structure having synapse processing units that contain instruction and data storage units, receive instructions and data, and execute instructions. the n neuron structure should contain communicating adder trees, neuron activation function units, and an arrangement for communicating both instructions, data, and the outputs of neuron activation function units back to the input synapse processing units by means of the communicating adder trees. the apparatus can be structured as a bit-serial or word parallel system. the preferred structure contains n.sup.2 synapse processing units, each associated with a connection weight in the n neural network to be emulated, placed in the form of a n by n matrix that has been folded along the diagonal and made up of diagonal cells and general cells. diagonal cells, each utilizing a single synapse processing unit, are associated with the diagonal connection weights of the folded n by n connection weight matrix and general cells, each of which has two synapse processing units merged together, and which are associated with the symmetric connection weights of the folded n by n connection weight matrix. the back-propagation learning algorithm is first discussed followed by a presentation of the learning machine synapse processor architecture. an example implementation of the back-propagation learning algorithm is then presented. this is followed by a boltzmann like machine example and data parallel examples mapped onto the architecture dated 1996-01-09"
5483804,defrost control apparatus for refrigerator,"a defrost control apparatus for refrigerator includes a microcomputer which counts the number of opening/closing times of a door of a storage room for each of time zones within a day so as to set indexes for every time zones on the basis of the number of opening/closing times. the microcomputer also counts operating hours of a compressor and total elapsed hours, and determines a sudden phenomenon and a season. the microcomputer further selectively fetches the indexes, and generates a single index by joining a plurality of indexes so as to apply the same to a neural network included in a defrost signal generating unit. the neural network generates a defrost on/off signal on the basis of inputted data. in addition, if feature amounts are generated on the basis of the indexes by a feature detecting unit, the neural network generates the defrost on/off signal on the basis of the feature amounts.",1996-01-16,"The title of the patent is defrost control apparatus for refrigerator and its abstract is a defrost control apparatus for refrigerator includes a microcomputer which counts the number of opening/closing times of a door of a storage room for each of time zones within a day so as to set indexes for every time zones on the basis of the number of opening/closing times. the microcomputer also counts operating hours of a compressor and total elapsed hours, and determines a sudden phenomenon and a season. the microcomputer further selectively fetches the indexes, and generates a single index by joining a plurality of indexes so as to apply the same to a neural network included in a defrost signal generating unit. the neural network generates a defrost on/off signal on the basis of inputted data. in addition, if feature amounts are generated on the basis of the indexes by a feature detecting unit, the neural network generates the defrost on/off signal on the basis of the feature amounts. dated 1996-01-16"
5485545,control method using neural networks and a voltage/reactive-power controller for a power system using the control method,"a neural network apparatus and method for use in applications such as in a voltage/reactive-power controller in which a neuro control-object simulator and a neuro controller pre-learn so as to make input-output relations of the controller match the input-output relations of a control unit and so as to make input-output relations of the simulator match input-output relations of a control object. the controller re-learns so as to make the output of the simulator match an input corresponding to a desired output of the control object. after re-learning, the controller controls the control-object.",1996-01-16,"The title of the patent is control method using neural networks and a voltage/reactive-power controller for a power system using the control method and its abstract is a neural network apparatus and method for use in applications such as in a voltage/reactive-power controller in which a neuro control-object simulator and a neuro controller pre-learn so as to make input-output relations of the controller match the input-output relations of a control unit and so as to make input-output relations of the simulator match input-output relations of a control object. the controller re-learns so as to make the output of the simulator match an input corresponding to a desired output of the control object. after re-learning, the controller controls the control-object. dated 1996-01-16"
5485546,discrimination and testing methods and apparatus employing adaptively changing network behavior based on spatial and heterocellular modification rules,"an apparatus for categorizing objects employs a neural network having a plurality of cells each having memory for storing a state variable, and a plurality of synaptic junctions connecting cells of the network and having memory for storing a synaptic strength variable. a computer is used to modify the synaptic strength variable in accordance with a heterocellular synaptic modification rule. that modification rule includes both the passage of time and the values of the state variables of each cell and of those other cells having specific spatial locations with respect to the cell in three dimensional space.",1996-01-16,"The title of the patent is discrimination and testing methods and apparatus employing adaptively changing network behavior based on spatial and heterocellular modification rules and its abstract is an apparatus for categorizing objects employs a neural network having a plurality of cells each having memory for storing a state variable, and a plurality of synaptic junctions connecting cells of the network and having memory for storing a synaptic strength variable. a computer is used to modify the synaptic strength variable in accordance with a heterocellular synaptic modification rule. that modification rule includes both the passage of time and the values of the state variables of each cell and of those other cells having specific spatial locations with respect to the cell in three dimensional space. dated 1996-01-16"
5485548,signal processing apparatus using a hierarchical neural network,"in a neural network configured of neuron model cells, each neuron model cell is adapted to hold an input signal when a forward process is performed and to hold an error signal inputted when a learning process is performed. the signal processing apparatus is arranged to execute the forward process and the learning process in parallel.",1996-01-16,"The title of the patent is signal processing apparatus using a hierarchical neural network and its abstract is in a neural network configured of neuron model cells, each neuron model cell is adapted to hold an input signal when a forward process is performed and to hold an error signal inputted when a learning process is performed. the signal processing apparatus is arranged to execute the forward process and the learning process in parallel. dated 1996-01-16"
5485908,pattern recognition using artificial neural network for coin validation,"a coin validation system for determining if a coin moving along a coin rail is a valid coin, and if so, its denomination the system including a rail along which coins move, at least one optical sensor located along the rail to sense the presence or absence of a coin moving therealong, at least one magnetic sensor associated with each optical sensor located in the vicinity of the respective optical sensor, each of the magnetic sensors including an inductive element and a circuit for exciting the magnetic sensor to produce a field that is coupled to the coin moving past so that the coin and the inductive element have mutual inductance therebetween, the circuit ringing the magnetic sensor a predetermined number of times while the coin is adjacent to the magnetic sensor whereby the magnetic sensor generates a damped wave signal having characteristics representative of the physical and magnetic characteristics of the coin, a signal preprocessor operatively connected to the magnetic sensor for producing output responses representative of distinguishing characteristics of the coin, a feature extraction circuit for extracting from the output responses of the signal preprocessor signal portions representative of predetermined distinguishing characteristics of the coin, a circuit for producing a multi-dimensional representation of the extracted features and for comparing the multi-dimensional representation with the center of an established ellipsoidal cluster of selected coin denominations to determine the extent of the comparison therebetween and to be used to determine whether the coin is an acceptable coin or not, and an artificial neural network classifier circuit having connections to the preprocessor and to the comparator circuit, the neural network classifier circuit having an output which identifies the denomination of coins that are determined by the comparator circuit to be acceptable.",1996-01-23,"The title of the patent is pattern recognition using artificial neural network for coin validation and its abstract is a coin validation system for determining if a coin moving along a coin rail is a valid coin, and if so, its denomination the system including a rail along which coins move, at least one optical sensor located along the rail to sense the presence or absence of a coin moving therealong, at least one magnetic sensor associated with each optical sensor located in the vicinity of the respective optical sensor, each of the magnetic sensors including an inductive element and a circuit for exciting the magnetic sensor to produce a field that is coupled to the coin moving past so that the coin and the inductive element have mutual inductance therebetween, the circuit ringing the magnetic sensor a predetermined number of times while the coin is adjacent to the magnetic sensor whereby the magnetic sensor generates a damped wave signal having characteristics representative of the physical and magnetic characteristics of the coin, a signal preprocessor operatively connected to the magnetic sensor for producing output responses representative of distinguishing characteristics of the coin, a feature extraction circuit for extracting from the output responses of the signal preprocessor signal portions representative of predetermined distinguishing characteristics of the coin, a circuit for producing a multi-dimensional representation of the extracted features and for comparing the multi-dimensional representation with the center of an established ellipsoidal cluster of selected coin denominations to determine the extent of the comparison therebetween and to be used to determine whether the coin is an acceptable coin or not, and an artificial neural network classifier circuit having connections to the preprocessor and to the comparator circuit, the neural network classifier circuit having an output which identifies the denomination of coins that are determined by the comparator circuit to be acceptable. dated 1996-01-23"
5486996,parameterized neurocontrollers,""" a controller based on a neural network whose output is responsive to input signals that represent user or designer defined control system parameters which may include process parameters, control parameters and/or disturbance parameters. the neural network can be """"trained"""" to mimic an existing controller which may or not receive inputs of control system parameters. the trained neural network controller may have advantages of faster execution and reduced code size. the neural network can also be trained to result in a nonlinear controller that is more powerful than an existing controller. """,1996-01-23,"The title of the patent is parameterized neurocontrollers and its abstract is "" a controller based on a neural network whose output is responsive to input signals that represent user or designer defined control system parameters which may include process parameters, control parameters and/or disturbance parameters. the neural network can be """"trained"""" to mimic an existing controller which may or not receive inputs of control system parameters. the trained neural network controller may have advantages of faster execution and reduced code size. the neural network can also be trained to result in a nonlinear controller that is more powerful than an existing controller. "" dated 1996-01-23"
5486999,apparatus and method for categorizing health care utilization,"an apparatus and method for categorizing health care utilization provides an efficient aid in identifying patients who are seeking inappropriate care. the invention involves a computer system having a neural network responsive to several input variables to categorize the utilization characteristics of the patient. the input variables define selected characteristics of a patient. in one embodiment, a screening process identifies patients who are at high risk to an immediate threat to their health and eliminates those least likely to be seeking inappropriate care.",1996-01-23,"The title of the patent is apparatus and method for categorizing health care utilization and its abstract is an apparatus and method for categorizing health care utilization provides an efficient aid in identifying patients who are seeking inappropriate care. the invention involves a computer system having a neural network responsive to several input variables to categorize the utilization characteristics of the patient. the input variables define selected characteristics of a patient. in one embodiment, a screening process identifies patients who are at high risk to an immediate threat to their health and eliminates those least likely to be seeking inappropriate care. dated 1996-01-23"
5487026,"multiplying device, linear algebraic processor, neuromorphic processor, and optical processor","a linear algebraic processor is provided, for instance for performing vector matrix multiplications as part of a neural network. light emitting diode strips 4 are formed on one substrate 1 and are illuminated in accordance with the values of elements of an input vector. photodiode strips 8 are arranged orthogonally to the light emitting diode strips 4 on another substrate 2. a ferro-electric liquid crystal layer 3 is disposed between the strips 4 and 8 and provided with polarisers and electrodes 7 and 9 to permit the light attenuation properties of each matrix element between facing portions of the light emitting diode strips 4 and photodiode strips 8 to be varied and stored in a non-volatile way. the optical attenuation represents the value of the elements of the matrix and the outputs of the photodiode strips 8 represent the value of the elements of an output vector formed as the product of the input vector and the matrix. the optical attenuation of the ferro-electric liquid crystal can be incremented or decremented by applying suitable voltages to the electrodes so as to create electric fields across the liquid crystal matrix elements.",1996-01-23,"The title of the patent is multiplying device, linear algebraic processor, neuromorphic processor, and optical processor and its abstract is a linear algebraic processor is provided, for instance for performing vector matrix multiplications as part of a neural network. light emitting diode strips 4 are formed on one substrate 1 and are illuminated in accordance with the values of elements of an input vector. photodiode strips 8 are arranged orthogonally to the light emitting diode strips 4 on another substrate 2. a ferro-electric liquid crystal layer 3 is disposed between the strips 4 and 8 and provided with polarisers and electrodes 7 and 9 to permit the light attenuation properties of each matrix element between facing portions of the light emitting diode strips 4 and photodiode strips 8 to be varied and stored in a non-volatile way. the optical attenuation represents the value of the elements of the matrix and the outputs of the photodiode strips 8 represent the value of the elements of an output vector formed as the product of the input vector and the matrix. the optical attenuation of the ferro-electric liquid crystal can be incremented or decremented by applying suitable voltages to the electrodes so as to create electric fields across the liquid crystal matrix elements. dated 1996-01-23"
5487133,distance calculating neural network classifier chip and system,"an adaptive distance calculating neural network classifier chip accepts high dimensionality input pattern vectors with up to 256 5-bit elements per vector and compares the input vector with up to 1024 prototype vectors stored on-chip by calculating the distance between the input vector and each of the prototype vectors. the classifier further provides for identifying up to 64 classes to which the prototype vectors belong. if the distance between input and prototype vector is less than a programmable threshold distance, the prototype fires and the class to which it belongs is identified. if prototype vectors belonging to more than one class fire, a probabilistic model based on parzen windows may be invoked to resolve the classification by providing the relative probabilities of various class membership. the classifier chip is trainable by supplying appropriate training vectors and associated class membership. distance and probability parameters are generated during training and are stored for use in the classification mode. incremental training is also provided so that additional prototypes may be added to an existing base. in order to extend the classifier capacity, multichip operation is provided under the supervision of a system administrator controller/cpu.",1996-01-23,"The title of the patent is distance calculating neural network classifier chip and system and its abstract is an adaptive distance calculating neural network classifier chip accepts high dimensionality input pattern vectors with up to 256 5-bit elements per vector and compares the input vector with up to 1024 prototype vectors stored on-chip by calculating the distance between the input vector and each of the prototype vectors. the classifier further provides for identifying up to 64 classes to which the prototype vectors belong. if the distance between input and prototype vector is less than a programmable threshold distance, the prototype fires and the class to which it belongs is identified. if prototype vectors belonging to more than one class fire, a probabilistic model based on parzen windows may be invoked to resolve the classification by providing the relative probabilities of various class membership. the classifier chip is trainable by supplying appropriate training vectors and associated class membership. distance and probability parameters are generated during training and are stored for use in the classification mode. incremental training is also provided so that additional prototypes may be added to an existing base. in order to extend the classifier capacity, multichip operation is provided under the supervision of a system administrator controller/cpu. dated 1996-01-23"
5487153,neural network sequencer and interface apparatus,"the sequencer (14) is part of a computational system (10) which includes a computational circuit component, or processor node array (16); the sequencer, or controller component (14); and a boundary interface (34) between the computational circuit component (16) and the controller component (14). the controller component (14) provides three main functions in the system: (one) it sequences computations in a computational component (16), which includes an array of processor nodes (74, 76, 78, 80, 82, 84); (two) it provides i/o processing (20) from several disparate sources between the processor node array (16) and a host processor (12); and (three) it synchronizes data flow from a substantially asynchronous portion of the system (12) with a substantially synchronous data flow in the processor node array portion of the system (16).",1996-01-23,"The title of the patent is neural network sequencer and interface apparatus and its abstract is the sequencer (14) is part of a computational system (10) which includes a computational circuit component, or processor node array (16); the sequencer, or controller component (14); and a boundary interface (34) between the computational circuit component (16) and the controller component (14). the controller component (14) provides three main functions in the system: (one) it sequences computations in a computational component (16), which includes an array of processor nodes (74, 76, 78, 80, 82, 84); (two) it provides i/o processing (20) from several disparate sources between the processor node array (16) and a host processor (12); and (three) it synchronizes data flow from a substantially asynchronous portion of the system (12) with a substantially synchronous data flow in the processor node array portion of the system (16). dated 1996-01-23"
5488589,neural network based three dimensional ocean modeler,"a method is described for providing an estimate of the state of a moving tact in a three dimensional ocean. the method comprises the steps of providing a device for estimating the state of the contact, inputting into the device information about a location of an observer during a sequence of time, information from at least one sensor about the position of the contact relative to the observer during the time sequence, and a knowledge vector, transforming the inputted information into a series of three dimensional geographical grids, and analyzing the geographical grids to produce an estimate of the state of the contact with respect to the location of the observer. the device for providing the estimate of the state of the moving contact is a neurally inspired contact estimation device. the device includes a grid stimulation block for transforming the inputted information into the three dimensional geographical grids, a fusion block where grids corresponding to similar time intervals are combined or fused, a correlation block for providing constraints such as constant speed and heading and for producing a path likelihood vector, and an estimate block for providing the estimate of the state of the moving contact. the device further includes a controller for providing knowledge to the aforementioned blocks.",1996-01-30,"The title of the patent is neural network based three dimensional ocean modeler and its abstract is a method is described for providing an estimate of the state of a moving tact in a three dimensional ocean. the method comprises the steps of providing a device for estimating the state of the contact, inputting into the device information about a location of an observer during a sequence of time, information from at least one sensor about the position of the contact relative to the observer during the time sequence, and a knowledge vector, transforming the inputted information into a series of three dimensional geographical grids, and analyzing the geographical grids to produce an estimate of the state of the contact with respect to the location of the observer. the device for providing the estimate of the state of the moving contact is a neurally inspired contact estimation device. the device includes a grid stimulation block for transforming the inputted information into the three dimensional geographical grids, a fusion block where grids corresponding to similar time intervals are combined or fused, a correlation block for providing constraints such as constant speed and heading and for producing a path likelihood vector, and an estimate block for providing the estimate of the state of the moving contact. the device further includes a controller for providing knowledge to the aforementioned blocks. dated 1996-01-30"
5490062,real-time neural network earthquake profile predictor,"a neural network has been developed that uses first-arrival energy to predict the characteristics of impending earthquake seismograph signals. the propagation of ground motion energy through the earth is a highly nonlinear function. this is due to different forms of ground motion as well as to changes in the elastic properties of the media throughout the propagation path. the neural network is trained using seismogram data from earthquakes. presented with a previously unseen earthquake, the neural network produces a profile of the complete earthquake signal using data from the first seconds of the signal. this offers a significant advance in the real-time monitoring, warning, and subsequent hazard minimization of catastrophic ground motion.",1996-02-06,"The title of the patent is real-time neural network earthquake profile predictor and its abstract is a neural network has been developed that uses first-arrival energy to predict the characteristics of impending earthquake seismograph signals. the propagation of ground motion energy through the earth is a highly nonlinear function. this is due to different forms of ground motion as well as to changes in the elastic properties of the media throughout the propagation path. the neural network is trained using seismogram data from earthquakes. presented with a previously unseen earthquake, the neural network produces a profile of the complete earthquake signal using data from the first seconds of the signal. this offers a significant advance in the real-time monitoring, warning, and subsequent hazard minimization of catastrophic ground motion. dated 1996-02-06"
5490164,apparatus for classifying and storing connection coefficients for a multi-layer neural network,"an apparatus for storing weight coefficients for a multi-layer neural network is disclosed, wherein information representing connection coefficients is stored which represent strengths of connections between neurons of a multi-layer neural network constituted of an input layer, an intermediate layer, and an output layer, each of the input layer, the intermediate layer, and the output layer being composed of at least a single neuron. the apparatus for storing weight coefficients for a multi-layer neural network is provided with a device for classifying connections, which classifies the connections as having a high degree of connection or having a low degree of connection by comparing the connection coefficients with a predetermined threshold value, and which classifies the connections on the input side of each of the neurons as having a low degree of connection in cases where all of the connections on the output side of each said neuron have been classified as having a low degree of connection, and a storage device for storing information representing the connection coefficients of the connections, which have been classified as having a high degree of connection, and information concerning the corresponding table, which indicates whether the connections between the neurons have been classified as having a high degree of connection or having a low degree of connection by the device for classifying connections.",1996-02-06,"The title of the patent is apparatus for classifying and storing connection coefficients for a multi-layer neural network and its abstract is an apparatus for storing weight coefficients for a multi-layer neural network is disclosed, wherein information representing connection coefficients is stored which represent strengths of connections between neurons of a multi-layer neural network constituted of an input layer, an intermediate layer, and an output layer, each of the input layer, the intermediate layer, and the output layer being composed of at least a single neuron. the apparatus for storing weight coefficients for a multi-layer neural network is provided with a device for classifying connections, which classifies the connections as having a high degree of connection or having a low degree of connection by comparing the connection coefficients with a predetermined threshold value, and which classifies the connections on the input side of each of the neurons as having a low degree of connection in cases where all of the connections on the output side of each said neuron have been classified as having a low degree of connection, and a storage device for storing information representing the connection coefficients of the connections, which have been classified as having a high degree of connection, and information concerning the corresponding table, which indicates whether the connections between the neurons have been classified as having a high degree of connection or having a low degree of connection by the device for classifying connections. dated 1996-02-06"
5490236,method of assigning initial values of connection parameters to a multilayered neural network,"a generator in a back propagation type neural network having an input layer, an output layer and an intermediate layer coupled the input and output layers forms initial values for connection parameters. a first generator produces an initial value w10 of a weight coefficient of each connection parameter of the intermediate layer from in-class covariant data sw and inter-class co-variant data sb over data inputted to the input layer. the produced values are set into respective synapses of the intermediate layer as connection parameters.",1996-02-06,"The title of the patent is method of assigning initial values of connection parameters to a multilayered neural network and its abstract is a generator in a back propagation type neural network having an input layer, an output layer and an intermediate layer coupled the input and output layers forms initial values for connection parameters. a first generator produces an initial value w10 of a weight coefficient of each connection parameter of the intermediate layer from in-class covariant data sw and inter-class co-variant data sb over data inputted to the input layer. the produced values are set into respective synapses of the intermediate layer as connection parameters. dated 1996-02-06"
5491627,method and system for the detection of microcalcifications in digital mammograms,"a method and system for the detection of microcalcifications in digital mammograms. digital mammograms are obtained and regions-of-interest (rois) are selected therefrom which contain suspected microcalcifications, either individual or clustered microcalcifications. the suspect rois are background-trend corrected, followed by fourier transformation and power spectrum calculation to perform detection in the frequency domain. detection can also be carried out in the spatial domain by omitting the fourier transformation and power spectrum calculation. the roi is then scaled for input into a neural network trained to detect microcalcifications. the neural network outputs rois with detected microcalcifications. the method and system can also include normalizing the background-trend corrected rois and imputing the normalized roi to a shift-invariant neural network trained to detect microcalcifications. the output roi of the shift-invariant neural network is thresholded to remove additional false positive detections, and then the thresholded roi undergoes a cluster detection to detect clustered microcalcifications. feature extraction techniques can be applied to the remaining roi to remove additional false positive detections.",1996-02-13,"The title of the patent is method and system for the detection of microcalcifications in digital mammograms and its abstract is a method and system for the detection of microcalcifications in digital mammograms. digital mammograms are obtained and regions-of-interest (rois) are selected therefrom which contain suspected microcalcifications, either individual or clustered microcalcifications. the suspect rois are background-trend corrected, followed by fourier transformation and power spectrum calculation to perform detection in the frequency domain. detection can also be carried out in the spatial domain by omitting the fourier transformation and power spectrum calculation. the roi is then scaled for input into a neural network trained to detect microcalcifications. the neural network outputs rois with detected microcalcifications. the method and system can also include normalizing the background-trend corrected rois and imputing the normalized roi to a shift-invariant neural network trained to detect microcalcifications. the output roi of the shift-invariant neural network is thresholded to remove additional false positive detections, and then the thresholded roi undergoes a cluster detection to detect clustered microcalcifications. feature extraction techniques can be applied to the remaining roi to remove additional false positive detections. dated 1996-02-13"
5491650,high precision computing with charge domain devices and a pseudo-spectral method therefor,"the present invention discloses increased bit resolution of a charge coupled device (ccd)/charge injection device (cid) matrix vector multiplication (mvm) processor by storing each bit of each matrix element as a separate ccd charge packet. the bits of each input vector are separately multiplied by each bit of each matrix element in massive parallelism and the resulting products are combined appropriately to synthesize the correct product. in addition, such arrays are employed in a pseudo-spectral method of the invention, in which partial differential equations are solved by expressing each derivative analytically as matrices, and the state function is updated at each computation cycle by multiplying it by the matrices. the matrices are treated as synaptic arrays of a neutral network and the state function vector elements are treated as neurons. further, moving target detection is performed by driving the soliton equation with a vector of detector outputs. the neural architecture consists of two synaptic arrays corresponding to the two differential terms of the soliton equation and an adder connected to the output thereof and to the output of the detector array to drive the soliton equation.",1996-02-13,"The title of the patent is high precision computing with charge domain devices and a pseudo-spectral method therefor and its abstract is the present invention discloses increased bit resolution of a charge coupled device (ccd)/charge injection device (cid) matrix vector multiplication (mvm) processor by storing each bit of each matrix element as a separate ccd charge packet. the bits of each input vector are separately multiplied by each bit of each matrix element in massive parallelism and the resulting products are combined appropriately to synthesize the correct product. in addition, such arrays are employed in a pseudo-spectral method of the invention, in which partial differential equations are solved by expressing each derivative analytically as matrices, and the state function is updated at each computation cycle by multiplying it by the matrices. the matrices are treated as synaptic arrays of a neutral network and the state function vector elements are treated as neurons. further, moving target detection is performed by driving the soliton equation with a vector of detector outputs. the neural architecture consists of two synaptic arrays corresponding to the two differential terms of the soliton equation and an adder connected to the output thereof and to the output of the detector array to drive the soliton equation. dated 1996-02-13"
5491776,signal processing apparatus and learning method therefor,"mathematical calculation processing is executed at a high speed for an input signal of image information or the like. weight coefficients corresponding to a dct (discrete cosine transform) are set in a neural network nn of an n-layer structure including an input layer i1, (an intermediate layer h1) and an output layer o1. the input layer i1 is constituted from a number of input interfaces equal to the number of picture elements included in a block 12, and the output layer o1 is constituted from a number of neurons equal to the number of dct coefficients to be outputted. when input information of the block 12 is inputted to the neural network nn, dct coefficients a(0) to a(63) which are results of dct processing of the image information are outputted immediately. the weight coefficient is determined by learning using, as teacher data, dct coefficients obtained by actual mathematical calculation of the same image information as that inputted to the neural network nn by a dct transform section.",1996-02-13,"The title of the patent is signal processing apparatus and learning method therefor and its abstract is mathematical calculation processing is executed at a high speed for an input signal of image information or the like. weight coefficients corresponding to a dct (discrete cosine transform) are set in a neural network nn of an n-layer structure including an input layer i1, (an intermediate layer h1) and an output layer o1. the input layer i1 is constituted from a number of input interfaces equal to the number of picture elements included in a block 12, and the output layer o1 is constituted from a number of neurons equal to the number of dct coefficients to be outputted. when input information of the block 12 is inputted to the neural network nn, dct coefficients a(0) to a(63) which are results of dct processing of the image information are outputted immediately. the weight coefficient is determined by learning using, as teacher data, dct coefficients obtained by actual mathematical calculation of the same image information as that inputted to the neural network nn by a dct transform section. dated 1996-02-13"
5493631,stabilized adaptive neural network based control system,"a neural network based control system includes a nominal control system augmented by adaptive control such as a neuro-controller which generates additional compensating control signals based on differences between a model and actual system output. the nominal control system provides basic stability and performance, while the adaptive controller provides performance enhancement. the adaptive control can rely on any neural network that can encode a-priori knowledge and employs a high resolution pattern discrimination capability suitable for real-time changes. to prevent unbounded adaptation, the output of the adaptive controller is constrained by a limiter, thus ensuring safety of the overall system.",1996-02-20,"The title of the patent is stabilized adaptive neural network based control system and its abstract is a neural network based control system includes a nominal control system augmented by adaptive control such as a neuro-controller which generates additional compensating control signals based on differences between a model and actual system output. the nominal control system provides basic stability and performance, while the adaptive controller provides performance enhancement. the adaptive control can rely on any neural network that can encode a-priori knowledge and employs a high resolution pattern discrimination capability suitable for real-time changes. to prevent unbounded adaptation, the output of the adaptive controller is constrained by a limiter, thus ensuring safety of the overall system. dated 1996-02-20"
5493632,neural network employing a location addressable memory and method for operating the same,"a neural network provides faster learning speed and simplified overall structure through use of the concept of indirect association and a method for operating the same. the neural network is constructed as a clcam comprising an input-side single layer perceptron adapted to realize direct associations (x.sub.i, z.sub.1i) as linearly separable problems with respect to given inputs (x.sub.i) and first intermediate states (z.sub.1i) derived by the user, an output-side single layer perceptron adapted to realize direct associations (z.sub.2i,y.sub.i) as linearly separable problems with respect to given outputs (y.sub.i) and second intermediate states (z.sub.2i) derived by the user, and a location addressable memory adapted to connect said first intermediate states (z.sub.1i) with said second intermediate states (z.sub.2i). the neural network is also constructed as hylcam comprising a single layer perceptron adapted to realize direct associations (x.sub.i,z.sub.i) as linearly separable problems with respect to given inputs (x.sub.i) and intermediate states (z.sub.i) manually derived by the user, and a location addressable memory adapted to receive the intermediate states (z.sub.i) from the single layer perceptron as addresses and store given output data (y.sub.i) as desired output values, correspondingly to the addresses.",1996-02-20,"The title of the patent is neural network employing a location addressable memory and method for operating the same and its abstract is a neural network provides faster learning speed and simplified overall structure through use of the concept of indirect association and a method for operating the same. the neural network is constructed as a clcam comprising an input-side single layer perceptron adapted to realize direct associations (x.sub.i, z.sub.1i) as linearly separable problems with respect to given inputs (x.sub.i) and first intermediate states (z.sub.1i) derived by the user, an output-side single layer perceptron adapted to realize direct associations (z.sub.2i,y.sub.i) as linearly separable problems with respect to given outputs (y.sub.i) and second intermediate states (z.sub.2i) derived by the user, and a location addressable memory adapted to connect said first intermediate states (z.sub.1i) with said second intermediate states (z.sub.2i). the neural network is also constructed as hylcam comprising a single layer perceptron adapted to realize direct associations (x.sub.i,z.sub.i) as linearly separable problems with respect to given inputs (x.sub.i) and intermediate states (z.sub.i) manually derived by the user, and a location addressable memory adapted to receive the intermediate states (z.sub.i) from the single layer perceptron as addresses and store given output data (y.sub.i) as desired output values, correspondingly to the addresses. dated 1996-02-20"
5495415,method and system for detecting a misfire of a reciprocating internal combustion engine,"a trainable, pattern recognition-based method and system for detecting misfire in a reciprocating internal combustion engine having an engine cycle frequency in the frequency domain from crankshaft angular velocity. preferably, a pattern recognition system including a neural network is utilized. crankshaft angular position is sensed to develop an electrical signal which is a function of the crankshaft angular velocity. the electrical signal contains data which is sampled. the sampled data is transformed to an equivalent frequency domain spectrum including frequency components of the engine cycle frequency and harmonics thereof. a load signal such as mass airflow and an rpm signal are generated. the magnitudes and phases of the frequency components and the load and rpm signals are supplied to the neural network to distinguish between a true misfire and normal cyclic variability which characterizes the combustion process.",1996-02-27,"The title of the patent is method and system for detecting a misfire of a reciprocating internal combustion engine and its abstract is a trainable, pattern recognition-based method and system for detecting misfire in a reciprocating internal combustion engine having an engine cycle frequency in the frequency domain from crankshaft angular velocity. preferably, a pattern recognition system including a neural network is utilized. crankshaft angular position is sensed to develop an electrical signal which is a function of the crankshaft angular velocity. the electrical signal contains data which is sampled. the sampled data is transformed to an equivalent frequency domain spectrum including frequency components of the engine cycle frequency and harmonics thereof. a load signal such as mass airflow and an rpm signal are generated. the magnitudes and phases of the frequency components and the load and rpm signals are supplied to the neural network to distinguish between a true misfire and normal cyclic variability which characterizes the combustion process. dated 1996-02-27"
5495430,process time estimating apparatus,"a process time estimating apparatus is disclosed for estimating the process time for manufacturing an object such as a metal die. the apparatus includes a process time estimating section, a process occupancy time measurement section and a process program scheduling section. the process time estimating section includes a neural network device as an estimating device. an estimation input factor extracting section extracts input factors such as drawing information for an object to be manufactured. a storing section stores input factors for later neural network learning to improve the estimation capability of the system. the process occupancy time measurement section reads the process code and automatically measures the actual time involved in performing the process for a particular object being manufactured. a selecting section selects a measured process time for neural network learning. the process program scheduling section receives output information from the time measurement section and stores time estimates which are compared with actual process times for selecting a measured process time for further neural network learning.",1996-02-27,"The title of the patent is process time estimating apparatus and its abstract is a process time estimating apparatus is disclosed for estimating the process time for manufacturing an object such as a metal die. the apparatus includes a process time estimating section, a process occupancy time measurement section and a process program scheduling section. the process time estimating section includes a neural network device as an estimating device. an estimation input factor extracting section extracts input factors such as drawing information for an object to be manufactured. a storing section stores input factors for later neural network learning to improve the estimation capability of the system. the process occupancy time measurement section reads the process code and automatically measures the actual time involved in performing the process for a particular object being manufactured. a selecting section selects a measured process time for neural network learning. the process program scheduling section receives output information from the time measurement section and stores time estimates which are compared with actual process times for selecting a measured process time for further neural network learning. dated 1996-02-27"
5495542,binary to multi-level image restoration using neural network,"use is made of a neural network in order to restore a binary image to an original multi-level image, by way of example. using the neural network makes it possible to raise the accuracy of restoration and the speed of processing.",1996-02-27,"The title of the patent is binary to multi-level image restoration using neural network and its abstract is use is made of a neural network in order to restore a binary image to an original multi-level image, by way of example. using the neural network makes it possible to raise the accuracy of restoration and the speed of processing. dated 1996-02-27"
5497196,video camera having an adaptive automatic iris control circuit,"a video camera includes a lens, an iris and an iris driving circuit, an image sensor, a circuit for dividing the picture into a plurality of sub-areas for extracting the luminance of each sub-area according to the luminance signal provided from the image sensor as a luminance distribution signal, a circuit for generating a signal defining a target value of an iris driving signal, an adaptive circuit using an artificial neural network to which the luminance distribution signal is input for carrying out adaptive conversion so that the offset between a provided teacher signal and its own output is minimized, and a switch for selecting either the target value signal or the output of the adaptive circuit to provide the same as a teacher signal to the adaptive circuit.",1996-03-05,"The title of the patent is video camera having an adaptive automatic iris control circuit and its abstract is a video camera includes a lens, an iris and an iris driving circuit, an image sensor, a circuit for dividing the picture into a plurality of sub-areas for extracting the luminance of each sub-area according to the luminance signal provided from the image sensor as a luminance distribution signal, a circuit for generating a signal defining a target value of an iris driving signal, an adaptive circuit using an artificial neural network to which the luminance distribution signal is input for carrying out adaptive conversion so that the offset between a provided teacher signal and its own output is minimized, and a switch for selecting either the target value signal or the output of the adaptive circuit to provide the same as a teacher signal to the adaptive circuit. dated 1996-03-05"
5497253,multi-layer opto-electronic neural network,"a pattern recognition apparatus and a method for operating same. the apparatus includes a volume holographic medium (4) having a plurality of fourier-space volume holograms representing pattern templates stored within. the apparatus further includes a spatial light modulator (1) and a phase encoder (2). the phase encoder has an output optically coupled to the medium by a first fourier transform lens (3). the spatial light modulator spatially modulates a spatially uniform laser beam (7) in accordance with an unknown pattern. the two-dimensional phase encoder causes the spatially modulated laser beam to be spatially distributed prior to application to the medium. the apparatus also includes a detector (6, 11) having an input optically coupled by a second fourier transform lens (5) means to an angular spectrum of plane waves generated by the medium in response to the output of the spatial modulator, phase encoder, and first fourier lens. the detector detects plane waves that correspond to vector inner products generated within medium (4) in response to the unknown pattern. the apparatus further contains a means (12) for nonlinearly processing the output of detector (6, 11) and a means (13) by which the output of nonlinear processing means (12) may be temporarily stored.",1996-03-05,"The title of the patent is multi-layer opto-electronic neural network and its abstract is a pattern recognition apparatus and a method for operating same. the apparatus includes a volume holographic medium (4) having a plurality of fourier-space volume holograms representing pattern templates stored within. the apparatus further includes a spatial light modulator (1) and a phase encoder (2). the phase encoder has an output optically coupled to the medium by a first fourier transform lens (3). the spatial light modulator spatially modulates a spatially uniform laser beam (7) in accordance with an unknown pattern. the two-dimensional phase encoder causes the spatially modulated laser beam to be spatially distributed prior to application to the medium. the apparatus also includes a detector (6, 11) having an input optically coupled by a second fourier transform lens (5) means to an angular spectrum of plane waves generated by the medium in response to the output of the spatial modulator, phase encoder, and first fourier lens. the detector detects plane waves that correspond to vector inner products generated within medium (4) in response to the unknown pattern. the apparatus further contains a means (12) for nonlinearly processing the output of detector (6, 11) and a means (13) by which the output of nonlinear processing means (12) may be temporarily stored. dated 1996-03-05"
5497430,method and apparatus for image recognition using invariant feature signals,"a method of operating an image recognition system including providing a neural network including a plurality of input neurons, a plurality of output neurons and an interconnection weight matrix; providing a display including an indicator; initializing the indicator to an initialized state; obtaining an image of a structure; digitizing the image so as to obtain a plurality of input intensity cells and define an input object space; transforming the input object space to a feature vector including a set of n scale-, position- and rotation- invariant feature signals, where n is a positive integer not greater than the plurality of input neurons, by extracting the set of n scale-, position- and rotation-invariant feature signals from the input object space according to a set of relationships i.sub.k =.intg..sub..omega. .intg.i(x,y)h[k,i(x,y)]dxdy, where i.sub.k is the set of n scale-, position- and rotation-invariant feature signals, k is a series of counting numbers from 1 to n inclusive, (x,y) are the coordinates of a given cell of the plurality of input intensity cells, i(x,y) is a function of an intensity of the given cell of the plurality of input intensity cells, .omega. is an area of integration of input intensity cells, and h[k,i(x,y)] is a data dependent kernel transform from a set of orthogonal functions, of i(x,y) and k; transmitting the set of n scale-, position- and rotation- invariant feature signals to the plurality of input neurons; transforming the set of n scale-, position- and rotation- invariant feature signals at the plurality of input neurons to a set of structure recognition output signals at the plurality of output neurons according to a set of relationships defined at least in part by the interconnection weight matrix of the neural network; transforming the set of structure recognition output signals to a structure classification signal; and transmitting the structure classification signal to the display so as to perceptively alter the initialized state of the indicator and display the structure recognition signal for the structure.",1996-03-05,"The title of the patent is method and apparatus for image recognition using invariant feature signals and its abstract is a method of operating an image recognition system including providing a neural network including a plurality of input neurons, a plurality of output neurons and an interconnection weight matrix; providing a display including an indicator; initializing the indicator to an initialized state; obtaining an image of a structure; digitizing the image so as to obtain a plurality of input intensity cells and define an input object space; transforming the input object space to a feature vector including a set of n scale-, position- and rotation- invariant feature signals, where n is a positive integer not greater than the plurality of input neurons, by extracting the set of n scale-, position- and rotation-invariant feature signals from the input object space according to a set of relationships i.sub.k =.intg..sub..omega. .intg.i(x,y)h[k,i(x,y)]dxdy, where i.sub.k is the set of n scale-, position- and rotation-invariant feature signals, k is a series of counting numbers from 1 to n inclusive, (x,y) are the coordinates of a given cell of the plurality of input intensity cells, i(x,y) is a function of an intensity of the given cell of the plurality of input intensity cells, .omega. is an area of integration of input intensity cells, and h[k,i(x,y)] is a data dependent kernel transform from a set of orthogonal functions, of i(x,y) and k; transmitting the set of n scale-, position- and rotation- invariant feature signals to the plurality of input neurons; transforming the set of n scale-, position- and rotation- invariant feature signals at the plurality of input neurons to a set of structure recognition output signals at the plurality of output neurons according to a set of relationships defined at least in part by the interconnection weight matrix of the neural network; transforming the set of structure recognition output signals to a structure classification signal; and transmitting the structure classification signal to the display so as to perceptively alter the initialized state of the indicator and display the structure recognition signal for the structure. dated 1996-03-05"
5498943,feedback control device,"a feedback control device at a certain control time, predicts a feedback quantity at the next control time on the basis of a feedback quantity fed back from a controlled object and then performs a control operation based on the predicted feedback quantity. the feedback control device includes a predictive control unit for producing such an actuating signal as to decrease a deviation between the predicted feedback quantity and a desired value. the actuating signal is used to control the controlled object. this eliminates the detrimental effects of time delays associated with the controlled object or the control device, thus ensuring good control. the use of the predictive control unit permits the controlled quantity from the controlled object to converge to the desired value monotonically and quickly. the predictive control unit may use a layered neural network having an input layer supplied with a feedback quantity and an input value corresponding to an actuating signal, and an output layer outputting a predicted feedback quantity. the deviation is back propagated from the output layer to the input layer by a relaxation algorithm in order to update the input value. the updated input value is applied to an actuating unit as an actuating signal, thereby controlling the controlled object.",1996-03-12,"The title of the patent is feedback control device and its abstract is a feedback control device at a certain control time, predicts a feedback quantity at the next control time on the basis of a feedback quantity fed back from a controlled object and then performs a control operation based on the predicted feedback quantity. the feedback control device includes a predictive control unit for producing such an actuating signal as to decrease a deviation between the predicted feedback quantity and a desired value. the actuating signal is used to control the controlled object. this eliminates the detrimental effects of time delays associated with the controlled object or the control device, thus ensuring good control. the use of the predictive control unit permits the controlled quantity from the controlled object to converge to the desired value monotonically and quickly. the predictive control unit may use a layered neural network having an input layer supplied with a feedback quantity and an input value corresponding to an actuating signal, and an output layer outputting a predicted feedback quantity. the deviation is back propagated from the output layer to the input layer by a relaxation algorithm in order to update the input value. the updated input value is applied to an actuating unit as an actuating signal, thereby controlling the controlled object. dated 1996-03-12"
5500905,pattern recognition neural network with saccade-like operation,"a multi-layered pattern recognition neural network (30) is disclosed that comprises an input layer (50) that is operable to be mapped onto an input space that includes a scan window (32). two hidden layers (54) and (58) map the input space to an output layer (34). the hidden layers utilize a local receptor field architecture and store representations of objects within the scan window (32) for mapping into one of a plurality of output nodes. further, the output layer (34) is also operable to store representations of desired distances between the center of the scan window (32) and the next adjacent object thereto and also the distance between the center of the scan window (32) and the center of the current object. a scanning system can then utilize the information regarding the distance to the next adjacent object, which is stored in an output vector (40) to incrementally jump to the center of the next adjacent character rather than scan the entire distance therebetween. this is referred to as a saccade operation. once the scan window ( 32) is disposed over the next object, a corrective saccade can be performed by utilizing the information output by the neural network (30) relating to the distance between the center of the scan window (32) and the current character. this information is output as an output vector (38) from the neural network (30).",1996-03-19,"The title of the patent is pattern recognition neural network with saccade-like operation and its abstract is a multi-layered pattern recognition neural network (30) is disclosed that comprises an input layer (50) that is operable to be mapped onto an input space that includes a scan window (32). two hidden layers (54) and (58) map the input space to an output layer (34). the hidden layers utilize a local receptor field architecture and store representations of objects within the scan window (32) for mapping into one of a plurality of output nodes. further, the output layer (34) is also operable to store representations of desired distances between the center of the scan window (32) and the next adjacent object thereto and also the distance between the center of the scan window (32) and the center of the current object. a scanning system can then utilize the information regarding the distance to the next adjacent object, which is stored in an output vector (40) to incrementally jump to the center of the next adjacent character rather than scan the entire distance therebetween. this is referred to as a saccade operation. once the scan window ( 32) is disposed over the next object, a corrective saccade can be performed by utilizing the information output by the neural network (30) relating to the distance between the center of the scan window (32) and the current character. this information is output as an output vector (38) from the neural network (30). dated 1996-03-19"
5502688,feedforward neural network system for the detection and characterization of sonar signals with characteristic spectrogram textures,the present invention provides a method and system for characterizing the sounds of ocean captured by passive sonar listening devices. the present invention accomplishes this by first generating a spectrogram from the received sonar signal. the spectrogram is characterized in terms of textural features and signal processing parameters. the textural features and signal processing parameters are fed into a neural network ensemble that has been trained to favor specific features and/or parameters. the trained neural network ensemble classifies the signal as either type-i or clutter.,1996-03-26,The title of the patent is feedforward neural network system for the detection and characterization of sonar signals with characteristic spectrogram textures and its abstract is the present invention provides a method and system for characterizing the sounds of ocean captured by passive sonar listening devices. the present invention accomplishes this by first generating a spectrogram from the received sonar signal. the spectrogram is characterized in terms of textural features and signal processing parameters. the textural features and signal processing parameters are fed into a neural network ensemble that has been trained to favor specific features and/or parameters. the trained neural network ensemble classifies the signal as either type-i or clutter. dated 1996-03-26
5502773,method and apparatus for automated processing of dna sequence data,"a method and apparatus for the processing of dna sequence image data in real time is implemented using a series of linked neural network processors. as raw image data is received from a sequencing machine, it is buffered and then separately transformed in real time in the processors to enhance the signals indicative of the unknown dna sequence. a fourth processor receives the transformed data and determines and reports the sequence indicating events.",1996-03-26,"The title of the patent is method and apparatus for automated processing of dna sequence data and its abstract is a method and apparatus for the processing of dna sequence image data in real time is implemented using a series of linked neural network processors. as raw image data is received from a sequencing machine, it is buffered and then separately transformed in real time in the processors to enhance the signals indicative of the unknown dna sequence. a fourth processor receives the transformed data and determines and reports the sequence indicating events. dated 1996-03-26"
5502775,method and apparatus for adjusting read-out and processing conditions for radiation images,"a first image signal representing a radiation image is detected from a stimulable phosphor sheet, on which the radiation image has been stored, its probability density function is created, and a characteristic value of the probability density function is detected. values of differences between the characteristic value and values of the first image signal in the probability density function are calculated. information representing the values of the differences is fed into a neural network, and information representing temporary read-out conditions and/or temporary image processing conditions is fed out from the neural network. the temporary read-out conditions and/or the temporary image processing conditions are corrected in accordance with the level of the characteristic value. read-out conditions, under which a second image signal representing the radiation image is to be obtained from the stimulable phosphor sheet, and/or image processing conditions, under which the second image signal having been obtained is to be image processed, are thereby adjusted.",1996-03-26,"The title of the patent is method and apparatus for adjusting read-out and processing conditions for radiation images and its abstract is a first image signal representing a radiation image is detected from a stimulable phosphor sheet, on which the radiation image has been stored, its probability density function is created, and a characteristic value of the probability density function is detected. values of differences between the characteristic value and values of the first image signal in the probability density function are calculated. information representing the values of the differences is fed into a neural network, and information representing temporary read-out conditions and/or temporary image processing conditions is fed out from the neural network. the temporary read-out conditions and/or the temporary image processing conditions are corrected in accordance with the level of the characteristic value. read-out conditions, under which a second image signal representing the radiation image is to be obtained from the stimulable phosphor sheet, and/or image processing conditions, under which the second image signal having been obtained is to be image processed, are thereby adjusted. dated 1996-03-26"
5504675,method and apparatus for automatic selection and presentation of sales promotion programs,"a sales promotion program is dynamically selected from a plurality of programs for presentation in a program presentation unit by a neural network that makes its selection based on first detecting if a person is in the area immediately around the program presentation unit, then either selecting a general attract loop sales promotion program with the trained neural network using a set of predetermined system criteria if no person is detected in the immediate area or selecting a specific loop sales promotion program if at least one person is detected in the immediate area. the neural network is trained by selecting general attract loop programs that are run and then collecting data indicative of the number of persons responding to the general attract loop and also by selecting specific loop programs that are run if a person is in the immediate area and then collecting data indicative of the responses to the specific loop programs. the collected data thereby represents the success of the various sales programs in attracting and holding the attention of persons. the collected data is provided to the neural network in any one of a plurality of training schemes typical for neural networks, after which the trained neural network is provided with current, real-time selection data such that the trained network can select the most appropriate sales promotion program for running. the network can be retrained at regular intervals or in response to sales data or changes in the collected data.",1996-04-02,"The title of the patent is method and apparatus for automatic selection and presentation of sales promotion programs and its abstract is a sales promotion program is dynamically selected from a plurality of programs for presentation in a program presentation unit by a neural network that makes its selection based on first detecting if a person is in the area immediately around the program presentation unit, then either selecting a general attract loop sales promotion program with the trained neural network using a set of predetermined system criteria if no person is detected in the immediate area or selecting a specific loop sales promotion program if at least one person is detected in the immediate area. the neural network is trained by selecting general attract loop programs that are run and then collecting data indicative of the number of persons responding to the general attract loop and also by selecting specific loop programs that are run if a person is in the immediate area and then collecting data indicative of the responses to the specific loop programs. the collected data thereby represents the success of the various sales programs in attracting and holding the attention of persons. the collected data is provided to the neural network in any one of a plurality of training schemes typical for neural networks, after which the trained neural network is provided with current, real-time selection data such that the trained network can select the most appropriate sales promotion program for running. the network can be retrained at regular intervals or in response to sales data or changes in the collected data. dated 1996-04-02"
5504780,adaptive equalizer using self-learning neural network,"a channel equalizer is formed using a self-learning neural network. during a training period, the neural network is taught the channel response function. the network is then used to equalize distortions introduced into signals by the channel. the neural network may be a boltzmann machine type of neural network comprising neurons arranged in an input layer, a hidden layer, and an output layer. the neurons are interconnected by bidirectional symmetric weighted synapses. each neuron is preferably implemented by an analog integrated circuit. direct communication between the input and output layers helps in faster channel acquisition. the scheme can very easily be extended to multilevel and multisymbol modulation schemes such as qam and psk.",1996-04-02,"The title of the patent is adaptive equalizer using self-learning neural network and its abstract is a channel equalizer is formed using a self-learning neural network. during a training period, the neural network is taught the channel response function. the network is then used to equalize distortions introduced into signals by the channel. the neural network may be a boltzmann machine type of neural network comprising neurons arranged in an input layer, a hidden layer, and an output layer. the neurons are interconnected by bidirectional symmetric weighted synapses. each neuron is preferably implemented by an analog integrated circuit. direct communication between the input and output layers helps in faster channel acquisition. the scheme can very easily be extended to multilevel and multisymbol modulation schemes such as qam and psk. dated 1996-04-02"
5504839,processor and processing element for use in a neural network,a synaptic processor for use in a neural network produces a result signal in response to an input signal. the synaptic processor initializes an input expectation and receives an input signal. the synaptic processor determines a net modification to the the input expectation. the net modification to the input expectation has an increase term and a decrease term. the increase term is determined as a function of the input signal. the decrease term is independent of the magnitude of the input signal and is a function of a decay constant. the synaptic processor produces a result signal in response to the input expectation.,1996-04-02,The title of the patent is processor and processing element for use in a neural network and its abstract is a synaptic processor for use in a neural network produces a result signal in response to an input signal. the synaptic processor initializes an input expectation and receives an input signal. the synaptic processor determines a net modification to the the input expectation. the net modification to the input expectation has an increase term and a decrease term. the increase term is determined as a function of the input signal. the decrease term is independent of the magnitude of the input signal and is a function of a decay constant. the synaptic processor produces a result signal in response to the input expectation. dated 1996-04-02
5504841,method of processing signals within a neural network to position arobot,"a signal processing method for efficiently searching an optimum solution in a neural network by including a term of a nonlinear resistance in an equation of motion and changing such nonlinear resistance periodically. according to the method, the range of absolute values of connection weights between units in the neural network is limited by the equation of motion, hence preventing a prolonged search time that may otherwise be caused by excessive extension of the search scope beyond the requisite. a plurality of patterns are previously embedded or stored in the neural network and, upon input of a predetermined key pattern, the nonlinear resistance is changed periodically to recall a pattern similar to the key pattern, whereby any desired pattern can be searched or retrieved with rapidity and facility out of the complicated patterns. a process of calculating the next position of an articulated robot corresponding to an optimum solution is repeated while periodically changing a nonlinear resistance included in another equation of the positional energy of the robot, thereby acquiring the data of the robot path up to a desired goal.",1996-04-02,"The title of the patent is method of processing signals within a neural network to position arobot and its abstract is a signal processing method for efficiently searching an optimum solution in a neural network by including a term of a nonlinear resistance in an equation of motion and changing such nonlinear resistance periodically. according to the method, the range of absolute values of connection weights between units in the neural network is limited by the equation of motion, hence preventing a prolonged search time that may otherwise be caused by excessive extension of the search scope beyond the requisite. a plurality of patterns are previously embedded or stored in the neural network and, upon input of a predetermined key pattern, the nonlinear resistance is changed periodically to recall a pattern similar to the key pattern, whereby any desired pattern can be searched or retrieved with rapidity and facility out of the complicated patterns. a process of calculating the next position of an articulated robot corresponding to an optimum solution is repeated while periodically changing a nonlinear resistance included in another equation of the positional energy of the robot, thereby acquiring the data of the robot path up to a desired goal. dated 1996-04-02"
5504884,information retrieval system,"an information retrieval system allowing the user to build a database or to retrieve data therefrom based on vague retrieval-designating data without becoming aware of the database structure. the system comprises a neural network, a memory, an interface part and a crt. the neural network stores data designating electronic still pictures contained in the memory. when data for designating retrieval are input, the interface part groups the data into such categories as the place where the desired picture was taken and the date on which it was taken, and supplies the neural network with the categorized input data. in turn, the neural network outputs by association the data corresponding to the input data. given the data from the network, the memory outputs the relevant electronic still picture to the crt for display.",1996-04-02,"The title of the patent is information retrieval system and its abstract is an information retrieval system allowing the user to build a database or to retrieve data therefrom based on vague retrieval-designating data without becoming aware of the database structure. the system comprises a neural network, a memory, an interface part and a crt. the neural network stores data designating electronic still pictures contained in the memory. when data for designating retrieval are input, the interface part groups the data into such categories as the place where the desired picture was taken and the date on which it was taken, and supplies the neural network with the categorized input data. in turn, the neural network outputs by association the data corresponding to the input data. given the data from the network, the memory outputs the relevant electronic still picture to the crt for display. dated 1996-04-02"
5506696,color image reproduction system having color analysis function performed with a neural network system,"a color separation section for converting colorimetric values into color separation values adopts a neural network. when a color image output device for outputting a color image on the basis of color separation value signals is used, the color image output device to be used outputs a standard color sample having known color separation values. the color sample is colorimetrically measured by a colorimetry device to obtain colorimetric values of the color sample. the neural network executes learning to have conversion characteristics for converting the colorimetric values into corresponding color separation values. an object to be reproduced which has a required color is colorimetrically measured by the colorimetry device. colorimetric values obtained by the colorimetry processing are converted into color separation values using the neural network of the color separation section. the color image output device outputs a target color based on the converted color separation values.",1996-04-09,"The title of the patent is color image reproduction system having color analysis function performed with a neural network system and its abstract is a color separation section for converting colorimetric values into color separation values adopts a neural network. when a color image output device for outputting a color image on the basis of color separation value signals is used, the color image output device to be used outputs a standard color sample having known color separation values. the color sample is colorimetrically measured by a colorimetry device to obtain colorimetric values of the color sample. the neural network executes learning to have conversion characteristics for converting the colorimetric values into corresponding color separation values. an object to be reproduced which has a required color is colorimetrically measured by the colorimetry device. colorimetric values obtained by the colorimetry processing are converted into color separation values using the neural network of the color separation section. the color image output device outputs a target color based on the converted color separation values. dated 1996-04-09"
5508203,apparatus and method for radio frequency spectroscopy using spectral analysis,"a source of high frequency electromagnetic radiation is coupled to a specimen containing a target chemical whose presence and/or concentration is to be ascertained. preferably the source radiation includes a plurality of high frequency spectra, at least one of which encourages energy transfer with the chemical of interest. the source radiation is coupled to the specimen via a probe pair that is also used to access a return signal representing an interaction between the source signal and the specimen. the return signal is processed to yield a spectral signature correlating to the target chemical and/or its concentration. preferably signal processing compares frequency spectra within the source signal to spectra within a sampled return signal. the sampled return signal is a signal obtained by sampling a return signal at the probe pair and by sampling the response to the source signal of a circuit that electrically approximates the specimen. the amplitude and/or phase difference between the sampled return signal and the source signal provides recognition of the spectral signature. target chemical concentration data may be obtained from the signature and can be displayed in a number of ways. operation of the signal processor may be optimized using a neural network. in the preferred embodiment, the specimen is a human finger that is pressed against the probe pair, and the chemical is glucose. the invention thus permits a lay user to non-invasively determine his or her glucose level.",1996-04-16,"The title of the patent is apparatus and method for radio frequency spectroscopy using spectral analysis and its abstract is a source of high frequency electromagnetic radiation is coupled to a specimen containing a target chemical whose presence and/or concentration is to be ascertained. preferably the source radiation includes a plurality of high frequency spectra, at least one of which encourages energy transfer with the chemical of interest. the source radiation is coupled to the specimen via a probe pair that is also used to access a return signal representing an interaction between the source signal and the specimen. the return signal is processed to yield a spectral signature correlating to the target chemical and/or its concentration. preferably signal processing compares frequency spectra within the source signal to spectra within a sampled return signal. the sampled return signal is a signal obtained by sampling a return signal at the probe pair and by sampling the response to the source signal of a circuit that electrically approximates the specimen. the amplitude and/or phase difference between the sampled return signal and the source signal provides recognition of the spectral signature. target chemical concentration data may be obtained from the signature and can be displayed in a number of ways. operation of the signal processor may be optimized using a neural network. in the preferred embodiment, the specimen is a human finger that is pressed against the probe pair, and the chemical is glucose. the invention thus permits a lay user to non-invasively determine his or her glucose level. dated 1996-04-16"
5509106,triangular scalable neural array processor,"a triangular scalable neural array processor unit for use in a neural network and having multipliers, communicating adder trees, sigmoid generators, and a reverse feedback loop for communicating the output of a sigmoid generator back to input multipliers of selected neurons.",1996-04-16,"The title of the patent is triangular scalable neural array processor and its abstract is a triangular scalable neural array processor unit for use in a neural network and having multipliers, communicating adder trees, sigmoid generators, and a reverse feedback loop for communicating the output of a sigmoid generator back to input multipliers of selected neurons. dated 1996-04-16"
5509424,continuous cardiac output monitoring system,"a cardiac catheter continuously monitors cardiac output within an artery. one temperature sensor measures native blood temperature within the artery, while another temperature sensor measures the temperature of a thermal coil which is in thermal contact with the blood stream. the temperature signals are provided as inputs to a monitoring system which includes isolators, filters, and data processing circuits. a temperature difference signal over time is generated between the native blood temperature and the thermal coil temperature. first and second derivatives are taken of the temperature difference signal and selected features are extracted from the three waveforms. the extracted features are used as to calculate cardiac output. in the present case, a neural network processor is utilized to provide accurate cardiac output measurements based upon the extracted features.",1996-04-23,"The title of the patent is continuous cardiac output monitoring system and its abstract is a cardiac catheter continuously monitors cardiac output within an artery. one temperature sensor measures native blood temperature within the artery, while another temperature sensor measures the temperature of a thermal coil which is in thermal contact with the blood stream. the temperature signals are provided as inputs to a monitoring system which includes isolators, filters, and data processing circuits. a temperature difference signal over time is generated between the native blood temperature and the thermal coil temperature. first and second derivatives are taken of the temperature difference signal and selected features are extracted from the three waveforms. the extracted features are used as to calculate cardiac output. in the present case, a neural network processor is utilized to provide accurate cardiac output measurements based upon the extracted features. dated 1996-04-23"
5510596,penetration sensor/controller arc welder,"an arc welding device apparatus for controlling an arc welder through use of a neural network in real-time. the invention can also record output from an arc welding apparatus, indicating whether penetration has occurred during the welding process, and can also activate an alarm when penetration occurs during the arc welding process.",1996-04-23,"The title of the patent is penetration sensor/controller arc welder and its abstract is an arc welding device apparatus for controlling an arc welder through use of a neural network in real-time. the invention can also record output from an arc welding apparatus, indicating whether penetration has occurred during the welding process, and can also activate an alarm when penetration occurs during the arc welding process. dated 1996-04-23"
5511133,method and apparatus of recognizing a moving object,"disclosed is a moving object recognizing system, resistant to noise, which precisely obtains velocities, positions and configurations of an object moving in three-dimensional space. the system employs a bidirectional neural network including velocity neurons, coupled by twos and disposed, for respectively obtaining x-and-y components of an optical flow at respective points of the moving object and also line processes, interposed between the velocity neurons, for taking analog values for detecting edges of the moving object.",1996-04-23,"The title of the patent is method and apparatus of recognizing a moving object and its abstract is disclosed is a moving object recognizing system, resistant to noise, which precisely obtains velocities, positions and configurations of an object moving in three-dimensional space. the system employs a bidirectional neural network including velocity neurons, coupled by twos and disposed, for respectively obtaining x-and-y components of an optical flow at respective points of the moving object and also line processes, interposed between the velocity neurons, for taking analog values for detecting edges of the moving object. dated 1996-04-23"
5511134,image recognition device and image recognition method,"a two-dimensional bit image generated by the input device is processed by a four-layer neural network and recognized. the neurons in the second layer are connected only to the neurons of the first layer aligned in a specific direction, thereby enabling the second layer to extract line components in specific directions. the second layer is further divided into plural regions, and all neurons in each region are connected to one corresponding neuron in the third layer. the output of the third layer neurons thus express the position and degree of the extracted line component in the image. all of the neurons in the third layer are connected to all of the neurons in the fourth layer, and image recognition is possible by learning.",1996-04-23,"The title of the patent is image recognition device and image recognition method and its abstract is a two-dimensional bit image generated by the input device is processed by a four-layer neural network and recognized. the neurons in the second layer are connected only to the neurons of the first layer aligned in a specific direction, thereby enabling the second layer to extract line components in specific directions. the second layer is further divided into plural regions, and all neurons in each region are connected to one corresponding neuron in the third layer. the output of the third layer neurons thus express the position and degree of the extracted line component in the image. all of the neurons in the third layer are connected to all of the neurons in the fourth layer, and image recognition is possible by learning. dated 1996-04-23"
5513097,method and control device for controlling a process including the use of a neural network having variable network parameters,"for the control of a process in a controlled system, a presetting of the system takes place at the start of each process sequence as a function of a precalculated process parameter which exhibits a system-induced dependence on faulty input variables. in this case, the description of the dependence takes place by a model of the process which is adapted during the course of the process. to prevent dependence on the creation of models, which as a rule are imprecise, the input variables are fed before the start of the process to a neural network having variable network parameters for the precalculation of the process parameter based on measurements of the input variables and of the process parameter. these variables are recalculated during the course of the process and utilized for the adaptation of the network parameters.",1996-04-30,"The title of the patent is method and control device for controlling a process including the use of a neural network having variable network parameters and its abstract is for the control of a process in a controlled system, a presetting of the system takes place at the start of each process sequence as a function of a precalculated process parameter which exhibits a system-induced dependence on faulty input variables. in this case, the description of the dependence takes place by a model of the process which is adapted during the course of the process. to prevent dependence on the creation of models, which as a rule are imprecise, the input variables are fed before the start of the process to a neural network having variable network parameters for the precalculation of the process parameter based on measurements of the input variables and of the process parameter. these variables are recalculated during the course of the process and utilized for the adaptation of the network parameters. dated 1996-04-30"
5513098,method for model-free control of general discrete-time systems,"a method of developing a controller for general (nonlinear) discrete-time systems, where the equations governing the system are unknown and where a controller is estimated without building or assuming a model for the system. the controller is constructed through the use of a function approximator (fa) such as a neural network or polynomial. this involves the estimation of the unknown parameters within the fa through the use of a stochastic approximation that is based on a simultaneous perturbation gradient approximation, which requires only system measurements (not a system model).",1996-04-30,"The title of the patent is method for model-free control of general discrete-time systems and its abstract is a method of developing a controller for general (nonlinear) discrete-time systems, where the equations governing the system are unknown and where a controller is estimated without building or assuming a model for the system. the controller is constructed through the use of a function approximator (fa) such as a neural network or polynomial. this involves the estimation of the unknown parameters within the fa through the use of a stochastic approximation that is based on a simultaneous perturbation gradient approximation, which requires only system measurements (not a system model). dated 1996-04-30"
5515189,neural network device and image recognition method employing photoconductive liquid crystal device with patterned electrode,"a light modulation device comprising a first transparent electrode layer, a photoconductive layer, a conductive electrode, a light modulation layer, and a second transparent electrode layer formed together in the preceding order, and characterized by the light modulation characteristic of the light modulation layer being a non-linear saturation function of an applied electrical field, and the conductive electrode comprising plural electrode patterns. the light modulation layer modulates the read light when the input light exceeds a specific threshold value wherein by forming the electrode pattern of the conductive electrode in the shape of the pattern to be extracted, the features of the input image corresponding to that shape can be quickly extracted.",1996-05-07,"The title of the patent is neural network device and image recognition method employing photoconductive liquid crystal device with patterned electrode and its abstract is a light modulation device comprising a first transparent electrode layer, a photoconductive layer, a conductive electrode, a light modulation layer, and a second transparent electrode layer formed together in the preceding order, and characterized by the light modulation characteristic of the light modulation layer being a non-linear saturation function of an applied electrical field, and the conductive electrode comprising plural electrode patterns. the light modulation layer modulates the read light when the input light exceeds a specific threshold value wherein by forming the electrode pattern of the conductive electrode in the shape of the pattern to be extracted, the features of the input image corresponding to that shape can be quickly extracted. dated 1996-05-07"
5515450,method and apparatus for adjusting read-out conditions and/or image,"a first image signal representing a radiation image of an object is obtained by exposing a stimulable phosphor sheet, on which the radiation image has been stored, to stimulating rays, which cause the stimulable phosphor sheet to emit light in proportion to the amount of energy stored thereon during its exposure to radiation, the emitted light being detected. a second image signal representing the radiation image is thereafter obtained by again exposing the stimulable phosphor sheet to stimulating rays, the light emitted by the stimulable phosphor sheet being detected. read-out conditions, under which the second image signal is to be obtained, and/or image processing conditions, under which the second image signal having been obtained is to be image processed, are adjusted on the basis of the first image signal. a storage device stores information representing a standard pattern of radiation images. a signal transforming device transforms the first image signal representing the radiation image into a transformed image signal representing the radiation image, which has been transformed into the standard pattern. a condition adjuster is provided with a neural network, which receives the transformed image signal and feeds out information representing the read-out conditions and/or the image processing conditions.",1996-05-07,"The title of the patent is method and apparatus for adjusting read-out conditions and/or image and its abstract is a first image signal representing a radiation image of an object is obtained by exposing a stimulable phosphor sheet, on which the radiation image has been stored, to stimulating rays, which cause the stimulable phosphor sheet to emit light in proportion to the amount of energy stored thereon during its exposure to radiation, the emitted light being detected. a second image signal representing the radiation image is thereafter obtained by again exposing the stimulable phosphor sheet to stimulating rays, the light emitted by the stimulable phosphor sheet being detected. read-out conditions, under which the second image signal is to be obtained, and/or image processing conditions, under which the second image signal having been obtained is to be image processed, are adjusted on the basis of the first image signal. a storage device stores information representing a standard pattern of radiation images. a signal transforming device transforms the first image signal representing the radiation image into a transformed image signal representing the radiation image, which has been transformed into the standard pattern. a condition adjuster is provided with a neural network, which receives the transformed image signal and feeds out information representing the read-out conditions and/or the image processing conditions. dated 1996-05-07"
5515477,neural networks,"the present invention relates to adaptive information processing systems, and in particular to associative memories utilizing confidence-mediated associations, and especially neural network systems comprising an auto-organizational apparatus and processes for dynamically mapping an input onto a semantically congruous and contemporaneously-valid, learned response. in particular the present invention relates to such an associative memory system in which provision is made for improving the congruence between an associative memory, by impressing a desired response on an associative memory mapping based on complex polar values.",1996-05-07,"The title of the patent is neural networks and its abstract is the present invention relates to adaptive information processing systems, and in particular to associative memories utilizing confidence-mediated associations, and especially neural network systems comprising an auto-organizational apparatus and processes for dynamically mapping an input onto a semantically congruous and contemporaneously-valid, learned response. in particular the present invention relates to such an associative memory system in which provision is made for improving the congruence between an associative memory, by impressing a desired response on an associative memory mapping based on complex polar values. dated 1996-05-07"
5517429,intelligent area monitoring system,"an intelligent area monitoring system having a plurality of sensors (11,12,13,14,15,16), a neural network computer (20), a data communications network (28,30,32,42), and multiple graphic display stations (40). the neural network computer (20) accepts the input signals from each sensor. any changes that occur within a monitored area are communicated to system users as symbols which appear in context of a graphic rendering of the monitored area. the sensors can be active or passive. each sensor provides an analog output (54). codes are communicated to graphic display stations (40) via a data communications network (28,30,32,42). based on these codes, the graphic display stations (40) select and place symbols on their display screens to accurately represent the identity and location of targets in the monitored area.",1996-05-14,"The title of the patent is intelligent area monitoring system and its abstract is an intelligent area monitoring system having a plurality of sensors (11,12,13,14,15,16), a neural network computer (20), a data communications network (28,30,32,42), and multiple graphic display stations (40). the neural network computer (20) accepts the input signals from each sensor. any changes that occur within a monitored area are communicated to system users as symbols which appear in context of a graphic rendering of the monitored area. the sensors can be active or passive. each sensor provides an analog output (54). codes are communicated to graphic display stations (40) via a data communications network (28,30,32,42). based on these codes, the graphic display stations (40) select and place symbols on their display screens to accurately represent the identity and location of targets in the monitored area. dated 1996-05-14"
5517537,integrated acoustic leak detection beamforming system,"a system for mapping absolute acoustic noise intensity in a three-dimensional acoustic noise field, and using three-dimensional absolute noise intensities to infer operational or performance characteristics of components or structures within the monitored field. localized noise sources are extracted using a distant array of transducers, and the absolute intensity can be measured even when totally masked by background noise at the transducer locations. the system includes an integrated sensor installation; a neural network detection system algorithm; a zoom system to precisely examine a small region in the steam generator vessel; a fuzzy logic system detection algorithm; and an expert detection system. the signal processing subsystem operates at three different levels of detection. at the top level of detection a trained neural network system monitors the vessel length for any indication of a leak. the proposed approach uses discriminators which do not require any beamforming of the sensor signals. the presence of a leak and its general location in the vessel is determined using a fuzzy logic expert system. the leak indications are passed to a second level of detection: a beamformed system which will monitor the indicated portion of the vessel. a third level of detection systematically monitors the tagged locations to determine if a leak is really present, remaining on the specific locations to detect a leak above the defined threshold to a high degree of accuracy. the actions to be taken will be controlled by a second fuzzy logic expert interface controller.",1996-05-14,"The title of the patent is integrated acoustic leak detection beamforming system and its abstract is a system for mapping absolute acoustic noise intensity in a three-dimensional acoustic noise field, and using three-dimensional absolute noise intensities to infer operational or performance characteristics of components or structures within the monitored field. localized noise sources are extracted using a distant array of transducers, and the absolute intensity can be measured even when totally masked by background noise at the transducer locations. the system includes an integrated sensor installation; a neural network detection system algorithm; a zoom system to precisely examine a small region in the steam generator vessel; a fuzzy logic system detection algorithm; and an expert detection system. the signal processing subsystem operates at three different levels of detection. at the top level of detection a trained neural network system monitors the vessel length for any indication of a leak. the proposed approach uses discriminators which do not require any beamforming of the sensor signals. the presence of a leak and its general location in the vessel is determined using a fuzzy logic expert system. the leak indications are passed to a second level of detection: a beamformed system which will monitor the indicated portion of the vessel. a third level of detection systematically monitors the tagged locations to determine if a leak is really present, remaining on the specific locations to detect a leak above the defined threshold to a high degree of accuracy. the actions to be taken will be controlled by a second fuzzy logic expert interface controller. dated 1996-05-14"
5517596,learning machine synapse processor system apparatus,"a neural synapse processor apparatus having a neuron architecture for the synapse processing elements of the apparatus. the apparatus which we prefer will have a n neuron structure having synapse processing units that contain instruction and data storage units, receive instructions and data, and execute instructions. the n neuron structure should contain communicating adder trees, neuron activation function units, and an arrangement for communicating both instructions, data, and the outputs of neuron activation function units back to the input synapse processing units by means of the communicating adder trees. the apparatus can be structured as a bit-serial or word parallel system. the preferred structure contains n.sup.2 synapse processing units, each associated with a connection weight in the n neural network to be emulated, placed in the form of a n by n matrix that has been folded along the diagonal and made up of diagonal cells and general cells. diagonal cells, each utilizing a single synapse processing unit, are associated with the diagonal connection weights of the folded n by n connection weight matrix and general cells, each of which has two synapse processing units merged together, and which are associated with the symmetric connection weights of the folded n by n connection weight matrix. the back-propagation learning algorithm is first discussed followed by a presentation of the learning machine synapse processor architecture. an example implementation of the back-propagation learning algorithm is then presented. this is followed by a boltzmann like machine example and data parallel examples mapped onto the architecture.",1996-05-14,"The title of the patent is learning machine synapse processor system apparatus and its abstract is a neural synapse processor apparatus having a neuron architecture for the synapse processing elements of the apparatus. the apparatus which we prefer will have a n neuron structure having synapse processing units that contain instruction and data storage units, receive instructions and data, and execute instructions. the n neuron structure should contain communicating adder trees, neuron activation function units, and an arrangement for communicating both instructions, data, and the outputs of neuron activation function units back to the input synapse processing units by means of the communicating adder trees. the apparatus can be structured as a bit-serial or word parallel system. the preferred structure contains n.sup.2 synapse processing units, each associated with a connection weight in the n neural network to be emulated, placed in the form of a n by n matrix that has been folded along the diagonal and made up of diagonal cells and general cells. diagonal cells, each utilizing a single synapse processing unit, are associated with the diagonal connection weights of the folded n by n connection weight matrix and general cells, each of which has two synapse processing units merged together, and which are associated with the symmetric connection weights of the folded n by n connection weight matrix. the back-propagation learning algorithm is first discussed followed by a presentation of the learning machine synapse processor architecture. an example implementation of the back-propagation learning algorithm is then presented. this is followed by a boltzmann like machine example and data parallel examples mapped onto the architecture. dated 1996-05-14"
5517598,error back-propagation method and neural network system,"method and apparatus of error back-propagation for use in a neural network system. a first group (11) of processing devices (13.sub.1, 13.sub.2, 13.sub.3) performs the resolving steps and a second group (12) of analogous processing devices (13.sub.4, 13.sub.5) performs the training steps while backpropagating errors calculated in a central processing device (10). the synaptic coefficient matrix c.sub.ij of the first group and the transposed matrix t.sub.ji of the second group are simultaneously updated. this updating of the synaptic coefficients can be performed by means of multipliers (34.sub.1 to 34.sub.n) and adders (37.sub.1 to 37.sub.n).",1996-05-14,"The title of the patent is error back-propagation method and neural network system and its abstract is method and apparatus of error back-propagation for use in a neural network system. a first group (11) of processing devices (13.sub.1, 13.sub.2, 13.sub.3) performs the resolving steps and a second group (12) of analogous processing devices (13.sub.4, 13.sub.5) performs the training steps while backpropagating errors calculated in a central processing device (10). the synaptic coefficient matrix c.sub.ij of the first group and the transposed matrix t.sub.ji of the second group are simultaneously updated. this updating of the synaptic coefficients can be performed by means of multipliers (34.sub.1 to 34.sub.n) and adders (37.sub.1 to 37.sub.n). dated 1996-05-14"
5517667,neural network that does not require repetitive training,"a neural network, which may be implemented either in hardware or software, is constructed of neurons or neuron circuits each having only one significant processing element in the form of a multiplier. a hidden neuron in the neural network generates an output based on the product of a plurality of functions. the neural network utilizes a training algorithm which does not require repetitive training and which yields a global minimum to each given set of input vectors.",1996-05-14,"The title of the patent is neural network that does not require repetitive training and its abstract is a neural network, which may be implemented either in hardware or software, is constructed of neurons or neuron circuits each having only one significant processing element in the form of a multiplier. a hidden neuron in the neural network generates an output based on the product of a plurality of functions. the neural network utilizes a training algorithm which does not require repetitive training and which yields a global minimum to each given set of input vectors. dated 1996-05-14"
5519610,control system for automatic transmissions with teaching and automatic modes using a neural network,"a control system for an automatic transmission, which can easily achieve a variety of shift patterns according to the taste of a driver and the road conditions. the control system includes: a neural network; and a mode switch for selecting either a teaching mode or an automatic mode. the neural network learns in the teaching mode and neuro-computes to output a gear ratio in the automatic mode. tolerance error is computed from at least one of a propriety of a teaching gear ratio, a compatibility of an input value and a compatibility of the teaching gear ratio. a learning pattern is produced during the teaching mode for correcting load factors of the neural network which, in the automatic mode, outputs a gear ratio used to determine shifting of the transmission.",1996-05-21,"The title of the patent is control system for automatic transmissions with teaching and automatic modes using a neural network and its abstract is a control system for an automatic transmission, which can easily achieve a variety of shift patterns according to the taste of a driver and the road conditions. the control system includes: a neural network; and a mode switch for selecting either a teaching mode or an automatic mode. the neural network learns in the teaching mode and neuro-computes to output a gear ratio in the automatic mode. tolerance error is computed from at least one of a propriety of a teaching gear ratio, a compatibility of an input value and a compatibility of the teaching gear ratio. a learning pattern is produced during the teaching mode for correcting load factors of the neural network which, in the automatic mode, outputs a gear ratio used to determine shifting of the transmission. dated 1996-05-21"
5519647,apparatus for and method of generating an approximation function,"an apparatus (5) for generating an approximation function based on first pairs ((x.sub.1, y.sub.1) to (x.sub.6, y.sub.6)) of values associating a dependent variable (y.sub.1 to y.sub.6) with an independent variable (x.sub.1 to x.sub.6), and for determining second pairs (x.sub.a, y'.sub.a) of values of said variables in accordance with said approximation function. the apparatus comprises: a) first means (10) for iteratively determining at least one current linear regression function, for selecting that one of the current linear functions which produces the approximation of all the pairs of said series with minimal errors, and for coding the selected linear regression function with the aid of specific codes (p, q), and b) second means (17) for determining said second pairs (x.sub.a, y'.sub.a) with the aid of said specific codes. the apparatus can also be used for calculating approximated values of mathematical functions, for example a in a neural network, or for determining a regression function forming an approximation to experimental measurement results, for example distributed measurements resulting from monitoring an industrial process. the invention also relates to a method of generating an approximation function.",1996-05-21,"The title of the patent is apparatus for and method of generating an approximation function and its abstract is an apparatus (5) for generating an approximation function based on first pairs ((x.sub.1, y.sub.1) to (x.sub.6, y.sub.6)) of values associating a dependent variable (y.sub.1 to y.sub.6) with an independent variable (x.sub.1 to x.sub.6), and for determining second pairs (x.sub.a, y'.sub.a) of values of said variables in accordance with said approximation function. the apparatus comprises: a) first means (10) for iteratively determining at least one current linear regression function, for selecting that one of the current linear functions which produces the approximation of all the pairs of said series with minimal errors, and for coding the selected linear regression function with the aid of specific codes (p, q), and b) second means (17) for determining said second pairs (x.sub.a, y'.sub.a) with the aid of said specific codes. the apparatus can also be used for calculating approximated values of mathematical functions, for example a in a neural network, or for determining a regression function forming an approximation to experimental measurement results, for example distributed measurements resulting from monitoring an industrial process. the invention also relates to a method of generating an approximation function. dated 1996-05-21"
5519784,apparatus for classifying movement of objects along a passage by type and direction employing time domain patterns,"an array of spaced parallel linear radiation beams are projected across a passage in sequentially spaced regions. detectors sense when the beams are interrupted by one or more persons moving in the passage in either of two opposite directions. the beams are interrupted at different times in a sequence corresponding to the number of and direction of movement of persons. the sequentially generated interrupted beam signals are stored as object movement historic information in memory and later processed to generate composite beam interrupt patterns manifesting the number of persons and direction of movement, the patterns being a function of time domain and sensor index, i.e., sensor identity and position in the passage. the resulting generated patterns are compared to reference patterns utilizing computerized pattern recognition analysis such as with an artificial neural network. the comparison classifies the persons in the passage into direction of movement and number.",1996-05-21,"The title of the patent is apparatus for classifying movement of objects along a passage by type and direction employing time domain patterns and its abstract is an array of spaced parallel linear radiation beams are projected across a passage in sequentially spaced regions. detectors sense when the beams are interrupted by one or more persons moving in the passage in either of two opposite directions. the beams are interrupted at different times in a sequence corresponding to the number of and direction of movement of persons. the sequentially generated interrupted beam signals are stored as object movement historic information in memory and later processed to generate composite beam interrupt patterns manifesting the number of persons and direction of movement, the patterns being a function of time domain and sensor index, i.e., sensor identity and position in the passage. the resulting generated patterns are compared to reference patterns utilizing computerized pattern recognition analysis such as with an artificial neural network. the comparison classifies the persons in the passage into direction of movement and number. dated 1996-05-21"
5519805,signal processing arrangements,""" a signal processing system for time varying band-limited signals such as speech comprises signal coding means for affording a time encoded signal symbol stream, a matrix generator for generating a fixed size, typically """"a"""" matrix from the symbol stream, and an artificial neural network to which the matrix elements are applied for network training and subsequently for affording an output indicative of the nature of an input signal. """,1996-05-21,"The title of the patent is signal processing arrangements and its abstract is "" a signal processing system for time varying band-limited signals such as speech comprises signal coding means for affording a time encoded signal symbol stream, a matrix generator for generating a fixed size, typically """"a"""" matrix from the symbol stream, and an artificial neural network to which the matrix elements are applied for network training and subsequently for affording an output indicative of the nature of an input signal. "" dated 1996-05-21"
5519811,"neural network, processor, and pattern recognition apparatus","apparatus for realizing a neural network of a complex structure, such as the neocognitron, in a neural network processor comprises processing elements corresponding to the neurons of a multilayer feed-forward neural network. each of the processing elements comprises an mos analog circuit that receives input voltage signals and provides output voltage signals. the mos analog circuits are arranged in a systolic array.",1996-05-21,"The title of the patent is neural network, processor, and pattern recognition apparatus and its abstract is apparatus for realizing a neural network of a complex structure, such as the neocognitron, in a neural network processor comprises processing elements corresponding to the neurons of a multilayer feed-forward neural network. each of the processing elements comprises an mos analog circuit that receives input voltage signals and provides output voltage signals. the mos analog circuits are arranged in a systolic array. dated 1996-05-21"
5519812,ferrelectric adaptive-learning type product-sum operation circuit element and circuit using such element,"the present invention relates to a product-sum operation circuit element and a circuit for addition by weighting a number of signals input in one neuron circuit in a neural network, and can provide an adaptive-learning neuron circuit for changing an interval of output pulses by learning by connecting a simple pulse generating circuit consisting of capacitance, resistance, unijunction transistor and the like. a product-sum operation circuit element according to the present invention, includes an insulator substrate, a single crystal semiconductor thin film having a p-n-p or n-p-n structure in a lateral direction formed in the shape of stripes on the insulator substrate, a ferroelectric thin film deposited thereon for covering at least the semiconductor stripe structure, and a stripe-like electrode consisting of a metal or a polycrystalline semiconductor further formed thereon for intersecting the semiconductor stripes at a right angle or suitable angle.",1996-05-21,"The title of the patent is ferrelectric adaptive-learning type product-sum operation circuit element and circuit using such element and its abstract is the present invention relates to a product-sum operation circuit element and a circuit for addition by weighting a number of signals input in one neuron circuit in a neural network, and can provide an adaptive-learning neuron circuit for changing an interval of output pulses by learning by connecting a simple pulse generating circuit consisting of capacitance, resistance, unijunction transistor and the like. a product-sum operation circuit element according to the present invention, includes an insulator substrate, a single crystal semiconductor thin film having a p-n-p or n-p-n structure in a lateral direction formed in the shape of stripes on the insulator substrate, a ferroelectric thin film deposited thereon for covering at least the semiconductor stripe structure, and a stripe-like electrode consisting of a metal or a polycrystalline semiconductor further formed thereon for intersecting the semiconductor stripes at a right angle or suitable angle. dated 1996-05-21"
5521813,system and method for the advanced prediction of weather impact on managerial planning applications,"a computer-based system and method which incorporates long-range weather forecasts in a predictive model which quantifies historical weather impact relationships between datasets, and uses the long-range weather forecasts to predict future weather impact on managerial plans. the predictive model can use multiple regression or a neural network. in a retail application, the computer-based system and method has a weather impact predictive model based on correlations of historical weather and point-of-sale store transactions data. the weather impact model is coupled with long-range weather forecasts to adjust managerial plans for buying, distribution, financial budgeting, promotional and advertising applications. a graphical user interface provides easy assimilation of analysis for specific managerial planning applications.",1996-05-28,"The title of the patent is system and method for the advanced prediction of weather impact on managerial planning applications and its abstract is a computer-based system and method which incorporates long-range weather forecasts in a predictive model which quantifies historical weather impact relationships between datasets, and uses the long-range weather forecasts to predict future weather impact on managerial plans. the predictive model can use multiple regression or a neural network. in a retail application, the computer-based system and method has a weather impact predictive model based on correlations of historical weather and point-of-sale store transactions data. the weather impact model is coupled with long-range weather forecasts to adjust managerial plans for buying, distribution, financial budgeting, promotional and advertising applications. a graphical user interface provides easy assimilation of analysis for specific managerial planning applications. dated 1996-05-28"
5521840,diagnostic system responsive to learned audio signatures,"loose parts are sensed moving in a conduit carrying a flowing material, such as the cooling circuit of a pressurized water nuclear reactor. an acoustic pickup produces an electrical signal with vibration of the conduit due to impact of the loose part, and background noise. a signal processor encodes the values of distinct parameters of the electrical signal such as amplitude, amplitude at particular frequencies, etc., in an ongoing manner, producing discrete output values. these outputs are coupled as inputs to a neural network with physical or logical neuron cells loaded with weighting factors affecting the strength and polarity of neural interconnections. the factors represent the acoustic signature of the loose part. products of the input values and the weighting factors are summed to produce one or more neural network outputs, compared to a threshold. the sum normally varies randomly, but has a strong swing when the pattern is encountered, due to the factors emphasizing the pattern over background noise. the threshold comparison operates a display or alarm. the weighting factors are learned by repeating empirical tests and correlating the factors to the signal to minimize error.",1996-05-28,"The title of the patent is diagnostic system responsive to learned audio signatures and its abstract is loose parts are sensed moving in a conduit carrying a flowing material, such as the cooling circuit of a pressurized water nuclear reactor. an acoustic pickup produces an electrical signal with vibration of the conduit due to impact of the loose part, and background noise. a signal processor encodes the values of distinct parameters of the electrical signal such as amplitude, amplitude at particular frequencies, etc., in an ongoing manner, producing discrete output values. these outputs are coupled as inputs to a neural network with physical or logical neuron cells loaded with weighting factors affecting the strength and polarity of neural interconnections. the factors represent the acoustic signature of the loose part. products of the input values and the weighting factors are summed to produce one or more neural network outputs, compared to a threshold. the sum normally varies randomly, but has a strong swing when the pattern is encountered, due to the factors emphasizing the pattern over background noise. the threshold comparison operates a display or alarm. the weighting factors are learned by repeating empirical tests and correlating the factors to the signal to minimize error. dated 1996-05-28"
5521985,apparatus for recognizing machine generated or handprinted text,"a computer system with an image reader and nerual network is provided for determining whether an image of text has been generated by a machine or by hand. this serves the useful purpose of allowing one to use speciallized recognition techiques that are more suited to one form of printing, thus achieving a higher recognition accuracy than by using a single recognition technique for both types of printing. the method is based on the premise that the spatial spectra for an image of machine text will have more higher frequency components than one generated by hand, because of the nonregular, nonuniform slant of the handprint. the method proposed generates this spectra by convolving spatial templates with vertical histograms from each line of text, and uses a neural network to classify the resulting spectra.",1996-05-28,"The title of the patent is apparatus for recognizing machine generated or handprinted text and its abstract is a computer system with an image reader and nerual network is provided for determining whether an image of text has been generated by a machine or by hand. this serves the useful purpose of allowing one to use speciallized recognition techiques that are more suited to one form of printing, thus achieving a higher recognition accuracy than by using a single recognition technique for both types of printing. the method is based on the premise that the spatial spectra for an image of machine text will have more higher frequency components than one generated by hand, because of the nonregular, nonuniform slant of the handprint. the method proposed generates this spectra by convolving spatial templates with vertical histograms from each line of text, and uses a neural network to classify the resulting spectra. dated 1996-05-28"
5522015,neural network and learning method for linearly unseparable patterns,"a neural network has an input layer, a hidden layer, and an output layer. the neural network includes a lower neural network model composed of hidden layer neurons and input layer neurons for learning a plurality of linearly separable patterns, and a higher neural network model composed of hidden layer neurons and output layer neurons for combining the linearly separable patterns into a linearly unseparable pattern.",1996-05-28,"The title of the patent is neural network and learning method for linearly unseparable patterns and its abstract is a neural network has an input layer, a hidden layer, and an output layer. the neural network includes a lower neural network model composed of hidden layer neurons and input layer neurons for learning a plurality of linearly separable patterns, and a higher neural network model composed of hidden layer neurons and output layer neurons for combining the linearly separable patterns into a linearly unseparable pattern. dated 1996-05-28"
5522863,pulsating behavior monitoring and modification system for neural networks,the pulsating behavioral activity of a neural network such as that embodied n a brain tissue slice is monitored by measurement of intervals between spontaneous events to identify the presence of a chaotic regime and determine by real-time calculation a waiting time for electrical pulse intervention pursuant to a behavioral modifying program having a control or anti-control strategy.,1996-06-04,The title of the patent is pulsating behavior monitoring and modification system for neural networks and its abstract is the pulsating behavioral activity of a neural network such as that embodied n a brain tissue slice is monitored by measurement of intervals between spontaneous events to identify the presence of a chaotic regime and determine by real-time calculation a waiting time for electrical pulse intervention pursuant to a behavioral modifying program having a control or anti-control strategy. dated 1996-06-04
5524086,dipole parameter estimation method and apparatus,"a trained neural network is used, for estimating the number, positions or moments of one or more dipoles which are assumed as sources of the electromagnetic field distribution based upon an electromagnetic field distribution of a living body or an object. at least either one of the dipole number, positions and moments or more than two of their combination is referred to as dipole parameters.",1996-06-04,"The title of the patent is dipole parameter estimation method and apparatus and its abstract is a trained neural network is used, for estimating the number, positions or moments of one or more dipoles which are assumed as sources of the electromagnetic field distribution based upon an electromagnetic field distribution of a living body or an object. at least either one of the dipole number, positions and moments or more than two of their combination is referred to as dipole parameters. dated 1996-06-04"
5524175,neuro-computer system for executing a plurality of controlling algorithms,"a general neuro-computer and system using it is capable of executing a plurality of learning algorithms, providing an instruction execution speed comparable with a hard wired system, and practically neglecting a time required for rewriting microprograms. the neuro-computer is constituted by a neuron array having a plurality of neurons, a control storage unit for storing microinstructions, a parameter register, a control logic, and a global memory. a host computer as a user interface inputs information necessary for the learning and execution of the neuro-computer to the system, the information including learning algorithms, neural network architecture, the number of learnings, the number of input patterns, input signals, and desired signals. the information inputted from the host computer is transferred via a scsi to the neuro-computer to perform a desired neural network operation.",1996-06-04,"The title of the patent is neuro-computer system for executing a plurality of controlling algorithms and its abstract is a general neuro-computer and system using it is capable of executing a plurality of learning algorithms, providing an instruction execution speed comparable with a hard wired system, and practically neglecting a time required for rewriting microprograms. the neuro-computer is constituted by a neuron array having a plurality of neurons, a control storage unit for storing microinstructions, a parameter register, a control logic, and a global memory. a host computer as a user interface inputs information necessary for the learning and execution of the neuro-computer to the system, the information including learning algorithms, neural network architecture, the number of learnings, the number of input patterns, input signals, and desired signals. the information inputted from the host computer is transferred via a scsi to the neuro-computer to perform a desired neural network operation. dated 1996-06-04"
5524176,fuzzy expert system learning network,""" in the present invention, prior art techniques are extended to allow application of the backpropagation learning technique to artificial neural networks derived from fuzzy expert system rule-bases. a method in accordance with the invention, referred to herein as a fuzzy expert network (fen), is implemented in a programmed machine such as a computer to provide automated learning of both """"fine"""" and """"coarse"""" knowledge in a network of artificial neural objects (anos) implementing fuzzy modeling rules. through application of the fen method, an event-driven fuzzy expert network comprising acyclically connected anos derived from fuzzy modelling rules may be implemented. neural objects implement one or more fuzzy combining and defuzzification rules and use backpropagation of error techniques to implement learning. as in prior art, the fen allows each ano to adjust its input weight parameters--""""fine"""" knowledge learning. unlike prior art, the fen allows each ano to modify its internal parameters--""""coarse"""" knowledge learning. this latter action means that individual anos have the capability to modify the parameters of the fuzzy rule's membership function upon which they are based. in this way the fen is able to change the structure of its encoded knowledge over time, making it a more adaptable architecture for autonomous and/or adaptable control systems. simulation results showing the fen's learning and adaptability behavior are given. """,1996-06-04,"The title of the patent is fuzzy expert system learning network and its abstract is "" in the present invention, prior art techniques are extended to allow application of the backpropagation learning technique to artificial neural networks derived from fuzzy expert system rule-bases. a method in accordance with the invention, referred to herein as a fuzzy expert network (fen), is implemented in a programmed machine such as a computer to provide automated learning of both """"fine"""" and """"coarse"""" knowledge in a network of artificial neural objects (anos) implementing fuzzy modeling rules. through application of the fen method, an event-driven fuzzy expert network comprising acyclically connected anos derived from fuzzy modelling rules may be implemented. neural objects implement one or more fuzzy combining and defuzzification rules and use backpropagation of error techniques to implement learning. as in prior art, the fen allows each ano to adjust its input weight parameters--""""fine"""" knowledge learning. unlike prior art, the fen allows each ano to modify its internal parameters--""""coarse"""" knowledge learning. this latter action means that individual anos have the capability to modify the parameters of the fuzzy rule's membership function upon which they are based. in this way the fen is able to change the structure of its encoded knowledge over time, making it a more adaptable architecture for autonomous and/or adaptable control systems. simulation results showing the fen's learning and adaptability behavior are given. "" dated 1996-06-04"
5524177,learning of associative memory in form of neural network suitable for connectionist model,"the learning of an associative memory suitable for the connectionist model which can deal with the patterns having the non-random frequencies of the appearances or the non-random correlations. in this invention, the learning of the associative memory in a form of a neural network, in which a plurality of nodes having activation values are connected by a plurality of links having link weight values, is achieved by entering a plurality of learning patterns sequentially, where each learning pattern has a plurality of elements in correspondence with the nodes, calculating an energy e of the entered learning pattern, determining a learning amount .delta. for the entered learning pattern according to a difference between the calculated energy e and a predetermined reference energy level eth, and updating the link weight values of the links according to the entered learning pattern and the determined learning amount .delta..",1996-06-04,"The title of the patent is learning of associative memory in form of neural network suitable for connectionist model and its abstract is the learning of an associative memory suitable for the connectionist model which can deal with the patterns having the non-random frequencies of the appearances or the non-random correlations. in this invention, the learning of the associative memory in a form of a neural network, in which a plurality of nodes having activation values are connected by a plurality of links having link weight values, is achieved by entering a plurality of learning patterns sequentially, where each learning pattern has a plurality of elements in correspondence with the nodes, calculating an energy e of the entered learning pattern, determining a learning amount .delta. for the entered learning pattern according to a difference between the calculated energy e and a predetermined reference energy level eth, and updating the link weight values of the links according to the entered learning pattern and the determined learning amount .delta.. dated 1996-06-04"
5524178,neural network learning system,"a neural network learning system is applied to extensive use in applications such as pattern and character recognizing operations, various controls, etc. the neural network learning system operates on, for example, a plurality of neural networks each having a different number of intermediate layer units to efficiently perform a learning process at a high speed with a reduced amount of hardware. a neural network system having a plurality of hierarchical neural networks each having an input layer, one or more intermediate layers and output layers is formed from a common input layer shared among two or more neural networks, or the common input layer and one or more intermediate layers and a learning controller for controlling a learning process performed by a plurality of neural networks.",1996-06-04,"The title of the patent is neural network learning system and its abstract is a neural network learning system is applied to extensive use in applications such as pattern and character recognizing operations, various controls, etc. the neural network learning system operates on, for example, a plurality of neural networks each having a different number of intermediate layer units to efficiently perform a learning process at a high speed with a reduced amount of hardware. a neural network system having a plurality of hierarchical neural networks each having an input layer, one or more intermediate layers and output layers is formed from a common input layer shared among two or more neural networks, or the common input layer and one or more intermediate layers and a learning controller for controlling a learning process performed by a plurality of neural networks. dated 1996-06-04"
5526281,machine-learning approach to modeling biological activity for molecular design and to modeling other characteristics,"explicit representation of molecular shape of molecules is combined with neural network learning methods to provide models with high predictive ability that generalize to different chemical classes where structurally diverse molecules exhibiting similar surface characteristics are treated as similar. a new machine-learning methodology that can accept multiple representations of objects and construct models that predict characteristics of those objects. an extension of this methodology can be applied in cases where the representations of the objects are determined by a set of adjustable parameters. an iterative process applies intermediate models to generate new representations of the objects by adjusting said parameters and repeatedly retrains the models to obtain better predictive models. this method can be applied to molecules because each molecule can have many orientations and conformations (representations) that are determined by a set of translation, rotation and torsion angle parameters.",1996-06-11,"The title of the patent is machine-learning approach to modeling biological activity for molecular design and to modeling other characteristics and its abstract is explicit representation of molecular shape of molecules is combined with neural network learning methods to provide models with high predictive ability that generalize to different chemical classes where structurally diverse molecules exhibiting similar surface characteristics are treated as similar. a new machine-learning methodology that can accept multiple representations of objects and construct models that predict characteristics of those objects. an extension of this methodology can be applied in cases where the representations of the objects are determined by a set of adjustable parameters. an iterative process applies intermediate models to generate new representations of the objects by adjusting said parameters and repeatedly retrains the models to obtain better predictive models. this method can be applied to molecules because each molecule can have many orientations and conformations (representations) that are determined by a set of translation, rotation and torsion angle parameters. dated 1996-06-11"
5526446,noise reduction system,"a technique is provided to remove noise from images and to enhance their visual appearance through the utilization of a technique which converts an image into a set of coefficients in a multi-scale image decomposition process, followed by modification of each coefficient based on its value and the value of coefficients of related orientation, position, or scale, which is in turn followed by a reconstruction or synthesis process to generate the enhanced image. also contributing to the improved enhancement is a set of orientation tuned filters of a specialized design to permit steering, with the analysis and synthesis filters also having a self-inverting characteristic. additionally, steerable pyramid architecture is used for image enhancement for the first time, with the steering being provided by the above orientation tuned filters. the utilization of related coefficients permits coefficient modification with multipliers derived through a statistical or neural-network analysis of coefficients derived through the utilization of clean and degraded images, with the modifiers corresponding to vectors which result in translating the degraded image coefficients into clean image coefficients, in essence by cancelling those portions of a coefficient due to noise. further improvements include an overlay of classical coring on single coefficients. thus, the subject technique provides improved image enhancement through the use of a multi-band or scale-oriented analysis and synthesis transform having improved coefficient modification, good orientation tuning, improved bandpass characteristics, and good spatial localization.",1996-06-11,"The title of the patent is noise reduction system and its abstract is a technique is provided to remove noise from images and to enhance their visual appearance through the utilization of a technique which converts an image into a set of coefficients in a multi-scale image decomposition process, followed by modification of each coefficient based on its value and the value of coefficients of related orientation, position, or scale, which is in turn followed by a reconstruction or synthesis process to generate the enhanced image. also contributing to the improved enhancement is a set of orientation tuned filters of a specialized design to permit steering, with the analysis and synthesis filters also having a self-inverting characteristic. additionally, steerable pyramid architecture is used for image enhancement for the first time, with the steering being provided by the above orientation tuned filters. the utilization of related coefficients permits coefficient modification with multipliers derived through a statistical or neural-network analysis of coefficients derived through the utilization of clean and degraded images, with the modifiers corresponding to vectors which result in translating the degraded image coefficients into clean image coefficients, in essence by cancelling those portions of a coefficient due to noise. further improvements include an overlay of classical coring on single coefficients. thus, the subject technique provides improved image enhancement through the use of a multi-band or scale-oriented analysis and synthesis transform having improved coefficient modification, good orientation tuning, improved bandpass characteristics, and good spatial localization. dated 1996-06-11"
5528700,character recognition system based on a neural network,"a character recognition system based on a neural network determines activation patterns in an input layer and output layer, increases weights of synapses in a middle layer so that neurons activate with more than a certain rate among those corresponding to neurons in the input layer and the output layer and repeats the same process for each neuron in the middle layer. the input layer and output layer possess a plurality of neurons which activate and output certain data according to a specific result and the middle layer is between the input layer and output layer. the middle layer also possesses a plurality of neurons which are connected to each neuron in the input layer and output layer.",1996-06-18,"The title of the patent is character recognition system based on a neural network and its abstract is a character recognition system based on a neural network determines activation patterns in an input layer and output layer, increases weights of synapses in a middle layer so that neurons activate with more than a certain rate among those corresponding to neurons in the input layer and the output layer and repeats the same process for each neuron in the middle layer. the input layer and output layer possess a plurality of neurons which activate and output certain data according to a specific result and the middle layer is between the input layer and output layer. the middle layer also possesses a plurality of neurons which are connected to each neuron in the input layer and output layer. dated 1996-06-18"
5528728,speaker independent speech recognition system and method using neural network and dtw matching technique,"improved speaker independent speech recognition system and method are disclosed in which an utterance by an unspecified person into an electrical signal is input through a device such as a telephone, the electrical signal from the input telephone converting the electrical signal into a time series of characteristic multidimensional vectors, the time series of characteristic multidimensional vectors are received, each of the vectors being converted into a plurality of candidates so that the plurality of phonemes constitutes a plurality of strings of phonemes in time series as a plurality of candidates, the plurality of candidates of phonemes are compared simultaneously (one at a time) with a reference pattern of a reference string of phonemes for each word previously stored in a dictionary to determine which string of phonemes derived from the phoneme recognition means has a highest similarity to one of the reference strings of the phonemes for the respective words stored in the dictionary using a predetermined word matching technique, and at least one candidate of the words as a result of word recognition on the basis of one of the plurality of the strings of phonemes which has the highest similarity to the corresponding one of the reference strings of the respective words is output as the result of speech recognition.",1996-06-18,"The title of the patent is speaker independent speech recognition system and method using neural network and dtw matching technique and its abstract is improved speaker independent speech recognition system and method are disclosed in which an utterance by an unspecified person into an electrical signal is input through a device such as a telephone, the electrical signal from the input telephone converting the electrical signal into a time series of characteristic multidimensional vectors, the time series of characteristic multidimensional vectors are received, each of the vectors being converted into a plurality of candidates so that the plurality of phonemes constitutes a plurality of strings of phonemes in time series as a plurality of candidates, the plurality of candidates of phonemes are compared simultaneously (one at a time) with a reference pattern of a reference string of phonemes for each word previously stored in a dictionary to determine which string of phonemes derived from the phoneme recognition means has a highest similarity to one of the reference strings of the phonemes for the respective words stored in the dictionary using a predetermined word matching technique, and at least one candidate of the words as a result of word recognition on the basis of one of the plurality of the strings of phonemes which has the highest similarity to the corresponding one of the reference strings of the respective words is output as the result of speech recognition. dated 1996-06-18"
5528729,neural network learning apparatus and learning method,"a neural network learning method firstly performs finding a linear operator which satisfies required conditions suited to the function for learning by using linear prediction, where the function characterizes the neural network of the function approximation type, and operating the linear operator on the teaching patterns which approximated by a linear sum of the function, and determining results of the operation, and secondly it performs replacing learning of a neural network with a linear predictive problem according to the operation result firstly obtained by the first, and for determining a solution with respect to the linear predictive problem, and thirdly it performs a determination of weights from the hidden layer to output layer so that an error between output data from the output layer and the teaching patterns is minimized, according to the solution with respect to the linear predictive problem.",1996-06-18,"The title of the patent is neural network learning apparatus and learning method and its abstract is a neural network learning method firstly performs finding a linear operator which satisfies required conditions suited to the function for learning by using linear prediction, where the function characterizes the neural network of the function approximation type, and operating the linear operator on the teaching patterns which approximated by a linear sum of the function, and determining results of the operation, and secondly it performs replacing learning of a neural network with a linear predictive problem according to the operation result firstly obtained by the first, and for determining a solution with respect to the linear predictive problem, and thirdly it performs a determination of weights from the hidden layer to output layer so that an error between output data from the output layer and the teaching patterns is minimized, according to the solution with respect to the linear predictive problem. dated 1996-06-18"
5529267,railway structure hazard predictor,"a hazard predictor that processes both rail and superstructure measurements to predict some potentially hazardous conditions on a railway structure. measurement is collected in real time with the aid of fiber optic sensor based linear array mesh, and processed with a neural network. sensors placed under the rail and sensors placed laterally of the rail provide data collection in real time both during occupied and unoccupied periods. in some embodiments the measurement data is compressed into two signatures which can be represented as two vectors. the collinearity of the vectors and the angle between the vectors are utilized to interpret the data as to track conditions. the angle between the descriptors can be used to predict the severity of degradation of the structure. the predictor can be used to manage maintenance of the structure and interface with existing railway signalling equipment to provide traffic management.",1996-06-25,"The title of the patent is railway structure hazard predictor and its abstract is a hazard predictor that processes both rail and superstructure measurements to predict some potentially hazardous conditions on a railway structure. measurement is collected in real time with the aid of fiber optic sensor based linear array mesh, and processed with a neural network. sensors placed under the rail and sensors placed laterally of the rail provide data collection in real time both during occupied and unoccupied periods. in some embodiments the measurement data is compressed into two signatures which can be represented as two vectors. the collinearity of the vectors and the angle between the vectors are utilized to interpret the data as to track conditions. the angle between the descriptors can be used to predict the severity of degradation of the structure. the predictor can be used to manage maintenance of the structure and interface with existing railway signalling equipment to provide traffic management. dated 1996-06-25"
5530243,formation density well logging tool with detector array for compensation of wellbore roughness and tool tilt,"a method for determining density of a formation. the method includes irradiating the formation with gamma rays having energy consistent with compton scattering. gamma rays are measured at axially spaced apart locations. two of the axially spaced apart locations are at an equal distance and in opposite directions relative to the source of gamma rays. the distance is smaller than the spacing of another one of the spaced apart locations. an apparent density for each spaced apart location is determined from the counts. differences are calculated in apparent density between each one of the spaced apart locations, and a correction factor is determined for apparent density at each spaced apart location, thereby determining the density of the earth formation. in a preferred embodiment, determining correction factors is performed by a neural network using the differences in apparent density as an input vector.",1996-06-25,"The title of the patent is formation density well logging tool with detector array for compensation of wellbore roughness and tool tilt and its abstract is a method for determining density of a formation. the method includes irradiating the formation with gamma rays having energy consistent with compton scattering. gamma rays are measured at axially spaced apart locations. two of the axially spaced apart locations are at an equal distance and in opposite directions relative to the source of gamma rays. the distance is smaller than the spacing of another one of the spaced apart locations. an apparent density for each spaced apart location is determined from the counts. differences are calculated in apparent density between each one of the spaced apart locations, and a correction factor is determined for apparent density at each spaced apart location, thereby determining the density of the earth formation. in a preferred embodiment, determining correction factors is performed by a neural network using the differences in apparent density as an input vector. dated 1996-06-25"
5530275,semiconductor device for summing weighted input signals,"the invention relates to a semiconductor device with which input signals can be weighted and the weighted input signals can be summed, and which in conjunction with a neuron can be used, for example, as a synapse in a neural network. the device comprises a number of switched capacitances with a common capacitor plate formed by a surface region 3 in a p-type substrate 1. the region 3 is connected to the inverting input of an amplifier 11 whose +input is connected to a reference voltage and whose output 12 supplies the summed output signal. the output 12 can be fed back to the input 3 via switch s. the other plate of the capacitances is formed by an electrode 6a, 6b, 6c, which can be switched between a reference voltage and an input source. the weight factors are stored in the form of electric charges on a floating gate 5a, 5b, 5c, which is provided between each input electrode 6 and the surface region 3. during operation, the input signals are each converted into a depletion charge in the surface region 3 whose value is dependent not only on the input signal but also on the charge on the associated floating gate. the sum of the depletion charges is subsequently stored in an output capacitance 13,3 and read out by means of the amplifier 11.",1996-06-25,"The title of the patent is semiconductor device for summing weighted input signals and its abstract is the invention relates to a semiconductor device with which input signals can be weighted and the weighted input signals can be summed, and which in conjunction with a neuron can be used, for example, as a synapse in a neural network. the device comprises a number of switched capacitances with a common capacitor plate formed by a surface region 3 in a p-type substrate 1. the region 3 is connected to the inverting input of an amplifier 11 whose +input is connected to a reference voltage and whose output 12 supplies the summed output signal. the output 12 can be fed back to the input 3 via switch s. the other plate of the capacitances is formed by an electrode 6a, 6b, 6c, which can be switched between a reference voltage and an input source. the weight factors are stored in the form of electric charges on a floating gate 5a, 5b, 5c, which is provided between each input electrode 6 and the surface region 3. during operation, the input signals are each converted into a depletion charge in the surface region 3 whose value is dependent not only on the input signal but also on the charge on the associated floating gate. the sum of the depletion charges is subsequently stored in an output capacitance 13,3 and read out by means of the amplifier 11. dated 1996-06-25"
5532950,dynamic digital filter using neural networks,""" the present invention provides an apparatus for decoding and classifying a digital audio input signal and for reconstructing the digital audio input signal, so that when the reconstructed signal is converted to an analog signal by a digital to analog converter (""""dac""""), the analog signal can drive a preamplifier, power amplifier or speakers directly. in particular, the present invention proposes a digital filter than can be adapted to have appropriate filtering characteristics based on the signal being filtered. the invention uses a neural network to adjust coefficients of a digital filter, depending on whether the digital audio input signal is more periodic or more aperiodic. if the digital audio input signal is more periodic, the coefficients will configure the digital filter so that the filter has the characteristics of an analog brickwall filter. whereas if the digital audio input signal is more aperiodic, the coefficients produced by the neural network will configure the digital filter to have more characteristics of an interpolation filter. the neural network is trained to recognize certain periodic and aperiodic signals and to produce digital filter parameters, preferably polynomial coefficients, correspondingly. the coefficients are selected to respond to the pure or blended periodic and aperiodic features of certain archetypal input signals. """,1996-07-02,"The title of the patent is dynamic digital filter using neural networks and its abstract is "" the present invention provides an apparatus for decoding and classifying a digital audio input signal and for reconstructing the digital audio input signal, so that when the reconstructed signal is converted to an analog signal by a digital to analog converter (""""dac""""), the analog signal can drive a preamplifier, power amplifier or speakers directly. in particular, the present invention proposes a digital filter than can be adapted to have appropriate filtering characteristics based on the signal being filtered. the invention uses a neural network to adjust coefficients of a digital filter, depending on whether the digital audio input signal is more periodic or more aperiodic. if the digital audio input signal is more periodic, the coefficients will configure the digital filter so that the filter has the characteristics of an analog brickwall filter. whereas if the digital audio input signal is more aperiodic, the coefficients produced by the neural network will configure the digital filter to have more characteristics of an interpolation filter. the neural network is trained to recognize certain periodic and aperiodic signals and to produce digital filter parameters, preferably polynomial coefficients, correspondingly. the coefficients are selected to respond to the pure or blended periodic and aperiodic features of certain archetypal input signals. "" dated 1996-07-02"
5533169,neural network system having dynamically reconfigurable connections,"a neural network system is provided that includes a network of neural operators pilot controlled by a control unit. the activities calculated by the operators are memorized in associated memory spaces that are addressable by an activity address. to facilitate reconfiguring the network, the memory space of at least one portion of the operators contains the activity addresses of other operators of the network. hence, the input activities of these operators may be issued from any other operators in a way that is modifiable by simply changing the values of the activity addresses. the invention is particularly suited to image and sound analysis and synthesis.",1996-07-02,"The title of the patent is neural network system having dynamically reconfigurable connections and its abstract is a neural network system is provided that includes a network of neural operators pilot controlled by a control unit. the activities calculated by the operators are memorized in associated memory spaces that are addressable by an activity address. to facilitate reconfiguring the network, the memory space of at least one portion of the operators contains the activity addresses of other operators of the network. hence, the input activities of these operators may be issued from any other operators in a way that is modifiable by simply changing the values of the activity addresses. the invention is particularly suited to image and sound analysis and synthesis. dated 1996-07-02"
5533383,integrated acoustic leak detection processing system,"a system for mapping absolute acoustic noise intensity in a three-dimensional acoustic noise field, and using three-dimensional absolute noise intensities to infer operational or performance characteristics of components or structures within the monitored field. localized noise sources are extracted using a distant array of transducers, and the absolute intensity can be measured even when totally masked by background noise at the transducer locations. the system includes an integrated sensor installation; a neural network detection system algorithm; a zoom system to precisely examine a small region in the steam generator vessel; a fuzzy logic system detection algorithm; and an expert detection system. the signal processing subsystem operates at three different levels of detection. at the top level of detection a trained neural network system monitors the vessel length for any indication of a leak. the proposed approach uses discriminators which do not require any beamforming of the sensor signals. the presence of a leak and the general location in the vessel is determined using a fuzzy logic expert system. the leak indications are passed to a second level of detection: a beamformed system which will monitor the indicated portion of the vessel. a third level of detection systematically monitors the tagged locations to determine if a leak is really present, remaining on the specific locations to detect a leak above the defined threshold to a high degree of accuracy. the actions to be taken will be controlled by a second fuzzy logic expert interface controller.",1996-07-09,"The title of the patent is integrated acoustic leak detection processing system and its abstract is a system for mapping absolute acoustic noise intensity in a three-dimensional acoustic noise field, and using three-dimensional absolute noise intensities to infer operational or performance characteristics of components or structures within the monitored field. localized noise sources are extracted using a distant array of transducers, and the absolute intensity can be measured even when totally masked by background noise at the transducer locations. the system includes an integrated sensor installation; a neural network detection system algorithm; a zoom system to precisely examine a small region in the steam generator vessel; a fuzzy logic system detection algorithm; and an expert detection system. the signal processing subsystem operates at three different levels of detection. at the top level of detection a trained neural network system monitors the vessel length for any indication of a leak. the proposed approach uses discriminators which do not require any beamforming of the sensor signals. the presence of a leak and the general location in the vessel is determined using a fuzzy logic expert system. the leak indications are passed to a second level of detection: a beamformed system which will monitor the indicated portion of the vessel. a third level of detection systematically monitors the tagged locations to determine if a leak is really present, remaining on the specific locations to detect a leak above the defined threshold to a high degree of accuracy. the actions to be taken will be controlled by a second fuzzy logic expert interface controller. dated 1996-07-09"
5533519,method and apparatus for diagnosing joints,"a method and apparatus for diagnosing joints based on sensed joint vibrations. accelerometers disposed on the skin adjacent to the joint detect vibrational patterns during movement of the joint. these patterns are then processed by one processor to generate a predetermined set of data parameters descriptive of the vibration pattern. also, the position and velocity of the joint during the vibration is recorded. this information from numerous patients with known joint conditions is used to train a adaptive interpreter, such as a neural network, to produce an output in response to these inputs which is indicative of the known joint condition. once trained, the adaptive interpreter can then interpret this set of parameters for an unknown joint to generate a fast and reliable diagnosis. the result is a non-subjective joint disorder classification system that can be utilized by persons without particular expertise in analyzing joint vibrational patterns.",1996-07-09,"The title of the patent is method and apparatus for diagnosing joints and its abstract is a method and apparatus for diagnosing joints based on sensed joint vibrations. accelerometers disposed on the skin adjacent to the joint detect vibrational patterns during movement of the joint. these patterns are then processed by one processor to generate a predetermined set of data parameters descriptive of the vibration pattern. also, the position and velocity of the joint during the vibration is recorded. this information from numerous patients with known joint conditions is used to train a adaptive interpreter, such as a neural network, to produce an output in response to these inputs which is indicative of the known joint condition. once trained, the adaptive interpreter can then interpret this set of parameters for an unknown joint to generate a fast and reliable diagnosis. the result is a non-subjective joint disorder classification system that can be utilized by persons without particular expertise in analyzing joint vibrational patterns. dated 1996-07-09"
5535148,method and apparatus for approximating a sigmoidal response using digital circuitry,"a method and apparatus for approximating a sigmoidal response using digital circuitry for neural network computations. the digital circuitry in processing element (16) produces the neuron output signal (110) by performing a squashing operation which determines an approximation of a sigmoid function. in one form, the present invention uses digital circuitry (16) in data processor (10) to approximate a sigmoid function of a neuron (100) using a plurality of parabolas. in an alternate embodiment, the sigmoid function of neuron (100) is approximated using a quasi-log.sub.2 function.",1996-07-09,"The title of the patent is method and apparatus for approximating a sigmoidal response using digital circuitry and its abstract is a method and apparatus for approximating a sigmoidal response using digital circuitry for neural network computations. the digital circuitry in processing element (16) produces the neuron output signal (110) by performing a squashing operation which determines an approximation of a sigmoid function. in one form, the present invention uses digital circuitry (16) in data processor (10) to approximate a sigmoid function of a neuron (100) using a plurality of parabolas. in an alternate embodiment, the sigmoid function of neuron (100) is approximated using a quasi-log.sub.2 function. dated 1996-07-09"
5535303,"""""""barometer"""" neuron for a neural network""",""" a """"barometer"""" neuron enhances stability in a neural network system that, when used as a track-while-scan system, assigns sensor plots to predicted track positions in a plot/track association situation. the """"barometer"""" neuron functions as a bench-mark or reference system node that equates a superimposed plot and track to a zero distance as a """"perfect"""" pairing of plot and track which has a measured/desired level of inhibition. the """"barometer"""" neuron responds to the system inputs, compares these inputs against the level of inhibition of the """"perfect"""" pair, and generates a supplied excitation or inhibition output signal to the system which adjusts the system to a desired value at or near 1.0; this the reference level of inhibition of the """"perfect"""" pair. """,1996-07-09,"The title of the patent is """"""barometer"""" neuron for a neural network"" and its abstract is "" a """"barometer"""" neuron enhances stability in a neural network system that, when used as a track-while-scan system, assigns sensor plots to predicted track positions in a plot/track association situation. the """"barometer"""" neuron functions as a bench-mark or reference system node that equates a superimposed plot and track to a zero distance as a """"perfect"""" pairing of plot and track which has a measured/desired level of inhibition. the """"barometer"""" neuron responds to the system inputs, compares these inputs against the level of inhibition of the """"perfect"""" pair, and generates a supplied excitation or inhibition output signal to the system which adjusts the system to a desired value at or near 1.0; this the reference level of inhibition of the """"perfect"""" pair. "" dated 1996-07-09"
5535309,single layer neural network circuit for performing linearly separable and non-linearly separable logical operations,"a neural network provides both linearly separable and non-linearly separable logic operations, including the exclusive-or operation, on input signals in a single layer of circuits. the circuit weights the input signals with complex weights by multiplication and addition, and provides weighted signals to a neuron circuit (a neuron body or some a) which provides an output corresponding to the desired logical operation.",1996-07-09,"The title of the patent is single layer neural network circuit for performing linearly separable and non-linearly separable logical operations and its abstract is a neural network provides both linearly separable and non-linearly separable logic operations, including the exclusive-or operation, on input signals in a single layer of circuits. the circuit weights the input signals with complex weights by multiplication and addition, and provides weighted signals to a neuron circuit (a neuron body or some a) which provides an output corresponding to the desired logical operation. dated 1996-07-09"
5536938,pulsed neutron decay logging,"a borehole logging tool having a pair of spaced-apart detectors records intensity signals representing the die-away of nuclear radiation in a subsurface formation. weighted moments of the intensity signals as well as of a model are produced. corresponding weighted intensity and model moments are equated and simultaneously solved to obtain values for a borehole decay constant, a formation decay constant, and a formation-to-borehole amplitude ratio for each of the pair of detectors. from these values a trained neural network produces intrinsic values of a formation macroscopic thermal neutron absorption cross section, a formation porosity, and a borehole fluid cross section. a log is generated of these intrinsic values versus depth as the logging tool traverses the borehole.",1996-07-16,"The title of the patent is pulsed neutron decay logging and its abstract is a borehole logging tool having a pair of spaced-apart detectors records intensity signals representing the die-away of nuclear radiation in a subsurface formation. weighted moments of the intensity signals as well as of a model are produced. corresponding weighted intensity and model moments are equated and simultaneously solved to obtain values for a borehole decay constant, a formation decay constant, and a formation-to-borehole amplitude ratio for each of the pair of detectors. from these values a trained neural network produces intrinsic values of a formation macroscopic thermal neutron absorption cross section, a formation porosity, and a borehole fluid cross section. a log is generated of these intrinsic values versus depth as the logging tool traverses the borehole. dated 1996-07-16"
5537094,method and apparatus for detecting an eas marker using a neural network processing device,a signal received by an electronic article surveillance system is processed using a neural network processing algorithm to determine whether an electronic surveillance marker of a predetermined kind is present. the raw received signal is processed to generate a small number of parameter values to be provided as inputs for the neural network. the neural network processing distinguishes between two different types of surveillance marker.,1996-07-16,The title of the patent is method and apparatus for detecting an eas marker using a neural network processing device and its abstract is a signal received by an electronic article surveillance system is processed using a neural network processing algorithm to determine whether an electronic surveillance marker of a predetermined kind is present. the raw received signal is processed to generate a small number of parameter values to be provided as inputs for the neural network. the neural network processing distinguishes between two different types of surveillance marker. dated 1996-07-16
5537327,method and apparatus for detecting high-impedance faults in electrical power systems,"the present invention features a method and apparatus for detecting and enabling the clearance of high impedance faults (hifs) in an electrical transmission or distribution system. current in at least one phase in a distribution system is monitored in real time by sensors. analog current signature information is then digitized for processing by a digital computer. zero crossings are identified and current maxima and minima located. the first derivatives of the maxima and minima are computed and a modified fast fourier transform (fft) is then performed to convert time domain to frequency domain information. the transformed data is formatted and normalized and then applied to a trained neural network, which provides an output trigger signal when an hif condition is probable. the trigger signal is made available to either a network administrator for manual intervention, or directly to switchgear to deactivate an affected portion of the network. the inventive method may be practiced using either conventional computer hardware and software or dedicated custom hardware such as a vlsi chip.",1996-07-16,"The title of the patent is method and apparatus for detecting high-impedance faults in electrical power systems and its abstract is the present invention features a method and apparatus for detecting and enabling the clearance of high impedance faults (hifs) in an electrical transmission or distribution system. current in at least one phase in a distribution system is monitored in real time by sensors. analog current signature information is then digitized for processing by a digital computer. zero crossings are identified and current maxima and minima located. the first derivatives of the maxima and minima are computed and a modified fast fourier transform (fft) is then performed to convert time domain to frequency domain information. the transformed data is formatted and normalized and then applied to a trained neural network, which provides an output trigger signal when an hif condition is probable. the trigger signal is made available to either a network administrator for manual intervention, or directly to switchgear to deactivate an affected portion of the network. the inventive method may be practiced using either conventional computer hardware and software or dedicated custom hardware such as a vlsi chip. dated 1996-07-16"
5537511,neural network based data fusion system for source localization,"a method is described for providing an estimate of the state of a moving contact. the method comprises providing a device for estimating the state of the contact, inputting information about a location of an observer platform at particular time intervals and information from at least one sensor about a position of the moving contact relative to the observer platform at each time interval into the device, transforming the inputted information into a series of geographical grids with one grid being formed for each reading of the at least one sensor; combining grids corresponding to similar time intervals into a series of consolidated grid representations; and analyzing the series of consolidated grid representations to produce an estimate of the state of the contact at a final point in time where an observation was made. the device of the present invention includes a grid stimulation block for forming the geographical grids, a fusion block for forming the consolidated grid representations, a correlation block for providing a path likelihood vector, and an estimation block for providing the desired estimate.",1996-07-16,"The title of the patent is neural network based data fusion system for source localization and its abstract is a method is described for providing an estimate of the state of a moving contact. the method comprises providing a device for estimating the state of the contact, inputting information about a location of an observer platform at particular time intervals and information from at least one sensor about a position of the moving contact relative to the observer platform at each time interval into the device, transforming the inputted information into a series of geographical grids with one grid being formed for each reading of the at least one sensor; combining grids corresponding to similar time intervals into a series of consolidated grid representations; and analyzing the series of consolidated grid representations to produce an estimate of the state of the contact at a final point in time where an observation was made. the device of the present invention includes a grid stimulation block for forming the geographical grids, a fusion block for forming the consolidated grid representations, a correlation block for providing a path likelihood vector, and an estimation block for providing the desired estimate. dated 1996-07-16"
5537512,neural network elements,"an analog neural network element includes one or more eeproms as analog, reprogrammable synapses applying weighted inputs to positive and negative term outputs which are combined in a comparator. in one embodiment a pair of eeproms is used in each synaptic connection to separately drive the positive and negative term outputs. in another embodiment, a single eeprom is used as a programmable current source to control the operation of a differential amplifier driving the positive and negative term outputs. in a still further embodiment, an mnos memory transistor replaces the eeprom or eeproms. these memory elements have limited retention or endurance which is used to simulate forgetfulness to emulate human brain function. multiple elements are combinable on a single chip to form neural net building blocks which are then combinable to form massively parallel neural nets.",1996-07-16,"The title of the patent is neural network elements and its abstract is an analog neural network element includes one or more eeproms as analog, reprogrammable synapses applying weighted inputs to positive and negative term outputs which are combined in a comparator. in one embodiment a pair of eeproms is used in each synaptic connection to separately drive the positive and negative term outputs. in another embodiment, a single eeprom is used as a programmable current source to control the operation of a differential amplifier driving the positive and negative term outputs. in a still further embodiment, an mnos memory transistor replaces the eeprom or eeproms. these memory elements have limited retention or endurance which is used to simulate forgetfulness to emulate human brain function. multiple elements are combinable on a single chip to form neural net building blocks which are then combinable to form massively parallel neural nets. dated 1996-07-16"
5538915,process for forming synapses in neural networks and resistor therefor,"customizable neural network in which one or more resistors form each synapse. all the resistors in the synaptic array are identical, thus simplifying the processing issues. highly doped, amorphous silicon is used as the resistor material, to create extremely high resistances occupying very small spaces. connected in series with each resistor in the array is at least one severable conductor whose uppermost layer has a lower reflectivity of laser energy than typical metal conductors at a desired laser wavelength.",1996-07-23,"The title of the patent is process for forming synapses in neural networks and resistor therefor and its abstract is customizable neural network in which one or more resistors form each synapse. all the resistors in the synaptic array are identical, thus simplifying the processing issues. highly doped, amorphous silicon is used as the resistor material, to create extremely high resistances occupying very small spaces. connected in series with each resistor in the array is at least one severable conductor whose uppermost layer has a lower reflectivity of laser energy than typical metal conductors at a desired laser wavelength. dated 1996-07-23"
5541417,quantative agglutination reaction analysis method,"embodiments described herein provide methods of analysis of a reaction, such as an agglutination reaction producing an agglutination pattern, and the like. in one method, an image of the agglutination pattern is captured. the image captured is digitized to form a digital image comprising a pixel. roughness is measured, reflecting pixel local environment, at each pixel comprising a region of interest of the digital image of the agglutination pattern. at least one of classification and quantification of the image captured of the agglutination pattern is performed based on the measured roughness. in another embodiment, roughness is derived using a shared weights neural network approach, the roughness reflecting pixel local environment at each pixel comprising a region of interest of the digital image of the agglutination pattern. at least one of classification and quantification of the image captured of the agglutination pattern is performed based on the derived roughness. in a further embodiment, a feature is derived using a shared weights neural network approach, the feature reflecting pixel local environment at each pixel comprising a region of interest of the digital image of the agglutination pattern. at least one of classification and quantification of the image captured of the agglutination pattern is performed based on the derived feature.",1996-07-30,"The title of the patent is quantative agglutination reaction analysis method and its abstract is embodiments described herein provide methods of analysis of a reaction, such as an agglutination reaction producing an agglutination pattern, and the like. in one method, an image of the agglutination pattern is captured. the image captured is digitized to form a digital image comprising a pixel. roughness is measured, reflecting pixel local environment, at each pixel comprising a region of interest of the digital image of the agglutination pattern. at least one of classification and quantification of the image captured of the agglutination pattern is performed based on the measured roughness. in another embodiment, roughness is derived using a shared weights neural network approach, the roughness reflecting pixel local environment at each pixel comprising a region of interest of the digital image of the agglutination pattern. at least one of classification and quantification of the image captured of the agglutination pattern is performed based on the derived roughness. in a further embodiment, a feature is derived using a shared weights neural network approach, the feature reflecting pixel local environment at each pixel comprising a region of interest of the digital image of the agglutination pattern. at least one of classification and quantification of the image captured of the agglutination pattern is performed based on the derived feature. dated 1996-07-30"
5541590,vehicle crash predictive and evasive operation system by neural networks,"a system for predicting and evading crash of a vehicle includes an image pick-up device mounted on the vehicle for picking up images of actual ever-changing views when the vehicle is on running to produce actual image data, a crash predicting device associated with said image pick-up device, said crash predicting device being successively supplied with the actual image data for predicting occurrence of crash between the vehicle and potentially dangerous objects on the roadway to produce an operational signal when there is possibility of crash and a safety drive ensuring device connected to said crash predicting device for actuating, in response to the operational signal, an occupant protecting mechanism which is operatively connected thereto and equipped in the vehicle. the crash predicting device includes a neural network which is previously trained with training data to predict the possibility of crash, the training data representing ever-changing views previously picked-up from said image picking-up device during driving of the vehicle for causing actual crash.",1996-07-30,"The title of the patent is vehicle crash predictive and evasive operation system by neural networks and its abstract is a system for predicting and evading crash of a vehicle includes an image pick-up device mounted on the vehicle for picking up images of actual ever-changing views when the vehicle is on running to produce actual image data, a crash predicting device associated with said image pick-up device, said crash predicting device being successively supplied with the actual image data for predicting occurrence of crash between the vehicle and potentially dangerous objects on the roadway to produce an operational signal when there is possibility of crash and a safety drive ensuring device connected to said crash predicting device for actuating, in response to the operational signal, an occupant protecting mechanism which is operatively connected thereto and equipped in the vehicle. the crash predicting device includes a neural network which is previously trained with training data to predict the possibility of crash, the training data representing ever-changing views previously picked-up from said image picking-up device during driving of the vehicle for causing actual crash. dated 1996-07-30"
5542006,neural network based character position detector for use in optical character recognition,""" apparatus, and an accompanying method, for use in an optical character recognition (ocr) system (5) for locating, e.g., center positions (""""hearts"""") of all desired characters within a field (310; 510) of characters such that the desired characters can be subsequently recognized using an appropriate classification process. specifically, a window (520) is slid in a step-wise convolutional-like fashion (520.sub.1, 520.sub.2, 520.sub.3) across a field of preprocessed, specifically uniformly scaled, characters. each pixel in the window is applied as an input to a positioning neural network (152) that has been trained to produce an output activation whenever a character """"heart"""" is spatially coincident with a pixel position within an array (430) centrally located within the window. as the window is successively moved across the field, in a stepped fashion, the activation outputs of the neural network are averaged, on a weighted basis, for each different window position and separately for each horizontal pixel position in the field. the resulting averaged activation output values, typically in the form of a gaussian distribution for each character, are then filtered, thresholded and then used, via a weighted average calculation with horizontal pixel positions being used as the weights, to determine the character """"heart"""" position as being the center pixel position in the distribution. """,1996-07-30,"The title of the patent is neural network based character position detector for use in optical character recognition and its abstract is "" apparatus, and an accompanying method, for use in an optical character recognition (ocr) system (5) for locating, e.g., center positions (""""hearts"""") of all desired characters within a field (310; 510) of characters such that the desired characters can be subsequently recognized using an appropriate classification process. specifically, a window (520) is slid in a step-wise convolutional-like fashion (520.sub.1, 520.sub.2, 520.sub.3) across a field of preprocessed, specifically uniformly scaled, characters. each pixel in the window is applied as an input to a positioning neural network (152) that has been trained to produce an output activation whenever a character """"heart"""" is spatially coincident with a pixel position within an array (430) centrally located within the window. as the window is successively moved across the field, in a stepped fashion, the activation outputs of the neural network are averaged, on a weighted basis, for each different window position and separately for each horizontal pixel position in the field. the resulting averaged activation output values, typically in the form of a gaussian distribution for each character, are then filtered, thresholded and then used, via a weighted average calculation with horizontal pixel positions being used as the weights, to determine the character """"heart"""" position as being the center pixel position in the distribution. "" dated 1996-07-30"
5542026,triangular scalable neural array processor,"a triangular scalable neural array processor unit for use in a neural network has an array of weight registers, multipliers, communicating adder trees, sigmoid generators, and a reverse feedback loop for communicating the output of a sigmoid generator back to input multipliers of selected neurons. the communicating adder trees provide the selectable feedback path.",1996-07-30,"The title of the patent is triangular scalable neural array processor and its abstract is a triangular scalable neural array processor unit for use in a neural network has an array of weight registers, multipliers, communicating adder trees, sigmoid generators, and a reverse feedback loop for communicating the output of a sigmoid generator back to input multipliers of selected neurons. the communicating adder trees provide the selectable feedback path. dated 1996-07-30"
5542054,artificial neurons using delta-sigma modulation,"an artificial neuron for use in a neural processing network comprises a plurality of input signal lines, an arrangement for computing a nonlinear function of the sum of the inputs multiplied by associated weights, and a saturating delta-sigma modulator which oversamples the computed value and produces an encoded neuron output signal. conversion of signals for use by these neurons preferably is performed by delta-sigma modulators at the inputs to the neurons, which may be incorporated directly into sensors. processing of the output signals from the neuron includes low-pass filtering and decimation. the present invention may be used in many diverse areas. for example, arrays of sensors with delta signal modulators may be coupled with a network of the neurons to form an intelligent vision system. linear signal processing, both conventional and adaptive, can be done by a simple neuronal system that operates linearly.",1996-07-30,"The title of the patent is artificial neurons using delta-sigma modulation and its abstract is an artificial neuron for use in a neural processing network comprises a plurality of input signal lines, an arrangement for computing a nonlinear function of the sum of the inputs multiplied by associated weights, and a saturating delta-sigma modulator which oversamples the computed value and produces an encoded neuron output signal. conversion of signals for use by these neurons preferably is performed by delta-sigma modulators at the inputs to the neurons, which may be incorporated directly into sensors. processing of the output signals from the neuron includes low-pass filtering and decimation. the present invention may be used in many diverse areas. for example, arrays of sensors with delta signal modulators may be coupled with a network of the neurons to form an intelligent vision system. linear signal processing, both conventional and adaptive, can be done by a simple neuronal system that operates linearly. dated 1996-07-30"
5542430,apparatus and method for discriminating between cardiac rhythms on the basis of their morphology using a neural network,"an apparatus and a method are provided for coupling to a patient's heart for discriminating between tachycardias of physiological origin, and those of pathological origin having similar rates; and also for discriminating amongst those of pathological origin having similar rates. the apparatus includes transducers and/or sensing electrodes in either or both the atrium and/or ventricle. also included are signal processing elements for determining the times of atrial and ventricular events and for extracting morphological features from the waveforms, and a neural network for classifying the heart rhythm. the method includes a step of discriminating between different types of heart rhythms having overlapping rates. the method utilizes atrial-atrial, ventricular-ventricular and atrio-ventricular intervals; integrated waveforms; sums of differences of waveform samples; rectified integrated bandpass filtered waveforms; numbers of zero crossings in the electrogram; area under the ventricular electrogram; and r wave slope, qr area and rs area of the electrogram.",1996-08-06,"The title of the patent is apparatus and method for discriminating between cardiac rhythms on the basis of their morphology using a neural network and its abstract is an apparatus and a method are provided for coupling to a patient's heart for discriminating between tachycardias of physiological origin, and those of pathological origin having similar rates; and also for discriminating amongst those of pathological origin having similar rates. the apparatus includes transducers and/or sensing electrodes in either or both the atrium and/or ventricle. also included are signal processing elements for determining the times of atrial and ventricular events and for extracting morphological features from the waveforms, and a neural network for classifying the heart rhythm. the method includes a step of discriminating between different types of heart rhythms having overlapping rates. the method utilizes atrial-atrial, ventricular-ventricular and atrio-ventricular intervals; integrated waveforms; sums of differences of waveform samples; rectified integrated bandpass filtered waveforms; numbers of zero crossings in the electrogram; area under the ventricular electrogram; and r wave slope, qr area and rs area of the electrogram. dated 1996-08-06"
5544280,unipolar terminal-attractor based neural associative memory with adaptive threshold,"a unipolar terminal-attractor based neural associative memory (tabam) system with adaptive threshold for perfect convergence is presented. by adaptively setting the threshold values for the dynamic iteration for the unipolar binary neuron states with terminal-attractors for the purpose of reducing the spurious states in a hopfield neural network for associative memory and using the inner-product approach, perfect convergence and correct retrieval is achieved. simulation is completed with a small number of stored states (m) and a small number of neurons (n) but a large m/n ratio. an experiment with optical exclusive-or logic operation using lctv slms shows the feasibility of optoelectronic implementation of the models. a complete inner-product tabam is implemented using a pc for calculation of adaptive threshold values to achieve a unipolar tabam (uit) in the case where there is no crosstalk, and a crosstalk model (crit) in the case where crosstalk corrupts the desired state.",1996-08-06,"The title of the patent is unipolar terminal-attractor based neural associative memory with adaptive threshold and its abstract is a unipolar terminal-attractor based neural associative memory (tabam) system with adaptive threshold for perfect convergence is presented. by adaptively setting the threshold values for the dynamic iteration for the unipolar binary neuron states with terminal-attractors for the purpose of reducing the spurious states in a hopfield neural network for associative memory and using the inner-product approach, perfect convergence and correct retrieval is achieved. simulation is completed with a small number of stored states (m) and a small number of neurons (n) but a large m/n ratio. an experiment with optical exclusive-or logic operation using lctv slms shows the feasibility of optoelectronic implementation of the models. a complete inner-product tabam is implemented using a pc for calculation of adaptive threshold values to achieve a unipolar tabam (uit) in the case where there is no crosstalk, and a crosstalk model (crit) in the case where crosstalk corrupts the desired state. dated 1996-08-06"
5546195,apparatus for reproducing color images,"in a reproduced color correction system, a photometer measures external illuminant light and sends a measured result to a workstation. the workstation determines the type of an illuminant on the basis of the measured result and sends a selection signal and corresponding colorimetric value data to a neural network management unit. the neural network management unit is provided with a plurality of neural networks associated with types of illuminants. the neural network selected according to the selection signal converts the colorimetric value to a color separation value according to a post-learning transformation function. a color printing device outputs a color image on the basis of the generated color separation value. thus, the neural network associated with the illuminant used for observation reference is selected, and appropriate color transformation is performed. accordingly, even if the viewing condition is changed, color matching can be performed so that observed reproduced colors are unchanged.",1996-08-13,"The title of the patent is apparatus for reproducing color images and its abstract is in a reproduced color correction system, a photometer measures external illuminant light and sends a measured result to a workstation. the workstation determines the type of an illuminant on the basis of the measured result and sends a selection signal and corresponding colorimetric value data to a neural network management unit. the neural network management unit is provided with a plurality of neural networks associated with types of illuminants. the neural network selected according to the selection signal converts the colorimetric value to a color separation value according to a post-learning transformation function. a color printing device outputs a color image on the basis of the generated color separation value. thus, the neural network associated with the illuminant used for observation reference is selected, and appropriate color transformation is performed. accordingly, even if the viewing condition is changed, color matching can be performed so that observed reproduced colors are unchanged. dated 1996-08-13"
5546503,apparatus for configuring neural network and pattern recognition apparatus using neural network,"in a neural network having neurons connected in a multi-layer, firstly, input signal sets are sequentially entered to statistically process the outputs of hidden neurons and determine the optimum number of hidden neurons. secondly, while changing the input signal entered to each input neuron to the maximum change limit, the change of output values of the other input neurons are checked to thereby determine an unnecessary input neuron. thirdly, the weights between input neurons and hidden neurons are set to be in correspondence with a hyperplane to enable pattern recognition.",1996-08-13,"The title of the patent is apparatus for configuring neural network and pattern recognition apparatus using neural network and its abstract is in a neural network having neurons connected in a multi-layer, firstly, input signal sets are sequentially entered to statistically process the outputs of hidden neurons and determine the optimum number of hidden neurons. secondly, while changing the input signal entered to each input neuron to the maximum change limit, the change of output values of the other input neurons are checked to thereby determine an unnecessary input neuron. thirdly, the weights between input neurons and hidden neurons are set to be in correspondence with a hyperplane to enable pattern recognition. dated 1996-08-13"
5546504,information processing device capable optically writing synapse strength matrix,"an information processing device having neural network functions for performing information processing includes a semiconductor integrated circuit section including a plurality of neuronic circuit regions having a neuronic function which is one of the neural network functions, and first and second molecular films provided on the integrated circuit section. the first molecular film has a photoelectric function and the second molecular film has a light-emitting function. coupling between the plurality of neurons is realized through a combination of the light-emitting and light-receiving functions of the first and second molecular films.",1996-08-13,"The title of the patent is information processing device capable optically writing synapse strength matrix and its abstract is an information processing device having neural network functions for performing information processing includes a semiconductor integrated circuit section including a plurality of neuronic circuit regions having a neuronic function which is one of the neural network functions, and first and second molecular films provided on the integrated circuit section. the first molecular film has a photoelectric function and the second molecular film has a light-emitting function. coupling between the plurality of neurons is realized through a combination of the light-emitting and light-receiving functions of the first and second molecular films. dated 1996-08-13"
5546505,program product for facilitating use of a neural network,"a neural network development utility assists a developer in generating one or more filters for data to be input to or output from a neural network. a filter is a device which translates data in accordance with a data transformation definition contained in a translate template. source data for the neural network may be expressed in any arbitrary combination of symbolic or numeric fields in a data base. the developer selects those fields to be used from an interactive menu. the utility scans the selected field entries in the source data base to identify the logical type of each field, and creates a default translate template based on this scan. numeric data is automatically scaled. the developer may use the default template, or edit it from an interactive editor. when editing the template, the developer may select from a menu of commonly used neural network data formats, and from a menu of commonly used primitive mathematical operations. the developer may interactively define additional filters to perform data transformations in series, thus achieving more complex mathematical operations on the data. templates may be edited at any time during the development process. if a network does not appear to be giving satisfactory results, the developer may easily alter the template to present inputs in some other format.",1996-08-13,"The title of the patent is program product for facilitating use of a neural network and its abstract is a neural network development utility assists a developer in generating one or more filters for data to be input to or output from a neural network. a filter is a device which translates data in accordance with a data transformation definition contained in a translate template. source data for the neural network may be expressed in any arbitrary combination of symbolic or numeric fields in a data base. the developer selects those fields to be used from an interactive menu. the utility scans the selected field entries in the source data base to identify the logical type of each field, and creates a default translate template based on this scan. numeric data is automatically scaled. the developer may use the default template, or edit it from an interactive editor. when editing the template, the developer may select from a menu of commonly used neural network data formats, and from a menu of commonly used primitive mathematical operations. the developer may interactively define additional filters to perform data transformations in series, thus achieving more complex mathematical operations on the data. templates may be edited at any time during the development process. if a network does not appear to be giving satisfactory results, the developer may easily alter the template to present inputs in some other format. dated 1996-08-13"
5548512,autonomous navigation apparatus with neural network for a mobile vehicle,"an autonomous navigation system for a mobile vehicle arranged to move within an environment includes a plurality of sensors arranged on the vehicle and at least one neural network including an input layer coupled to the sensors, a hidden layer coupled to the input layer, and an output layer coupled to the hidden layer. the neural network produces output signals representing respective positions of the vehicle, such as the x coordinate, the y coordinate, and the angular orientation of the vehicle. a plurality of patch locations within the environment are used to train the neural networks to produce the correct outputs in response to the distances sensed.",1996-08-20,"The title of the patent is autonomous navigation apparatus with neural network for a mobile vehicle and its abstract is an autonomous navigation system for a mobile vehicle arranged to move within an environment includes a plurality of sensors arranged on the vehicle and at least one neural network including an input layer coupled to the sensors, a hidden layer coupled to the input layer, and an output layer coupled to the hidden layer. the neural network produces output signals representing respective positions of the vehicle, such as the x coordinate, the y coordinate, and the angular orientation of the vehicle. a plurality of patch locations within the environment are used to train the neural networks to produce the correct outputs in response to the distances sensed. dated 1996-08-20"
5548662,edge extracting method and apparatus using diffusion neural network,"edge extracting method and apparatus using a neural network performing a function of diffusing an excitation. the method and apparatus continuously detect a variety of intensity changes of an image via a function having a variety of frequency characteristics. an edge of a fixed object is detected from images continuously input, and an edge of a moving object is selectively detected from the images. the edge extracting apparatus includes a first neural network which receives an image signal. the first neural network derives a gaussian function representing the regularity of an excitatory response and an inhibitory response to a spot excitation of the image signal. the apparatus also includes a second neural network which detects edges of an image represented by the image signal by convolving the gaussian function and the image signal.",1996-08-20,"The title of the patent is edge extracting method and apparatus using diffusion neural network and its abstract is edge extracting method and apparatus using a neural network performing a function of diffusing an excitation. the method and apparatus continuously detect a variety of intensity changes of an image via a function having a variety of frequency characteristics. an edge of a fixed object is detected from images continuously input, and an edge of a moving object is selectively detected from the images. the edge extracting apparatus includes a first neural network which receives an image signal. the first neural network derives a gaussian function representing the regularity of an excitatory response and an inhibitory response to a spot excitation of the image signal. the apparatus also includes a second neural network which detects edges of an image represented by the image signal by convolving the gaussian function and the image signal. dated 1996-08-20"
5548683,data fusion neural network,an information processing system and method forms a fast optimal or near optimal association based on satisfying global constraints expressed in an association matrix by simulating the behavior of a network of interconnected processing elements resembling neurons in a brain.,1996-08-20,The title of the patent is data fusion neural network and its abstract is an information processing system and method forms a fast optimal or near optimal association based on satisfying global constraints expressed in an association matrix by simulating the behavior of a network of interconnected processing elements resembling neurons in a brain. dated 1996-08-20
5548684,artificial neural network viterbi decoding system and method,"an artificial neural network (ann) decoding system decodes a convolutionally-encoded data stream at high speed and with high efficiency. the ann decoding system implements the viterbi algorithm and is significantly faster than comparable digital-only designs due to its fully parallel architecture. several modifications to the fully analog system are described, including an analog/digital hybrid design that results in an extremely fast and efficient viterbi decoding system. a complexity and analysis shows that the modified ann decoding system is much simpler and easier to implement than its fully digital counterpart. the structure of the ann decoding system of the invention provides a natural fit for vlsi implementation. simulation results show that the performance of the ann decoding system exactly matches that of an ideal viterbi decoding system.",1996-08-20,"The title of the patent is artificial neural network viterbi decoding system and method and its abstract is an artificial neural network (ann) decoding system decodes a convolutionally-encoded data stream at high speed and with high efficiency. the ann decoding system implements the viterbi algorithm and is significantly faster than comparable digital-only designs due to its fully parallel architecture. several modifications to the fully analog system are described, including an analog/digital hybrid design that results in an extremely fast and efficient viterbi decoding system. a complexity and analysis shows that the modified ann decoding system is much simpler and easier to implement than its fully digital counterpart. the structure of the ann decoding system of the invention provides a natural fit for vlsi implementation. simulation results show that the performance of the ann decoding system exactly matches that of an ideal viterbi decoding system. dated 1996-08-20"
5548697,non-linear color corrector having a neural network and using fuzzy membership values to correct color and a method thereof,a color corrector for changing pixels in an image where the color corrector includes a neural fuzzy classifier to generate a membership value which defines a degree of membership of each pixel in a group of pixels to be transformed. a pixel color changer is also provided to transform the pixel according to its membership in the group of pixels to be changed. the color corrector can also include a pixel group classifier for identifying groups of pixels in the image to train the neural fuzzy classifier to generate the membership value.,1996-08-20,The title of the patent is non-linear color corrector having a neural network and using fuzzy membership values to correct color and a method thereof and its abstract is a color corrector for changing pixels in an image where the color corrector includes a neural fuzzy classifier to generate a membership value which defines a degree of membership of each pixel in a group of pixels to be transformed. a pixel color changer is also provided to transform the pixel according to its membership in the group of pixels to be changed. the color corrector can also include a pixel group classifier for identifying groups of pixels in the image to train the neural fuzzy classifier to generate the membership value. dated 1996-08-20
5550951,metrics for specifying and/or testing neural networks,"a method for testing a neural network, and apparatus for carrying out the method. a first embodiment of the method includes the steps of (a) providing a neural network having a set of connection values, (b) stimulating at least one input of the neural network with an input vector to obtain output signals at an output of the neural network, (c) obtaining a plurality of samples of the output signals, wherein at least one of the plurality of samples is delayed in time from another one of the samples, and wherein at least one of the plurality of samples may represent a difference between two samples, (d) generating an image from the plurality of samples, and (e) comparing the image to a reference image to determine an operational characteristic of the neural network. the step of generating generates an image that is comprised of a plurality of points, wherein each of the points is referenced to an x-y coordinate system, wherein a distance along the x-axis is a function of a value of the output of the neural network for a given input vector applied to a first input of the neural network, and wherein a distance along the y-axis is function of a value of the output of the neural network for the given input vector applied a second input of the neural network.",1996-08-27,"The title of the patent is metrics for specifying and/or testing neural networks and its abstract is a method for testing a neural network, and apparatus for carrying out the method. a first embodiment of the method includes the steps of (a) providing a neural network having a set of connection values, (b) stimulating at least one input of the neural network with an input vector to obtain output signals at an output of the neural network, (c) obtaining a plurality of samples of the output signals, wherein at least one of the plurality of samples is delayed in time from another one of the samples, and wherein at least one of the plurality of samples may represent a difference between two samples, (d) generating an image from the plurality of samples, and (e) comparing the image to a reference image to determine an operational characteristic of the neural network. the step of generating generates an image that is comprised of a plurality of points, wherein each of the points is referenced to an x-y coordinate system, wherein a distance along the x-axis is a function of a value of the output of the neural network for a given input vector applied to a first input of the neural network, and wherein a distance along the y-axis is function of a value of the output of the neural network for the given input vector applied a second input of the neural network. dated 1996-08-27"
5553159,radiation image processing method utilizing neural networks,"in a radiation image processing method utilizing a neural network, an image signal representing a radiation image is fed into a neural network, image processing is carried out on the image signal by the neural network, and an output representing the results of the image processing is obtained from the neural network. image processing, with respect to the whole region of the radiation image, is carried out on the image signal by a first group of neurons of an intermediate layer of the neural network. image processing, with respect to parts of the region of the radiation image, is carried out on the image signal by a second group of neurons of the intermediate layer of the neural network. regardless of set values of initial conditions, the output of the neural network becomes converged to a global minimum corresponding to the stored information, and the results of operation obtained from the neural network are not trapped at a local minimum.",1996-09-03,"The title of the patent is radiation image processing method utilizing neural networks and its abstract is in a radiation image processing method utilizing a neural network, an image signal representing a radiation image is fed into a neural network, image processing is carried out on the image signal by the neural network, and an output representing the results of the image processing is obtained from the neural network. image processing, with respect to the whole region of the radiation image, is carried out on the image signal by a first group of neurons of an intermediate layer of the neural network. image processing, with respect to parts of the region of the radiation image, is carried out on the image signal by a second group of neurons of the intermediate layer of the neural network. regardless of set values of initial conditions, the output of the neural network becomes converged to a global minimum corresponding to the stored information, and the results of operation obtained from the neural network are not trapped at a local minimum. dated 1996-09-03"
5553196,method for processing data using a neural network having a number of layers equal to an abstraction degree of the pattern to be processed,"there is provided a layer construction of neural layers according to the abstraction degree of data to be processed, and data is inputted to a neural layer corresponding to its abstraction degree.",1996-09-03,"The title of the patent is method for processing data using a neural network having a number of layers equal to an abstraction degree of the pattern to be processed and its abstract is there is provided a layer construction of neural layers according to the abstraction degree of data to be processed, and data is inputted to a neural layer corresponding to its abstraction degree. dated 1996-09-03"
5553616,determination of concentrations of biological substances using raman spectroscopy and artificial neural network discriminator,"the concentration of a substance, such as glucose, in a biological sample, such as human tissue (e.g. the skin of an index finger) is non-invasively determined by directing the output beam of a laser diode onto and into the skin so as to cause raman scattering. the output of a charge coupled device, upon which the scattered light is spatially dispersed according to frequency is digitized and applied to a processor. the processor compares the raman scattering intensity characteristics of the sample with a comparative model, in particular, an artificial neural network discriminator (annd). the annd is trained with a plurality of raman spectral characteristics from biological fluids or tissue, possessing known raman scattered light intensities versus wavelength characteristics at known concentrations. a preferred implementation of the annd employs fuzzy adaptive resonance theory-mapping (artmap), which has robust noise rejection capabilities and can readily handle nonlinear phenomena.",1996-09-10,"The title of the patent is determination of concentrations of biological substances using raman spectroscopy and artificial neural network discriminator and its abstract is the concentration of a substance, such as glucose, in a biological sample, such as human tissue (e.g. the skin of an index finger) is non-invasively determined by directing the output beam of a laser diode onto and into the skin so as to cause raman scattering. the output of a charge coupled device, upon which the scattered light is spatially dispersed according to frequency is digitized and applied to a processor. the processor compares the raman scattering intensity characteristics of the sample with a comparative model, in particular, an artificial neural network discriminator (annd). the annd is trained with a plurality of raman spectral characteristics from biological fluids or tissue, possessing known raman scattered light intensities versus wavelength characteristics at known concentrations. a preferred implementation of the annd employs fuzzy adaptive resonance theory-mapping (artmap), which has robust noise rejection capabilities and can readily handle nonlinear phenomena. dated 1996-09-10"
5554273,neural network compensation for sensors,"a method for correcting an output from an electrochemical cell sensor, having an output responsive to a concentration of sensed species in a stream, a pressure and a temperature in an interactive non-linear relationship, comprising the steps of providing a pressure sensor for producing an output responsive to a pressure in proximity to the electrochemical sensor; providing a temperature sensor for producing an output responsive to a temperature of said electrochemical sensor; and processing the outputs of the electrochemical sensor, pressure sensor and temperature sensor in a neural network having an output function which compensates said electrochemical sensor for changes in pressure and temperature to indicate a concentration of the sensed species. an apparatus is also provided, for compensating an electrochemical sensing apparatus, comprising an electrochemical sensor, being responsive to a sensed species, and an environmental variable; an environmental variable sensor, being responsive to said environmental variable; and a compensation network for producing a compensated output based on an output of said electrochemical sensor based and an output of said environmental variable sensor, said network comprising a neural network.",1996-09-10,"The title of the patent is neural network compensation for sensors and its abstract is a method for correcting an output from an electrochemical cell sensor, having an output responsive to a concentration of sensed species in a stream, a pressure and a temperature in an interactive non-linear relationship, comprising the steps of providing a pressure sensor for producing an output responsive to a pressure in proximity to the electrochemical sensor; providing a temperature sensor for producing an output responsive to a temperature of said electrochemical sensor; and processing the outputs of the electrochemical sensor, pressure sensor and temperature sensor in a neural network having an output function which compensates said electrochemical sensor for changes in pressure and temperature to indicate a concentration of the sensed species. an apparatus is also provided, for compensating an electrochemical sensing apparatus, comprising an electrochemical sensor, being responsive to a sensed species, and an environmental variable; an environmental variable sensor, being responsive to said environmental variable; and a compensation network for producing a compensated output based on an output of said electrochemical sensor based and an output of said environmental variable sensor, said network comprising a neural network. dated 1996-09-10"
5555345,learning method of neural network,"the present invention is a learning method of a neural network for identifying n category using a data set consisted of n categories, in which one learning sample is extracted from a learning sample set in step sp1, and the distances between the sample and all the learning samples are obtained in step sp2. the closest n samples are obtained for each category in step sp3, and similarity for each category is obtained using the distances from the samples and a similarity conversion function f(d)=exp (-.alpha..multidot.d.sup.2). in step sp4, the similarity for each category is used as a target signal for the extracted learning sample, and it returns to an initial state until target signals for all the learning samples are determined. when target signals are determined for all the learning samples, in step sp5, the neural network is subjected to learning by the back-propagation using the learning samples and the obtained target signals.",1996-09-10,"The title of the patent is learning method of neural network and its abstract is the present invention is a learning method of a neural network for identifying n category using a data set consisted of n categories, in which one learning sample is extracted from a learning sample set in step sp1, and the distances between the sample and all the learning samples are obtained in step sp2. the closest n samples are obtained for each category in step sp3, and similarity for each category is obtained using the distances from the samples and a similarity conversion function f(d)=exp (-.alpha..multidot.d.sup.2). in step sp4, the similarity for each category is used as a target signal for the extracted learning sample, and it returns to an initial state until target signals for all the learning samples are determined. when target signals are determined for all the learning samples, in step sp5, the neural network is subjected to learning by the back-propagation using the learning samples and the obtained target signals. dated 1996-09-10"
5555347,method and apparatus for controlling a robot using a neural network,"a robot controller for an articulated robot, in which a joint angle vector indicating target joint angles of respective joints of the robot is calculated based on a target position matrix indicating a desired target position, and the robot is controlled ill accordance with the joint angle vector. the robot controller is further provided with a neural network for compensating the target position so as to reduce the positioning error. the learning operation of the neural network is repeatedly carried out at predetermined intervals during the operation of the robot. the target position may be preliminarily compensated using a mathematical model.",1996-09-10,"The title of the patent is method and apparatus for controlling a robot using a neural network and its abstract is a robot controller for an articulated robot, in which a joint angle vector indicating target joint angles of respective joints of the robot is calculated based on a target position matrix indicating a desired target position, and the robot is controlled ill accordance with the joint angle vector. the robot controller is further provided with a neural network for compensating the target position so as to reduce the positioning error. the learning operation of the neural network is repeatedly carried out at predetermined intervals during the operation of the robot. the target position may be preliminarily compensated using a mathematical model. dated 1996-09-10"
5555439,learning system and a learning pattern showing method for a neural network,"a neural network learning system using back propagation and a method of learning pattern showing are disclosed. in the case where a supervised signal contains an error or a pattern difficult to learn, the error is detected and the particular pattern is automatically removed during calculations of learning iterations to conduct rightly and accelerate the learning. the learning history for each pattern is stored to detect inconsistent and difficult-to-learn patterns, which are prevented from being shown to the network by a pattern showing control during the next learning iteration. as a result, an inconsistent or difficult-to-learn pattern which may be contained in a learning pattern set of input and supervised patterns is removed during learning iterations thereby to permit early completion of the learning process.",1996-09-10,"The title of the patent is learning system and a learning pattern showing method for a neural network and its abstract is a neural network learning system using back propagation and a method of learning pattern showing are disclosed. in the case where a supervised signal contains an error or a pattern difficult to learn, the error is detected and the particular pattern is automatically removed during calculations of learning iterations to conduct rightly and accelerate the learning. the learning history for each pattern is stored to detect inconsistent and difficult-to-learn patterns, which are prevented from being shown to the network by a pattern showing control during the next learning iteration. as a result, an inconsistent or difficult-to-learn pattern which may be contained in a learning pattern set of input and supervised patterns is removed during learning iterations thereby to permit early completion of the learning process. dated 1996-09-10"
5555512,picture processing apparatus for processing infrared pictures obtained with an infrared ray sensor and applied apparatus utilizing the picture processing apparatus,"a picture processing apparatus is composed of an infrared ray sensor for producing an infrared picture in which a heat distribution of a measured area is recorded, a picture processing section for processing the infrared picture to extract pieces of personal characteristic data and environmental characteristic data from the infrared picture, and a picture information detecting section composed of a neural network for detecting pieces of personal information and environmental information according to the characteristic data. to be concrete, one or more human-areas respectively recording one or more persons are picked out from the infrared picture, and representative points of the human-areas, the number of human-areas, the number of pixels in each of the human-areas and shapes of the human-areas are extracted as the personal characteristic data. also, the number of persons in each of the human-areas, foot positions of the persons, person's postures, skin temperatures of the persons and the volume of person's clothes are detected as the personal information according to the personal and environmental characteristic data.",1996-09-10,"The title of the patent is picture processing apparatus for processing infrared pictures obtained with an infrared ray sensor and applied apparatus utilizing the picture processing apparatus and its abstract is a picture processing apparatus is composed of an infrared ray sensor for producing an infrared picture in which a heat distribution of a measured area is recorded, a picture processing section for processing the infrared picture to extract pieces of personal characteristic data and environmental characteristic data from the infrared picture, and a picture information detecting section composed of a neural network for detecting pieces of personal information and environmental information according to the characteristic data. to be concrete, one or more human-areas respectively recording one or more persons are picked out from the infrared picture, and representative points of the human-areas, the number of human-areas, the number of pixels in each of the human-areas and shapes of the human-areas are extracted as the personal characteristic data. also, the number of persons in each of the human-areas, foot positions of the persons, person's postures, skin temperatures of the persons and the volume of person's clothes are detected as the personal information according to the personal and environmental characteristic data. dated 1996-09-10"
5557686,"method and apparatus for verification of a computer user's identification, based on keystroke characteristics","a method and apparatus for determining whether a user of a system is an authorized user or an imposter by examining the keystroke characteristics of the user. the authorized user initially enters a number of user training samples on a keyboard. the user training samples are then purified to eliminate training samples which are different from other training samples. the purification can be performed by a self-organizing neural network which has input thereto, authorized user training samples, or both authorized training samples and imposter training samples. the purified user training samples are then compared to a sample to be tested to determine whether the sample is from an authorized user or an imposter. the comparison of the purified samples with the sample to be tested can be performed by a neural network such as a back propagation trained network, an adaline unit, a distance method or a linear classifier, discriminate function, or piecewise linear classifier. the result of this testing step indicates whether the user is authorized or an imposter and the user can be granted or denied access to the system.",1996-09-17,"The title of the patent is method and apparatus for verification of a computer user's identification, based on keystroke characteristics and its abstract is a method and apparatus for determining whether a user of a system is an authorized user or an imposter by examining the keystroke characteristics of the user. the authorized user initially enters a number of user training samples on a keyboard. the user training samples are then purified to eliminate training samples which are different from other training samples. the purification can be performed by a self-organizing neural network which has input thereto, authorized user training samples, or both authorized training samples and imposter training samples. the purified user training samples are then compared to a sample to be tested to determine whether the sample is from an authorized user or an imposter. the comparison of the purified samples with the sample to be tested can be performed by a neural network such as a back propagation trained network, an adaline unit, a distance method or a linear classifier, discriminate function, or piecewise linear classifier. the result of this testing step indicates whether the user is authorized or an imposter and the user can be granted or denied access to the system. dated 1996-09-17"
5558790,method and laser system for the thermal analysis of a substance,"provided is a method and apparatus for the thermal analysis of a substance. the method comprises suspending a substance sample on a substrate in a reactor using a temperature sensor positioned at the center of the reactor. the reactor is then heated by two laser beams focused on the reactor. the resulting temperature dependence of the sample/substance during heating is measured. additionally, the sample/substrate is heated to a temperature above the reactor temperature with a third laser beam. the rate at which the sample/substance temperature relaxes to the temperature of the reactor is measured. this additional heating of the sample/substance is preferably achieved by a laser focused on the sample/substance itself. all of the measured information can then be fed into a computer through an electronic interface to provide data on the particular substance undergoing thermal analysis. the method and the system used to effect the practice of the method can be used to study samples of various substances to aid in the design and process control of manufacturing processes through integration with a neural network. the method has also been demonstrated to provide data necessary to optimize the burning of high sulfur fuel oils. the present invention is particularly applicable to aiding in the analysis and design of thermal processes used to destroy hazardous chemicals in waste materials.",1996-09-24,"The title of the patent is method and laser system for the thermal analysis of a substance and its abstract is provided is a method and apparatus for the thermal analysis of a substance. the method comprises suspending a substance sample on a substrate in a reactor using a temperature sensor positioned at the center of the reactor. the reactor is then heated by two laser beams focused on the reactor. the resulting temperature dependence of the sample/substance during heating is measured. additionally, the sample/substrate is heated to a temperature above the reactor temperature with a third laser beam. the rate at which the sample/substance temperature relaxes to the temperature of the reactor is measured. this additional heating of the sample/substance is preferably achieved by a laser focused on the sample/substance itself. all of the measured information can then be fed into a computer through an electronic interface to provide data on the particular substance undergoing thermal analysis. the method and the system used to effect the practice of the method can be used to study samples of various substances to aid in the design and process control of manufacturing processes through integration with a neural network. the method has also been demonstrated to provide data necessary to optimize the burning of high sulfur fuel oils. the present invention is particularly applicable to aiding in the analysis and design of thermal processes used to destroy hazardous chemicals in waste materials. dated 1996-09-24"
5559604,color matching apparatus for reproducing the same color under different illuminants,"colorimetric values such as l*a*b* values are transformed into color separation values such as cmyk values dependent on the characteristics of a color output device by using a multilayered feedforward neural network. the neural network learns in advance the relationships between a multiplicity of colorimetric values under different illuminants and color separation values. when colorimetric values of a standard illuminant and one of the color temperature of an observational illuminant, the spectral distribution of the observational illuminant, and colorimetric values at the time of illumination by the observational illuminant are inputted to the neural network, color separation values corresponding to one of the color temperature of the observational illuminant, the spectral distribution of the observational illuminant, and the colorimetric values obtained when illumination is provided by the observational illuminant are outputted. consequently, colorimetric values are transformed into color separation values such that the color reproduced under the observational illuminant will visually assume the same color as that under the standard illuminant.",1996-09-24,"The title of the patent is color matching apparatus for reproducing the same color under different illuminants and its abstract is colorimetric values such as l*a*b* values are transformed into color separation values such as cmyk values dependent on the characteristics of a color output device by using a multilayered feedforward neural network. the neural network learns in advance the relationships between a multiplicity of colorimetric values under different illuminants and color separation values. when colorimetric values of a standard illuminant and one of the color temperature of an observational illuminant, the spectral distribution of the observational illuminant, and colorimetric values at the time of illumination by the observational illuminant are inputted to the neural network, color separation values corresponding to one of the color temperature of the observational illuminant, the spectral distribution of the observational illuminant, and the colorimetric values obtained when illumination is provided by the observational illuminant are outputted. consequently, colorimetric values are transformed into color separation values such that the color reproduced under the observational illuminant will visually assume the same color as that under the standard illuminant. dated 1996-09-24"
5559690,residual activation neural network,"a plant (72) is operable to receive control inputs c(t) and provide an output y(t). the plant (72) has associated therewith state variables s(t) that are not variable. a control network (74) is provided that accurately models the plant (72). the output of the control network (74) provides a predicted output which is combined with a desired output to generate an error. this error is back propagated through an inverse control network (76), which is the inverse of the control network (74) to generate a control error signal that is input to a distributed control system (73) to vary the control inputs to the plant (72) in order to change the output y(t) to meet the desired output. the control network (74) is comprised of a first network net 1 that is operable to store a representation of the dependency of the control variables on the state variables. the predicted result is subtracted from the actual state variable input and stored as a residual in a residual layer (102). the output of the residual layer (102) is input to a hidden layer (108) which also receives the control inputs to generate a predicted output in an output layer (106). during back propagation of error, the residual values in the residual layer (102) are latched and only the control inputs allowed to vary.",1996-09-24,"The title of the patent is residual activation neural network and its abstract is a plant (72) is operable to receive control inputs c(t) and provide an output y(t). the plant (72) has associated therewith state variables s(t) that are not variable. a control network (74) is provided that accurately models the plant (72). the output of the control network (74) provides a predicted output which is combined with a desired output to generate an error. this error is back propagated through an inverse control network (76), which is the inverse of the control network (74) to generate a control error signal that is input to a distributed control system (73) to vary the control inputs to the plant (72) in order to change the output y(t) to meet the desired output. the control network (74) is comprised of a first network net 1 that is operable to store a representation of the dependency of the control variables on the state variables. the predicted result is subtracted from the actual state variable input and stored as a residual in a residual layer (102). the output of the residual layer (102) is input to a hidden layer (108) which also receives the control inputs to generate a predicted output in an output layer (106). during back propagation of error, the residual values in the residual layer (102) are latched and only the control inputs allowed to vary. dated 1996-09-24"
5559929,method of enhancing the selection of a training set for use in training of a neural network,a method for enhancing the performance of an artificially intelligent system employing a neural network by proving an optimized training set for training the neural network. the optimized training set is produced by identifying and permanently removing inaccurate training pairs in the training set. the permanent removal of inaccurate training pairs is performed in a worst-error order.,1996-09-24,The title of the patent is method of enhancing the selection of a training set for use in training of a neural network and its abstract is a method for enhancing the performance of an artificially intelligent system employing a neural network by proving an optimized training set for training the neural network. the optimized training set is produced by identifying and permanently removing inaccurate training pairs in the training set. the permanent removal of inaccurate training pairs is performed in a worst-error order. dated 1996-09-24
5561741,method of enhancing the performance of a neural network,a method for enhancing the performance of an artificially intelligent system employing a neural network by proving an optimized training set for training the neural network. the optimized training set is produced by identifying and removing inaccurate training pairs in the training set.,1996-10-01,The title of the patent is method of enhancing the performance of a neural network and its abstract is a method for enhancing the performance of an artificially intelligent system employing a neural network by proving an optimized training set for training the neural network. the optimized training set is produced by identifying and removing inaccurate training pairs in the training set. dated 1996-10-01
5562243,method an apparatus for obtaining reflow oven settings for soldering a pcb,"an artificial neural network is trained to recognize inputted thermal and physical features of a printed circuit board, for providing settings for a reflow oven for obtaining acceptable soldering of the printed circuit board.",1996-10-08,"The title of the patent is method an apparatus for obtaining reflow oven settings for soldering a pcb and its abstract is an artificial neural network is trained to recognize inputted thermal and physical features of a printed circuit board, for providing settings for a reflow oven for obtaining acceptable soldering of the printed circuit board. dated 1996-10-08"
5563982,apparatus and method for detection of molecular vapors in an atmospheric region,apparatus for detecting molecular vapors in an atmospheric region includes an interferometer which monitors light parameter data signals received and provides an interferometer light parameter signal corresponding to the light parameter data signals at a plurality of frequencies. the apparatus further includes an interferogram detector/converter which records and digitizes the interferometer light parameter signal to generate a plurality of discrete data points wherein each discrete data point corresponds to the interferometer light parameter signal at a specific frequency. the apparatus also includes a fourier transform circuit for receiving the discrete interferometer light parameter signal and providing a fourier transformed molecular parameter data signal. the apparatus further includes a probabilistic neural network for receiving and sorting the fourier transformed molecular parameter data signals without the use of a priori training data.,1996-10-08,The title of the patent is apparatus and method for detection of molecular vapors in an atmospheric region and its abstract is apparatus for detecting molecular vapors in an atmospheric region includes an interferometer which monitors light parameter data signals received and provides an interferometer light parameter signal corresponding to the light parameter data signals at a plurality of frequencies. the apparatus further includes an interferogram detector/converter which records and digitizes the interferometer light parameter signal to generate a plurality of discrete data points wherein each discrete data point corresponds to the interferometer light parameter signal at a specific frequency. the apparatus also includes a fourier transform circuit for receiving the discrete interferometer light parameter signal and providing a fourier transformed molecular parameter data signal. the apparatus further includes a probabilistic neural network for receiving and sorting the fourier transformed molecular parameter data signals without the use of a priori training data. dated 1996-10-08
5563983,learning system operated through a layered neural network,"the present invention predicts an output result in response to unknown input data using a layered neural network. for example, learning data collected in time series with rate of changes included are learned through the layered neural network. as a result, a predicted value can be obtained with a smaller amount of learning data within a shorter learning time. the present invention comprises a rate of change calculating unit for calculating the rate of change of time-series data at two different time points, and a network generating unit for controlling the learning steps performed by the layered neural network using at least the rate of change of data as learning data so as to establish a neural network in which a weight value is determined for learning data. as a predicting system, it further comprises at least a neural network recognizing unit for applying to the neural network established by the network generating unit an output of the rate of change calculating unit outputted in response to input predicting data, and for obtaining a prediction result.",1996-10-08,"The title of the patent is learning system operated through a layered neural network and its abstract is the present invention predicts an output result in response to unknown input data using a layered neural network. for example, learning data collected in time series with rate of changes included are learned through the layered neural network. as a result, a predicted value can be obtained with a smaller amount of learning data within a shorter learning time. the present invention comprises a rate of change calculating unit for calculating the rate of change of time-series data at two different time points, and a network generating unit for controlling the learning steps performed by the layered neural network using at least the rate of change of data as learning data so as to establish a neural network in which a weight value is determined for learning data. as a predicting system, it further comprises at least a neural network recognizing unit for applying to the neural network established by the network generating unit an output of the rate of change calculating unit outputted in response to input predicting data, and for obtaining a prediction result. dated 1996-10-08"
5564079,method for locating mobile stations in a digital telephone network,"the invention relates to a method for locating mobile stations in a digital telecommunication network, especially the gsm network. according to the invention, reference measurements are carried out on relevant traffic routes with the aid of a measuring mobile in order to provide position information related to measured signals. with the aid of these reference data and the position information, an adaptive neural network is trained, which network, with the aid of corresponding measurement data which are transmitted from the mobile station to a respective base station, carries out the localization of the mobile station. use of the adaptive neural network provides a more accurate position determination than earlier systems which were only based on the ta (timing advance) value.",1996-10-08,"The title of the patent is method for locating mobile stations in a digital telephone network and its abstract is the invention relates to a method for locating mobile stations in a digital telecommunication network, especially the gsm network. according to the invention, reference measurements are carried out on relevant traffic routes with the aid of a measuring mobile in order to provide position information related to measured signals. with the aid of these reference data and the position information, an adaptive neural network is trained, which network, with the aid of corresponding measurement data which are transmitted from the mobile station to a respective base station, carries out the localization of the mobile station. use of the adaptive neural network provides a more accurate position determination than earlier systems which were only based on the ta (timing advance) value. dated 1996-10-08"
5564115,neural network architecture with connection pointers,"a neural network unit is described which has a plurality of neurons. the network comprises a ram, which provides a plurality of storage locations for each of the neurons and an integrated circuit. the integrated circuit including means for defining an algorithm for the operation of the neurons and a control unit for causing the neurons to produce outputs on the basis of data stored in the storage locations and the algorithm. the integrated circuit may have a random number generator and a comparator. in effect, the neurons are virtual prams (probabilistic rams).",1996-10-08,"The title of the patent is neural network architecture with connection pointers and its abstract is a neural network unit is described which has a plurality of neurons. the network comprises a ram, which provides a plurality of storage locations for each of the neurons and an integrated circuit. the integrated circuit including means for defining an algorithm for the operation of the neurons and a control unit for causing the neurons to produce outputs on the basis of data stored in the storage locations and the algorithm. the integrated circuit may have a random number generator and a comparator. in effect, the neurons are virtual prams (probabilistic rams). dated 1996-10-08"
5566065,method and apparatus for controlling multivariable nonlinear processes,"a method and apparatus for a robust process control system which utilizes a neural-network based multivariable inner-loop pd controller cascaded with decoupled outer-loop controllers with integral action, the combination providing a multivariable nonlinear pid and feedforward controller. the inner-loop pd controller employs a quasi-newton iterative feedback loop structure whereby the manipulated variables are computed in an iterative fashion as a function of the difference between the inner loop setpoint and the predicted controlled variable as advanced by the optimum prediction time, in order to incorporate the downstream limiting effects on the non-limited control loops. the outer-loop controllers compensate for unmodeled process changes, unmeasured disturbances, and modeling errors by adjusting the inner-loop target values.",1996-10-15,"The title of the patent is method and apparatus for controlling multivariable nonlinear processes and its abstract is a method and apparatus for a robust process control system which utilizes a neural-network based multivariable inner-loop pd controller cascaded with decoupled outer-loop controllers with integral action, the combination providing a multivariable nonlinear pid and feedforward controller. the inner-loop pd controller employs a quasi-newton iterative feedback loop structure whereby the manipulated variables are computed in an iterative fashion as a function of the difference between the inner loop setpoint and the predicted controlled variable as advanced by the optimum prediction time, in order to incorporate the downstream limiting effects on the non-limited control loops. the outer-loop controllers compensate for unmodeled process changes, unmeasured disturbances, and modeling errors by adjusting the inner-loop target values. dated 1996-10-15"
5566072,method and apparatus for estimating a road traffic condition and method and apparatus for controlling a vehicle running characteristic,"an apparatus for estimating a road traffic condition and for controlling a vehicle running characteristic includes a controller having a fuzzy inference function and a neural network function. the controller carries out frequency analysis on vehicle driving parameters such as vehicle speed, steering angle, accelerator opening degree, and longitudinal acceleration and lateral acceleration of a vehicle, to thereby determine a mean value and variance of each parameter. it implements fuzzy inference based on a traveling time ratio, an average speed, and an average lateral acceleration, which are obtained from vehicle speed and/or steering angle, to thereby calculate road traffic condition parameters, including a city area degree, a jammed road degree, and a mountainous road degree. according to the neural network function, the controller further determines an output parameter, indicative of the vehicle maneuvering state, by subjecting the mean value and variance of the vehicle driving parameters and the weighted total sum of the road traffic condition parameters to nonlinear conversion. then, it variably controls the operating characteristic of a vehicle-mounted apparatus such as a rear-wheel steering controlling unit in accordance with the output parameter, thereby variably controlling the vehicle running characteristic.",1996-10-15,"The title of the patent is method and apparatus for estimating a road traffic condition and method and apparatus for controlling a vehicle running characteristic and its abstract is an apparatus for estimating a road traffic condition and for controlling a vehicle running characteristic includes a controller having a fuzzy inference function and a neural network function. the controller carries out frequency analysis on vehicle driving parameters such as vehicle speed, steering angle, accelerator opening degree, and longitudinal acceleration and lateral acceleration of a vehicle, to thereby determine a mean value and variance of each parameter. it implements fuzzy inference based on a traveling time ratio, an average speed, and an average lateral acceleration, which are obtained from vehicle speed and/or steering angle, to thereby calculate road traffic condition parameters, including a city area degree, a jammed road degree, and a mountainous road degree. according to the neural network function, the controller further determines an output parameter, indicative of the vehicle maneuvering state, by subjecting the mean value and variance of the vehicle driving parameters and the weighted total sum of the road traffic condition parameters to nonlinear conversion. then, it variably controls the operating characteristic of a vehicle-mounted apparatus such as a rear-wheel steering controlling unit in accordance with the output parameter, thereby variably controlling the vehicle running characteristic. dated 1996-10-15"
5566092,machine fault diagnostics system and method,"the invention provides a machine fault diagnostic system to help ensure effective equipment maintenance. the major technique used for fault diagnostics is a fault diagnostic network (fdn) which is based on a modified artmap neural network architecture. a hypothesis and test procedure based on fuzzy logic and physical bearing models is disclosed to operate with the fdn for detecting faults that cannot be recognized by the fdn and for analyzing complex machine conditions. the procedure described herein is able to provide accurate fault diagnosis for both one and multiple-fault conditions. furthermore, a transputer-based parallel processing technique is used in which the fdn is implemented on a network of four t800-25 transputers.",1996-10-15,"The title of the patent is machine fault diagnostics system and method and its abstract is the invention provides a machine fault diagnostic system to help ensure effective equipment maintenance. the major technique used for fault diagnostics is a fault diagnostic network (fdn) which is based on a modified artmap neural network architecture. a hypothesis and test procedure based on fuzzy logic and physical bearing models is disclosed to operate with the fdn for detecting faults that cannot be recognized by the fdn and for analyzing complex machine conditions. the procedure described herein is able to provide accurate fault diagnosis for both one and multiple-fault conditions. furthermore, a transputer-based parallel processing technique is used in which the fdn is implemented on a network of four t800-25 transputers. dated 1996-10-15"
5566270,speaker independent isolated word recognition system using neural networks,"a speech recognition apparatus in which the speech signal is digitalized and subjected to special analysis, word end detection is effected by energy analysis of the speech signal and the recognition system utilizes a markov model in combination with a neural network learning by specific training steps.",1996-10-15,"The title of the patent is speaker independent isolated word recognition system using neural networks and its abstract is a speech recognition apparatus in which the speech signal is digitalized and subjected to special analysis, word end detection is effected by energy analysis of the speech signal and the recognition system utilizes a markov model in combination with a neural network learning by specific training steps. dated 1996-10-15"
5566273,supervised training of a neural network,"the present invention provides a system and method for supervised training of a neural network. a neural network architecture and training method is disclosed that is a modification of an artmap architecture. the modified artmap network is an efficient and robust paradigm which has the unique property of incremental supervised learning. furthermore, the modified artmap network has the capability of removing undesired knowledge that has previously been learned by the network.",1996-10-15,"The title of the patent is supervised training of a neural network and its abstract is the present invention provides a system and method for supervised training of a neural network. a neural network architecture and training method is disclosed that is a modification of an artmap architecture. the modified artmap network is an efficient and robust paradigm which has the unique property of incremental supervised learning. furthermore, the modified artmap network has the capability of removing undesired knowledge that has previously been learned by the network. dated 1996-10-15"
5566275,control method and apparatus using two neural networks,"a control method of controlling a controlled system according to the invention comprises the first step of inputting a current and future target controlled variable to a first neural network model which performs learning using a past target controlled variable for the controlled system as an input signal and a past manipulated variable as a teacher signal, thereby obtaining a current virtual manipulated variable, the second step of causing a second neural network model, which have learnt to predict a behavior of the controlled system, to receive the virtual manipulated variable obtained in the first step and a controlled variable obtained from the controlled system at a current time, thereby obtaining a predicted controlled variable, the third step of obtaining an error of the predicted controlled variable obtained in the second step with respect to the target controlled variable, the fourth step of obtaining a correction amount for the virtual manipulated variable in accordance with a back propagation calculation of the second neural network model, using the error obtained in the third step, thereby correcting the virtual manipulated variable with the correction amount, and the fifth step of outputting the virtual manipulated variable corrected in the fourth step to the controlled system.",1996-10-15,"The title of the patent is control method and apparatus using two neural networks and its abstract is a control method of controlling a controlled system according to the invention comprises the first step of inputting a current and future target controlled variable to a first neural network model which performs learning using a past target controlled variable for the controlled system as an input signal and a past manipulated variable as a teacher signal, thereby obtaining a current virtual manipulated variable, the second step of causing a second neural network model, which have learnt to predict a behavior of the controlled system, to receive the virtual manipulated variable obtained in the first step and a controlled variable obtained from the controlled system at a current time, thereby obtaining a predicted controlled variable, the third step of obtaining an error of the predicted controlled variable obtained in the second step with respect to the target controlled variable, the fourth step of obtaining a correction amount for the virtual manipulated variable in accordance with a back propagation calculation of the second neural network model, using the error obtained in the third step, thereby correcting the virtual manipulated variable with the correction amount, and the fifth step of outputting the virtual manipulated variable corrected in the fourth step to the controlled system. dated 1996-10-15"
5568126,providing an alarm in response to a determination that a person may have suddenly experienced fear,"a physiological-condition monitoring system is attached to a person for providing monitored physiological data signals. a computer system processes the physiological data signals to determine whether the person may have suddenly experienced fear by comparing the monitored physiological data with stored stress profile data for the person. the stress profile data is based upon prior measurements of the monitored physiological conditions of the person during situations of stress. when such processing by the computer system determines that the person may have suddenly experienced fear, the computer system provides an alarm-indication signal to an initial alarm indicator. if the initial alarm indication was false or if the condition causing the person to suddenly experience fear was very brief and no longer poses a threat, the person can operate an alarm deactuation switch to cause the computer system to discontinue the alarm indication signal. the computer system includes a neural network for modifying the stored stress profile data in response to an input signal indicating that the computer system provided a false alarm indication. such an input signal is provided when the alarm deactuation switch is operated within a given interval .delta.t following commencement of the initial alarm indication signal. the modified stress profile data is stored for comparison with monitored physiological data subsequently received by the computer system. if the alarm deactuation switch is not operated within the given interval .delta.t following commencement of the initial alarm indication signal, the computer system provides a second alarm indication signal to an audio alarm indicator and to a remote alarm indicator. the surrounding conditions that may have caused the determination of probable fear are recorded and transmitted to a remote location.",1996-10-22,"The title of the patent is providing an alarm in response to a determination that a person may have suddenly experienced fear and its abstract is a physiological-condition monitoring system is attached to a person for providing monitored physiological data signals. a computer system processes the physiological data signals to determine whether the person may have suddenly experienced fear by comparing the monitored physiological data with stored stress profile data for the person. the stress profile data is based upon prior measurements of the monitored physiological conditions of the person during situations of stress. when such processing by the computer system determines that the person may have suddenly experienced fear, the computer system provides an alarm-indication signal to an initial alarm indicator. if the initial alarm indication was false or if the condition causing the person to suddenly experience fear was very brief and no longer poses a threat, the person can operate an alarm deactuation switch to cause the computer system to discontinue the alarm indication signal. the computer system includes a neural network for modifying the stored stress profile data in response to an input signal indicating that the computer system provided a false alarm indication. such an input signal is provided when the alarm deactuation switch is operated within a given interval .delta.t following commencement of the initial alarm indication signal. the modified stress profile data is stored for comparison with monitored physiological data subsequently received by the computer system. if the alarm deactuation switch is not operated within the given interval .delta.t following commencement of the initial alarm indication signal, the computer system provides a second alarm indication signal to an audio alarm indicator and to a remote alarm indicator. the surrounding conditions that may have caused the determination of probable fear are recorded and transmitted to a remote location. dated 1996-10-22"
5568590,image processing using genetic mutation of neural network parameters,"a system and method of image processing using neural networks to control image processing elements. neural network parameters are defined by genotypes consisting of network vectors. genotypes may be selectively mutated and cross-bred to provide a mechanism for modifying the behavior of the neural networks, or phenotypes. genetic modeling processes are used to perform such mutation and cross-over. user feedback concerning output images, is used to select particular genotypes for further mutation and exploration. preconditioning is employed to extract structural information from source images prior to network processing. genetic morphing and subnet fusion are also available, to provide additional variations on image processing operations.",1996-10-22,"The title of the patent is image processing using genetic mutation of neural network parameters and its abstract is a system and method of image processing using neural networks to control image processing elements. neural network parameters are defined by genotypes consisting of network vectors. genotypes may be selectively mutated and cross-bred to provide a mechanism for modifying the behavior of the neural networks, or phenotypes. genetic modeling processes are used to perform such mutation and cross-over. user feedback concerning output images, is used to select particular genotypes for further mutation and exploration. preconditioning is employed to extract structural information from source images prior to network processing. genetic morphing and subnet fusion are also available, to provide additional variations on image processing operations. dated 1996-10-22"
5568591,method and device using a neural network for classifying data,"method and device having a neural network for classifying data, and verification device for signatures. the device includes a neural network with an input layer 3, an internal layer 4, and an output layer 5. this network is designed to classify data vectors to classes, the synaptic weights in the network being determined through programming on the basis of specimens whose classes are known. each class is defined during programming as corresponding to a set of neurons of which each represents a domain which contains a fixed number of specimens. the network includes a number of neurons and synaptic weights which have been determined as a function of the classes thus defined.",1996-10-22,"The title of the patent is method and device using a neural network for classifying data and its abstract is method and device having a neural network for classifying data, and verification device for signatures. the device includes a neural network with an input layer 3, an internal layer 4, and an output layer 5. this network is designed to classify data vectors to classes, the synaptic weights in the network being determined through programming on the basis of specimens whose classes are known. each class is defined during programming as corresponding to a set of neurons of which each represents a domain which contains a fixed number of specimens. the network includes a number of neurons and synaptic weights which have been determined as a function of the classes thus defined. dated 1996-10-22"
5570282,multivariable nonlinear process controller,"a method and apparatus for a robust process control system that utilizes a neural-network multivariable inner-loop pd controller cascaded with decoupled outer-loop controllers with integral action, the combination providing a multivariable nonlinear pid and feedforward controller. the inner-loop neural-network controller is trained to achieve optimal performance behavior when future process behavior repeats the training experience. the outer-loop controllers compensate for process changes, unmeasured disturbances, and modeling errors. in the first and second embodiments, the neural network is used as an inner-loop controller in a process control system having a constraint management scheme which prevents integral windup by controlling the action of the outer-loop controllers when limiting is detected in the associated manipulated-variable control path. in the second and third embodiments, the neural-network controller is used without the integral controllers or the constraint management scheme as a simple pd feedforward controller.",1996-10-29,"The title of the patent is multivariable nonlinear process controller and its abstract is a method and apparatus for a robust process control system that utilizes a neural-network multivariable inner-loop pd controller cascaded with decoupled outer-loop controllers with integral action, the combination providing a multivariable nonlinear pid and feedforward controller. the inner-loop neural-network controller is trained to achieve optimal performance behavior when future process behavior repeats the training experience. the outer-loop controllers compensate for process changes, unmeasured disturbances, and modeling errors. in the first and second embodiments, the neural network is used as an inner-loop controller in a process control system having a constraint management scheme which prevents integral windup by controlling the action of the outer-loop controllers when limiting is detected in the associated manipulated-variable control path. in the second and third embodiments, the neural-network controller is used without the integral controllers or the constraint management scheme as a simple pd feedforward controller. dated 1996-10-29"
5570457,visual information processing device,"a visual information processing device having a neural network function and capable of visual information processing comprises a semiconductor integrated circuit section equipped with a plurality of neuronic circuit regions realizing a neuron function included in the neural network function, and first and second molecular film sections provided on the integrated circuit section. the first molecular film section comprises a light-receiving molecular film section including tij input elements having a photoelectric function and to which coupling strength levels (tij) between the plurality of neuronic circuit regions are optically written to realize electric connection between the neuronic circuit regions and image input elements for sensing visual images, each neuronic circuit region corresponding to one pixel. the second molecular film section comprises a light-emitting molecular film section including tij signal output elements having a light emitting function to output tij matrix signals as matrix light emission patterns.",1996-10-29,"The title of the patent is visual information processing device and its abstract is a visual information processing device having a neural network function and capable of visual information processing comprises a semiconductor integrated circuit section equipped with a plurality of neuronic circuit regions realizing a neuron function included in the neural network function, and first and second molecular film sections provided on the integrated circuit section. the first molecular film section comprises a light-receiving molecular film section including tij input elements having a photoelectric function and to which coupling strength levels (tij) between the plurality of neuronic circuit regions are optically written to realize electric connection between the neuronic circuit regions and image input elements for sensing visual images, each neuronic circuit region corresponding to one pixel. the second molecular film section comprises a light-emitting molecular film section including tij signal output elements having a light emitting function to output tij matrix signals as matrix light emission patterns. dated 1996-10-29"
5571969,vibration detection and reduction system and vibration sensors for use in micro-gravity environment,"the invention is to detect vibrations which would destroy a micro-gravity environment, to grasp factors of the vibrations, to detect and grasp a position and a scale of meteoroid/debris collisions, and to establish a counter-measure. vibration sensors are disposed in a matrix array on a vibrating body placed in a micro-gravity environment. a computer analyzes a spectrum of the vibration, in a neural network section a vibration source is specified on the basis of the analyzed spectra, in a fuzzy control section, actuators are driven so as to reduce harmful vibrations in response to the vibration energy and the energy source specified by the neural network section, and if necessary, a vibration factor is eliminated. the computer takes in the result of driving for the actuators, the results are learnt in the neural network section and in the fuzzy control section to be ready for generation of vibrations at the next time. the vibration sensor comprises a reflector or a transparent refractor disposed as floating in a micro-gravity space, output means fixed to the vibrating body for emitting energy towards the reflector or transparent refractor, and a receiver fixed to the vibrating body for receiving energy sent from the reflector or transparent refractor and measuring the nature of vibrations of the vibrating body on the basis of movement of the reflected or permeated energy.",1996-11-05,"The title of the patent is vibration detection and reduction system and vibration sensors for use in micro-gravity environment and its abstract is the invention is to detect vibrations which would destroy a micro-gravity environment, to grasp factors of the vibrations, to detect and grasp a position and a scale of meteoroid/debris collisions, and to establish a counter-measure. vibration sensors are disposed in a matrix array on a vibrating body placed in a micro-gravity environment. a computer analyzes a spectrum of the vibration, in a neural network section a vibration source is specified on the basis of the analyzed spectra, in a fuzzy control section, actuators are driven so as to reduce harmful vibrations in response to the vibration energy and the energy source specified by the neural network section, and if necessary, a vibration factor is eliminated. the computer takes in the result of driving for the actuators, the results are learnt in the neural network section and in the fuzzy control section to be ready for generation of vibrations at the next time. the vibration sensor comprises a reflector or a transparent refractor disposed as floating in a micro-gravity space, output means fixed to the vibrating body for emitting energy towards the reflector or transparent refractor, and a receiver fixed to the vibrating body for receiving energy sent from the reflector or transparent refractor and measuring the nature of vibrations of the vibrating body on the basis of movement of the reflected or permeated energy. dated 1996-11-05"
5572028,multi-element dosimetry system using neural network,a dosimetry system and method characterized by use a plurality of radiation sensitive elements to monitor exposure to a radiation field composed of one or more types of radiation at one or more different energies; reading the radiation sensitive elements in a reader after irradiation by the radiation field to obtain element outputs; and supplying the element outputs to a trained neural network computer apparatus wherein the element outputs are analyzed to provide an output indicative of the radiation field.,1996-11-05,The title of the patent is multi-element dosimetry system using neural network and its abstract is a dosimetry system and method characterized by use a plurality of radiation sensitive elements to monitor exposure to a radiation field composed of one or more types of radiation at one or more different energies; reading the radiation sensitive elements in a reader after irradiation by the radiation field to obtain element outputs; and supplying the element outputs to a trained neural network computer apparatus wherein the element outputs are analyzed to provide an output indicative of the radiation field. dated 1996-11-05
5572030,method for determining parameter of hydrocarbon,"a method for evaluating a hydrocarbon so as to determine a desired parameter of the hydrocarbon includes the steps of providing a hydrocarbon to be evaluated, obtaining a near-infrared signal from the hydrocarbon, codifying the near-infrared signal so as to reduce the signal to a number of points, providing a neural network trained for correlating the number of points to the desired parameter and processing the number of points with the neural network so as to determine the desired parameter.",1996-11-05,"The title of the patent is method for determining parameter of hydrocarbon and its abstract is a method for evaluating a hydrocarbon so as to determine a desired parameter of the hydrocarbon includes the steps of providing a hydrocarbon to be evaluated, obtaining a near-infrared signal from the hydrocarbon, codifying the near-infrared signal so as to reduce the signal to a number of points, providing a neural network trained for correlating the number of points to the desired parameter and processing the number of points with the neural network so as to determine the desired parameter. dated 1996-11-05"
5572628,training system for neural networks,""" in order for neural network technology to make useful determinations of the identity of letters and numbers that are processed in real time at a postal service sorting center, it is necessary for the neural network to """"learn"""" to recognize accurately the many shapes and sizes in which each letter or number are formed on the address surface of the envelope by postal service users. it has been realized that accuracy in the recognition of many letters and numbers is not appreciably sacrificed if the neural network is instructed to identify those characteristics of each letter or number which are in the category """"invariant."""" then, rather than requiring the neural network to recognize all gradations of shape, location, size, etc. of the identified invariant characteristic, a generalized and bounded description of the invariant segments is used which requires far less inputting of sample data and less processing of information relating to an unknown letter or number. """,1996-11-05,"The title of the patent is training system for neural networks and its abstract is "" in order for neural network technology to make useful determinations of the identity of letters and numbers that are processed in real time at a postal service sorting center, it is necessary for the neural network to """"learn"""" to recognize accurately the many shapes and sizes in which each letter or number are formed on the address surface of the envelope by postal service users. it has been realized that accuracy in the recognition of many letters and numbers is not appreciably sacrificed if the neural network is instructed to identify those characteristics of each letter or number which are in the category """"invariant."""" then, rather than requiring the neural network to recognize all gradations of shape, location, size, etc. of the identified invariant characteristic, a generalized and bounded description of the invariant segments is used which requires far less inputting of sample data and less processing of information relating to an unknown letter or number. "" dated 1996-11-05"
5574387,radial basis function neural network autoassociator and method for induction motor monitoring,""" a method for detecting a departure from normal operation of an electric motor comprises obtaining a set of normal current measurements for a motor being monitored; forming clusters of the normal current measurements; training a neural network auto-associator using the set of normal current measurements; making current measurements for the motor in operation; comparing the input and output of the auto-associator; and indicating abnormal operation whenever the current measurements deviate more than a predetermined amount from the normal current measurements. the method models a set of normal current measurements for the motor being monitored, and indicates a potential failure whenever measurements from the motor deviate significantly from a model. the model takes the form of an neural network auto-associator which is """"trained""""--using clusters of current measurements collected while the motor is known to be in a normal operating condition--to reproduce the inputs on the output. a new set of fft's of current measurements are classified as """"good"""" or """"bad"""" by first transforming the measurement using a fast fourier transform (fft) and an internal scaling procedure, and then applying a subset of the transformed measurements as inputs to the neural network auto-associator. a decision is generated based on the difference between the input and output of the network. """,1996-11-12,"The title of the patent is radial basis function neural network autoassociator and method for induction motor monitoring and its abstract is "" a method for detecting a departure from normal operation of an electric motor comprises obtaining a set of normal current measurements for a motor being monitored; forming clusters of the normal current measurements; training a neural network auto-associator using the set of normal current measurements; making current measurements for the motor in operation; comparing the input and output of the auto-associator; and indicating abnormal operation whenever the current measurements deviate more than a predetermined amount from the normal current measurements. the method models a set of normal current measurements for the motor being monitored, and indicates a potential failure whenever measurements from the motor deviate significantly from a model. the model takes the form of an neural network auto-associator which is """"trained""""--using clusters of current measurements collected while the motor is known to be in a normal operating condition--to reproduce the inputs on the output. a new set of fft's of current measurements are classified as """"good"""" or """"bad"""" by first transforming the measurement using a fast fourier transform (fft) and an internal scaling procedure, and then applying a subset of the transformed measurements as inputs to the neural network auto-associator. a decision is generated based on the difference between the input and output of the network. "" dated 1996-11-12"
5574827,method of operating a neural network,"a method is implemented in hardware or software type neural network, the neural network is constructed of neurons or neuron circuits each having only one significant processing element in the form of a multiplier. each neuron applying a gating function to each of neural network inputs. the neural network utilizes a training algorithm which does not require repetitive training and which yields a global minimum to each given set of input vectors.",1996-11-12,"The title of the patent is method of operating a neural network and its abstract is a method is implemented in hardware or software type neural network, the neural network is constructed of neurons or neuron circuits each having only one significant processing element in the form of a multiplier. each neuron applying a gating function to each of neural network inputs. the neural network utilizes a training algorithm which does not require repetitive training and which yields a global minimum to each given set of input vectors. dated 1996-11-12"
5576548,nuclear imaging enhancer,a wavelet based neural network (wnn) is proposed for image restoration using a nuclear gamma camera. the wnn first segments the image noise from the detected signal using a wavelet or multi-resolution method and then restores the detector resolution using the neural network. an example of methods is described for detecting beta particles (bremsstrahlung radiation) but the proposed wnn can be applied to either the detection of positrons or gamma rays using the gamma camera.,1996-11-19,The title of the patent is nuclear imaging enhancer and its abstract is a wavelet based neural network (wnn) is proposed for image restoration using a nuclear gamma camera. the wnn first segments the image noise from the detected signal using a wavelet or multi-resolution method and then restores the detector resolution using the neural network. an example of methods is described for detecting beta particles (bremsstrahlung radiation) but the proposed wnn can be applied to either the detection of positrons or gamma rays using the gamma camera. dated 1996-11-19
5576632,neural network auto-associator and method for induction motor monitoring,""" a method for detecting a departure from normal operation of an electric motor comprises obtaining a set of normal current measurements for a motor being monitored; training a neural network auto-associator using the set of normal current measurements; making current measurements for the motor in operation; comparing the current measurements with the normal current measurements; and indicating abnormal operation whenever the current measurements deviate more than a predetermined amount from the normal current measurements. the method models a set of normal current measurements for the motor being monitored, and indicates a potential failure whenever measurements from the motor deviate significantly from a model. the model takes the form of an neural network auto-associator which is """"trained""""--using current measurements collected while the motor is known to be in a normal operating condition--to reproduce the inputs on the output. a new set of current measurements are classified as """"good"""" or """"bad"""" by first transforming the measurement using a fast fourier transform (fft) and an internal scaling procedure, and then applying a subset of the transformed measurements as inputs to the neural network auto-associator. a decision is generated based on the difference between the input and output of the network. """,1996-11-19,"The title of the patent is neural network auto-associator and method for induction motor monitoring and its abstract is "" a method for detecting a departure from normal operation of an electric motor comprises obtaining a set of normal current measurements for a motor being monitored; training a neural network auto-associator using the set of normal current measurements; making current measurements for the motor in operation; comparing the current measurements with the normal current measurements; and indicating abnormal operation whenever the current measurements deviate more than a predetermined amount from the normal current measurements. the method models a set of normal current measurements for the motor being monitored, and indicates a potential failure whenever measurements from the motor deviate significantly from a model. the model takes the form of an neural network auto-associator which is """"trained""""--using current measurements collected while the motor is known to be in a normal operating condition--to reproduce the inputs on the output. a new set of current measurements are classified as """"good"""" or """"bad"""" by first transforming the measurement using a fast fourier transform (fft) and an internal scaling procedure, and then applying a subset of the transformed measurements as inputs to the neural network auto-associator. a decision is generated based on the difference between the input and output of the network. "" dated 1996-11-19"
5576972,intelligent area monitoring system,"an intelligent area monitoring system having a field sensor, a neural network computer, and a communications apparatus is disclosed. the system has the capability of detecting and monitoring the location and identity of people, animals, and objects within an indoor or outdoor area for the purpose of intrusion detection, theft deterrence, and accident prevention. the neural network computer accepts the input signals from the field sensor and forms a virtual model of the monitored area from the input. any changes that occur within the monitored area are communicated to system users. the sensors can be active or passive, analog or binary, and the system is optimally configured with a mix of different sensor types such as vibration, sound, infrared, optical, microwave, and ultrasonic. each analog sensor provides an analog output which varies in proportion to the size and distance of its target. after monitoring changes in the space being monitored and identifying various objects within the space, the neural network computer communicates such information to the user.",1996-11-19,"The title of the patent is intelligent area monitoring system and its abstract is an intelligent area monitoring system having a field sensor, a neural network computer, and a communications apparatus is disclosed. the system has the capability of detecting and monitoring the location and identity of people, animals, and objects within an indoor or outdoor area for the purpose of intrusion detection, theft deterrence, and accident prevention. the neural network computer accepts the input signals from the field sensor and forms a virtual model of the monitored area from the input. any changes that occur within the monitored area are communicated to system users. the sensors can be active or passive, analog or binary, and the system is optimally configured with a mix of different sensor types such as vibration, sound, infrared, optical, microwave, and ultrasonic. each analog sensor provides an analog output which varies in proportion to the size and distance of its target. after monitoring changes in the space being monitored and identifying various objects within the space, the neural network computer communicates such information to the user. dated 1996-11-19"
5577028,routing system using a neural network,"a routing system in a multimedia integrated network formed of nodes connected by links is provided for transmitting various media such as voice, image and data in a packet format. the respective nodes forming the integrated network output the packets in an optimum output direction so that conditions required by various media and reliability of communication are satisfied. each node includes an interconnection type neural network for determining the packet output direction. an external stimulus input unit outputs an external stimulus to the neurons in the neural network in response to a present state of the integrated network, such as a packet delay time and a packet loss ratio for respective links, and a condition required by the media such as an allowable packet loss ratio. therefore, the packet is output in an optimum direction which is adaptive to the present state of the integrated network and which satisfies a condition required by the media.",1996-11-19,"The title of the patent is routing system using a neural network and its abstract is a routing system in a multimedia integrated network formed of nodes connected by links is provided for transmitting various media such as voice, image and data in a packet format. the respective nodes forming the integrated network output the packets in an optimum output direction so that conditions required by various media and reliability of communication are satisfied. each node includes an interconnection type neural network for determining the packet output direction. an external stimulus input unit outputs an external stimulus to the neurons in the neural network in response to a present state of the integrated network, such as a packet delay time and a packet loss ratio for respective links, and a condition required by the media such as an allowable packet loss ratio. therefore, the packet is output in an optimum direction which is adaptive to the present state of the integrated network and which satisfies a condition required by the media. dated 1996-11-19"
5577166,method and apparatus for classifying patterns by use of neural network,"a method of and an apparatus for classifying an input pattern as a classification object by use of a neural network and transforming the input pattern into an output pattern as a result of classification, wherein there is obtained from the neural network having received an input pattern an output pattern for the input pattern such that the obtained output pattern is compared with a correct output pattern thereafter attained for the input pattern, a result of the comparison is sequentially stored in a memory, and the stored result of comparison is monitored to thereby detect an abnormality thereof.",1996-11-19,"The title of the patent is method and apparatus for classifying patterns by use of neural network and its abstract is a method of and an apparatus for classifying an input pattern as a classification object by use of a neural network and transforming the input pattern into an output pattern as a result of classification, wherein there is obtained from the neural network having received an input pattern an output pattern for the input pattern such that the obtained output pattern is compared with a correct output pattern thereafter attained for the input pattern, a result of the comparison is sequentially stored in a memory, and the stored result of comparison is monitored to thereby detect an abnormality thereof. dated 1996-11-19"
5577178,neural network for color translations,a neural network for converting pixels represented in one color representation to pixels of a second color representation. one realization of the neural network utilizes an input layer that includes either a summing node or a product generating node. the neural network may be implemented in analog circuitry or as a simulation on a general purpose data processor. the analog circuit implementation requires a relatively small number of nodes and is less expensive than implementations based on general purpose data processors.,1996-11-19,The title of the patent is neural network for color translations and its abstract is a neural network for converting pixels represented in one color representation to pixels of a second color representation. one realization of the neural network utilizes an input layer that includes either a summing node or a product generating node. the neural network may be implemented in analog circuitry or as a simulation on a general purpose data processor. the analog circuit implementation requires a relatively small number of nodes and is less expensive than implementations based on general purpose data processors. dated 1996-11-19
5579232,system and method including neural net for tool break detection,"a system and method for monitoring vibrations of a cutting tool uses a neural network for classifying signal features as break or non-break or, in another embodiment, as non-break or abnormal. a vibration signal is produced by an accelerometer, positioned to sense vibrations at the tool-workpiece interface. the signal is pre-processed to extract low frequency machining noise and detect the energy in a higher frequency band. the signal is then sampled and segments of the digitized signals are processed by digital logic into feature vectors for input to a trained neural net having two output nodes for classification. the use of a neural net provides performance improvement and economies over previously known heuristic methods of signal analysis.",1996-11-26,"The title of the patent is system and method including neural net for tool break detection and its abstract is a system and method for monitoring vibrations of a cutting tool uses a neural network for classifying signal features as break or non-break or, in another embodiment, as non-break or abnormal. a vibration signal is produced by an accelerometer, positioned to sense vibrations at the tool-workpiece interface. the signal is pre-processed to extract low frequency machining noise and detect the energy in a higher frequency band. the signal is then sampled and segments of the digitized signals are processed by digital logic into feature vectors for input to a trained neural net having two output nodes for classification. the use of a neural net provides performance improvement and economies over previously known heuristic methods of signal analysis. dated 1996-11-26"
5579245,vehicle slip angle measuring method and a device therefor,"a vehicle slip angle measuring device includes an approximate calculation block for deriving an approximate value of a vehicle centroid slip angle by use of an approximate expression derived from a linear two-degree-of-freedom vehicle model. the approximate value is derived based on outputs from a steering wheel sensor, vehicle velocity sensor and yaw angular velocity sensor. the device further includes a preprocessing block for preprocessing outputs from the above three sensors, longitudinal acceleration sensor and lateral acceleration sensor to create input information. finally, the device includes a learned neural network for receiving the approximate value and input information and for outputting a correction value corresponding to a deviation between an actual slip angle and the approximate value. the correction value from the neural network is added to the approximate value from the approximate calculation block to derive a precise vehicle centroid slip angle.",1996-11-26,"The title of the patent is vehicle slip angle measuring method and a device therefor and its abstract is a vehicle slip angle measuring device includes an approximate calculation block for deriving an approximate value of a vehicle centroid slip angle by use of an approximate expression derived from a linear two-degree-of-freedom vehicle model. the approximate value is derived based on outputs from a steering wheel sensor, vehicle velocity sensor and yaw angular velocity sensor. the device further includes a preprocessing block for preprocessing outputs from the above three sensors, longitudinal acceleration sensor and lateral acceleration sensor to create input information. finally, the device includes a learned neural network for receiving the approximate value and input information and for outputting a correction value corresponding to a deviation between an actual slip angle and the approximate value. the correction value from the neural network is added to the approximate value from the approximate calculation block to derive a precise vehicle centroid slip angle. dated 1996-11-26"
5579439,fuzzy logic design generator using a neural network to generate fuzzy logic rules and membership functions for use in intelligent systems,"a fuzzy logic design generator for providing a fuzzy logic design for an intelligent controller in a plant control system includes an artificial neural network for generating fuzzy logic rules and membership functions data. these fuzzy logic rules and membership functions data can be stored for use in a fuzzy logic system for neural network based fuzzy antecedent processing, rule evaluation and defuzzification, thereby avoiding heuristics associated with conventional fuzzy logic algorithms. the neural network, used as a fuzzy rule generator to generate fuzzy logic rules and membership functions for the system's plant controller, is a multilayered feed-forward neural network based upon a modified version of a back-propagation neural network and learns the system behavior in accordance with input and output data and then maps the acquired knowledge into a new non-heuristic fuzzy logic system. interlayer weights of the neural network are mapped into fuzzy logic rules and membership functions. antecedent processing is performed according to a weighted product of the antecedents. one layer of the neural network is used for performing rule evaluation and defuzzification.",1996-11-26,"The title of the patent is fuzzy logic design generator using a neural network to generate fuzzy logic rules and membership functions for use in intelligent systems and its abstract is a fuzzy logic design generator for providing a fuzzy logic design for an intelligent controller in a plant control system includes an artificial neural network for generating fuzzy logic rules and membership functions data. these fuzzy logic rules and membership functions data can be stored for use in a fuzzy logic system for neural network based fuzzy antecedent processing, rule evaluation and defuzzification, thereby avoiding heuristics associated with conventional fuzzy logic algorithms. the neural network, used as a fuzzy rule generator to generate fuzzy logic rules and membership functions for the system's plant controller, is a multilayered feed-forward neural network based upon a modified version of a back-propagation neural network and learns the system behavior in accordance with input and output data and then maps the acquired knowledge into a new non-heuristic fuzzy logic system. interlayer weights of the neural network are mapped into fuzzy logic rules and membership functions. antecedent processing is performed according to a weighted product of the antecedents. one layer of the neural network is used for performing rule evaluation and defuzzification. dated 1996-11-26"
5579484,system for performing fast data accessing in multiply/accumulate operations while using a vram,"a system and method are described which provide for fast sequential access of stored data, as required, for example, in the performance of multiply/accumulate operations in neural network calculations and other sequential computations. the system includes a video dynamic random access memory (vram) with a shift register, a digital signal processor (dsp), and a data/address/control bus for coupling the dsp and vram. the vram has a parallel access port and a serial access port, and the dsp has a first input-output (i/o) port and a second i/o port. the second i/o port of the dsp is coupled, via the bus, to the parallel- and serial-access ports of the vram. in response to data applied via the first i/o port, the dsp transfers the applied data, via the second i/o port, the data bus, and the parallel (or serial) access port, to the vram for storage. the stored data is then accessed (by the dsp) as a serial sequence via the serial access port. the method employed by the system in providing such access includes the steps of transferring the stored data in parallel to the shift register, and outputting the data as a synchronized (clocked) serial sequence from the shift register.",1996-11-26,"The title of the patent is system for performing fast data accessing in multiply/accumulate operations while using a vram and its abstract is a system and method are described which provide for fast sequential access of stored data, as required, for example, in the performance of multiply/accumulate operations in neural network calculations and other sequential computations. the system includes a video dynamic random access memory (vram) with a shift register, a digital signal processor (dsp), and a data/address/control bus for coupling the dsp and vram. the vram has a parallel access port and a serial access port, and the dsp has a first input-output (i/o) port and a second i/o port. the second i/o port of the dsp is coupled, via the bus, to the parallel- and serial-access ports of the vram. in response to data applied via the first i/o port, the dsp transfers the applied data, via the second i/o port, the data bus, and the parallel (or serial) access port, to the vram for storage. the stored data is then accessed (by the dsp) as a serial sequence via the serial access port. the method employed by the system in providing such access includes the steps of transferring the stored data in parallel to the shift register, and outputting the data as a synchronized (clocked) serial sequence from the shift register. dated 1996-11-26"
5579778,method and apparatus for producing thermodilution cardiac output measurements utilizing a neural network,"a method and device for indirect, quantitative estimation of cardiac output utilizing invasive, indirect techniques. the method of practice includes (i) generating a sequence of signals which are quantitatively dependent upon cardiac output, (ii) transmitting and processing the signals within a computer system and associated neural network capable of generating a single output signal for the combined input signals, (iii) directly determining an actual value for the parameter concurrent with the invasive generation of signals of the previous steps, (iv) applying weighting factors within the neural network at interconnecting nodes to force the output signal of the neural network to match the known value of the parameter as determined invasively, (v) recording the input signals, weighting factors and known value as training data within memory of the computer, and (vi) repeating the previous steps to develop sufficient training data to enable the neural network to accurately estimate parameter value upon future receipt of on-line input signals. procedures are also described for preclassification of signals and artifact rejection. following training of the neural network, further direct measurement is unnecessary and the system is ready for diagnostic application and invasive estimation of parameter values.",1996-12-03,"The title of the patent is method and apparatus for producing thermodilution cardiac output measurements utilizing a neural network and its abstract is a method and device for indirect, quantitative estimation of cardiac output utilizing invasive, indirect techniques. the method of practice includes (i) generating a sequence of signals which are quantitatively dependent upon cardiac output, (ii) transmitting and processing the signals within a computer system and associated neural network capable of generating a single output signal for the combined input signals, (iii) directly determining an actual value for the parameter concurrent with the invasive generation of signals of the previous steps, (iv) applying weighting factors within the neural network at interconnecting nodes to force the output signal of the neural network to match the known value of the parameter as determined invasively, (v) recording the input signals, weighting factors and known value as training data within memory of the computer, and (vi) repeating the previous steps to develop sufficient training data to enable the neural network to accurately estimate parameter value upon future receipt of on-line input signals. procedures are also described for preclassification of signals and artifact rejection. following training of the neural network, further direct measurement is unnecessary and the system is ready for diagnostic application and invasive estimation of parameter values. dated 1996-12-03"
5579993,hvac distribution system identification,"a controller implemented in a heating, ventilation and air-conditioning (hvac) distribution system provides improved control by implementing a general regression neural network to generate a control signal based on identified characteristics of components utilized within the hvac system. the general regression neural network utilizes past characteristics and desired (calculated) characteristics to generate an output control signal. the general regression neural network is implemented in a feedforward process, which is combined with a feedback process to generate an improved control signal to control components within the hvac distribution system. implementation of the general regression neural network is simple and accurate, requires no input from an operator supervising the controller, and provides adaptive, real-time control of components within the hvac distribution system.",1996-12-03,"The title of the patent is hvac distribution system identification and its abstract is a controller implemented in a heating, ventilation and air-conditioning (hvac) distribution system provides improved control by implementing a general regression neural network to generate a control signal based on identified characteristics of components utilized within the hvac system. the general regression neural network utilizes past characteristics and desired (calculated) characteristics to generate an output control signal. the general regression neural network is implemented in a feedforward process, which is combined with a feedback process to generate an improved control signal to control components within the hvac distribution system. implementation of the general regression neural network is simple and accurate, requires no input from an operator supervising the controller, and provides adaptive, real-time control of components within the hvac distribution system. dated 1996-12-03"
5581459,plant operation support system,"when amounts of operation control variables and/or operation state evaluation variables of a plant are calculated through a multilayered neural network based on data indicating a plant state, and the calculated results are indicated to an operator, a plant operation support system has functions to analyze an internal causality between neurons in the neural network, and display quantitative guidance by association in the neural network and also the analyzed result of the internal causality between the neurons as an association reason.",1996-12-03,"The title of the patent is plant operation support system and its abstract is when amounts of operation control variables and/or operation state evaluation variables of a plant are calculated through a multilayered neural network based on data indicating a plant state, and the calculated results are indicated to an operator, a plant operation support system has functions to analyze an internal causality between neurons in the neural network, and display quantitative guidance by association in the neural network and also the analyzed result of the internal causality between the neurons as an association reason. dated 1996-12-03"
5581658,adaptive system for broadcast program identification and reporting,""" a computer-implemented method and system for monitoring, identifying, classifying and logging musical work performance broadcasts over the public airwaves. the system uses a neural network to classify specially-processed """"retinal"""" signatures of the musical work performance. the neural network is trained for each musical work using a single noise-biased retinal sample of the spectral distribution of preselected dynamic features of the corresponding audio signal. a detection decision is made at the neural network output using fuzzy logic circuitry to compare results of predetermined thresholding. the system of this invention fully automates the real-time identification of broadcast musical work performances. """,1996-12-03,"The title of the patent is adaptive system for broadcast program identification and reporting and its abstract is "" a computer-implemented method and system for monitoring, identifying, classifying and logging musical work performance broadcasts over the public airwaves. the system uses a neural network to classify specially-processed """"retinal"""" signatures of the musical work performance. the neural network is trained for each musical work using a single noise-biased retinal sample of the spectral distribution of preselected dynamic features of the corresponding audio signal. a detection decision is made at the neural network output using fuzzy logic circuitry to compare results of predetermined thresholding. the system of this invention fully automates the real-time identification of broadcast musical work performances. "" dated 1996-12-03"
5581659,system for solving an adaptation problem by adjusting weights of a neural network based on changes in a cost function,"a system and method for solving adaptation problems including a neural network having multiple neurons. each neuron includes outputs that are input to every other neuron excepting itself, and a cost calculation portion for calculating cost based on a formula that uses outputs from each neuron as variables with respect to the given adaptation problem. adaptation problems are solved by i) randomly setting a coupling weight for each neuron based on other neuron operations and an output of each neuron in the initial state, ii) calculating a net value with respect to at least one neuron, iii) calculating cost by transmitting the output of each neuron after the net value calculation to the cost calculation portion, iv) simultaneously feeding back the difference between the cost before and after the net calculation to a neural network of mutual connection, v) correcting the coupling weight of at least one neuron according to the difference, and repeatedly changing the state of the neural network until a predetermined condition is established, and vi) changing the state of the neural network after the predetermined condition is satisfied, realizing the solution to the adaptation problem based on the output of each neuron. it is unnecessary to determine the coupling weight of synapse preliminarily when using the present invention. the present invention can solve adaptation problems having various cost functions to which conventional models cannot be applied.",1996-12-03,"The title of the patent is system for solving an adaptation problem by adjusting weights of a neural network based on changes in a cost function and its abstract is a system and method for solving adaptation problems including a neural network having multiple neurons. each neuron includes outputs that are input to every other neuron excepting itself, and a cost calculation portion for calculating cost based on a formula that uses outputs from each neuron as variables with respect to the given adaptation problem. adaptation problems are solved by i) randomly setting a coupling weight for each neuron based on other neuron operations and an output of each neuron in the initial state, ii) calculating a net value with respect to at least one neuron, iii) calculating cost by transmitting the output of each neuron after the net value calculation to the cost calculation portion, iv) simultaneously feeding back the difference between the cost before and after the net calculation to a neural network of mutual connection, v) correcting the coupling weight of at least one neuron according to the difference, and repeatedly changing the state of the neural network until a predetermined condition is established, and vi) changing the state of the neural network after the predetermined condition is satisfied, realizing the solution to the adaptation problem based on the output of each neuron. it is unnecessary to determine the coupling weight of synapse preliminarily when using the present invention. the present invention can solve adaptation problems having various cost functions to which conventional models cannot be applied. dated 1996-12-03"
5581660,neural network structure having lateral interconnections,"a neural network layer (1) is made up of nodes or neurons which each comprise a pair of physically separate and optically coupled sub-units (x.sub.1, y.sub.1). one sub-unit broadcasts excitatory and receives inhibitory signals, whereas the other sub-unit broadcasts inhibitory and receives the excitatory signals. an electrical feedback connection (20) is provided between corresponding sub-units for determination of net node activation. diffractive or holographic optical elements (2) are used for optical coupling.",1996-12-03,"The title of the patent is neural network structure having lateral interconnections and its abstract is a neural network layer (1) is made up of nodes or neurons which each comprise a pair of physically separate and optically coupled sub-units (x.sub.1, y.sub.1). one sub-unit broadcasts excitatory and receives inhibitory signals, whereas the other sub-unit broadcasts inhibitory and receives the excitatory signals. an electrical feedback connection (20) is provided between corresponding sub-units for determination of net node activation. diffractive or holographic optical elements (2) are used for optical coupling. dated 1996-12-03"
5583769,automatic train operation apparatus incorporating security function with improved reliability,"an automatic train operation apparatus capable of realizing an optimal train operation with an improved reliability. the apparatus includes: an ato/c system including a plurality of ato/c units redundantly provided, each ato/c unit having a fail safe configuration formed by a plurality of execution processors and a supervisor processor for monitoring normal operations of the execution processors, each execution processor being capable of executing an automatic train operation program, and all of the execution processors in the plurality of ato/c units executing an identical automatic train operation program simultaneously; and a majority logic circuit for selecting an output obtained by a majority of the execution processors in the ato/c units of the ato/c system as a control command output for controlling a train operation. each execution processor in the ato/c system operates as a neural network with a learning function.",1996-12-10,"The title of the patent is automatic train operation apparatus incorporating security function with improved reliability and its abstract is an automatic train operation apparatus capable of realizing an optimal train operation with an improved reliability. the apparatus includes: an ato/c system including a plurality of ato/c units redundantly provided, each ato/c unit having a fail safe configuration formed by a plurality of execution processors and a supervisor processor for monitoring normal operations of the execution processors, each execution processor being capable of executing an automatic train operation program, and all of the execution processors in the plurality of ato/c units executing an identical automatic train operation program simultaneously; and a majority logic circuit for selecting an output obtained by a majority of the execution processors in the ato/c units of the ato/c system as a control command output for controlling a train operation. each execution processor in the ato/c system operates as a neural network with a learning function. dated 1996-12-10"
5583771,method and apparatus for distinguishing between deployment events and non-deployment events in an sir system,"a pattern recognition system is utilized in a supplementary inflatable restraint (sir) system to distinguish between deployment and non-deployment events. the pattern recognition system preferably includes dedicated hardware or a microprocessor programmed to perform a neural network simulation utilizing crash data in the form of vehicle acceleration data. training and trial vectors are generated from the data to train and, subsequently, test the neural network.",1996-12-10,"The title of the patent is method and apparatus for distinguishing between deployment events and non-deployment events in an sir system and its abstract is a pattern recognition system is utilized in a supplementary inflatable restraint (sir) system to distinguish between deployment and non-deployment events. the pattern recognition system preferably includes dedicated hardware or a microprocessor programmed to perform a neural network simulation utilizing crash data in the form of vehicle acceleration data. training and trial vectors are generated from the data to train and, subsequently, test the neural network. dated 1996-12-10"
5583964,computer utilizing neural network and method of using same,"a computing device, which may be implemented as an integrated circuit, is constructed of a microprocessor and one or more neural network co-processors. the microprocessor normally executes programs which transfer data to the neural network co-processors, which are used to compute complicated mathematical functions. direct memory access (dma) is also used to transfer data. each neural network co-processor interfaces to the microprocessor in a manner substantially similar to that of a conventional memory device. the co-processor does not require any instructions and is configured to execute mathematical operations simply by being pre-loaded with gating functions and weight values. in addition, the co-processor executes a plurality of arithmetic operations in parallel, and the results of such operations are simply read from the co-processor.",1996-12-10,"The title of the patent is computer utilizing neural network and method of using same and its abstract is a computing device, which may be implemented as an integrated circuit, is constructed of a microprocessor and one or more neural network co-processors. the microprocessor normally executes programs which transfer data to the neural network co-processors, which are used to compute complicated mathematical functions. direct memory access (dma) is also used to transfer data. each neural network co-processor interfaces to the microprocessor in a manner substantially similar to that of a conventional memory device. the co-processor does not require any instructions and is configured to execute mathematical operations simply by being pre-loaded with gating functions and weight values. in addition, the co-processor executes a plurality of arithmetic operations in parallel, and the results of such operations are simply read from the co-processor. dated 1996-12-10"
5583968,noise reduction for speech recognition,"a neural network for noise reduction for speech recognition in a noisy environment uses an algorithm for automatic network generation which automatically selects a suitable signal representation. nodes may be added to the input layer of the neural network successively, with a new node being trained by calculating and minimizing a mapping error. a squared mapping error may be formed and the mapping error may be assigned a weight dependent on the importance of the vectors. in addition, a neural network that performs neural noise reduction by reducing, in a training phase, a mapping error between noise-free vectors at an output of the neural network and noise-reduced vectors at the output of the neural network using an iterative process, has the mapping error further reduced by additional information which is selected from a suitable signal representation at the input of the neural network.",1996-12-10,"The title of the patent is noise reduction for speech recognition and its abstract is a neural network for noise reduction for speech recognition in a noisy environment uses an algorithm for automatic network generation which automatically selects a suitable signal representation. nodes may be added to the input layer of the neural network successively, with a new node being trained by calculating and minimizing a mapping error. a squared mapping error may be formed and the mapping error may be assigned a weight dependent on the importance of the vectors. in addition, a neural network that performs neural noise reduction by reducing, in a training phase, a mapping error between noise-free vectors at an output of the neural network and noise-reduced vectors at the output of the neural network using an iterative process, has the mapping error further reduced by additional information which is selected from a suitable signal representation at the input of the neural network. dated 1996-12-10"
5585890,neural network controlled image copying apparatus,"an image forming apparatus is disclosed which, by using a neural network to have preliminarily learned input conditions for a number of characteristic originals and having only the coupling coefficient after the learning and the output calculating portion to make calculations when necessary for the controlling in the apparatus itself, is capable of automatically performing an optimum copying control corresponding to various conditions of each of various originals.",1996-12-17,"The title of the patent is neural network controlled image copying apparatus and its abstract is an image forming apparatus is disclosed which, by using a neural network to have preliminarily learned input conditions for a number of characteristic originals and having only the coupling coefficient after the learning and the output calculating portion to make calculations when necessary for the controlling in the apparatus itself, is capable of automatically performing an optimum copying control corresponding to various conditions of each of various originals. dated 1996-12-17"
5586028,road surface condition-detecting system and anti-lock brake system employing same,"a road surface condition-detecting system for a vehicle detects a road surface condition from road noise generated by a vehicle wheel. the road surface condition is determined based on parameter data of frequency components of the road noise, by a neural network. the road noise may be corrected by eliminating therefrom a disturbance, such as audio output and exhaust noise. a present state of the road surface condition may be determined based on at least two consecutive determinations made based on the road noise detected at regular time intervals. exclusive neural networks may be used for respective road surface condition types. one of a plurality of neural networks provided for respective vehicle speed ranges may be selected according to an actual vehicle speed. detected sound pressure levels of the road noise extracted by frequency analysis may be normalized within respective ranges defined by upper and lower limits set corresponding to predetermined frequency ranges before being supplied to the neural network.",1996-12-17,"The title of the patent is road surface condition-detecting system and anti-lock brake system employing same and its abstract is a road surface condition-detecting system for a vehicle detects a road surface condition from road noise generated by a vehicle wheel. the road surface condition is determined based on parameter data of frequency components of the road noise, by a neural network. the road noise may be corrected by eliminating therefrom a disturbance, such as audio output and exhaust noise. a present state of the road surface condition may be determined based on at least two consecutive determinations made based on the road noise detected at regular time intervals. exclusive neural networks may be used for respective road surface condition types. one of a plurality of neural networks provided for respective vehicle speed ranges may be selected according to an actual vehicle speed. detected sound pressure levels of the road noise extracted by frequency analysis may be normalized within respective ranges defined by upper and lower limits set corresponding to predetermined frequency ranges before being supplied to the neural network. dated 1996-12-17"
5586033,control system with neural network trained as general and local models,"a neural network is trained with a general set of data to function as a general model of a machine or process with local condition inputs set equal to zero. the network is then retrained or receives additional training on an extentd data set containing the general set of data, characterized by zero values for the local condition inputs, and data on specific local conditions, characterized by non-zero values for the local condition inputs. the result is a trained neural network which functions as a general model when the inputs for the local conditions inputs are set equal to zero, and which functions as a model of some specific local condition when the local condition inputs match the encoding of the some local data set contained within the training data. the neural network has an architecture and a number of neurons such that its functioning as the local model is partially dependent upon its functioning as the general model. this trained neural network is combined with sensors, actuators, a control and communications computer and with a user interface to function as combine control system.",1996-12-17,"The title of the patent is control system with neural network trained as general and local models and its abstract is a neural network is trained with a general set of data to function as a general model of a machine or process with local condition inputs set equal to zero. the network is then retrained or receives additional training on an extentd data set containing the general set of data, characterized by zero values for the local condition inputs, and data on specific local conditions, characterized by non-zero values for the local condition inputs. the result is a trained neural network which functions as a general model when the inputs for the local conditions inputs are set equal to zero, and which functions as a model of some specific local condition when the local condition inputs match the encoding of the some local data set contained within the training data. the neural network has an architecture and a number of neurons such that its functioning as the local model is partially dependent upon its functioning as the general model. this trained neural network is combined with sensors, actuators, a control and communications computer and with a user interface to function as combine control system. dated 1996-12-17"
5586215,neural network acoustic and visual speech recognition system,"the apparatus for the recognition of speech comprises an acoustic preprocessor, a visual preprocessor, and a speech classifier that operates on the acoustic and visual preprocessed data. the acoustic preprocessor comprises a log mel spectrum analyzer that produces an equal mel bandwidth log power spectrum. the visual processor detects the motion of a set of fiducial markers on the speaker's face and extracts a set of normalized distance vectors describing lip and mouth movement. the speech classifier uses a multilevel time-delay neural network operating on the preprocessed acoustic and visual data to form an output probability distribution that indicates the probability of each candidate utterance having been spoken, based on the acoustic and visual data.",1996-12-17,"The title of the patent is neural network acoustic and visual speech recognition system and its abstract is the apparatus for the recognition of speech comprises an acoustic preprocessor, a visual preprocessor, and a speech classifier that operates on the acoustic and visual preprocessed data. the acoustic preprocessor comprises a log mel spectrum analyzer that produces an equal mel bandwidth log power spectrum. the visual processor detects the motion of a set of fiducial markers on the speaker's face and extracts a set of normalized distance vectors describing lip and mouth movement. the speech classifier uses a multilevel time-delay neural network operating on the preprocessed acoustic and visual data to form an output probability distribution that indicates the probability of each candidate utterance having been spoken, based on the acoustic and visual data. dated 1996-12-17"
5586220,safe system provided with neural circuit,"disclosed is an electronic control circuit of the fail-safe emergency shut-down type which includes an input circuit, a first signal processing circuit, in which information is stored beforehand relating to extreme values input signals are permitted to reach, and an output circuit. the first signal processing circuit includes at least one neural network configured from very large numbers of neurons operating as integrators and in real-time which operate in parallel and which are mutually connected on a large scale. the networks are implemented in hardware and the extreme values are distributed over the neurons. when one of the extreme values is exceeded, the output circuit generates a shut-off signal and a component controlled thereby is placed in a safe state.",1996-12-17,"The title of the patent is safe system provided with neural circuit and its abstract is disclosed is an electronic control circuit of the fail-safe emergency shut-down type which includes an input circuit, a first signal processing circuit, in which information is stored beforehand relating to extreme values input signals are permitted to reach, and an output circuit. the first signal processing circuit includes at least one neural network configured from very large numbers of neurons operating as integrators and in real-time which operate in parallel and which are mutually connected on a large scale. the networks are implemented in hardware and the extreme values are distributed over the neurons. when one of the extreme values is exceeded, the output circuit generates a shut-off signal and a component controlled thereby is placed in a safe state. dated 1996-12-17"
5586221,predictive control of rolling mills using neural network gauge estimation,"a system for controlling the output of a rolling mill. an intelligent control system is part of a control loop between the mill and a pid controller. the control loop does not rely on the output of an exit gauge sensor in normal operation. the intelligent control system can be an artificial neural network or a parallel cascade network, and has an output node for generating an output signal that is predictive of the exit gauge at a future time. a comparator coupled to the artificial neural network output signal and to a reference signal derives an error signal which is fed to the pid controller for modulating the metal thickness.",1996-12-17,"The title of the patent is predictive control of rolling mills using neural network gauge estimation and its abstract is a system for controlling the output of a rolling mill. an intelligent control system is part of a control loop between the mill and a pid controller. the control loop does not rely on the output of an exit gauge sensor in normal operation. the intelligent control system can be an artificial neural network or a parallel cascade network, and has an output node for generating an output signal that is predictive of the exit gauge at a future time. a comparator coupled to the artificial neural network output signal and to a reference signal derives an error signal which is fed to the pid controller for modulating the metal thickness. dated 1996-12-17"
5586223,high speed segmented neural network and fabrication method,"a high speed, feed forward, segmented neural network and fabrication technique are described. the segmented network includes a plurality of network layers stacked in an ascending pyramid fashion. the network layers are structured with a plurality of subnetworks, and within each subnetwork exists a plurality of nodes structured in a fully interconnected and/or partially interconnected layered neural network arrangement. the inputs and outputs of each subnetwork are one bit digital values constrained to `0` or `1`, while any number of nodes with any number of layers may be modeled for each subnetwork. each subnetwork is independent of all other subnetworks in a given network layer, and thus, each network layer is segmented. in hardware implementation, each subnetwork comprises a simple memory device, such as a ram or prom look-up table. the speed of the neural network system is high and largely dictated by the access time of the memory devices used.",1996-12-17,"The title of the patent is high speed segmented neural network and fabrication method and its abstract is a high speed, feed forward, segmented neural network and fabrication technique are described. the segmented network includes a plurality of network layers stacked in an ascending pyramid fashion. the network layers are structured with a plurality of subnetworks, and within each subnetwork exists a plurality of nodes structured in a fully interconnected and/or partially interconnected layered neural network arrangement. the inputs and outputs of each subnetwork are one bit digital values constrained to `0` or `1`, while any number of nodes with any number of layers may be modeled for each subnetwork. each subnetwork is independent of all other subnetworks in a given network layer, and thus, each network layer is segmented. in hardware implementation, each subnetwork comprises a simple memory device, such as a ram or prom look-up table. the speed of the neural network system is high and largely dictated by the access time of the memory devices used. dated 1996-12-17"
5588091,dynamically stable associative learning neural network system,""" a dynamically stable associative learning neural network system includes, in its basic architectural unit, at least one each of a conditioned signal input, an unconditioned signal input and an output. interposed between input and output elements are """"patches,"""" or storage areas of dynamic interaction between conditioned and unconditioned signals which process information to achieve associative learning locally under rules designed for application-related goals of the system. patches may be fixed or variable in size. adjustments to a patch radius may be by """"pruning"""" or """"budding."""" the neural network is taught by successive application of training sets of input signals to the input terminals until a dynamic equilibrium is reached. enhancements and expansions of the basic unit result in multilayered (multi-subnetworked) systems having increased capabilities for complex pattern classification and feature recognition. """,1996-12-24,"The title of the patent is dynamically stable associative learning neural network system and its abstract is "" a dynamically stable associative learning neural network system includes, in its basic architectural unit, at least one each of a conditioned signal input, an unconditioned signal input and an output. interposed between input and output elements are """"patches,"""" or storage areas of dynamic interaction between conditioned and unconditioned signals which process information to achieve associative learning locally under rules designed for application-related goals of the system. patches may be fixed or variable in size. adjustments to a patch radius may be by """"pruning"""" or """"budding."""" the neural network is taught by successive application of training sets of input signals to the input terminals until a dynamic equilibrium is reached. enhancements and expansions of the basic unit result in multilayered (multi-subnetworked) systems having increased capabilities for complex pattern classification and feature recognition. "" dated 1996-12-24"
5590218,unsupervised neural network classification with back propagation,"an unsupervised back propagation method for training neural networks. for a set of inputs, target outputs are assigned l's and o's randomly or arbitrarily for a small number of outputs. the learning process is initiated and the convergence of outputs towards targets is monitored. at intervals, the learning is paused, and the values for those targets for the outputs which are converging at a less than average rate, are changed (e.g., 0.fwdarw.1, or 1.fwdarw.0), and the learning is then resumed with the new targets. the process is continuously iterated and the outputs converge on a stable classification, thereby providing unsupervised back propagation. in a further embodiment, samples classified with the trained network may serve as the training sets for additional subdivisions to grow additional layers of a hierarchical classification tree which converges to indivisible branch tips. after training is completed, such a tree may be used to classify new unlabelled samples with high efficiency. in yet another embodiment, the unsupervised back propagation method of the present invention may be adapted to classify fuzzy sets.",1996-12-31,"The title of the patent is unsupervised neural network classification with back propagation and its abstract is an unsupervised back propagation method for training neural networks. for a set of inputs, target outputs are assigned l's and o's randomly or arbitrarily for a small number of outputs. the learning process is initiated and the convergence of outputs towards targets is monitored. at intervals, the learning is paused, and the values for those targets for the outputs which are converging at a less than average rate, are changed (e.g., 0.fwdarw.1, or 1.fwdarw.0), and the learning is then resumed with the new targets. the process is continuously iterated and the outputs converge on a stable classification, thereby providing unsupervised back propagation. in a further embodiment, samples classified with the trained network may serve as the training sets for additional subdivisions to grow additional layers of a hierarchical classification tree which converges to indivisible branch tips. after training is completed, such a tree may be used to classify new unlabelled samples with high efficiency. in yet another embodiment, the unsupervised back propagation method of the present invention may be adapted to classify fuzzy sets. dated 1996-12-31"
5590243,neural network system with sampling data,"a neural network system includes input, intermediate and output layers, each layer containing at least one neural network element, each having an input and output, for simulating a neuron; and a plurality of inter-layer connections between neural elements wherein each input layer element has a connection to at least one intermediate layer element, and each intermediate layer element has a connection to at least one output layer element. each inter-layer connection has a connecting weight. the system further includes sampling data and teaming data. the sampling data has pairs of values, each pair including an input value and corresponding output value, the input value having regular intervals. the learning data has at least three pairs of values, each pair including an input value and corresponding desired output value, the input values having irregular intervals. the intermediate layer elements are assigned unique sampling data value pairs and have unique sampling functions derived by translating original sampling functions by sampling data input values assigned to the neural elements. the sampling function defines a relationship between the input and output of the neural element. the connecting weight for each connection between an intermediate layer element and an output layer element is set to the sampling data output value assigned to the intermediate layer element. the system further includes a training mechanism that adjusts the connecting weights to minimize errors between learning data output values and actual output values obtained by applying the learning data input values to the neural network.",1996-12-31,"The title of the patent is neural network system with sampling data and its abstract is a neural network system includes input, intermediate and output layers, each layer containing at least one neural network element, each having an input and output, for simulating a neuron; and a plurality of inter-layer connections between neural elements wherein each input layer element has a connection to at least one intermediate layer element, and each intermediate layer element has a connection to at least one output layer element. each inter-layer connection has a connecting weight. the system further includes sampling data and teaming data. the sampling data has pairs of values, each pair including an input value and corresponding output value, the input value having regular intervals. the learning data has at least three pairs of values, each pair including an input value and corresponding desired output value, the input values having irregular intervals. the intermediate layer elements are assigned unique sampling data value pairs and have unique sampling functions derived by translating original sampling functions by sampling data input values assigned to the neural elements. the sampling function defines a relationship between the input and output of the neural element. the connecting weight for each connection between an intermediate layer element and an output layer element is set to the sampling data output value assigned to the intermediate layer element. the system further includes a training mechanism that adjusts the connecting weights to minimize errors between learning data output values and actual output values obtained by applying the learning data input values to the neural network. dated 1996-12-31"
5590665,method of diagnosing cerebral infarction,"a novel method of diagnosing cerebral infarction using a neural network, wherein plural sets of data previously obtained from healthy and sick persons, each including an age, measured values of coagulo-fibrinolytic molecular markers ( e.g., d-dimer, tat and pap) , an index indicative of the state of cerebral infarction (e.g., 0 for healthy persons and 1 for sick persons) and the like, are repeatedly input into a neural network to let it learn the correlation of these characteristics and, thereafter, a set of data of a person to be diagnosed, including his age, measured values of the coagulo-fibrinolytic molecular markers and the like, are input in the neural network to obtain an index indicative of his state of cerebral infarction as a degree of dangerousness of cerebral infarction. this method is significantly higher in accuracy as compared with the prior art methods using the same data.",1997-01-07,"The title of the patent is method of diagnosing cerebral infarction and its abstract is a novel method of diagnosing cerebral infarction using a neural network, wherein plural sets of data previously obtained from healthy and sick persons, each including an age, measured values of coagulo-fibrinolytic molecular markers ( e.g., d-dimer, tat and pap) , an index indicative of the state of cerebral infarction (e.g., 0 for healthy persons and 1 for sick persons) and the like, are repeatedly input into a neural network to let it learn the correlation of these characteristics and, thereafter, a set of data of a person to be diagnosed, including his age, measured values of the coagulo-fibrinolytic molecular markers and the like, are input in the neural network to obtain an index indicative of his state of cerebral infarction as a degree of dangerousness of cerebral infarction. this method is significantly higher in accuracy as compared with the prior art methods using the same data. dated 1997-01-07"
5594597,neural network disk drive read channel pattern detector,a magnetic data recording system includes an electronically trainable neural network in the read channel. the neural network recognizes reproduced patterns of data bits regardless of distortion introduced by the recording and reproduction process. an electronically trainable analog neural network has weights which are set so that the network recognizes a limited number of bit patterns which are stored on a magnetic disk.,1997-01-14,The title of the patent is neural network disk drive read channel pattern detector and its abstract is a magnetic data recording system includes an electronically trainable neural network in the read channel. the neural network recognizes reproduced patterns of data bits regardless of distortion introduced by the recording and reproduction process. an electronically trainable analog neural network has weights which are set so that the network recognizes a limited number of bit patterns which are stored on a magnetic disk. dated 1997-01-14
5594916,neural network processing system using semiconductor memories and processing paired data in parallel,"a data processing system has a memory for realizing large-scale and high-speed parallel distributed processing and, especially, a data processing system for neural network processing. the neural network processing system comprises: a memory circuit for storing neuron output values, connection weights, the desired values of outputs, and data necessary for learning; an input/output circuit for writing or reading data in or out of said memory circuit; a processing circuit for performing a processing for determining the neuron outputs such as the product, sum and nonlinear conversion of the data stored in said memory circuit, a comparison of the output value and its desired value, and a processing necessary for learning; and a control circuit for controlling the operation of the memory circuit, the input/output circuit and the processing circuit. the processing circuit includes at least one of an address, a multiplier, a nonlinear transfer function circuit and a comparator so that at least a portion of the processing necessary for determining the neuron output values such as the product of sum may be accomplished in parallel. moreover, these circuits are shared among a plurality of neurons and are operated in a time sharing manner to determine the plural neuron output values. still moreover, the aforementioned comparator compares the neuron output value determined and the desired value of the output in parallel.",1997-01-14,"The title of the patent is neural network processing system using semiconductor memories and processing paired data in parallel and its abstract is a data processing system has a memory for realizing large-scale and high-speed parallel distributed processing and, especially, a data processing system for neural network processing. the neural network processing system comprises: a memory circuit for storing neuron output values, connection weights, the desired values of outputs, and data necessary for learning; an input/output circuit for writing or reading data in or out of said memory circuit; a processing circuit for performing a processing for determining the neuron outputs such as the product, sum and nonlinear conversion of the data stored in said memory circuit, a comparison of the output value and its desired value, and a processing necessary for learning; and a control circuit for controlling the operation of the memory circuit, the input/output circuit and the processing circuit. the processing circuit includes at least one of an address, a multiplier, a nonlinear transfer function circuit and a comparator so that at least a portion of the processing necessary for determining the neuron output values such as the product of sum may be accomplished in parallel. moreover, these circuits are shared among a plurality of neurons and are operated in a time sharing manner to determine the plural neuron output values. still moreover, the aforementioned comparator compares the neuron output value determined and the desired value of the output in parallel. dated 1997-01-14"
5596681,method of determining an optimal number of neurons contained in hidden layers of a neural network,"an object of the present invention is to determine the optimal number of neurons in the hidden layers of a feed-forward neural network. the number of the neurons in the hidden layers corresponds to the number of the independent variables of a linear question and the minimum number of the variables required for solving a linear question can be obtained from the rank value in the matrix theory. therefore, the rank value corresponds to the minimum number of the neurons required for the hidden layers. accordingly, when the relation between the neural network constructed and trained is memorized in matrix and the rank value is obtained from this matrix, if the number of the neurons in use is larger than the rank value, as it means that redundant neurons exist which correspond to dependent variables, such redundant neurons can be eliminated. in many cases, due to errors in calculation, the diagonal elements of the matrix are not reduced to 0 and consequently the rank value can not be determined. by neglecting the diagonal elements that are smaller than the specific value e, however, the rank value can be estimated within the range of the error.",1997-01-21,"The title of the patent is method of determining an optimal number of neurons contained in hidden layers of a neural network and its abstract is an object of the present invention is to determine the optimal number of neurons in the hidden layers of a feed-forward neural network. the number of the neurons in the hidden layers corresponds to the number of the independent variables of a linear question and the minimum number of the variables required for solving a linear question can be obtained from the rank value in the matrix theory. therefore, the rank value corresponds to the minimum number of the neurons required for the hidden layers. accordingly, when the relation between the neural network constructed and trained is memorized in matrix and the rank value is obtained from this matrix, if the number of the neurons in use is larger than the rank value, as it means that redundant neurons exist which correspond to dependent variables, such redundant neurons can be eliminated. in many cases, due to errors in calculation, the diagonal elements of the matrix are not reduced to 0 and consequently the rank value can not be determined. by neglecting the diagonal elements that are smaller than the specific value e, however, the rank value can be estimated within the range of the error. dated 1997-01-21"
5598354,motion video compression system with neural network having winner-take-all function,"a motion video data system includes a compression system, including an image compressor, an image decompressor correlative to the image compressor having an input connected to an output of the image compressor, a feedback summing node having one input connected to an output of the image decompressor, a picture memory having an input connected to an output of the feedback summing node, apparatus for comparing an image stored in the picture memory with a received input image and deducing therefrom pixels having differences between the stored image and the received image and for retrieving from the picture memory a partial image including the pixels only and applying the partial image to another input of the feedback summing node, whereby to produce at the output of the feedback summing node an updated decompressed image, a subtraction node having one input connected to received the received image and another input connected to receive the partial image so as to generate a difference image, the image compressor having an input connected to receive the difference image whereby to produce a compressed difference image at the output of the image compressor.",1997-01-28,"The title of the patent is motion video compression system with neural network having winner-take-all function and its abstract is a motion video data system includes a compression system, including an image compressor, an image decompressor correlative to the image compressor having an input connected to an output of the image compressor, a feedback summing node having one input connected to an output of the image decompressor, a picture memory having an input connected to an output of the feedback summing node, apparatus for comparing an image stored in the picture memory with a received input image and deducing therefrom pixels having differences between the stored image and the received image and for retrieving from the picture memory a partial image including the pixels only and applying the partial image to another input of the feedback summing node, whereby to produce at the output of the feedback summing node an updated decompressed image, a subtraction node having one input connected to received the received image and another input connected to receive the partial image so as to generate a difference image, the image compressor having an input connected to receive the difference image whereby to produce a compressed difference image at the output of the image compressor. dated 1997-01-28"
5598508,real-time waveform analysis using artificial neural networks,"a real-time waveform analysis system utilizes neural networks to perform various stages of the analysis. the signal containing the waveform is first stored in a buffer and the buffer contents transmitted to a first and second neural network which have been previously trained to recognize the start point and the end point of the waveform respectively. a third neural network receives the signal occurring between the start and end points and classifies that waveform as comprising either an incomplete waveform, a normal waveform or one of a variety of predetermined characteristic classifications. ambiguities in the output of the third neural network are arbitrated by a fourth neural network which may be given additional information which serves to resolve these ambiguities. in accordance with the preferred embodiment, the present invention is applied to a system analyzing respiratory waveforms of a patient undergoing anesthesia and the classifications of the waveform correspond to normal or various categories of abnormal features functioning in the respiratory signal. the system performs the analysis rapidly enough to be used in realtime systems and can be operated with relatively low cost hardware and with minimal software development required.",1997-01-28,"The title of the patent is real-time waveform analysis using artificial neural networks and its abstract is a real-time waveform analysis system utilizes neural networks to perform various stages of the analysis. the signal containing the waveform is first stored in a buffer and the buffer contents transmitted to a first and second neural network which have been previously trained to recognize the start point and the end point of the waveform respectively. a third neural network receives the signal occurring between the start and end points and classifies that waveform as comprising either an incomplete waveform, a normal waveform or one of a variety of predetermined characteristic classifications. ambiguities in the output of the third neural network are arbitrated by a fourth neural network which may be given additional information which serves to resolve these ambiguities. in accordance with the preferred embodiment, the present invention is applied to a system analyzing respiratory waveforms of a patient undergoing anesthesia and the classifications of the waveform correspond to normal or various categories of abnormal features functioning in the respiratory signal. the system performs the analysis rapidly enough to be used in realtime systems and can be operated with relatively low cost hardware and with minimal software development required. dated 1997-01-28"
5598509,method of configuring a neural network and a diagnosis/recognition system using the same,"in a recognition/diagnosis method, a relation between total sums of inputs of respective cases and teacher data is listed in the order of magnitude of the total sums of inputs. based on the value of the teacher data for the input having the maximum total sum and the number of times of change of teacher data in the table are considered, a configuration of a neural network (the number of hidden layers and the number of neurons thereof) is determined. coupling weights are analytically calculated based on the table.",1997-01-28,"The title of the patent is method of configuring a neural network and a diagnosis/recognition system using the same and its abstract is in a recognition/diagnosis method, a relation between total sums of inputs of respective cases and teacher data is listed in the order of magnitude of the total sums of inputs. based on the value of the teacher data for the input having the maximum total sum and the number of times of change of teacher data in the table are considered, a configuration of a neural network (the number of hidden layers and the number of neurons thereof) is determined. coupling weights are analytically calculated based on the table. dated 1997-01-28"
5600753,speech recognition by neural network adapted to reference pattern learning,"a speech recognition method according to the present invention uses distances calculated through a variance weighting process using covariance matrixes as the local distances (prediction residuals) between the feature vectors of input syllables/sound elements and predicted vectors formed by different statuses of reference neural prediction models (npm's) using finite status transition networks. the category to minimize the accumulated value of these local distances along the status transitions of all the prediction models is figured out by dynamic programming, and used as the recognition output. learning of the reference prediction models used in this recognition method is accomplished by repeating said distance calculating process and the process to correct the parameters of the different statuses and the covariance matrixes of said prediction models in the direction of reducing the distance between the learning patterns whose category is known and the prediction models of the same category as this known category, and what have satisfied prescribed conditions of convergence through these calculating and correcting processes are determined as reference pattern models.",1997-02-04,"The title of the patent is speech recognition by neural network adapted to reference pattern learning and its abstract is a speech recognition method according to the present invention uses distances calculated through a variance weighting process using covariance matrixes as the local distances (prediction residuals) between the feature vectors of input syllables/sound elements and predicted vectors formed by different statuses of reference neural prediction models (npm's) using finite status transition networks. the category to minimize the accumulated value of these local distances along the status transitions of all the prediction models is figured out by dynamic programming, and used as the recognition output. learning of the reference prediction models used in this recognition method is accomplished by repeating said distance calculating process and the process to correct the parameters of the different statuses and the covariance matrixes of said prediction models in the direction of reducing the distance between the learning patterns whose category is known and the prediction models of the same category as this known category, and what have satisfied prescribed conditions of convergence through these calculating and correcting processes are determined as reference pattern models. dated 1997-02-04"
5600758,method and device for conducting a process in a controlled system with at least one precomputed process parameter.,known methods for conducting a process in an automatically controlled system preset the system at the beginning of each process run based on at least one process parameter. the process parameter is precomputed with a model of the process which is supplied with input values. during the process the input values and the process parameter are measured and are used to adaptively improve the precomputed process parameter after the process run. the present invention simplifies and improves the precomputed value of the process parameter by supplying at least part of the input values to a neural network. the network response of the neural network forms a correction value for the approximate value delivered by the model for the process parameter to be precomputed. the network parameters of the neural network are modified after each process run to adapt the precomputed value to the actual process events.,1997-02-04,The title of the patent is method and device for conducting a process in a controlled system with at least one precomputed process parameter. and its abstract is known methods for conducting a process in an automatically controlled system preset the system at the beginning of each process run based on at least one process parameter. the process parameter is precomputed with a model of the process which is supplied with input values. during the process the input values and the process parameter are measured and are used to adaptively improve the precomputed process parameter after the process run. the present invention simplifies and improves the precomputed value of the process parameter by supplying at least part of the input values to a neural network. the network response of the neural network forms a correction value for the approximate value delivered by the model for the process parameter to be precomputed. the network parameters of the neural network are modified after each process run to adapt the precomputed value to the actual process events. dated 1997-02-04
5601090,method and apparatus for automatically determining somatic state,"an apparatus and a method for automatically determining the present somatic state of a human subject. the characteristic values of the subject (e.g., scalp potential, muscle potential, heart-rate, eye-movement and frequency of eye blinks, or any combination thereof) are detected and output signals corresponding to the detected characteristic values are produced, amplified and digitized. the fourier transformation is performed on the output signals. a set of state variables for each selected frequency sub-band of a selected frequency band for each of the output signals is determined. sets of reference weights and sets of reference biases for a neural network from sets of state reference variables corresponding to known somatic states are formed. each of the sets of state variables, the sets of reference weights and the sets of reference biases are applied to the neural network to determine present somatic state of the subject. the present somatic state of the subject is displayed.",1997-02-11,"The title of the patent is method and apparatus for automatically determining somatic state and its abstract is an apparatus and a method for automatically determining the present somatic state of a human subject. the characteristic values of the subject (e.g., scalp potential, muscle potential, heart-rate, eye-movement and frequency of eye blinks, or any combination thereof) are detected and output signals corresponding to the detected characteristic values are produced, amplified and digitized. the fourier transformation is performed on the output signals. a set of state variables for each selected frequency sub-band of a selected frequency band for each of the output signals is determined. sets of reference weights and sets of reference biases for a neural network from sets of state reference variables corresponding to known somatic states are formed. each of the sets of state variables, the sets of reference weights and the sets of reference biases are applied to the neural network to determine present somatic state of the subject. the present somatic state of the subject is displayed. dated 1997-02-11"
5602964,automata networks and methods for obtaining optimized dynamically reconfigurable computational architectures and controls,"a system for obtaining optimum performance and optimum graceful degradation from lie algebra descriptions of a spectrum of reconfigurable network architectures, including, neural nets and cellular automata comprised of interconnected nodes. the dynamic performance of the computational process is monitored by continued extraction of liapounov exponent indicators, reconfiguring said reconfigurable network architecture when said indicators predict non-optimum performance. the reconfigurable networks are reconfigured and compensatory adjustments are made of signal sampling performance and operating system performance of said reconfigurable network architecture, and the operating system architecture is optimized to the computational task by reconfiguration of nodal capabilities and degree of interconnectedness between nodes to obtain any lie algebra description architectural form between ideal neural net with maximum interconnectedness and ideal cellular automata with maximum nodal capability.",1997-02-11,"The title of the patent is automata networks and methods for obtaining optimized dynamically reconfigurable computational architectures and controls and its abstract is a system for obtaining optimum performance and optimum graceful degradation from lie algebra descriptions of a spectrum of reconfigurable network architectures, including, neural nets and cellular automata comprised of interconnected nodes. the dynamic performance of the computational process is monitored by continued extraction of liapounov exponent indicators, reconfiguring said reconfigurable network architecture when said indicators predict non-optimum performance. the reconfigurable networks are reconfigured and compensatory adjustments are made of signal sampling performance and operating system performance of said reconfigurable network architecture, and the operating system architecture is optimized to the computational task by reconfiguration of nodal capabilities and degree of interconnectedness between nodes to obtain any lie algebra description architectural form between ideal neural net with maximum interconnectedness and ideal cellular automata with maximum nodal capability. dated 1997-02-11"
5602965,laser programmable integrated curcuit for forming synapses in neural networks,"customizable neural network in which one or more resistors form each synapse. all the resistors in the synaptic array are identical, thus simplifying the processing issues. highly doped, amorphous silicon is used as the resistor material, to create extremely high resistances occupying very small spaces. connected in series with each resistor in the array is at least one severable conductor whose uppermost layer has a lower reflectivity of laser energy than typical metal conductors at a desired laser wavelength.",1997-02-11,"The title of the patent is laser programmable integrated curcuit for forming synapses in neural networks and its abstract is customizable neural network in which one or more resistors form each synapse. all the resistors in the synaptic array are identical, thus simplifying the processing issues. highly doped, amorphous silicon is used as the resistor material, to create extremely high resistances occupying very small spaces. connected in series with each resistor in the array is at least one severable conductor whose uppermost layer has a lower reflectivity of laser energy than typical metal conductors at a desired laser wavelength. dated 1997-02-11"
5604529,three-dimensional vision camera,"to provide a three-dimensional vision camera where no disagreement is caused between two-dimensional image information and distance information, the three-dimensional vision camera is provided with a memory (11), a neural network (12) and a three-dimensional image synthesizer (13). in the memory (11), image data of a photographic object shot from different directions are stored. information necessary for converting the image sensed by an image sensing device (2) into a three-dimensional image is read out from the image data of the memory (11) by the neural network (12). a three-dimensional image is produced by the three-dimensional image synthesizer (13).",1997-02-18,"The title of the patent is three-dimensional vision camera and its abstract is to provide a three-dimensional vision camera where no disagreement is caused between two-dimensional image information and distance information, the three-dimensional vision camera is provided with a memory (11), a neural network (12) and a three-dimensional image synthesizer (13). in the memory (11), image data of a photographic object shot from different directions are stored. information necessary for converting the image sensed by an image sensing device (2) into a three-dimensional image is read out from the image data of the memory (11) by the neural network (12). a three-dimensional image is produced by the three-dimensional image synthesizer (13). dated 1997-02-18"
5604820,method for extracting object images and method for detecting movements thereof,"in a method for extracting an object image, an extraction area for extraction of a candidate for a predetermined object image from an image is determined. the center point of a view window, which has a predetermined size, is caused to travel to the position of the candidate for the predetermined object image. the extraction area is determined in accordance with the size and/or the shape of the candidate for the predetermined object image, the center point of the view window being taken as a reference during the determination of the extraction area. the extraction of the candidate for the predetermined object image is carried out by using a neural network. even if a plurality of object images, which are to be extracted, are embedded in a given image, the object images are extracted efficiently such that an object image, which has already been extracted, may not be extracted again.",1997-02-18,"The title of the patent is method for extracting object images and method for detecting movements thereof and its abstract is in a method for extracting an object image, an extraction area for extraction of a candidate for a predetermined object image from an image is determined. the center point of a view window, which has a predetermined size, is caused to travel to the position of the candidate for the predetermined object image. the extraction area is determined in accordance with the size and/or the shape of the candidate for the predetermined object image, the center point of the view window being taken as a reference during the determination of the extraction area. the extraction of the candidate for the predetermined object image is carried out by using a neural network. even if a plurality of object images, which are to be extracted, are embedded in a given image, the object images are extracted efficiently such that an object image, which has already been extracted, may not be extracted again. dated 1997-02-18"
5604823,method for extracting object images and method for detecting movements thereof,"in a method for extracting an object image, an extraction area for extraction of a candidate for a predetermined object image from an image is determined. the center point of a view window, which has a predetermined size, is caused to travel to the position of the candidate for the predetermined object image. the extraction area is determined in accordance with the size and/or the shape of the candidate for the predetermined object image, the center point of the view window being taken as a reference during the determination of the extraction area. the extraction of the candidate for the predetermined object image is carried out by using a neural network. even if a plurality of object images, which are to be extracted, are embedded in a given image, the object images are extracted efficiently such that an object image, which has already been extracted, may not be extracted again.",1997-02-18,"The title of the patent is method for extracting object images and method for detecting movements thereof and its abstract is in a method for extracting an object image, an extraction area for extraction of a candidate for a predetermined object image from an image is determined. the center point of a view window, which has a predetermined size, is caused to travel to the position of the candidate for the predetermined object image. the extraction area is determined in accordance with the size and/or the shape of the candidate for the predetermined object image, the center point of the view window being taken as a reference during the determination of the extraction area. the extraction of the candidate for the predetermined object image is carried out by using a neural network. even if a plurality of object images, which are to be extracted, are embedded in a given image, the object images are extracted efficiently such that an object image, which has already been extracted, may not be extracted again. dated 1997-02-18"
5606646,recurrent neural network-based fuzzy logic system,"a recurrent, neural network-based fuzzy logic system includes neurons in a rule base layer which each have a recurrent architecture with an output-to-input feedback path including a time delay element and a neural weight. further included is a neural network-based, fuzzy logic finite state machine wherein the neural network-based, fuzzy logic system has a recurrent architecture with an output-to-input feedback path including at least a time delay element. still further included is a recurrent, neural network-based fuzzy logic rule generator wherein a neural network receives and fuzzifies input data and provides data corresponding to fuzzy logic membership functions and recurrent fuzzy logic rules.",1997-02-25,"The title of the patent is recurrent neural network-based fuzzy logic system and its abstract is a recurrent, neural network-based fuzzy logic system includes neurons in a rule base layer which each have a recurrent architecture with an output-to-input feedback path including a time delay element and a neural weight. further included is a neural network-based, fuzzy logic finite state machine wherein the neural network-based, fuzzy logic system has a recurrent architecture with an output-to-input feedback path including at least a time delay element. still further included is a recurrent, neural network-based fuzzy logic rule generator wherein a neural network receives and fuzzifies input data and provides data corresponding to fuzzy logic membership functions and recurrent fuzzy logic rules. dated 1997-02-25"
5608819,image processing system utilizing neural network for discrimination between text data and other image data,"an image processing system, which operates on an input image data stream consisting of successive multi-level values, effecting a plurality of respectively different types of image data processing in accordance with a plurality of different categories of the input image data, utilizes a neural network to assign each datum to a specific category, with resultant output signals from the neural network being used to select the appropriate type of image data processing for that datum.",1997-03-04,"The title of the patent is image processing system utilizing neural network for discrimination between text data and other image data and its abstract is an image processing system, which operates on an input image data stream consisting of successive multi-level values, effecting a plurality of respectively different types of image data processing in accordance with a plurality of different categories of the input image data, utilizes a neural network to assign each datum to a specific category, with resultant output signals from the neural network being used to select the appropriate type of image data processing for that datum. dated 1997-03-04"
5608842,method and device for conducting a process in a controlled system with at least one precomputed parameter based on a plurality of results from partial mathematical models combined by a neural network,"in known methods for conducting a process in an automatically controlled system, the system is preset at the beginning of each process run according to at least one process parameter. the at least one process parameter is precomputed with a model of the process which is supplied with input values. during the process, the input values and the process parameters are measured and are used after the process run to adaptively improve the precomputed value of the process parameters. to simplify and improve the precomputed value of a model having a plurality of partial models, computed results of the partial models are supplied to a neural network. the neural network produces the process parameters to be precomputed as a network response. the network parameters of the neural network are modified after each process run to adapt the precomputed value to the actual process events.",1997-03-04,"The title of the patent is method and device for conducting a process in a controlled system with at least one precomputed parameter based on a plurality of results from partial mathematical models combined by a neural network and its abstract is in known methods for conducting a process in an automatically controlled system, the system is preset at the beginning of each process run according to at least one process parameter. the at least one process parameter is precomputed with a model of the process which is supplied with input values. during the process, the input values and the process parameters are measured and are used after the process run to adaptively improve the precomputed value of the process parameters. to simplify and improve the precomputed value of a model having a plurality of partial models, computed results of the partial models are supplied to a neural network. the neural network produces the process parameters to be precomputed as a network response. the network parameters of the neural network are modified after each process run to adapt the precomputed value to the actual process events. dated 1997-03-04"
5608846,fuzzy rule generator,"this invention relates to an apparatus and method for generating membership functions and rules for a fuzzy system whereby fuzzy rules and membership functions are synthesized by observing a sample output/input data array and by creating new fuzzy sets which closely approximate various data associations in accordance with maximum inference error calculations. applications of this system are provided in a temperature controller, a vehicle suspension controller, and a neural network/fuzzy rule converter device.",1997-03-04,"The title of the patent is fuzzy rule generator and its abstract is this invention relates to an apparatus and method for generating membership functions and rules for a fuzzy system whereby fuzzy rules and membership functions are synthesized by observing a sample output/input data array and by creating new fuzzy sets which closely approximate various data associations in accordance with maximum inference error calculations. applications of this system are provided in a temperature controller, a vehicle suspension controller, and a neural network/fuzzy rule converter device. dated 1997-03-04"
5611020,neural network shell for application programs,"a neural network shell has a defined interface to an application program. by interfacing with the neural network shell, any application program becomes a neural network application program. the neural network shell contains a set of utility programs that transfers data into and out of a neural network data structure. this set of utility programs allows an application program to define a new neural network model, create a neural network data structure, train a neural network, and run a neural network. once trained, the neural network data structure can be transported to other computer systems or to application programs written in different computing languages running on similar or different computer systems.",1997-03-11,"The title of the patent is neural network shell for application programs and its abstract is a neural network shell has a defined interface to an application program. by interfacing with the neural network shell, any application program becomes a neural network application program. the neural network shell contains a set of utility programs that transfers data into and out of a neural network data structure. this set of utility programs allows an application program to define a new neural network model, create a neural network data structure, train a neural network, and run a neural network. once trained, the neural network data structure can be transported to other computer systems or to application programs written in different computing languages running on similar or different computer systems. dated 1997-03-11"
5611753,speed change control method for controlling changeover between gearshift positions of an automotive automatic transmission utilizing a detected degree of necessity of engine braking and learning correction,"a speed change control method controls changeover between gearshift positions of an automotive automatic transmission using a detected running condition parameter and a detected degree of necessity of engine braking. the degree of necessity of engine braking is detected using a neural network receiving the detected parameter as an input. then, a shift pattern is preferably selected by fuzzy inference based on the detected parameter and degree of necessity of engine braking.",1997-03-18,"The title of the patent is speed change control method for controlling changeover between gearshift positions of an automotive automatic transmission utilizing a detected degree of necessity of engine braking and learning correction and its abstract is a speed change control method controls changeover between gearshift positions of an automotive automatic transmission using a detected running condition parameter and a detected degree of necessity of engine braking. the degree of necessity of engine braking is detected using a neural network receiving the detected parameter as an input. then, a shift pattern is preferably selected by fuzzy inference based on the detected parameter and degree of necessity of engine braking. dated 1997-03-18"
5612928,method and apparatus for classifying objects in sonar images,"a method and apparatus for classifying objects in images utilize means for selecting portions of those images which contain objects and means for classifying those objects based upon parameters of the selected portions, which parameters are useful for classifying the objects. the selecting means preferably is a shadow and highlight detector, a statistical window detector and a neural network window detector whose output is combined in a combined cuer. the parameters are determined from the greylevels and positions of pixels using one or more modules which perform certain mathematical operations on this data. such parameters include edge parameters, smoothness, clutter, presence and characteristics of highlights and shadows, and texture. the invention is particularly useful for classifying objects in sonar images as natural or man-made.",1997-03-18,"The title of the patent is method and apparatus for classifying objects in sonar images and its abstract is a method and apparatus for classifying objects in images utilize means for selecting portions of those images which contain objects and means for classifying those objects based upon parameters of the selected portions, which parameters are useful for classifying the objects. the selecting means preferably is a shadow and highlight detector, a statistical window detector and a neural network window detector whose output is combined in a combined cuer. the parameters are determined from the greylevels and positions of pixels using one or more modules which perform certain mathematical operations on this data. such parameters include edge parameters, smoothness, clutter, presence and characteristics of highlights and shadows, and texture. the invention is particularly useful for classifying objects in sonar images as natural or man-made. dated 1997-03-18"
5613039,apparatus and method for motion detection and tracking of objects in a region for collision avoidance utilizing a real-time adaptive probabilistic neural network,apparatus for motion detection and tracking of objects in a region for collision avoidance includes a signal transmitter which provides first and second detection signals for at least partial reflection by an object located in a spatial region. the apparatus further includes a signal receiver for receiving the deflected first and second detection signals corresponding to first and second object parameter data signals. the apparatus further includes a fourier transform circuit for receiving the first and second object parameter data signals and providing first and second fourier transform object parameter data signals. the apparatus further includes a probabilistic neural network for receiving and sorting the first and second fourier transform object parameter data signals without the use of a priori training data.,1997-03-18,The title of the patent is apparatus and method for motion detection and tracking of objects in a region for collision avoidance utilizing a real-time adaptive probabilistic neural network and its abstract is apparatus for motion detection and tracking of objects in a region for collision avoidance includes a signal transmitter which provides first and second detection signals for at least partial reflection by an object located in a spatial region. the apparatus further includes a signal receiver for receiving the deflected first and second detection signals corresponding to first and second object parameter data signals. the apparatus further includes a fourier transform circuit for receiving the first and second object parameter data signals and providing first and second fourier transform object parameter data signals. the apparatus further includes a probabilistic neural network for receiving and sorting the first and second fourier transform object parameter data signals without the use of a priori training data. dated 1997-03-18
5613040,neural network shell for application programs,"a neural network shell has a defined interface to an application program. by interfacing with the neural network shell, any application program becomes a neural network application program. the neural network shell contains a set of utility programs that transfers data into and out of a neural network data structure. this set of utility programs allows an application program to define a new neural network model, create a neural network data structure, train a neural network, and run a neural network. once trained, the neural network data structure can be transported to other computer systems or to application programs written in different computing languages running on similar or different computer systems.",1997-03-18,"The title of the patent is neural network shell for application programs and its abstract is a neural network shell has a defined interface to an application program. by interfacing with the neural network shell, any application program becomes a neural network application program. the neural network shell contains a set of utility programs that transfers data into and out of a neural network data structure. this set of utility programs allows an application program to define a new neural network model, create a neural network data structure, train a neural network, and run a neural network. once trained, the neural network data structure can be transported to other computer systems or to application programs written in different computing languages running on similar or different computer systems. dated 1997-03-18"
5613041,method and apparatus for operating neural network with missing and/or incomplete data,"a neural network system is provided that models the system in a system model (12) with the output thereof providing a predicted output. this predicted output is modified or controlled by an output control (14). input data is processed in a data preprocess step (10) to reconcile the data for input to the system model (12). additionally, the error resulted from the reconciliation is input to an uncertainty model to predict the uncertainty in the predicted output. this is input to a decision processor (20) which is utilized to control the output control (14). the output control (14) is controlled to either vary the predicted output or to inhibit the predicted output whenever the output of the uncertainty model (18) exceeds a predetermined decision threshold, input by a decision threshold block (22). additionally, a validity model (16) is also provided which represents the reliability or validity of the output as a function of the number of data points in a given data region during training of the system model (12). this predicts the confidence in the predicted output which is also input to the decision processor (20). the decision processor (20) therefore bases its decision on the predicted confidence and the predicted uncertainty. additionally, the uncertainty output by the data preprocess block (10) can be utilized to train the system model (12).",1997-03-18,"The title of the patent is method and apparatus for operating neural network with missing and/or incomplete data and its abstract is a neural network system is provided that models the system in a system model (12) with the output thereof providing a predicted output. this predicted output is modified or controlled by an output control (14). input data is processed in a data preprocess step (10) to reconcile the data for input to the system model (12). additionally, the error resulted from the reconciliation is input to an uncertainty model to predict the uncertainty in the predicted output. this is input to a decision processor (20) which is utilized to control the output control (14). the output control (14) is controlled to either vary the predicted output or to inhibit the predicted output whenever the output of the uncertainty model (18) exceeds a predetermined decision threshold, input by a decision threshold block (22). additionally, a validity model (16) is also provided which represents the reliability or validity of the output as a function of the number of data points in a given data region during training of the system model (12). this predicts the confidence in the predicted output which is also input to the decision processor (20). the decision processor (20) therefore bases its decision on the predicted confidence and the predicted uncertainty. additionally, the uncertainty output by the data preprocess block (10) can be utilized to train the system model (12). dated 1997-03-18"
5613042,chaotic recurrent neural network and learning method therefor,""" a chaotic recurrent neural network includes n chaotic neural networks for receiving an external input and the outputs of n-1 chaotic neural networks among said n chaotic neural networks and performing an operation according to the following dynamic equation ##equ1## wherein w.sub.ij is a synapse connection coefficient of the feedback input from the """"j""""th neuron to the """"i""""th neuron, x.sub.i (t) is the output of the """"i""""th neuron at time t, and .gamma..sub.i, .alpha. and and k are a time-delaying constant, a non-negative parameter and a refractory time attenuation constant, respectively, and wherein z.sub.i (t) represents x.sub.i (t) when i belongs to the neuron group i and represents a.sub.i (t) when i belongs to the external input group e. also, a learning algorithm for the chaotic recurrent neural network increases its learning efficiency. """,1997-03-18,"The title of the patent is chaotic recurrent neural network and learning method therefor and its abstract is "" a chaotic recurrent neural network includes n chaotic neural networks for receiving an external input and the outputs of n-1 chaotic neural networks among said n chaotic neural networks and performing an operation according to the following dynamic equation ##equ1## wherein w.sub.ij is a synapse connection coefficient of the feedback input from the """"j""""th neuron to the """"i""""th neuron, x.sub.i (t) is the output of the """"i""""th neuron at time t, and .gamma..sub.i, .alpha. and and k are a time-delaying constant, a non-negative parameter and a refractory time attenuation constant, respectively, and wherein z.sub.i (t) represents x.sub.i (t) when i belongs to the neuron group i and represents a.sub.i (t) when i belongs to the external input group e. also, a learning algorithm for the chaotic recurrent neural network increases its learning efficiency. "" dated 1997-03-18"
5613043,neural network shell for application programs,"a neural network shell has a defined interface to an application program. by interfacing with the neural network shell, any application program becomes a neural network application program. the neural network shell contains a set of utility programs that transfers data into and out of a neural network data structure. this set of utility programs allows an application program to define a new neural network model, create a neural network data structure, train a neural network, and run a neural network. once trained, the neural network data structure can be transported to other computer systems or to application programs written in different computing languages running on similar or different computer systems.",1997-03-18,"The title of the patent is neural network shell for application programs and its abstract is a neural network shell has a defined interface to an application program. by interfacing with the neural network shell, any application program becomes a neural network application program. the neural network shell contains a set of utility programs that transfers data into and out of a neural network data structure. this set of utility programs allows an application program to define a new neural network model, create a neural network data structure, train a neural network, and run a neural network. once trained, the neural network data structure can be transported to other computer systems or to application programs written in different computing languages running on similar or different computer systems. dated 1997-03-18"
5613044,learning machine synapse processor system apparatus,"a neural synapse processor apparatus having a neuron architecture for the synapse processing elements of the apparatus. the apparatus which we prefer will have a n neuron structure having synapse processing units that contain instruction and data storage units, receive instructions and data, and execute instructions. the n neuron structure should contain communicating adder trees, neuron activation function units, and an arrangement for communicating both instructions, data, and the outputs of neuron activation function units back to the input synapse processing units by means of the communicating adder trees. the apparatus can be structured as a bit-serial or word parallel system. the preferred structure contains n.sup.2 synapse processing units, each associated with a connection weight in the n neural network to be emulated, placed in the form of a n by n matrix that has been folded along the diagonal and made up of diagonal cells and general cells. diagonal cells, each utilizing a single synapse processing unit, are associated with the diagonal connection weights of the folded n by n connection weight matrix and general cells, each of which has two synapse processing units merged together, and which are associated with the symmetric connection weights of the folded n by n connection weight matrix. the back-propagation learning algorithm is first discussed followed by a presentation of the learning machine synapse processor architecture. an example implementation of the back-propagation learning algorithm is then presented. this is followed by a boltzmann like machine example and data parallel examples mapped onto the architecture.",1997-03-18,"The title of the patent is learning machine synapse processor system apparatus and its abstract is a neural synapse processor apparatus having a neuron architecture for the synapse processing elements of the apparatus. the apparatus which we prefer will have a n neuron structure having synapse processing units that contain instruction and data storage units, receive instructions and data, and execute instructions. the n neuron structure should contain communicating adder trees, neuron activation function units, and an arrangement for communicating both instructions, data, and the outputs of neuron activation function units back to the input synapse processing units by means of the communicating adder trees. the apparatus can be structured as a bit-serial or word parallel system. the preferred structure contains n.sup.2 synapse processing units, each associated with a connection weight in the n neural network to be emulated, placed in the form of a n by n matrix that has been folded along the diagonal and made up of diagonal cells and general cells. diagonal cells, each utilizing a single synapse processing unit, are associated with the diagonal connection weights of the folded n by n connection weight matrix and general cells, each of which has two synapse processing units merged together, and which are associated with the symmetric connection weights of the folded n by n connection weight matrix. the back-propagation learning algorithm is first discussed followed by a presentation of the learning machine synapse processor architecture. an example implementation of the back-propagation learning algorithm is then presented. this is followed by a boltzmann like machine example and data parallel examples mapped onto the architecture. dated 1997-03-18"
5615305,neural processor element,"a neural processor element for use in a neural network includes a synaptic base weighting circuit which responds to an excitory input in the form of a stream of pulses of a first polarity which represents a sum of synaptic weights and an inhibitory input in the form of a stream of pulses of a second polarity which represents a sum of synaptic weights. the synaptic weight of the neural processor element is generated by summing the excitory and inhibitory pulse streams to produce a signal of varying level magnitude. this signal is subjected to a thresholding function which generates a pulse having a width corresponding to the sum of excitory and inhibitory pulse streams, this pulse width representing the synaptic weight for the neural processor element.",1997-03-25,"The title of the patent is neural processor element and its abstract is a neural processor element for use in a neural network includes a synaptic base weighting circuit which responds to an excitory input in the form of a stream of pulses of a first polarity which represents a sum of synaptic weights and an inhibitory input in the form of a stream of pulses of a second polarity which represents a sum of synaptic weights. the synaptic weight of the neural processor element is generated by summing the excitory and inhibitory pulse streams to produce a signal of varying level magnitude. this signal is subjected to a thresholding function which generates a pulse having a width corresponding to the sum of excitory and inhibitory pulse streams, this pulse width representing the synaptic weight for the neural processor element. dated 1997-03-25"
5615306,neural network shell for application programs,"a neural network shell has a defined interface to an application program. by interfacing with the neural network shell, any application program becomes a neural network application program. the neural network shell contains a set of utility programs that transfers data into and out of a neural network data structure. this set of utility programs allows an application program to define a new neural network model, create a neural network data structure, train a neural network, and run a neural network. once trained, the neural network data structure can be transported to other computer systems or to application programs written in different computing languages running on similar or different computer systems.",1997-03-25,"The title of the patent is neural network shell for application programs and its abstract is a neural network shell has a defined interface to an application program. by interfacing with the neural network shell, any application program becomes a neural network application program. the neural network shell contains a set of utility programs that transfers data into and out of a neural network data structure. this set of utility programs allows an application program to define a new neural network model, create a neural network data structure, train a neural network, and run a neural network. once trained, the neural network data structure can be transported to other computer systems or to application programs written in different computing languages running on similar or different computer systems. dated 1997-03-25"
5615307,neural network shell for application programs,"a neural network shell has a defined interface to an application program. by interfacing with the neural network shell, any application program becomes a neural network application program. the neural network shell contains a set of utility programs that transfers data into and out of a neural network data structure. this set of utility programs allows an application program to define a new neural network model, create a neural network data structure, train a neural network, and run a neural network. once trained, the neural network data structure can be transported to other computer systems or to application programs written in different computing languages running on similar or different computer systems.",1997-03-25,"The title of the patent is neural network shell for application programs and its abstract is a neural network shell has a defined interface to an application program. by interfacing with the neural network shell, any application program becomes a neural network application program. the neural network shell contains a set of utility programs that transfers data into and out of a neural network data structure. this set of utility programs allows an application program to define a new neural network model, create a neural network data structure, train a neural network, and run a neural network. once trained, the neural network data structure can be transported to other computer systems or to application programs written in different computing languages running on similar or different computer systems. dated 1997-03-25"
5617483,pattern learning method,"in a pattern learning method, each of pieces of information representing a plurality of different fundamental patterns is presented to a large number of cells of a neural network, and the cells are thereby caused to learn a large number of feature patterns. the method comprises the steps of causing a cell, which best matches with a fundamental pattern having been presented to the neural network, to learn the fundamental pattern. for neighboring cells that fall within a neighboring region having a predetermined range and neighboring with the cell, which best matches with the fundamental pattern having been presented to the neural network, spatial interpolating operations are carried out from the fundamental pattern, which has been presented to the neural network, and a fundamental pattern, which is other than the fundamental pattern having been presented to the neural network and which has been learned by a cell that is among the large number of the cells of the neural network and that is other than the cell best matching with the fundamental pattern having been presented to the neural network. the neighboring cells are caused to learn the results of the spatial interpolating operations.",1997-04-01,"The title of the patent is pattern learning method and its abstract is in a pattern learning method, each of pieces of information representing a plurality of different fundamental patterns is presented to a large number of cells of a neural network, and the cells are thereby caused to learn a large number of feature patterns. the method comprises the steps of causing a cell, which best matches with a fundamental pattern having been presented to the neural network, to learn the fundamental pattern. for neighboring cells that fall within a neighboring region having a predetermined range and neighboring with the cell, which best matches with the fundamental pattern having been presented to the neural network, spatial interpolating operations are carried out from the fundamental pattern, which has been presented to the neural network, and a fundamental pattern, which is other than the fundamental pattern having been presented to the neural network and which has been learned by a cell that is among the large number of the cells of the neural network and that is other than the cell best matching with the fundamental pattern having been presented to the neural network. the neighboring cells are caused to learn the results of the spatial interpolating operations. dated 1997-04-01"
5617484,image binarizing apparatus,""" an image binarizing apparatus, which comprises an image sensor for inputting a character image or a line image, an a/d converter for digitizing the output of the image sensor, a frame memory for temporary storage of the digital image, a window circuit for generating address information for dividing the stored image into a predetermined number of partial images, a brightness extractor for obtaining the highest brightness, lowest brightness and average brightness for the pixels for each block, a neural network which provides the optimum threshold value based on pre-learned data when receiving the highest brightness, lowest brightness and average brightness, a binarizer for binarizing each pixel of a partial image block from the frame memory based on the optimum threshold value (""""white pixel"""" when the value of the brightness of the pixel is larger than the output value of the neural network, and """"black pixel"""" when the value of the brightness of the pixel is smaller than the output value of the neural network), and another frame memory for storing the binarized image at a predetermined address. """,1997-04-01,"The title of the patent is image binarizing apparatus and its abstract is "" an image binarizing apparatus, which comprises an image sensor for inputting a character image or a line image, an a/d converter for digitizing the output of the image sensor, a frame memory for temporary storage of the digital image, a window circuit for generating address information for dividing the stored image into a predetermined number of partial images, a brightness extractor for obtaining the highest brightness, lowest brightness and average brightness for the pixels for each block, a neural network which provides the optimum threshold value based on pre-learned data when receiving the highest brightness, lowest brightness and average brightness, a binarizer for binarizing each pixel of a partial image block from the frame memory based on the optimum threshold value (""""white pixel"""" when the value of the brightness of the pixel is larger than the output value of the neural network, and """"black pixel"""" when the value of the brightness of the pixel is smaller than the output value of the neural network), and another frame memory for storing the binarized image at a predetermined address. "" dated 1997-04-01"
5617490,camera system with neural network compensator for measuring 3-d position,"a camera system has a neural network for calibration of image distortion. the neural network learns the conversion from image coordinates with distortion to image coordinates with substantially reduced distortion, whereby the neural network provides image coordinates having substantially reduced distortion. in a learning process of the neural network, a relatively simple camera model is used to provide an instruction signal to the neural network according to sample data provided from the real camera.",1997-04-01,"The title of the patent is camera system with neural network compensator for measuring 3-d position and its abstract is a camera system has a neural network for calibration of image distortion. the neural network learns the conversion from image coordinates with distortion to image coordinates with substantially reduced distortion, whereby the neural network provides image coordinates having substantially reduced distortion. in a learning process of the neural network, a relatively simple camera model is used to provide an instruction signal to the neural network according to sample data provided from the real camera. dated 1997-04-01"
5617511,neural network shell for application programs,"a neural network shell has a defined interface to an application program. by interfacing with the neural network shell, any application program becomes a neural network application program. the neural network shell contains a set of utility programs that transfers data into and out of a neural network data structure. this set of utility programs allows an application program to define a new neural network model, create a neural network data structure, train a neural network, and run a neural network. once trained, the neural network data structure can be transported to other computer systems or to application programs written in different computing languages running on similar or different computer systems.",1997-04-01,"The title of the patent is neural network shell for application programs and its abstract is a neural network shell has a defined interface to an application program. by interfacing with the neural network shell, any application program becomes a neural network application program. the neural network shell contains a set of utility programs that transfers data into and out of a neural network data structure. this set of utility programs allows an application program to define a new neural network model, create a neural network data structure, train a neural network, and run a neural network. once trained, the neural network data structure can be transported to other computer systems or to application programs written in different computing languages running on similar or different computer systems. dated 1997-04-01"
5619593,method for extracting object images and method for detecting movements thereof,"in a method for extracting an object image an extraction area for extraction of a candidate for a predetermined object image from an image is determined. the center point of a view window, which has a predetermined size, is caused to travel to the position of the candidate for the predetermined object image. the extraction area is determined in accordance with the size and/or the shape of the candidate for the predetermined object image, the center point of the view window being taken as a reference during the determination of the extraction area. the extraction of the candidate for the predetermined object image is carried out by using a neural network. even if a plurality of object images, which are to be extracted, are embedded in a given image, the object images are extracted efficiently such that an object image, which has already been extracted, may not be extracted again.",1997-04-08,"The title of the patent is method for extracting object images and method for detecting movements thereof and its abstract is in a method for extracting an object image an extraction area for extraction of a candidate for a predetermined object image from an image is determined. the center point of a view window, which has a predetermined size, is caused to travel to the position of the candidate for the predetermined object image. the extraction area is determined in accordance with the size and/or the shape of the candidate for the predetermined object image, the center point of the view window being taken as a reference during the determination of the extraction area. the extraction of the candidate for the predetermined object image is carried out by using a neural network. even if a plurality of object images, which are to be extracted, are embedded in a given image, the object images are extracted efficiently such that an object image, which has already been extracted, may not be extracted again. dated 1997-04-08"
5619616,vehicle classification system using a passive audio input to a neural network,"a system for classifying vehicles based on the sound waved produced by the vehicles receives analog sound pressure levels and converts them to a power spectrum. fuzzification functions, such as asymmetric wedge shaped functions, are convoluted with the power spectrum to create a vector that characterizes the power spectrum while reducing the dimensionality of the characterizing vector. a neural network analyzes the characterizing vector and produces a classification designator indicative of the class of the object associated with the analog sound pressure levels received by the system.",1997-04-08,"The title of the patent is vehicle classification system using a passive audio input to a neural network and its abstract is a system for classifying vehicles based on the sound waved produced by the vehicles receives analog sound pressure levels and converts them to a power spectrum. fuzzification functions, such as asymmetric wedge shaped functions, are convoluted with the power spectrum to create a vector that characterizes the power spectrum while reducing the dimensionality of the characterizing vector. a neural network analyzes the characterizing vector and produces a classification designator indicative of the class of the object associated with the analog sound pressure levels received by the system. dated 1997-04-08"
5619617,"neuron unit, neural network and signal processing method","a neuron unit processes a plurality of input signals and outputs an output signal which is indicative of a result of the processing. the neuron unit includes input lines for receiving the input signals, a forward process part including a supplying part for supplying weight functions and an operation part for carrying out an operation on each of the input signals using one of the weight functions and for outputting the output signal, and a self-learning part including a generating part for generating new weight functions based on errors between the output signal of the forward process part and teaching signals and a varying part for varying the weight functions supplied by the supplying part of the forward process part to the new weight functions generated by the generating part.",1997-04-08,"The title of the patent is neuron unit, neural network and signal processing method and its abstract is a neuron unit processes a plurality of input signals and outputs an output signal which is indicative of a result of the processing. the neuron unit includes input lines for receiving the input signals, a forward process part including a supplying part for supplying weight functions and an operation part for carrying out an operation on each of the input signals using one of the weight functions and for outputting the output signal, and a self-learning part including a generating part for generating new weight functions based on errors between the output signal of the forward process part and teaching signals and a varying part for varying the weight functions supplied by the supplying part of the forward process part to the new weight functions generated by the generating part. dated 1997-04-08"
5619618,neural network shell for application programs,"a neural network shell has a defined interface to an application program. by interfacing with the neural network shell, any application program becomes a neural network application program. the neural network shell contains a set of utility programs that transfers data into and out of a neural network data structure. this set of utility programs allows an application program to define a new neural network model, create a neural network data structure, train a neural network, and run a neural network. once trained, the neural network data structure can be transported to other computer systems or to application programs written in different computing languages running on similar or different computer systems.",1997-04-08,"The title of the patent is neural network shell for application programs and its abstract is a neural network shell has a defined interface to an application program. by interfacing with the neural network shell, any application program becomes a neural network application program. the neural network shell contains a set of utility programs that transfers data into and out of a neural network data structure. this set of utility programs allows an application program to define a new neural network model, create a neural network data structure, train a neural network, and run a neural network. once trained, the neural network data structure can be transported to other computer systems or to application programs written in different computing languages running on similar or different computer systems. dated 1997-04-08"
5619619,information recognition system and control system using same,"an information recognition circuit comprises a plurality of recognition processing units each composed of a neural network. teacher signals and information signals to be processed are supplied to a plurality of the units, individually so as to obtain output signals by executing individual learning. thereafter, the plural units are connected to each other so as to construct a large scale information recognition system. further, in the man-machine interface system, a plurality of operating instruction data are prepared. an operator's face is sensed by a tv camera to extract the factors related to the operator's facial expression. the neural network analogizes operator's feeling on the basis of the extracted factors. in accordance with the guessed results, a specific sort of the operating instruction is selected from a plurality of sorts of the operating instructions, and the selected instruction is displayed as an appropriate instruction for the operator. further, the one-loop controller for automatizing operation comprises an input interface section for acquiring image information, an image recognition section for recognizing the image using the acquired image information, a control section for calculating control commands according to the image recognition results, and an output interface for outputting control commands to process actuators or subordinate controllers, respectively.",1997-04-08,"The title of the patent is information recognition system and control system using same and its abstract is an information recognition circuit comprises a plurality of recognition processing units each composed of a neural network. teacher signals and information signals to be processed are supplied to a plurality of the units, individually so as to obtain output signals by executing individual learning. thereafter, the plural units are connected to each other so as to construct a large scale information recognition system. further, in the man-machine interface system, a plurality of operating instruction data are prepared. an operator's face is sensed by a tv camera to extract the factors related to the operator's facial expression. the neural network analogizes operator's feeling on the basis of the extracted factors. in accordance with the guessed results, a specific sort of the operating instruction is selected from a plurality of sorts of the operating instructions, and the selected instruction is displayed as an appropriate instruction for the operator. further, the one-loop controller for automatizing operation comprises an input interface section for acquiring image information, an image recognition section for recognizing the image using the acquired image information, a control section for calculating control commands according to the image recognition results, and an output interface for outputting control commands to process actuators or subordinate controllers, respectively. dated 1997-04-08"
5619620,neural network for banknote recognition and authentication,""" a probabilistic neural network (pnn) comprises a layer l1 of input nodes, a layer l2 of exemplar nodes, a layer l3 of primary parzen nodes, a layer l4 of sum nodes, and optionally a layer l5 of output nodes. each exemplar node determines the degree of match between a respective exemplar vector and an input vector and feeds a respective primary parzen node. the exemplar and primary parzen nodes are grouped into design classes, with a sum node for each class which combines the outputs of the primary parzen nodes for that class and feeds a corresponding output node. the network includes for each primary parzen node (e.g. l3-2-3p) for the design classes a secondary parzen node (l3-2-3s), the secondary parzen nodes all feeding a null class sum node (l4-0). each secondary parzen node has a parzen function with a lower peak amplitude and a broader spread than the corresponding primary parzen node, and is fed from the exemplar node for that primary parzen node. the secondary parzen nodes in effect detect input vectors which are """"sufficiently different"""" from the design classes--that is, null class vectors. the network is applicable to banknote recognition and authentication, the null class corresponding to counterfeit banknotes. """,1997-04-08,"The title of the patent is neural network for banknote recognition and authentication and its abstract is "" a probabilistic neural network (pnn) comprises a layer l1 of input nodes, a layer l2 of exemplar nodes, a layer l3 of primary parzen nodes, a layer l4 of sum nodes, and optionally a layer l5 of output nodes. each exemplar node determines the degree of match between a respective exemplar vector and an input vector and feeds a respective primary parzen node. the exemplar and primary parzen nodes are grouped into design classes, with a sum node for each class which combines the outputs of the primary parzen nodes for that class and feeds a corresponding output node. the network includes for each primary parzen node (e.g. l3-2-3p) for the design classes a secondary parzen node (l3-2-3s), the secondary parzen nodes all feeding a null class sum node (l4-0). each secondary parzen node has a parzen function with a lower peak amplitude and a broader spread than the corresponding primary parzen node, and is fed from the exemplar node for that primary parzen node. the secondary parzen nodes in effect detect input vectors which are """"sufficiently different"""" from the design classes--that is, null class vectors. the network is applicable to banknote recognition and authentication, the null class corresponding to counterfeit banknotes. "" dated 1997-04-08"
5619709,system and method of context vector generation and retrieval,""" a system and method for generating context vectors for use in storage and retrieval of documents and other information items. context vectors represent conceptual relationships among information items by quantitative means. a neural network operates on a training corpus of records to develop relationship-based context vectors based on word proximity and co-importance using a technique of """"windowed co-occurrence"""". relationships among context vectors are deterministic, so that a context vector set has one logical solution, although it may have a plurality of physical solutions. no human knowledge, thesaurus, synonym list, knowledge base, or conceptual hierarchy, is required. summary vectors of records may be clustered to reduce searching time, by forming a tree of clustered nodes. once the context vectors are determined, records may be retrieved using a query interface that allows a user to specify content terms, boolean terms, and/or document feedback. the present invention further facilitates visualization of textual information by translating context vectors into visual and graphical representations. thus, a user can explore visual representations of meaning, and can apply human visual pattern recognition skills to document searches. """,1997-04-08,"The title of the patent is system and method of context vector generation and retrieval and its abstract is "" a system and method for generating context vectors for use in storage and retrieval of documents and other information items. context vectors represent conceptual relationships among information items by quantitative means. a neural network operates on a training corpus of records to develop relationship-based context vectors based on word proximity and co-importance using a technique of """"windowed co-occurrence"""". relationships among context vectors are deterministic, so that a context vector set has one logical solution, although it may have a plurality of physical solutions. no human knowledge, thesaurus, synonym list, knowledge base, or conceptual hierarchy, is required. summary vectors of records may be clustered to reduce searching time, by forming a tree of clustered nodes. once the context vectors are determined, records may be retrieved using a query interface that allows a user to specify content terms, boolean terms, and/or document feedback. the present invention further facilitates visualization of textual information by translating context vectors into visual and graphical representations. thus, a user can explore visual representations of meaning, and can apply human visual pattern recognition skills to document searches. "" dated 1997-04-08"
5621172,method and apparatus for testing material strengths,"an apparatus and a non-destructive method for evaluating material strengths is described. one embodiment of the apparatus comprises a waveform generator that generates either (1) a sinusoidal waveform having a frequency that sweeps from a low frequency to a high frequency, or a high frequency to a low frequency, wherein the low frequency is from about 10 hz to about 150 hz, and the high frequency is from about 6,000 hz to about 24,000 hz, or (2) a pseudo random within the frequency range of from about 150 hz to about 6,000 hz. an electromechanical driver is electronically linked to the waveform generator and mechanically coupled to a test material, particularly in-service utility poles, at a drive position. force and acceleration sensors are coupled to the test material at various positions. a microprocessor is used to collect digitized data from the force sensor and the accelerometers, perform a transfer function and determine the strength remaining in the test material using a neural network model.",1997-04-15,"The title of the patent is method and apparatus for testing material strengths and its abstract is an apparatus and a non-destructive method for evaluating material strengths is described. one embodiment of the apparatus comprises a waveform generator that generates either (1) a sinusoidal waveform having a frequency that sweeps from a low frequency to a high frequency, or a high frequency to a low frequency, wherein the low frequency is from about 10 hz to about 150 hz, and the high frequency is from about 6,000 hz to about 24,000 hz, or (2) a pseudo random within the frequency range of from about 150 hz to about 6,000 hz. an electromechanical driver is electronically linked to the waveform generator and mechanically coupled to a test material, particularly in-service utility poles, at a drive position. force and acceleration sensors are coupled to the test material at various positions. a microprocessor is used to collect digitized data from the force sensor and the accelerometers, perform a transfer function and determine the strength remaining in the test material using a neural network model. dated 1997-04-15"
5621724,echo cancelling device capable of coping with deterioration of acoustic echo condition in a short time,"in an echo cancelling device including adaptive echo cancellers for cancelling an echo signal to a microphone amplifier and an echo signal to a receiver amplifier, respectively, there is provided a neural network controller responsive to cancelling errors at every taps in each of the cancellers for searching, with reference to a tap producing a maximum error and to the maximum error, a library containing past echo signal patterns to find an optimum echo signal pattern as a searched pattern, changing weighting factors for the taps in each of the cancellers in accordance with the searched pattern, modifying the searched pattern to eliminate the error when the error becomes small, finely adjusting the weighting factors for the taps, and registering the modified pattern in the library.",1997-04-15,"The title of the patent is echo cancelling device capable of coping with deterioration of acoustic echo condition in a short time and its abstract is in an echo cancelling device including adaptive echo cancellers for cancelling an echo signal to a microphone amplifier and an echo signal to a receiver amplifier, respectively, there is provided a neural network controller responsive to cancelling errors at every taps in each of the cancellers for searching, with reference to a tap producing a maximum error and to the maximum error, a library containing past echo signal patterns to find an optimum echo signal pattern as a searched pattern, changing weighting factors for the taps in each of the cancellers in accordance with the searched pattern, modifying the searched pattern to eliminate the error when the error becomes small, finely adjusting the weighting factors for the taps, and registering the modified pattern in the library. dated 1997-04-15"
5621815,global threshold method and apparatus,"the problem of thresholding is considered from a clustering point of view and a novel weight-based clustering method (wcthresh) is implemented in a neural network image processor 50. the neural network image processor 50 uses weights 51-53, representing clusters of gray scale pixels of an image of document 43, to provide a threshold for the image of document 43. the processor 50 modifies weights 51-53 with the input pixels and comparator 60 using a nearest value criterion to provide the threshold.",1997-04-15,"The title of the patent is global threshold method and apparatus and its abstract is the problem of thresholding is considered from a clustering point of view and a novel weight-based clustering method (wcthresh) is implemented in a neural network image processor 50. the neural network image processor 50 uses weights 51-53, representing clusters of gray scale pixels of an image of document 43, to provide a threshold for the image of document 43. the processor 50 modifies weights 51-53 with the input pixels and comparator 60 using a nearest value criterion to provide the threshold. dated 1997-04-15"
5621857,method and system for identifying and recognizing speech,"improved system and method for speaker-independent speech token recognition are described. the system is neural network-based and involves processing a sequence of spoken utterances, e.g. separately articulated letters of a name, to identify the same based upon a highest probability match of each utterance with learned speech tokens, e.g. the letters of the english language alphabet, and based upon a highest probability match of the uttered sequence with a defined utterance library, e.g. a list of names. first, the spoken utterance is digitized or captured and processed into a spectral representation. second, discrete time frames of the dft are classified phonetically. third, the time-frame outputs are used by a modified viterbi search to locate segment boundaries, near which such segment boundaries lies the information that is needed to discriminate letters. fourth, the segmented or bounded representation is reclassified using such information into individual hypothesized letters. fifth, successive, hypothesized letter scores are analyzed to obtain a high probability match with a spelled word within the utterance library. the system and method comprehend finer distinctions near points of interest used to discriminate difficult-to-recognize letter pair differences such as m/n, b/d, etc. the system is described in the context of phone line reception of names spelled by remote users.",1997-04-15,"The title of the patent is method and system for identifying and recognizing speech and its abstract is improved system and method for speaker-independent speech token recognition are described. the system is neural network-based and involves processing a sequence of spoken utterances, e.g. separately articulated letters of a name, to identify the same based upon a highest probability match of each utterance with learned speech tokens, e.g. the letters of the english language alphabet, and based upon a highest probability match of the uttered sequence with a defined utterance library, e.g. a list of names. first, the spoken utterance is digitized or captured and processed into a spectral representation. second, discrete time frames of the dft are classified phonetically. third, the time-frame outputs are used by a modified viterbi search to locate segment boundaries, near which such segment boundaries lies the information that is needed to discriminate letters. fourth, the segmented or bounded representation is reclassified using such information into individual hypothesized letters. fifth, successive, hypothesized letter scores are analyzed to obtain a high probability match with a spelled word within the utterance library. the system and method comprehend finer distinctions near points of interest used to discriminate difficult-to-recognize letter pair differences such as m/n, b/d, etc. the system is described in the context of phone line reception of names spelled by remote users. dated 1997-04-15"
5621858,neural network acoustic and visual speech recognition system training method and apparatus,"the apparatus for the recognition of speech includes an acoustic preprocessor, a visual preprocessor, and a speech classifier that operates on the acoustic and visual preprocessed data. the acoustic preprocessor comprises a log mel spectrum analyzer that produces an equal mel bandwidth log power spectrum. the visual processor detects the motion of a set of fiducial markers on the speaker's face and extracts a set of normalized distance vectors describing lip and mouth movement. the speech classifier uses a multilevel time-delay neural network operating on the preprocessed acoustic and visual data to form an output probability distribution that indicates the probability of each candidate utterance having been spoken, based on the acoustic and visual data. the training system includes the speech recognition apparatus and a control processor with an associated memory. noisy acoustic input training data together with visual data is used to generate acoustic and visual feature training vectors for processing by the speech classifier. a control computer adjusts the synaptic weights of the speech classifier based upon the noisy input training data and exemplar output vectors for producing a robustly trained classifier based on the analogous visual counterpart of the lombard effect.",1997-04-15,"The title of the patent is neural network acoustic and visual speech recognition system training method and apparatus and its abstract is the apparatus for the recognition of speech includes an acoustic preprocessor, a visual preprocessor, and a speech classifier that operates on the acoustic and visual preprocessed data. the acoustic preprocessor comprises a log mel spectrum analyzer that produces an equal mel bandwidth log power spectrum. the visual processor detects the motion of a set of fiducial markers on the speaker's face and extracts a set of normalized distance vectors describing lip and mouth movement. the speech classifier uses a multilevel time-delay neural network operating on the preprocessed acoustic and visual data to form an output probability distribution that indicates the probability of each candidate utterance having been spoken, based on the acoustic and visual data. the training system includes the speech recognition apparatus and a control processor with an associated memory. noisy acoustic input training data together with visual data is used to generate acoustic and visual feature training vectors for processing by the speech classifier. a control computer adjusts the synaptic weights of the speech classifier based upon the noisy input training data and exemplar output vectors for producing a robustly trained classifier based on the analogous visual counterpart of the lombard effect. dated 1997-04-15"
5621861,method of reducing amount of data required to achieve neural network learning,"a method of reducing the amount of learning data required to execute a neural network learning procedure, whereby an original entire set of learning sample data are divided, using cluster analysis of the original entire learning sample data, into a plurality of sub-groups, with the sub-groups being respectively applied to a neural network as learning data and with respective values of recognition index obtained thereby for the neural network being judged, to select the smallest sub-group which will provide a value of recognition index that is at least equal to the recognition index obtainable by using the original entire learning data.",1997-04-15,"The title of the patent is method of reducing amount of data required to achieve neural network learning and its abstract is a method of reducing the amount of learning data required to execute a neural network learning procedure, whereby an original entire set of learning sample data are divided, using cluster analysis of the original entire learning sample data, into a plurality of sub-groups, with the sub-groups being respectively applied to a neural network as learning data and with respective values of recognition index obtained thereby for the neural network being judged, to select the smallest sub-group which will provide a value of recognition index that is at least equal to the recognition index obtainable by using the original entire learning data. dated 1997-04-15"
5621862,information processing apparatus for implementing neural network,"in an information processing apparatus for implementing a neural network, if an input vector is inputted to a calculating unit, a neuron which responds to the input vector is retrieved in accordance with network interconnection information stored in a first storage unit and the neuron number indicating the retrieved neuron is written in a first register. the calculating unit reads out the internal information of the neuron stored in a second storage unit by using the neuron number, writes it in a second register, and calculates the sum of products of the outputs of the neurons and the connection loads of synapses connected to the neurons. by repeating the sequence of operations by the number of times corresponding to the total number of input vectors, a recognition process is executed. the neural network can easily be expanded by rewriting the contents of the first and second storage units.",1997-04-15,"The title of the patent is information processing apparatus for implementing neural network and its abstract is in an information processing apparatus for implementing a neural network, if an input vector is inputted to a calculating unit, a neuron which responds to the input vector is retrieved in accordance with network interconnection information stored in a first storage unit and the neuron number indicating the retrieved neuron is written in a first register. the calculating unit reads out the internal information of the neuron stored in a second storage unit by using the neuron number, writes it in a second register, and calculates the sum of products of the outputs of the neurons and the connection loads of synapses connected to the neurons. by repeating the sequence of operations by the number of times corresponding to the total number of input vectors, a recognition process is executed. the neural network can easily be expanded by rewriting the contents of the first and second storage units. dated 1997-04-15"
5621863,neuron circuit,"in a neural network comprised of a plurality of neuron circuits, an improved neuron circuit that generates local result signals, e.g. of the fire type, and a local output signal of the distance or category type. the neuron circuit which is connected to buses that transport input data (e.g. the input category) and control signals. a multi-norm distance evaluation circuit calculates the distance d between the input vector and a prototype vector stored in a r/w memory circuit. a distance compare circuit compares this distance d with either the stored prototype vector's actual influence field or the lower limit thereof to generate first and second comparison signals. an identification circuit processes the comparison signals, the input category signal, the local category signal and a feedback signal to generate local result signals that represent the neuron circuit's response to the input vector. a minimum distance determination circuit determines the minimum distance dmin among all the calculated distances from all of the neuron circuits of the neural network and generates a local output signal of the distance type. the circuit may be used to search and sort categories. the feed-back signal is collectively generated by all the neuron circuits by oring all the local distances/categories. a daisy chain circuit is serially connected to corresponding daisy chain circuits of two adjacent neuron circuits to chain the neurons together. the daisy chain circuit also determines the neuron circuit state as free or engaged. finally, a context circuitry enables or inhibits neuron participation with other neuron circuits in generation of the feedback signal.",1997-04-15,"The title of the patent is neuron circuit and its abstract is in a neural network comprised of a plurality of neuron circuits, an improved neuron circuit that generates local result signals, e.g. of the fire type, and a local output signal of the distance or category type. the neuron circuit which is connected to buses that transport input data (e.g. the input category) and control signals. a multi-norm distance evaluation circuit calculates the distance d between the input vector and a prototype vector stored in a r/w memory circuit. a distance compare circuit compares this distance d with either the stored prototype vector's actual influence field or the lower limit thereof to generate first and second comparison signals. an identification circuit processes the comparison signals, the input category signal, the local category signal and a feedback signal to generate local result signals that represent the neuron circuit's response to the input vector. a minimum distance determination circuit determines the minimum distance dmin among all the calculated distances from all of the neuron circuits of the neural network and generates a local output signal of the distance type. the circuit may be used to search and sort categories. the feed-back signal is collectively generated by all the neuron circuits by oring all the local distances/categories. a daisy chain circuit is serially connected to corresponding daisy chain circuits of two adjacent neuron circuits to chain the neurons together. the daisy chain circuit also determines the neuron circuit state as free or engaged. finally, a context circuitry enables or inhibits neuron participation with other neuron circuits in generation of the feedback signal. dated 1997-04-15"
5622171,method and system for differential diagnosis based on clinical and radiological information using artificial neural networks,"a method and system for computer-aided differential diagnosis of diseases, and in particular, computer-aided differential diagnosis using neural networks. a first embodiment of the neural network distinguishes between a plurality of interstitial lung diseases on the basis of inputted clinical parameters and radiographic information. a second embodiment distinguishes between malignant and benign mammographic cases based upon similar inputted clinical and radiographic information. the neural networks were first trained using a hypothetical data base made up of hypothetical cases for each of the interstitial lung diseases and for malignant and benign cases. the performance of the neural network was evaluated using receiver operating characteristics (roc) analysis. the decision performance of the neural network was compared to experienced radiologists and achieved a high performance comparable to that of the experienced radiologists. the neural network according to the invention can be made up of a single network or a plurality of successive or parallel networks. the neural network according to the invention can also be interfaced to a computer which provides computerized automated lung texture analysis to supply radiographic input data in an objective and automated manner.",1997-04-22,"The title of the patent is method and system for differential diagnosis based on clinical and radiological information using artificial neural networks and its abstract is a method and system for computer-aided differential diagnosis of diseases, and in particular, computer-aided differential diagnosis using neural networks. a first embodiment of the neural network distinguishes between a plurality of interstitial lung diseases on the basis of inputted clinical parameters and radiographic information. a second embodiment distinguishes between malignant and benign mammographic cases based upon similar inputted clinical and radiographic information. the neural networks were first trained using a hypothetical data base made up of hypothetical cases for each of the interstitial lung diseases and for malignant and benign cases. the performance of the neural network was evaluated using receiver operating characteristics (roc) analysis. the decision performance of the neural network was compared to experienced radiologists and achieved a high performance comparable to that of the experienced radiologists. the neural network according to the invention can be made up of a single network or a plurality of successive or parallel networks. the neural network according to the invention can also be interfaced to a computer which provides computerized automated lung texture analysis to supply radiographic input data in an objective and automated manner. dated 1997-04-22"
5623579,automated method for the systematic interpretation of resonance peaks in spectrum data,"a method for spectral signature interpretation. the method includes the creation of a mathematical model of a system or process. a neural network training set is then developed based upon the mathematical model. the neural network training set is developed by using the mathematical model to generate measurable phenomena of the system or process based upon model input parameter that correspond to the physical condition of the system or process. the neural network training set is then used to adjust internal parameters of a neural network. the physical condition of an actual system or process represented by the mathematical model is then monitored by extracting spectral features from measured spectra of the actual process or system. the spectral features are then input into said neural network to determine the physical condition of the system or process represented by the mathematical. more specifically, the neural network correlates the spectral features (i.e. measurable phenomena) of the actual process or system with the corresponding model input parameters. the model input parameters relate to specific components of the system or process, and, consequently, correspond to the physical condition of the process or system.",1997-04-22,"The title of the patent is automated method for the systematic interpretation of resonance peaks in spectrum data and its abstract is a method for spectral signature interpretation. the method includes the creation of a mathematical model of a system or process. a neural network training set is then developed based upon the mathematical model. the neural network training set is developed by using the mathematical model to generate measurable phenomena of the system or process based upon model input parameter that correspond to the physical condition of the system or process. the neural network training set is then used to adjust internal parameters of a neural network. the physical condition of an actual system or process represented by the mathematical model is then monitored by extracting spectral features from measured spectra of the actual process or system. the spectral features are then input into said neural network to determine the physical condition of the system or process represented by the mathematical. more specifically, the neural network correlates the spectral features (i.e. measurable phenomena) of the actual process or system with the corresponding model input parameters. the model input parameters relate to specific components of the system or process, and, consequently, correspond to the physical condition of the process or system. dated 1997-04-22"
5625552,closed loop neural network automatic tuner,"a closed loop neural network based autotuner develops optimized proportional, integral and/or derivative parameters based on the outputs of other elements in the loop. adjustments are initiated by making a step change in the setpoint which may be done by a user or automatically. a smith predictor may also be employed.",1997-04-29,"The title of the patent is closed loop neural network automatic tuner and its abstract is a closed loop neural network based autotuner develops optimized proportional, integral and/or derivative parameters based on the outputs of other elements in the loop. adjustments are initiated by making a step change in the setpoint which may be done by a user or automatically. a smith predictor may also be employed. dated 1997-04-29"
5625636,integration of photoactive and electroactive components with vertical cavity surface emitting lasers,"a monolithically integrated optoelectronic device is provided which integrates a vertical cavity surface emitting laser and either a photosensitive or an electrosensitive device either as input or output to the vertical cavity surface emitting laser either in parallel or series connection. both vertical and side-by-side arrangements are disclosed, and optical and electronic feedback means are provided. arrays of these devices can be configured to enable optical computing and neural network applications.",1997-04-29,"The title of the patent is integration of photoactive and electroactive components with vertical cavity surface emitting lasers and its abstract is a monolithically integrated optoelectronic device is provided which integrates a vertical cavity surface emitting laser and either a photosensitive or an electrosensitive device either as input or output to the vertical cavity surface emitting laser either in parallel or series connection. both vertical and side-by-side arrangements are disclosed, and optical and electronic feedback means are provided. arrays of these devices can be configured to enable optical computing and neural network applications. dated 1997-04-29"
5625707,training a neural network using centroid dithering by randomly displacing a template,"pattern recognition, for instance optical character recognition, is achieved by training a neural network, scanning an image, segmenting the image to detect a pattern, preprocessing the detected pattern, and applying the preprocessed detected pattern to the trained neural network. the preprocessing includes determining a centroid of the pattern and centrally positioning the centroid in a frame containing the pattern. the training of the neural network includes randomly displacing template patterns within frames before applying the template patterns to the neural network.",1997-04-29,"The title of the patent is training a neural network using centroid dithering by randomly displacing a template and its abstract is pattern recognition, for instance optical character recognition, is achieved by training a neural network, scanning an image, segmenting the image to detect a pattern, preprocessing the detected pattern, and applying the preprocessed detected pattern to the trained neural network. the preprocessing includes determining a centroid of the pattern and centrally positioning the centroid in a frame containing the pattern. the training of the neural network includes randomly displacing template patterns within frames before applying the template patterns to the neural network. dated 1997-04-29"
5625708,method and apparatus for symbol recognition using multidimensional preprocessing,"data samples describing a plurality of micro-segments that compose a symbol to be recognized are received from a device such as an electronic pad. a preprocessor maps the micro-segments into cells of an array that has several feature dimensions. the preprocessor assigns values to the cells based on the length of a micro-segment associated with the cell, and how well the features of the associated micro-segment correspond to the feature label of the cell. the cell values are used as inputs to a neural network that identifies the symbol. in one embodiment, recognizing symbols from large groups of symbols is facilitated by using a plurality of neural networks. each neural network is trained to recognize symbols from a different subgroup of symbols, and each neural network can determine whether the symbol belongs to its subgroup.",1997-04-29,"The title of the patent is method and apparatus for symbol recognition using multidimensional preprocessing and its abstract is data samples describing a plurality of micro-segments that compose a symbol to be recognized are received from a device such as an electronic pad. a preprocessor maps the micro-segments into cells of an array that has several feature dimensions. the preprocessor assigns values to the cells based on the length of a micro-segment associated with the cell, and how well the features of the associated micro-segment correspond to the feature label of the cell. the cell values are used as inputs to a neural network that identifies the symbol. in one embodiment, recognizing symbols from large groups of symbols is facilitated by using a plurality of neural networks. each neural network is trained to recognize symbols from a different subgroup of symbols, and each neural network can determine whether the symbol belongs to its subgroup. dated 1997-04-29"
5625750,catalyst monitor with direct prediction of hydrocarbon conversion efficiency by dynamic neural networks,"a process and apparatus for monitoring catalyst conversion activity includes a predictor of feedgas emissions and a predictor of tailpipe emissions, each predictor providing an output for generating a ratio of conversion activity. each predictor comprises a trained neural network receiving at least one of, and preferably a plurality of, the engine operating condition signals available from an electronic engine control. preferably, each neural network is trained by inputting accumulated data acquired from performance evaluation of a plurality of vehicles having consistent powertrains but with different degrees of deterioration.",1997-04-29,"The title of the patent is catalyst monitor with direct prediction of hydrocarbon conversion efficiency by dynamic neural networks and its abstract is a process and apparatus for monitoring catalyst conversion activity includes a predictor of feedgas emissions and a predictor of tailpipe emissions, each predictor providing an output for generating a ratio of conversion activity. each predictor comprises a trained neural network receiving at least one of, and preferably a plurality of, the engine operating condition signals available from an electronic engine control. preferably, each neural network is trained by inputting accumulated data acquired from performance evaluation of a plurality of vehicles having consistent powertrains but with different degrees of deterioration. dated 1997-04-29"
5625751,neural network for contingency ranking dynamic security indices for use under fault conditions in a power distribution system,"analysis and evaluation of outage effects on the dynamic security of power systems is made with a neural network using composite contingency severity indices. a preferably small number of indices describes the power system characteristics immediately post-contingency. these indices are then used as classifiers of the safety of the power system. using the values of the severity indices, an artificial neural network distinguishes between safe, stable contingencies and potentially unstable contingencies. the severity of the contingency is evaluated based upon a relatively small fixed set of severity indices that are calculated based on a partial time domain simulation. because a fixed set of severity indices is used, the size and architecture of the neural network is problem independent, thus permitting its use with large scale power systems. further, the amount of required time domain simulation for the selection of the potentially harmful unstable contingencies is reduced by screening out benign, stable appearing contingencies. the network is trained off-line using training cases that concentrate around the security boundary to reduce the number of cases required to train the neural network.",1997-04-29,"The title of the patent is neural network for contingency ranking dynamic security indices for use under fault conditions in a power distribution system and its abstract is analysis and evaluation of outage effects on the dynamic security of power systems is made with a neural network using composite contingency severity indices. a preferably small number of indices describes the power system characteristics immediately post-contingency. these indices are then used as classifiers of the safety of the power system. using the values of the severity indices, an artificial neural network distinguishes between safe, stable contingencies and potentially unstable contingencies. the severity of the contingency is evaluated based upon a relatively small fixed set of severity indices that are calculated based on a partial time domain simulation. because a fixed set of severity indices is used, the size and architecture of the neural network is problem independent, thus permitting its use with large scale power systems. further, the amount of required time domain simulation for the selection of the potentially harmful unstable contingencies is reduced by screening out benign, stable appearing contingencies. the network is trained off-line using training cases that concentrate around the security boundary to reduce the number of cases required to train the neural network. dated 1997-04-29"
5625752,artificial neural system with binary weighting by equal resistor network,"an artificial neural system has input operational amplifiers providing differential voltage input signals to a neuron including a voltage divider network having a plurality of substantially equal resistances selectably connectable to the components of the input signals so as to define the bits of binary weights for each of the input signals and to generate unweighted network output voltage signals corresponding to each bit position of the weights and representing the sums of the products of each input signal and the bit at each bit position. the unweighted bit position signals are provided to a bit position weighting device which is common to all of the weights of a neuron and which weights the unweighted signals by the binary positional values of the bit positions. the unweighted bit position signals are differential signals having one component generated by reference resistances of the network, and the sign of each weight may be selected by connection of the reference resistances to one or the other of the input signal components. a preferred embodiment has only one reference resistance for each weight. this reference resistance corresponds to a sign bit position, and the reference resistances for all of the weights are connected to provide a common reference voltage component for all of the unweighted bit position signals. differential voltage output signals from a system utilized as one artificial neural layer may be directly connected as differential voltage input signals for the voltage divider network of a system utilized as a second layer.",1997-04-29,"The title of the patent is artificial neural system with binary weighting by equal resistor network and its abstract is an artificial neural system has input operational amplifiers providing differential voltage input signals to a neuron including a voltage divider network having a plurality of substantially equal resistances selectably connectable to the components of the input signals so as to define the bits of binary weights for each of the input signals and to generate unweighted network output voltage signals corresponding to each bit position of the weights and representing the sums of the products of each input signal and the bit at each bit position. the unweighted bit position signals are provided to a bit position weighting device which is common to all of the weights of a neuron and which weights the unweighted signals by the binary positional values of the bit positions. the unweighted bit position signals are differential signals having one component generated by reference resistances of the network, and the sign of each weight may be selected by connection of the reference resistances to one or the other of the input signal components. a preferred embodiment has only one reference resistance for each weight. this reference resistance corresponds to a sign bit position, and the reference resistances for all of the weights are connected to provide a common reference voltage component for all of the unweighted bit position signals. differential voltage output signals from a system utilized as one artificial neural layer may be directly connected as differential voltage input signals for the voltage divider network of a system utilized as a second layer. dated 1997-04-29"
5627941,method of configuring a neural network and a diagnosis/control system using the neural network,"the number of hidden layers is no larger than 2 and a sum is determined for each case. a relationship between the sum of the inputs and teaching data for each case is expressed in a table in a descending order of the sum of inputs for each output of an output layer, and the teacher data for the maximum sum and the number of times of change in the teacher data are considered. a configuration (the number of hidden layers and the number of neurons thereof) is determined based on those data, and the coupling weights can be analytically calculated by using the table. where the number of times of change of the teacher data is odd, some inputs do not route a second hidden layer.",1997-05-06,"The title of the patent is method of configuring a neural network and a diagnosis/control system using the neural network and its abstract is the number of hidden layers is no larger than 2 and a sum is determined for each case. a relationship between the sum of the inputs and teaching data for each case is expressed in a table in a descending order of the sum of inputs for each output of an output layer, and the teacher data for the maximum sum and the number of times of change in the teacher data are considered. a configuration (the number of hidden layers and the number of neurons thereof) is determined based on those data, and the coupling weights can be analytically calculated by using the table. where the number of times of change of the teacher data is odd, some inputs do not route a second hidden layer. dated 1997-05-06"
5627942,trainable neural network having short-term memory for altering input layer topology during training,"a neutral net in which new nodes and connections are created in both input and intermediate layers during training, which is by punishment, reward and teaching. this can use a small increase in memory requirement to preclude the necessity for long training times applicable problems in speech and natural language processing, video recognition and simple logic functions.",1997-05-06,"The title of the patent is trainable neural network having short-term memory for altering input layer topology during training and its abstract is a neutral net in which new nodes and connections are created in both input and intermediate layers during training, which is by punishment, reward and teaching. this can use a small increase in memory requirement to preclude the necessity for long training times applicable problems in speech and natural language processing, video recognition and simple logic functions. dated 1997-05-06"
5627943,neural network processor including systolic array of two-dimensional layers,the invention provides a pattern recognition processing apparatus and a technique for realizing a neural network of a complex structure within the processing apparatus. the apparatus includes a neural network having two-dimensional layers connected to form a feed-forward systolic array. each two dimensional layer includes a feature extraction layer connected with a positional error absorbing layer. a host system provides inputs to the network. each layer within the network includes processing elements such as a mos analog circuit that receives input voltage signals and provides output voltage signals.,1997-05-06,The title of the patent is neural network processor including systolic array of two-dimensional layers and its abstract is the invention provides a pattern recognition processing apparatus and a technique for realizing a neural network of a complex structure within the processing apparatus. the apparatus includes a neural network having two-dimensional layers connected to form a feed-forward systolic array. each two dimensional layer includes a feature extraction layer connected with a positional error absorbing layer. a host system provides inputs to the network. each layer within the network includes processing elements such as a mos analog circuit that receives input voltage signals and provides output voltage signals. dated 1997-05-06
5629870,method and apparatus for predicting electric induction machine failure during operation,"methods and apparatus for identifying in real time an operating condition of an in-service electrical device which draws a power load by monitoring spectral content of the power signature and associating spectral components with device operating conditions. in the preferred embodiment, motor (12) power line current signature is sampled by data conditioning sampler (18). preprocessor (30) spectrally separates the current signature into plural frequency domain components having respective frequency and magnitude values. spectral characteristic component selector filter (40) selects for analysis at least one specific spectral component by referencing a base of stored knowledge of operational characteristics of the motor. a neural network (50) associates without supervision the selected detected spectral component with an operating condition of the motor (12). the base of stored knowledge is updated with new operating condition associations which have been made during real-time monitoring. a post processor (60) may enunciate the association to a user via an output device and it may generate a control signal to operate electrical distribution system protection or control apparatus.",1997-05-13,"The title of the patent is method and apparatus for predicting electric induction machine failure during operation and its abstract is methods and apparatus for identifying in real time an operating condition of an in-service electrical device which draws a power load by monitoring spectral content of the power signature and associating spectral components with device operating conditions. in the preferred embodiment, motor (12) power line current signature is sampled by data conditioning sampler (18). preprocessor (30) spectrally separates the current signature into plural frequency domain components having respective frequency and magnitude values. spectral characteristic component selector filter (40) selects for analysis at least one specific spectral component by referencing a base of stored knowledge of operational characteristics of the motor. a neural network (50) associates without supervision the selected detected spectral component with an operating condition of the motor (12). the base of stored knowledge is updated with new operating condition associations which have been made during real-time monitoring. a post processor (60) may enunciate the association to a user via an output device and it may generate a control signal to operate electrical distribution system protection or control apparatus. dated 1997-05-13"
5630018,fuzzy inference device using neural network,"a fuzzy inference device determines an inference operational quantity in accordance with inference rules each constituted by an antecedent and a consequent. an inference rule division determiner which receives data of input variables and output variables so as to determine the number of the inference rules. an antecedent neural element obtains a membership value corresponding to an antecedent of a specific inference rule from the divided data of the input variables and the output variables. a situational change processor adaptively determines an inference quantity of a consequent of each inference rule in the case of a change of an initial state or inference situations. an inference operational quantity determiner receives outputs from the antecedent neural element and the situational change processor and performs fuzzy inference in accordance with the inference rules so as to determine the inference operational quantity. an evaluator evaluates, on the basis of an evaluation reference, the inference operational quantity outputted by the inference operational quantity determiner.",1997-05-13,"The title of the patent is fuzzy inference device using neural network and its abstract is a fuzzy inference device determines an inference operational quantity in accordance with inference rules each constituted by an antecedent and a consequent. an inference rule division determiner which receives data of input variables and output variables so as to determine the number of the inference rules. an antecedent neural element obtains a membership value corresponding to an antecedent of a specific inference rule from the divided data of the input variables and the output variables. a situational change processor adaptively determines an inference quantity of a consequent of each inference rule in the case of a change of an initial state or inference situations. an inference operational quantity determiner receives outputs from the antecedent neural element and the situational change processor and performs fuzzy inference in accordance with the inference rules so as to determine the inference operational quantity. an evaluator evaluates, on the basis of an evaluation reference, the inference operational quantity outputted by the inference operational quantity determiner. dated 1997-05-13"
5630019,waveform evaluating apparatus using neural network,"disclosed is a waveform evaluating apparatus for evaluating and adjusting a waveform measured by a measurement apparatus such as a synchroscope and, more particularly, a waveform evaluating apparatus having a plurality of neural network modules formed independently for each object of judgment, in which the neural weight ratio of each neural network module is determined by causing the module to learn with a first ideal waveform module as an ideal signal and the like. such an arrangement is also made, in which a signal in phase with the learned teacher signal extracted from the signal from an object of judgment signal is input to the input layer, and in which phasic information is detected in a phase detecting portion and a waveform is sliced in a waveform slicing portion on the basis of the phasic information. further, such an arrangement is made, in which signal waveform data as an object of evaluation is input to the input layer and an analog output is output from the output layer. therefore, waveform adjustments can also be achieved by causing the module to learn with a waveform to be adjusted.",1997-05-13,"The title of the patent is waveform evaluating apparatus using neural network and its abstract is disclosed is a waveform evaluating apparatus for evaluating and adjusting a waveform measured by a measurement apparatus such as a synchroscope and, more particularly, a waveform evaluating apparatus having a plurality of neural network modules formed independently for each object of judgment, in which the neural weight ratio of each neural network module is determined by causing the module to learn with a first ideal waveform module as an ideal signal and the like. such an arrangement is also made, in which a signal in phase with the learned teacher signal extracted from the signal from an object of judgment signal is input to the input layer, and in which phasic information is detected in a phase detecting portion and a waveform is sliced in a waveform slicing portion on the basis of the phasic information. further, such an arrangement is made, in which signal waveform data as an object of evaluation is input to the input layer and an analog output is output from the output layer. therefore, waveform adjustments can also be achieved by causing the module to learn with a waveform to be adjusted. dated 1997-05-13"
5630020,learning method and neural network structure,"the invention relates to a method of learning which is carried out in a neural network operating on the basis of the gradient back-propagation algorithm. in order to determine the new synaptic coefficients with a minimum learning period, the invention introduces parameters which privilege corrections based on the sign of the error at the start of learning and which gradually induce less coarse corrections. this can be complemented by other parameters which favor a layer-wise strategy, accelerating the learning in the input layers with respect to the output layers. it is also possible to add a strategy which acts on the entire neural network.",1997-05-13,"The title of the patent is learning method and neural network structure and its abstract is the invention relates to a method of learning which is carried out in a neural network operating on the basis of the gradient back-propagation algorithm. in order to determine the new synaptic coefficients with a minimum learning period, the invention introduces parameters which privilege corrections based on the sign of the error at the start of learning and which gradually induce less coarse corrections. this can be complemented by other parameters which favor a layer-wise strategy, accelerating the learning in the input layers with respect to the output layers. it is also possible to add a strategy which acts on the entire neural network. dated 1997-05-13"
5630021,hamming neural network circuit,"a hamming neural network circuit is provided with n binary inputs and m exemplar template outputs, and has a template matching calculation subnet and a winner-take-all subnet. the template matching calculation subnet includes m first neurons in which m exemplar templates are stored respectively. each first neuron includes n pull-up and pull-down transistor pairs connected in parallel with each other, and connected to and controlled by the n binary inputs, respectively, so that the m first neurons generate m template matching signals depending on the matching degrees between the n binary inputs and the m exemplar templates. the winner-take-all subnet includes m second neurons, each having a template competition node, a load element connected between a power source and the template competition node, and a competition circuit connected between the template competition node and ground. the m template competition nodes are connected to the m template matching signals respectively for generating the m exemplar template outputs. the competition circuit of each second neuron includes m -1 parallel-connected transistors controlled respectively by the template competition nodes of all second neurons except the template competition node of itself so that the template competition node connecting with the largest template matching signal is eventually at a relatively high voltage level, and the other template competition nodes are at a relatively low voltage level, after competition.",1997-05-13,"The title of the patent is hamming neural network circuit and its abstract is a hamming neural network circuit is provided with n binary inputs and m exemplar template outputs, and has a template matching calculation subnet and a winner-take-all subnet. the template matching calculation subnet includes m first neurons in which m exemplar templates are stored respectively. each first neuron includes n pull-up and pull-down transistor pairs connected in parallel with each other, and connected to and controlled by the n binary inputs, respectively, so that the m first neurons generate m template matching signals depending on the matching degrees between the n binary inputs and the m exemplar templates. the winner-take-all subnet includes m second neurons, each having a template competition node, a load element connected between a power source and the template competition node, and a competition circuit connected between the template competition node and ground. the m template competition nodes are connected to the m template matching signals respectively for generating the m exemplar template outputs. the competition circuit of each second neuron includes m -1 parallel-connected transistors controlled respectively by the template competition nodes of all second neurons except the template competition node of itself so that the template competition node connecting with the largest template matching signal is eventually at a relatively high voltage level, and the other template competition nodes are at a relatively low voltage level, after competition. dated 1997-05-13"
5630022,self-organizing pattern learning optical system,"a pattern learning system uses a self-organizing optical neural network of the kohonen variety for unsupervised learning. the learning system comprises an optical pattern correlation degree detector, which comprises a pattern storing device for storing a plurality of patterns, an input pattern displaying device for displaying a presented input pattern, and a photo detecting device. two or more of the input pattern displaying devices, the pattern storing device, and the photo detecting device are located at positions adjacent to each other. the photo detecting device optically detects a degree of pattern correlation between the input pattern displayed on the input pattern displaying device and each of the memory patterns stored in the pattern storing device. a learning pattern creating device creates a group of learning patterns in accordance with the degrees of pattern correlation, which have been detected by the optical pattern correlation degree detecting device, and the input pattern. a memory pattern updating device updates the memory patterns, which are stored in the pattern storing device, in accordance with the learning patterns, which have been created by the learning pattern creating devices.",1997-05-13,"The title of the patent is self-organizing pattern learning optical system and its abstract is a pattern learning system uses a self-organizing optical neural network of the kohonen variety for unsupervised learning. the learning system comprises an optical pattern correlation degree detector, which comprises a pattern storing device for storing a plurality of patterns, an input pattern displaying device for displaying a presented input pattern, and a photo detecting device. two or more of the input pattern displaying devices, the pattern storing device, and the photo detecting device are located at positions adjacent to each other. the photo detecting device optically detects a degree of pattern correlation between the input pattern displayed on the input pattern displaying device and each of the memory patterns stored in the pattern storing device. a learning pattern creating device creates a group of learning patterns in accordance with the degrees of pattern correlation, which have been detected by the optical pattern correlation degree detecting device, and the input pattern. a memory pattern updating device updates the memory patterns, which are stored in the pattern storing device, in accordance with the learning patterns, which have been created by the learning pattern creating devices. dated 1997-05-13"
5630024,method and apparatus for processing using neural network with reduced calculation amount,"a neural network circuit and a processing scheme using the neural network circuit in which a synapse calculation for each input value and a corresponding synapse weight of each input value which are expressed by binary bit sequences is carried out by using a sequentially specified bit of the corresponding synapse weight, a summation calculation for sequentially summing synapse calculation results for the input values is carried out to obtain a summation value, a prescribed nonlinear processing is applied to the obtained summation value so as to determine the output value, whether the obtained summation value reached to a saturation region of a transfer characteristic of the prescribed nonlinear processing is judged, the synapse calculation and the summation calculation are controlled to sequentially carry out the synapse calculation from upper bits of the corresponding synapse weight, and to stop the synapse calculation and the summation calculation whenever it is judged that the obtained summation value reached to the saturation region.",1997-05-13,"The title of the patent is method and apparatus for processing using neural network with reduced calculation amount and its abstract is a neural network circuit and a processing scheme using the neural network circuit in which a synapse calculation for each input value and a corresponding synapse weight of each input value which are expressed by binary bit sequences is carried out by using a sequentially specified bit of the corresponding synapse weight, a summation calculation for sequentially summing synapse calculation results for the input values is carried out to obtain a summation value, a prescribed nonlinear processing is applied to the obtained summation value so as to determine the output value, whether the obtained summation value reached to a saturation region of a transfer characteristic of the prescribed nonlinear processing is judged, the synapse calculation and the summation calculation are controlled to sequentially carry out the synapse calculation from upper bits of the corresponding synapse weight, and to stop the synapse calculation and the summation calculation whenever it is judged that the obtained summation value reached to the saturation region. dated 1997-05-13"
5630957,control of power to an inductively heated part,"a process for induction hardening a part to a desired depth with an ac signal applied to the part from a closely coupled induction coil includes measuring the voltage of the ac signal at the coil and the current passing through the coil; and controlling the depth of hardening of the part from the measured voltage and current. the control system determines parameters of the part that are functions of applied voltage and current to the induction coil, and uses a neural network to control the application of the ac signal based on the detected functions for each part.",1997-05-20,"The title of the patent is control of power to an inductively heated part and its abstract is a process for induction hardening a part to a desired depth with an ac signal applied to the part from a closely coupled induction coil includes measuring the voltage of the ac signal at the coil and the current passing through the coil; and controlling the depth of hardening of the part from the measured voltage and current. the control system determines parameters of the part that are functions of applied voltage and current to the induction coil, and uses a neural network to control the application of the ac signal based on the detected functions for each part. dated 1997-05-20"
5631469,neural network computing system for pattern recognition of thermoluminescence signature spectra and chemical defense,"a four-layer neural network is trained with data of midinfrared absorption by nerve and blister agent compounds (and simulants of this chemical group) in a standoff detection application. known infrared absorption spectra by these analyte compounds and their computed first derivative are scaled and then transformed into binary or decimal arrays for network training by a backward-error-propagation (bep) algorithm with gradient descent paradigm. the neural network transfer function gain and learning rate are adjusted on occasion per training session so that a global minimum in final epoch convergence is attained. three successful neural network filters have been built around an architecture design containing: (1) an input layer of 350 neurons, one neuron per absorption intensity spanning 700.ltoreq..nu..ltoreq.1400 wavenumbers with resolution .delta..nu.=2; (2) two hidden layers in 256- and 128-neuron groups, respectively, providing good training convergence and adaptable for downloading to a configured group of neural ic chips; and (3) an output layer of one neuron per analyte--each analyte defined by a singular vector in the training data set. such a neural network is preferably implemented with a network of known microprocessor chips.",1997-05-20,"The title of the patent is neural network computing system for pattern recognition of thermoluminescence signature spectra and chemical defense and its abstract is a four-layer neural network is trained with data of midinfrared absorption by nerve and blister agent compounds (and simulants of this chemical group) in a standoff detection application. known infrared absorption spectra by these analyte compounds and their computed first derivative are scaled and then transformed into binary or decimal arrays for network training by a backward-error-propagation (bep) algorithm with gradient descent paradigm. the neural network transfer function gain and learning rate are adjusted on occasion per training session so that a global minimum in final epoch convergence is attained. three successful neural network filters have been built around an architecture design containing: (1) an input layer of 350 neurons, one neuron per absorption intensity spanning 700.ltoreq..nu..ltoreq.1400 wavenumbers with resolution .delta..nu.=2; (2) two hidden layers in 256- and 128-neuron groups, respectively, providing good training convergence and adaptable for downloading to a configured group of neural ic chips; and (3) an output layer of one neuron per analyte--each analyte defined by a singular vector in the training data set. such a neural network is preferably implemented with a network of known microprocessor chips. dated 1997-05-20"
5632006,circuit and method of error correcting with an artificial neural network,"an artificial neural network performs error correction on an input signal vector. the input signal vector is process in a forward direction through synapses in each of a plurality of neurons for providing an output signal from each of the neurons. the output signals from the neurons are monitored until the one having the greatest activity level is identified. a reverse flow signal having a predetermined magnitude is processed in the reverse direction through the neuron having the greatest activity level for updating the input signal vector. alternately, the output signals of competing neurons may be applied through synapses weighted to favor the neuron having the greatest output signal activity. thus, the neuron with synapses most closely matched to the elements of the input signal vector overpowers the remaining neurons and wins the competition. once the winning neuron is identified, its output signal is reverse processed through the respective neuron for providing an improved input signal vector with reduced noise and error. the improved signal is reprocessed in the forward direction through the neural network for providing higher confidence output signals.",1997-05-20,"The title of the patent is circuit and method of error correcting with an artificial neural network and its abstract is an artificial neural network performs error correction on an input signal vector. the input signal vector is process in a forward direction through synapses in each of a plurality of neurons for providing an output signal from each of the neurons. the output signals from the neurons are monitored until the one having the greatest activity level is identified. a reverse flow signal having a predetermined magnitude is processed in the reverse direction through the neuron having the greatest activity level for updating the input signal vector. alternately, the output signals of competing neurons may be applied through synapses weighted to favor the neuron having the greatest output signal activity. thus, the neuron with synapses most closely matched to the elements of the input signal vector overpowers the remaining neurons and wins the competition. once the winning neuron is identified, its output signal is reverse processed through the respective neuron for providing an improved input signal vector with reduced noise and error. the improved signal is reprocessed in the forward direction through the neural network for providing higher confidence output signals. dated 1997-05-20"
5634063,neural network and method for operating the same,"a neural network provides faster learning speed and simplified overall structure through use of the concept of indirect association and a method for operating the same. the neural network is constructed as a clcam comprising an input-side single layer perceptron adapted to realize direct associations (x.sub.i,z.sub.1i) as linearly separable problems with respect to given inputs (x.sub.i) and first intermediate states (z.sub.1i) derived by the user, an output-side single layer perceptron adapted to realize direct associations (z.sub.2i,y.sub.i) as linearly separable problems with respect to given outputs (y.sub.i) and second intermediate states (z.sub.2i) derived by the user, and a location addressable memory adapted to connect said first intermediate states (z.sub.1i) with said second intermediate states (z.sub.2i). the neural network is also constructed as hylcam comprising a single layer perceptron adapted to realize direct associations (x.sub.i,z.sub.i) as linearly separable problems with respect to given inputs (x.sub.i) and intermediate states (z.sub.i) manually derived by the user, and a location addressable memory adapted to receive the intermediate states (z.sub.i) from the single layer perceptron as addresses and store given output data (y.sub.i) as desired output values, correspondingly to the addresses.",1997-05-27,"The title of the patent is neural network and method for operating the same and its abstract is a neural network provides faster learning speed and simplified overall structure through use of the concept of indirect association and a method for operating the same. the neural network is constructed as a clcam comprising an input-side single layer perceptron adapted to realize direct associations (x.sub.i,z.sub.1i) as linearly separable problems with respect to given inputs (x.sub.i) and first intermediate states (z.sub.1i) derived by the user, an output-side single layer perceptron adapted to realize direct associations (z.sub.2i,y.sub.i) as linearly separable problems with respect to given outputs (y.sub.i) and second intermediate states (z.sub.2i) derived by the user, and a location addressable memory adapted to connect said first intermediate states (z.sub.1i) with said second intermediate states (z.sub.2i). the neural network is also constructed as hylcam comprising a single layer perceptron adapted to realize direct associations (x.sub.i,z.sub.i) as linearly separable problems with respect to given inputs (x.sub.i) and intermediate states (z.sub.i) manually derived by the user, and a location addressable memory adapted to receive the intermediate states (z.sub.i) from the single layer perceptron as addresses and store given output data (y.sub.i) as desired output values, correspondingly to the addresses. dated 1997-05-27"
5634087,rapidly trainable neural tree network,"an apparatus and methods characterized by an electric neural network including a node having multipliers respectively receiving signals representing feature vector elements and signals representing weight vector elements to produce product signals, a summer to add the product signals with a bias signal and output a sum signal to a hard limiter, the hard limiter for outputting a preliminary output signal of polarity. in response to the output signal of polarity, one of at least two logic branches is enabled. in response to such enabling, weight elements are assigned to a next weight vector to be used in subsequent processing by the one of the at least two logic branches until a label is to be produced.",1997-05-27,"The title of the patent is rapidly trainable neural tree network and its abstract is an apparatus and methods characterized by an electric neural network including a node having multipliers respectively receiving signals representing feature vector elements and signals representing weight vector elements to produce product signals, a summer to add the product signals with a bias signal and output a sum signal to a hard limiter, the hard limiter for outputting a preliminary output signal of polarity. in response to the output signal of polarity, one of at least two logic branches is enabled. in response to such enabling, weight elements are assigned to a next weight vector to be used in subsequent processing by the one of the at least two logic branches until a label is to be produced. dated 1997-05-27"
5636326,method for operating an optimal weight pruning apparatus for designing artificial neural networks,"a method and apparatus for designing a multilayer feed forward neural network that produces a design having a minimum number of connecting weights is based on a novel iterative procedure for inverting the full hessian matrix of the neural network. the inversion of the full hessian matrix results in a practical strategy for pruning weights of a trained neural network. the error caused by pruning is minimized by a correction that is applied to remaining (un-pruned) weights thus reducing the need for retraining. however, retraining may be applied to the network possibly leading to further the simplification of the network design.",1997-06-03,"The title of the patent is method for operating an optimal weight pruning apparatus for designing artificial neural networks and its abstract is a method and apparatus for designing a multilayer feed forward neural network that produces a design having a minimum number of connecting weights is based on a novel iterative procedure for inverting the full hessian matrix of the neural network. the inversion of the full hessian matrix results in a practical strategy for pruning weights of a trained neural network. the error caused by pruning is minimized by a correction that is applied to remaining (un-pruned) weights thus reducing the need for retraining. however, retraining may be applied to the network possibly leading to further the simplification of the network design. dated 1997-06-03"
5636327,neural network circuit,"in a multilayered neural network for recognizing and processing characteristic data of images and the like by carrying out network arithmetical operations, characteristic data memories store the characteristic data of the layers. coefficient memories store respective coupling coefficients of the layers other than the last layer. a weight memory stores weights of neurons of the last layer. address converters carry out arithmetical operations to find out addresses of nets of the network whose coupling coefficients are significant. a table memory outputs a total coupling coefficient obtained by inter-multiplying the significant coupling coefficients read out from the coefficient memories of the layers. a cumulative operation unit performs cumulative additions of the product of the total coupling coefficient times the weight of the weight memory. arithmetical operations are carried out only on particular nets with a significant coupling coefficient value. the speed of operation and recognition can be improved.",1997-06-03,"The title of the patent is neural network circuit and its abstract is in a multilayered neural network for recognizing and processing characteristic data of images and the like by carrying out network arithmetical operations, characteristic data memories store the characteristic data of the layers. coefficient memories store respective coupling coefficients of the layers other than the last layer. a weight memory stores weights of neurons of the last layer. address converters carry out arithmetical operations to find out addresses of nets of the network whose coupling coefficients are significant. a table memory outputs a total coupling coefficient obtained by inter-multiplying the significant coupling coefficients read out from the coefficient memories of the layers. a cumulative operation unit performs cumulative additions of the product of the total coupling coefficient times the weight of the weight memory. arithmetical operations are carried out only on particular nets with a significant coupling coefficient value. the speed of operation and recognition can be improved. dated 1997-06-03"
5638125,quantization step size control apparatus using neural networks,"an apparatus for controlling a quantization step size for use in an encoder which divides one image frame into first blocks and encodes the divided first blocks and transmits the encoded data at a constant transmission rate. the apparatus includes a forward analyzer for detecting image complexity with respect to each first block to be quantized, a luminance analyzer for outputting a luminance value representative of each first block, a picture quality estimator for restoring quantized data by using a quantization step size corresponding to the quantized data and generating a judgement reference value corresponding to a minimum blocking effect which cannot be visually recognized on the basis of the restored data of every second block composed of first blocks, a buffer for storing the quantized data and outputting buffer occupancy of the stored data, a neural network having stored weight values which are updated according to the judgement reference value of the picture quality estimator with respect to the quantized previous second block and the stored update rule, generating a quantization step size for the present second block on the basis of the motion vectors, and/or the image complexity and/or the luminance values with respect to the present second block, the buffer occupancy and the updated weight values, and supplying the quantization step size to the picture quality estimator, and a quantizer for quantizing data of the second blocks encoded by the encoder according to a corresponding quantization step size supplied from the neural network.",1997-06-10,"The title of the patent is quantization step size control apparatus using neural networks and its abstract is an apparatus for controlling a quantization step size for use in an encoder which divides one image frame into first blocks and encodes the divided first blocks and transmits the encoded data at a constant transmission rate. the apparatus includes a forward analyzer for detecting image complexity with respect to each first block to be quantized, a luminance analyzer for outputting a luminance value representative of each first block, a picture quality estimator for restoring quantized data by using a quantization step size corresponding to the quantized data and generating a judgement reference value corresponding to a minimum blocking effect which cannot be visually recognized on the basis of the restored data of every second block composed of first blocks, a buffer for storing the quantized data and outputting buffer occupancy of the stored data, a neural network having stored weight values which are updated according to the judgement reference value of the picture quality estimator with respect to the quantized previous second block and the stored update rule, generating a quantization step size for the present second block on the basis of the motion vectors, and/or the image complexity and/or the luminance values with respect to the present second block, the buffer occupancy and the updated weight values, and supplying the quantization step size to the picture quality estimator, and a quantizer for quantizing data of the second blocks encoded by the encoder according to a corresponding quantization step size supplied from the neural network. dated 1997-06-10"
5638281,target prediction and collision warning system,a device for target prediction and collision warning for tracking objects in a region proximate to a vehicle includes a signal transmitter which provides first and second detection signals for at least partial reflection by an object located in a spatial region. the device further includes a signal receiver for receiving the deflected first and second detection signals corresponding to first and second parameter signals. a fourier transform circuit is provided for receiving the first and second object parameter signals and generating first and second fourier transform object parameter signals corresponding to relative range and velocity data of a target being tracked. the device includes a probabilistic neural network which preferably sorts the first and second fourier transform object parameter signals corresponding to the relative range and velocity of a target being tracked. operatively coupled to the probabilistic neural network is a target tracker circuit which receives the sorted first and second fourier transform object parameter signals after at least three samples of relative range and velocity data have been measured. the target tracker generates an output signal indicative of a prediction of regression parameters of a second order or higher order equation that characterizes the change in relative range and velocity of the target being tracked.,1997-06-10,The title of the patent is target prediction and collision warning system and its abstract is a device for target prediction and collision warning for tracking objects in a region proximate to a vehicle includes a signal transmitter which provides first and second detection signals for at least partial reflection by an object located in a spatial region. the device further includes a signal receiver for receiving the deflected first and second detection signals corresponding to first and second parameter signals. a fourier transform circuit is provided for receiving the first and second object parameter signals and generating first and second fourier transform object parameter signals corresponding to relative range and velocity data of a target being tracked. the device includes a probabilistic neural network which preferably sorts the first and second fourier transform object parameter signals corresponding to the relative range and velocity of a target being tracked. operatively coupled to the probabilistic neural network is a target tracker circuit which receives the sorted first and second fourier transform object parameter signals after at least three samples of relative range and velocity data have been measured. the target tracker generates an output signal indicative of a prediction of regression parameters of a second order or higher order equation that characterizes the change in relative range and velocity of the target being tracked. dated 1997-06-10
5638487,automatic speech recognition,"a scheme for recognizing speech represented by a sequence of frames of acoustic events separated by boundaries, according to which the frames of speech are processed to assign to received frames respective boundary probabilities representative of the degree to which the frames of speech correspond to stored representations of boundaries between acoustic events. the assigned boundary probabilities are used in subsequent processing steps to enhance recognition of speech. the assignment of boundary probabilities and further adjustments of the assigned probabilities are preferably conducted by an artificial neural network (ann).",1997-06-10,"The title of the patent is automatic speech recognition and its abstract is a scheme for recognizing speech represented by a sequence of frames of acoustic events separated by boundaries, according to which the frames of speech are processed to assign to received frames respective boundary probabilities representative of the degree to which the frames of speech correspond to stored representations of boundaries between acoustic events. the assigned boundary probabilities are used in subsequent processing steps to enhance recognition of speech. the assignment of boundary probabilities and further adjustments of the assigned probabilities are preferably conducted by an artificial neural network (ann). dated 1997-06-10"
5638491,method and apparatus for hierarchical input classification using a neural network,"an input is classified into one of a plurality of possible outputs. a top-level classifier generates an approximate identification for the input as one of the possible outputs and selects two or more neural networks corresponding to the approximate identification. the selected neural networks generate two or more identifications for the input as one or more of the possible outputs. a postprocessor classifies the input as one of the possible outputs in accordance with the two or more identifications of the selected neural networks. according to an alternative embodiment, a top-level classifier selects a subset of neurons of a neural network and the subset of neurons identifies the input as one of the possible outputs.",1997-06-10,"The title of the patent is method and apparatus for hierarchical input classification using a neural network and its abstract is an input is classified into one of a plurality of possible outputs. a top-level classifier generates an approximate identification for the input as one of the possible outputs and selects two or more neural networks corresponding to the approximate identification. the selected neural networks generate two or more identifications for the input as one or more of the possible outputs. a postprocessor classifies the input as one of the possible outputs in accordance with the two or more identifications of the selected neural networks. according to an alternative embodiment, a top-level classifier selects a subset of neurons of a neural network and the subset of neurons identifies the input as one of the possible outputs. dated 1997-06-10"
5640103,radial basis function neural network autoassociator and method for induction motor monitoring,""" a method for detecting a departure from normal operation of an electric motor comprises obtaining a set of normal current measurements for a motor being monitored; forming clusters of the normal current measurements; training a neural network auto-associator using the set of normal current measurements; making current measurements for the motor in operation; comparing the input and output of the auto-associator; and indicating abnormal operation whenever the current measurements deviate more than a predetermined amount from the normal current measurements. the method models a set of normal current measurements for the motor being monitored, and indicates a potential failure whenever measurements from the motor deviate significantly from a model. the model takes the form of an neural network auto-associator which is """"trained""""--using clusters of current measurements collected while the motor is known to be in a normal operating condition--to reproduce the inputs on the output. a new set of fft's of current measurements are classified as """"good"""" or """"bad"""" by first transforming the measurement using a fast fourier transform (fft) and an internal scaling procedure, and then applying a subset of the transformed measurements as inputs to the neural network auto-associator. a decision is generated based on the difference between the input and output of the network. """,1997-06-17,"The title of the patent is radial basis function neural network autoassociator and method for induction motor monitoring and its abstract is "" a method for detecting a departure from normal operation of an electric motor comprises obtaining a set of normal current measurements for a motor being monitored; forming clusters of the normal current measurements; training a neural network auto-associator using the set of normal current measurements; making current measurements for the motor in operation; comparing the input and output of the auto-associator; and indicating abnormal operation whenever the current measurements deviate more than a predetermined amount from the normal current measurements. the method models a set of normal current measurements for the motor being monitored, and indicates a potential failure whenever measurements from the motor deviate significantly from a model. the model takes the form of an neural network auto-associator which is """"trained""""--using clusters of current measurements collected while the motor is known to be in a normal operating condition--to reproduce the inputs on the output. a new set of fft's of current measurements are classified as """"good"""" or """"bad"""" by first transforming the measurement using a fast fourier transform (fft) and an internal scaling procedure, and then applying a subset of the transformed measurements as inputs to the neural network auto-associator. a decision is generated based on the difference between the input and output of the network. "" dated 1997-06-17"
5640468,method for identifying objects and features in an image,"the present invention features the use of the fundamental concept of color perception and multi-level resolution to perform scene segmentation and object/feature extraction in the context of self-determining and self-calibration modes. the technique uses only a single image, instead of multiple images as the input to generate segmented images. moreover, a flexible and arbitrary scheme is incorporated, rather than a fixed scheme of segmentation analysis. the process allows users to perform digital analysis using any appropriate means for object extraction after an image is segmented. first, an image is retrieved. the image is then transformed into at least two distinct bands. each transformed image is then projected into a color domain or a multi-level resolution setting. a segmented image is then created from all of the transformed images. the segmented image is analyzed to identify objects. object identification is achieved by matching a segmented region against an image library. a featureless library contains full shape, partial shape and real-world images in a dual library system. the depth contours and height-above-ground structural components constitute a dual library. also provided is a mathematical model called a parzen window-based statistical/neural network classifier, which forms an integral part of this featureless dual library object identification system. all images are considered three-dimensional. laser radar based 3-d images represent a special case.",1997-06-17,"The title of the patent is method for identifying objects and features in an image and its abstract is the present invention features the use of the fundamental concept of color perception and multi-level resolution to perform scene segmentation and object/feature extraction in the context of self-determining and self-calibration modes. the technique uses only a single image, instead of multiple images as the input to generate segmented images. moreover, a flexible and arbitrary scheme is incorporated, rather than a fixed scheme of segmentation analysis. the process allows users to perform digital analysis using any appropriate means for object extraction after an image is segmented. first, an image is retrieved. the image is then transformed into at least two distinct bands. each transformed image is then projected into a color domain or a multi-level resolution setting. a segmented image is then created from all of the transformed images. the segmented image is analyzed to identify objects. object identification is achieved by matching a segmented region against an image library. a featureless library contains full shape, partial shape and real-world images in a dual library system. the depth contours and height-above-ground structural components constitute a dual library. also provided is a mathematical model called a parzen window-based statistical/neural network classifier, which forms an integral part of this featureless dual library object identification system. all images are considered three-dimensional. laser radar based 3-d images represent a special case. dated 1997-06-17"
5640491,"control system using an adaptive neural network for target and path optimization for a multivariable, nonlinear process","a control system having four major components: a target optimizer, a path optimizer, a neural network adaptation controller and a neural network. in the target optimizer, the controlled variables are optimized to provide the most economically desirable outputs, subject to operating constraints. various manipulated variable and disturbance values are provided for modeling purposes. the neural network receives as inputs a plurality of settings for each manipulated and disturbance variable. for target optimization all the neural network input values are set equal to produce a steady state controlled variable value. the entire process is repeated with differing manipulated variable values until an optimal solution develops. the resulting target controlled and manipulated variable values are provided to the path optimizer to allow the manipulated variables to be adjusted to obtain the target output. various manipulated variable values are developed to model moves from current to desired values. in this case trend indicating values of the manipulated and disturbance variables are provided to produce time varying values of the controlled variables. the process is repeated until an optimal path is obtained, at which time the manipulated variable values are applied to the actual process. on a periodic basis all of the disturbance, manipulated and controlled variables are sampled to find areas where the training of the neural network is sparse or where high dynamic conditions are indicated. these values are added to the set of values used to train the neural network.",1997-06-17,"The title of the patent is control system using an adaptive neural network for target and path optimization for a multivariable, nonlinear process and its abstract is a control system having four major components: a target optimizer, a path optimizer, a neural network adaptation controller and a neural network. in the target optimizer, the controlled variables are optimized to provide the most economically desirable outputs, subject to operating constraints. various manipulated variable and disturbance values are provided for modeling purposes. the neural network receives as inputs a plurality of settings for each manipulated and disturbance variable. for target optimization all the neural network input values are set equal to produce a steady state controlled variable value. the entire process is repeated with differing manipulated variable values until an optimal solution develops. the resulting target controlled and manipulated variable values are provided to the path optimizer to allow the manipulated variables to be adjusted to obtain the target output. various manipulated variable values are developed to model moves from current to desired values. in this case trend indicating values of the manipulated and disturbance variables are provided to produce time varying values of the controlled variables. the process is repeated until an optimal path is obtained, at which time the manipulated variable values are applied to the actual process. on a periodic basis all of the disturbance, manipulated and controlled variables are sampled to find areas where the training of the neural network is sparse or where high dynamic conditions are indicated. these values are added to the set of values used to train the neural network. dated 1997-06-17"
5640493,historical database training method for neural networks,"an on-line training neural network for controlling a process for producing a product having at least one product property that trains by retrieving training sets from a stream of process data. the neural network detects the availability of new training data, and constructs a training set by retrieving the corresponding input data. the neural network is trained using the training set. over time, many training sets are presented to the neural network. when multiple presentations are needed to effectively train, a buffer of training sets is filled and updated as new training data becomes available. the size of the buffer is selected in accordance with the training needs of the neural network. once the buffer is full, a new training set bumps the oldest training set off the top of the buffer stack. the training sets in the buffer stack can be presented one or more times each time a new training set is constructed. an historical database of timestamped data can be used to construct training sets when training input data has a time delay from sample time to availability for the neural network. the timestamps of the training input data are used to select input data for use in the training set. using the historical database, the neural network can be trained retrospectively by searching the historical database and constructing training sets based on past data.",1997-06-17,"The title of the patent is historical database training method for neural networks and its abstract is an on-line training neural network for controlling a process for producing a product having at least one product property that trains by retrieving training sets from a stream of process data. the neural network detects the availability of new training data, and constructs a training set by retrieving the corresponding input data. the neural network is trained using the training set. over time, many training sets are presented to the neural network. when multiple presentations are needed to effectively train, a buffer of training sets is filled and updated as new training data becomes available. the size of the buffer is selected in accordance with the training needs of the neural network. once the buffer is full, a new training set bumps the oldest training set off the top of the buffer stack. the training sets in the buffer stack can be presented one or more times each time a new training set is constructed. an historical database of timestamped data can be used to construct training sets when training input data has a time delay from sample time to availability for the neural network. the timestamps of the training input data are used to select input data for use in the training set. using the historical database, the neural network can be trained retrospectively by searching the historical database and constructing training sets based on past data. dated 1997-06-17"
5640494,neural network with training by perturbation,"a neural network comprises an input port connected to an output port by one or more paths, each of which comprises an alternating series of weights and neurons. the weights amplify passing signals by a strength factor. the network can be trained by finding a set of strength factor values for the weights such that the network produces the correct output pattern from a given input pattern. during training, a strength factor perturbating and refresh means applies perturbations to the strength factors of weights in the network, and updates the values of the strength factors depending on the difference between signals appearing at the output port, for a given pair of input and training patterns, when the weight is perturbed, and when it is not.",1997-06-17,"The title of the patent is neural network with training by perturbation and its abstract is a neural network comprises an input port connected to an output port by one or more paths, each of which comprises an alternating series of weights and neurons. the weights amplify passing signals by a strength factor. the network can be trained by finding a set of strength factor values for the weights such that the network produces the correct output pattern from a given input pattern. during training, a strength factor perturbating and refresh means applies perturbations to the strength factors of weights in the network, and updates the values of the strength factors depending on the difference between signals appearing at the output port, for a given pair of input and training patterns, when the weight is perturbed, and when it is not. dated 1997-06-17"
5640966,medical apparatus for analyzing electrical signals from a patient,an analysis apparatus such as an ecg apparatus has a control unit which includes an artificial neural network for discovering signal-recording electrodes which are erroneously attached to a patient. at least one artificial neural network is taught by being fed measurement signals from both correctly recorded measurements and from erroneously recorded measurements. the artificial neural network is then able to identify erroneous attachments with great accuracy from recorded measurement signals.,1997-06-24,The title of the patent is medical apparatus for analyzing electrical signals from a patient and its abstract is an analysis apparatus such as an ecg apparatus has a control unit which includes an artificial neural network for discovering signal-recording electrodes which are erroneously attached to a patient. at least one artificial neural network is taught by being fed measurement signals from both correctly recorded measurements and from erroneously recorded measurements. the artificial neural network is then able to identify erroneous attachments with great accuracy from recorded measurement signals. dated 1997-06-24
5642341,cd rom apparatus for improved tracking and signal sensing,"a compact disc (cd) read sensor apparatus for reading data from a cd track uses multiple laser beams for performing both tracking and reading data. in one embodiment, two laser beams are used to track by controlling the radial position of the two beams that are nominally radially disposed towards opposite edges of the recorded track. a differential signal is used to obtain tracking information and a sum signal is used to produce a read signal. in another embodiment, three laser beams are used corresponding to prior art three beam systems using two beams nominally disposed towards opposite edges for tracking and a center beam for reading. the three beams are combined to form an enhanced read signal with an improved signal-to-noise ratio over that obtained by any one beam. the combining operation includes adjusting for any relative delays between the multiple beam signals detected by the photocells to produce synchronous laser beam signals followed by a summing operation that may optionally include optimal combining operations such as weighted summing, filtering, and neural network processing.",1997-06-24,"The title of the patent is cd rom apparatus for improved tracking and signal sensing and its abstract is a compact disc (cd) read sensor apparatus for reading data from a cd track uses multiple laser beams for performing both tracking and reading data. in one embodiment, two laser beams are used to track by controlling the radial position of the two beams that are nominally radially disposed towards opposite edges of the recorded track. a differential signal is used to obtain tracking information and a sum signal is used to produce a read signal. in another embodiment, three laser beams are used corresponding to prior art three beam systems using two beams nominally disposed towards opposite edges for tracking and a center beam for reading. the three beams are combined to form an enhanced read signal with an improved signal-to-noise ratio over that obtained by any one beam. the combining operation includes adjusting for any relative delays between the multiple beam signals detected by the photocells to produce synchronous laser beam signals followed by a summing operation that may optionally include optimal combining operations such as weighted summing, filtering, and neural network processing. dated 1997-06-24"
5642734,method and apparatus for noninvasively determining hematocrit,"a method and apparatus for noninvasively determining hematocrit utilizing the frequency-dependent electrical impedance characteristics of whole blood by electrically stimulating a patient body portion containing a vascular compartment with a current source over a range of frequencies. a hematocrit measurement system includes a signal generator and demodulator (sgd) that sends an applied signal to an electrode pod that applies a current to a limb of a patient. the electrode pod receives resulting measured voltage signals and provides them to the sgd. the sgd provides to a personal computer (pc) signals indicative of the current passing through the limb of a patient and the resulting voltage. the voltage and current may be measured for various frequencies over, for example, a range from about 10 khz to about 10 mhz. the electrical impedance from the blood alone is isolated from the total limb impedance from the blood, tissue, bone, etc. by determining the difference between measurements at different blood volumes. the hematocrit is determined by the pc based on inphase and quadrature data provided by the sgd. a neural network may be useful in determining the hematocrit from the blood impedance patterns.",1997-07-01,"The title of the patent is method and apparatus for noninvasively determining hematocrit and its abstract is a method and apparatus for noninvasively determining hematocrit utilizing the frequency-dependent electrical impedance characteristics of whole blood by electrically stimulating a patient body portion containing a vascular compartment with a current source over a range of frequencies. a hematocrit measurement system includes a signal generator and demodulator (sgd) that sends an applied signal to an electrode pod that applies a current to a limb of a patient. the electrode pod receives resulting measured voltage signals and provides them to the sgd. the sgd provides to a personal computer (pc) signals indicative of the current passing through the limb of a patient and the resulting voltage. the voltage and current may be measured for various frequencies over, for example, a range from about 10 khz to about 10 mhz. the electrical impedance from the blood alone is isolated from the total limb impedance from the blood, tissue, bone, etc. by determining the difference between measurements at different blood volumes. the hematocrit is determined by the pc based on inphase and quadrature data provided by the sgd. a neural network may be useful in determining the hematocrit from the blood impedance patterns. dated 1997-07-01"
5644647,user-interactive reduction of scene balance failures,"a photofinishing image processing operator detects failures in the performance of a scene balance mechanism on a digitized image and enables a photofinisher to interactively correct for such failures before the failed image has been processed and an unacceptable output image printed. whenever a failed image is identified, one or more reasonably low resolution versions of the image as processed by the scene balance mechanism are displayed to the photofinishing operator, together with a request for image adjustment information that is used to modify or correct the color balance of the image. in response to this user input information, the image processor adjusts color balance parameters of the scene balance mechanism, so that the digitized image processed by the modified scene balance mechanism will yield a print of acceptable color balance quality. this interaction of the photofinishing operator with the image processing system may be accomplished by way of several modes, including displaying differently processed versions of the image, or a single version of the image processed by the scene balance mechanism, together with a question regarding the type of image. the failure detection mechanism preferably employs a non-linear discriminator, such as a neural network.",1997-07-01,"The title of the patent is user-interactive reduction of scene balance failures and its abstract is a photofinishing image processing operator detects failures in the performance of a scene balance mechanism on a digitized image and enables a photofinisher to interactively correct for such failures before the failed image has been processed and an unacceptable output image printed. whenever a failed image is identified, one or more reasonably low resolution versions of the image as processed by the scene balance mechanism are displayed to the photofinishing operator, together with a request for image adjustment information that is used to modify or correct the color balance of the image. in response to this user input information, the image processor adjusts color balance parameters of the scene balance mechanism, so that the digitized image processed by the modified scene balance mechanism will yield a print of acceptable color balance quality. this interaction of the photofinishing operator with the image processing system may be accomplished by way of several modes, including displaying differently processed versions of the image, or a single version of the image processed by the scene balance mechanism, together with a question regarding the type of image. the failure detection mechanism preferably employs a non-linear discriminator, such as a neural network. dated 1997-07-01"
5644681,learning method for neural network having discrete interconnection strengths,"a learning method for a neural network, in which at least a portion of the interconnection strength between neurons takes discrete values, includes the steps of updating an imaginary interconnection strength taking continuous values by using the quantity of update of the interconnection strength which has been calculated by using the discrete interconnection strength, and discretizing the imaginary interconnection strength so as to provide a novel interconnection strength.",1997-07-01,"The title of the patent is learning method for neural network having discrete interconnection strengths and its abstract is a learning method for a neural network, in which at least a portion of the interconnection strength between neurons takes discrete values, includes the steps of updating an imaginary interconnection strength taking continuous values by using the quantity of update of the interconnection strength which has been calculated by using the discrete interconnection strength, and discretizing the imaginary interconnection strength so as to provide a novel interconnection strength. dated 1997-07-01"
5647022,method and apparatus for symbol recognition using multidimensional preprocessing and symbol sorting,"data samples describing a plurality of micro-segments that compose a symbol to be recognized are received from a device such as an electronic pad. a preprocessor maps the micro-segments into cells of an array that has several feature dimensions. the preprocessor assigns values to the cells based on the length of a micro-segment associated with the cell, and how well the features of the associated micro-segment correspond to the feature label of the cell. the cell values are used as inputs to at least one of a plurality of neural networks, where each neural network is trained to identify symbols from a different group of symbols. a sorter examines the symbol to determine which neural network should be used to recognize the symbol. an output produced by the sorter controls a switching means that communicates the cell values to the proper neural network.",1997-07-08,"The title of the patent is method and apparatus for symbol recognition using multidimensional preprocessing and symbol sorting and its abstract is data samples describing a plurality of micro-segments that compose a symbol to be recognized are received from a device such as an electronic pad. a preprocessor maps the micro-segments into cells of an array that has several feature dimensions. the preprocessor assigns values to the cells based on the length of a micro-segment associated with the cell, and how well the features of the associated micro-segment correspond to the feature label of the cell. the cell values are used as inputs to at least one of a plurality of neural networks, where each neural network is trained to identify symbols from a different group of symbols. a sorter examines the symbol to determine which neural network should be used to recognize the symbol. an output produced by the sorter controls a switching means that communicates the cell values to the proper neural network. dated 1997-07-08"
5648627,musical performance control apparatus for processing a user's swing motion with fuzzy inference or a neural network,"a performance control apparatus is provided to control a manner of performance played by an electronic musical apparatus. herein, sensors are provided to sense a swing motion of a baton which is swung by a human operator in response to time of music to be played (e.g., triple time). then, a peak is detected from outputs of the sensors in accordance with a peak detection process using a fuzzy inference process. a kind of the swing motion is discriminated by effecting another fuzzy inference process on a result of the peak detection process. concretely, the kind of the swing motion is discriminated as one of predetermined motions which are determined specifically with respect to time of the music. performance control information is created based on the discriminated kind of the swing motion. thus, a tempo and/or dynamics of performance is controlled in response to the performance control information. incidentally, the fuzzy inference processes can be replaced by a neural network whose structure is determined in advance to calculate probabilities with respect to the swing motion so that the kind of the swing motion is discriminated. moreover, the sensors can be constructed by angular velocity sensors, preferably piezoelectric-vibration gyro sensors, to detect angular velocities of the swing motion of the baton in axial directions.",1997-07-15,"The title of the patent is musical performance control apparatus for processing a user's swing motion with fuzzy inference or a neural network and its abstract is a performance control apparatus is provided to control a manner of performance played by an electronic musical apparatus. herein, sensors are provided to sense a swing motion of a baton which is swung by a human operator in response to time of music to be played (e.g., triple time). then, a peak is detected from outputs of the sensors in accordance with a peak detection process using a fuzzy inference process. a kind of the swing motion is discriminated by effecting another fuzzy inference process on a result of the peak detection process. concretely, the kind of the swing motion is discriminated as one of predetermined motions which are determined specifically with respect to time of the music. performance control information is created based on the discriminated kind of the swing motion. thus, a tempo and/or dynamics of performance is controlled in response to the performance control information. incidentally, the fuzzy inference processes can be replaced by a neural network whose structure is determined in advance to calculate probabilities with respect to the swing motion so that the kind of the swing motion is discriminated. moreover, the sensors can be constructed by angular velocity sensors, preferably piezoelectric-vibration gyro sensors, to detect angular velocities of the swing motion of the baton in axial directions. dated 1997-07-15"
5648938,seismic data acquisition,"a system for acquiring and processing seismic data comprises source-means for generating sound waves and receiver-means for recording as data those waves as reflected from sub-surface interfaces, and means for processing the recorded data operable to generate sets of actual data each individually associated with specific sub-surface reflection points, order the data-sets in accordance with receiver and source-means separation, process each data-set to generate additional data intermediate the recorded data, and re-order each data-set and additional data in accordance with receiver and source-means separation for further processing. each data-set is applied to a filter to generate said additional data intermediate of the recorded actual data. the filter is selected from the group comprising linear, quadratic or spline interpolation filters, frequency space (f-x) filters, tau-p filters, smart filters artificially intelligent filters, neural network filters and (preferably) sinc filters. the recorded actual data are also subject to traveltime correction (for example nmo). data collection may be by use of a single shot source and a plurality of receivers which are equi-spaced, linked in a linear array, and movable above a surface to be surveyed. the data-sets may comprise common receiver or common shot gathers and the described techniques are applicable to other forms of seismic data gathers.",1997-07-15,"The title of the patent is seismic data acquisition and its abstract is a system for acquiring and processing seismic data comprises source-means for generating sound waves and receiver-means for recording as data those waves as reflected from sub-surface interfaces, and means for processing the recorded data operable to generate sets of actual data each individually associated with specific sub-surface reflection points, order the data-sets in accordance with receiver and source-means separation, process each data-set to generate additional data intermediate the recorded data, and re-order each data-set and additional data in accordance with receiver and source-means separation for further processing. each data-set is applied to a filter to generate said additional data intermediate of the recorded actual data. the filter is selected from the group comprising linear, quadratic or spline interpolation filters, frequency space (f-x) filters, tau-p filters, smart filters artificially intelligent filters, neural network filters and (preferably) sinc filters. the recorded actual data are also subject to traveltime correction (for example nmo). data collection may be by use of a single shot source and a plurality of receivers which are equi-spaced, linked in a linear array, and movable above a surface to be surveyed. the data-sets may comprise common receiver or common shot gathers and the described techniques are applicable to other forms of seismic data gathers. dated 1997-07-15"
5649061,device and method for estimating a mental decision,"a device and method for estimating a mental decision to select a visual cue from the viewer's eye fixation and corresponding single event evoked cerebral potential. the device comprises an eyetracker, an electronic biosignal processor and a digital computer. the eyetracker determines the instantaneous viewing direction from oculometric measurements and a head position and orientation sensor. the electronic processor continually estimates the cerebral electroencephalogramic potential from scalp surface measurements following corrections for electrooculogramic, electromyogramic and electrocardiogramic artifacts. the digital computer analyzes the viewing direction data for a fixation and then extracts the corresponding single event evoked cerebral potential. the fixation properties, such as duration, start and end pupil sizes, end state (saccade or blink) and gaze fixation count, and the parametric representation of the evoked potential are all inputs to an artificial neural network for outputting an estimate of the selection interest in the gaze point of regard. the artificial neural network is trained off-line prior to application to represent the mental decisions of the viewer. the device can be used to control computerized machinery from a video display by ocular gaze point of regard alone, by determining which visual cue the viewer is looking at and then using the estimation of the task-related selection as a selector switch.",1997-07-15,"The title of the patent is device and method for estimating a mental decision and its abstract is a device and method for estimating a mental decision to select a visual cue from the viewer's eye fixation and corresponding single event evoked cerebral potential. the device comprises an eyetracker, an electronic biosignal processor and a digital computer. the eyetracker determines the instantaneous viewing direction from oculometric measurements and a head position and orientation sensor. the electronic processor continually estimates the cerebral electroencephalogramic potential from scalp surface measurements following corrections for electrooculogramic, electromyogramic and electrocardiogramic artifacts. the digital computer analyzes the viewing direction data for a fixation and then extracts the corresponding single event evoked cerebral potential. the fixation properties, such as duration, start and end pupil sizes, end state (saccade or blink) and gaze fixation count, and the parametric representation of the evoked potential are all inputs to an artificial neural network for outputting an estimate of the selection interest in the gaze point of regard. the artificial neural network is trained off-line prior to application to represent the mental decisions of the viewer. the device can be used to control computerized machinery from a video display by ocular gaze point of regard alone, by determining which visual cue the viewer is looking at and then using the estimation of the task-related selection as a selector switch. dated 1997-07-15"
5649063,feedback process control using a neural network parameter estimator,"feedback control of a process to reduce process variations is advantageously accomplished by the combination of a signal processor (26) and an artificial neural network (27). the signal processor (26) first determines which of a plurality of process outputs has the greatest deviation from a corresponding desired value for that output. having determined which of the process outputs has the greatest deviation from its corresponding desired value, the process controller (25) then adjusts the output having the greatest deviation to yield an estimated process output vector t.sub.m.sup.n supplied to the artificial neural network (27) trained to represent an inverse model of the process. in response to the estimated process output vector t.sub.m.sup.n, the artificial neural network (27) generates a process control vector c.sub.n that controls the process in accordance with the first order variation between the actual process output and a desired value therefor to reduce process variations.",1997-07-15,"The title of the patent is feedback process control using a neural network parameter estimator and its abstract is feedback control of a process to reduce process variations is advantageously accomplished by the combination of a signal processor (26) and an artificial neural network (27). the signal processor (26) first determines which of a plurality of process outputs has the greatest deviation from a corresponding desired value for that output. having determined which of the process outputs has the greatest deviation from its corresponding desired value, the process controller (25) then adjusts the output having the greatest deviation to yield an estimated process output vector t.sub.m.sup.n supplied to the artificial neural network (27) trained to represent an inverse model of the process. in response to the estimated process output vector t.sub.m.sup.n, the artificial neural network (27) generates a process control vector c.sub.n that controls the process in accordance with the first order variation between the actual process output and a desired value therefor to reduce process variations. dated 1997-07-15"
5649064,system and method for modeling the flow performance features of an object,""" the method and apparatus includes a neural network for generating a model of an object in a wind tunnel from performance data on the object. the network is trained from test input signals (e.g., leading edge flap position, trailing edge flap position, angle of attack, and other geometric configurations, and power settings) and test output signals (e.g., lift, drag, pitching moment, or other performance features). in one embodiment, the neural network training method employs a modified levenberg-marquardt optimization technique. the model can be generated """"real time"""" as wind tunnel testing proceeds. once trained, the model is used to estimate performance features associated with the aircraft given geometric configuration and/or power setting input. the invention can also be applied in other similar static flow modeling applications in aerodynamics, hydrodynamics, fluid dynamics, and other such disciplines. for example, the static testing of cars, sails, and foils, propellers, keels, rudders, turbines, fins, and the like, in a wind tunnel, water trough, or other flowing medium. """,1997-07-15,"The title of the patent is system and method for modeling the flow performance features of an object and its abstract is "" the method and apparatus includes a neural network for generating a model of an object in a wind tunnel from performance data on the object. the network is trained from test input signals (e.g., leading edge flap position, trailing edge flap position, angle of attack, and other geometric configurations, and power settings) and test output signals (e.g., lift, drag, pitching moment, or other performance features). in one embodiment, the neural network training method employs a modified levenberg-marquardt optimization technique. the model can be generated """"real time"""" as wind tunnel testing proceeds. once trained, the model is used to estimate performance features associated with the aircraft given geometric configuration and/or power setting input. the invention can also be applied in other similar static flow modeling applications in aerodynamics, hydrodynamics, fluid dynamics, and other such disciplines. for example, the static testing of cars, sails, and foils, propellers, keels, rudders, turbines, fins, and the like, in a wind tunnel, water trough, or other flowing medium. "" dated 1997-07-15"
5649065,optimal filtering by neural networks with range extenders and/or reducers,"a method and apparatus is provided for processing a measurement process to estimate a signal process, even if the signal and/or measurement processes have large and/or expanding ranges. the method synthesizes training data comprising realizations of the signal and measurement processes into a primary filter for estimating the signal process and, if required, an ancillary filter for providing the primary filter's estimation error statistics. the primary and ancillary filters each comprise an artificial recurrent neural network (rnn) and at least one range extender or reducer. their implementation results in the filtering apparatus. many types of range extender and reducer are disclosed, which have different degrees of effectiveness and computational cost. for a neural filter under design, range extenders and/or reducers are selected from those types jointly with the architecture of the rnn in consideration of the filtering accuracy, the rnn size and the computational cost of each selected range extender and reducer so as to maximize the cost effectiveness of the neural filter. the aforementioned synthesis is performed through training rnns together with range extenders and/or reducers.",1997-07-15,"The title of the patent is optimal filtering by neural networks with range extenders and/or reducers and its abstract is a method and apparatus is provided for processing a measurement process to estimate a signal process, even if the signal and/or measurement processes have large and/or expanding ranges. the method synthesizes training data comprising realizations of the signal and measurement processes into a primary filter for estimating the signal process and, if required, an ancillary filter for providing the primary filter's estimation error statistics. the primary and ancillary filters each comprise an artificial recurrent neural network (rnn) and at least one range extender or reducer. their implementation results in the filtering apparatus. many types of range extender and reducer are disclosed, which have different degrees of effectiveness and computational cost. for a neural filter under design, range extenders and/or reducers are selected from those types jointly with the architecture of the rnn in consideration of the filtering accuracy, the rnn size and the computational cost of each selected range extender and reducer so as to maximize the cost effectiveness of the neural filter. the aforementioned synthesis is performed through training rnns together with range extenders and/or reducers. dated 1997-07-15"
5650722,using resin age factor to obtain measurements of improved accuracy of one or more polymer properties with an on-line nmr system,"a real-time, on-line nuclear magnetic resonance (nmr) system, and related method, can predict various properties of interest of a sample of polymer material. a regression or neural network technique is used to develop a model based upon manipulated nmr output and a resin age factor which compensates for time dependent aging phenomena and enhance predictive accuracy of the model. in a preferred embodiment, the resin age factor is a function of elapsed cycle time prior to sample measurement, sample temperature at time of measurement, and/or sample form. the polymer can be a plastic such as polyethylene, polypropylene, or polystyrene, or a rubber such as ethylene propylene rubber.",1997-07-22,"The title of the patent is using resin age factor to obtain measurements of improved accuracy of one or more polymer properties with an on-line nmr system and its abstract is a real-time, on-line nuclear magnetic resonance (nmr) system, and related method, can predict various properties of interest of a sample of polymer material. a regression or neural network technique is used to develop a model based upon manipulated nmr output and a resin age factor which compensates for time dependent aging phenomena and enhance predictive accuracy of the model. in a preferred embodiment, the resin age factor is a function of elapsed cycle time prior to sample measurement, sample temperature at time of measurement, and/or sample form. the polymer can be a plastic such as polyethylene, polypropylene, or polystyrene, or a rubber such as ethylene propylene rubber. dated 1997-07-22"
5653894,active neural network determination of endpoint in a plasma etch process,"the present invention is predicated upon the fact that a process signature from a plasma process used in fabricating integrated circuits contains information about phenomena which cause variations in the fabrication process such as age of the plasma reactor, densities of the wafers exposed to the plasma, chemistry of the plasma, and concentration of the remaining material. in accordance with the present invention, a method for using neural networks to determine plasma etch end-point times in an integrated circuit fabrication process is disclosed. the end-point time is based on in-situ monitoring of at least two parameters during the plasma etch process. after the neural network is trained to associate a certain condition or set of conditions with the endpoint of the process, the neural network is used to control the process.",1997-08-05,"The title of the patent is active neural network determination of endpoint in a plasma etch process and its abstract is the present invention is predicated upon the fact that a process signature from a plasma process used in fabricating integrated circuits contains information about phenomena which cause variations in the fabrication process such as age of the plasma reactor, densities of the wafers exposed to the plasma, chemistry of the plasma, and concentration of the remaining material. in accordance with the present invention, a method for using neural networks to determine plasma etch end-point times in an integrated circuit fabrication process is disclosed. the end-point time is based on in-situ monitoring of at least two parameters during the plasma etch process. after the neural network is trained to associate a certain condition or set of conditions with the endpoint of the process, the neural network is used to control the process. dated 1997-08-05"
5655031,method for determining attributes using neural network and fuzzy logic,""" an attribute decision method includes steps of making a comparison between one feature value, selected from seven feature values, of an input pattern contained in an input image and membership functions determined from 37 standard patterns including alphanumeric characters of """"0"""" to """"9"""" and """"a"""" to """"z"""" and a hyphen of """"-"""" with respect to the selected feature value and then obtaining output values of the membership functions for the individual standard patterns with respect to the input pattern. the seven feature values are, when the input pattern is a mesh pattern, vertical structure vector sums, horizontal structure vector sums, up-down and left-right area differences, and a vertical cross number. the method also includes steps of calculating a total value of the output value for the feature value for each of the standard patterns, and determining a standard pattern of the highest total value among the calculated total values and thus deciding one of the standard patterns which has the highest similarity to the input pattern. """,1997-08-05,"The title of the patent is method for determining attributes using neural network and fuzzy logic and its abstract is "" an attribute decision method includes steps of making a comparison between one feature value, selected from seven feature values, of an input pattern contained in an input image and membership functions determined from 37 standard patterns including alphanumeric characters of """"0"""" to """"9"""" and """"a"""" to """"z"""" and a hyphen of """"-"""" with respect to the selected feature value and then obtaining output values of the membership functions for the individual standard patterns with respect to the input pattern. the seven feature values are, when the input pattern is a mesh pattern, vertical structure vector sums, horizontal structure vector sums, up-down and left-right area differences, and a vertical cross number. the method also includes steps of calculating a total value of the output value for the feature value for each of the standard patterns, and determining a standard pattern of the highest total value among the calculated total values and thus deciding one of the standard patterns which has the highest similarity to the input pattern. "" dated 1997-08-05"
5657737,air-fuel ratio control system,"an air-fuel ratio control system which calculates a basic injection amount of fuel by state detecting sensors (21) each of which detects an operating state of an internal combustion engine, air-amount detecting sensors (22) each of which detects an intake air amount, an air-fuel ratio sensor (23), and a predetermined data group. further, the air-fuel ratio control system stores past amount data of injected fuel and air-fuel ratio data at each control cycle. a neuro-computation unit (29) reads values of the data detected by the sensors (21,22) and the stored data to obtain an air-fuel ratio estimate to calculate a correction injection amount of fuel by a neural network of a forward neuro-computing unit (210), which has learned beforehand about relations of the injected fuel amount, the detected air-fuel ratio, parameters, and dead time.",1997-08-19,"The title of the patent is air-fuel ratio control system and its abstract is an air-fuel ratio control system which calculates a basic injection amount of fuel by state detecting sensors (21) each of which detects an operating state of an internal combustion engine, air-amount detecting sensors (22) each of which detects an intake air amount, an air-fuel ratio sensor (23), and a predetermined data group. further, the air-fuel ratio control system stores past amount data of injected fuel and air-fuel ratio data at each control cycle. a neuro-computation unit (29) reads values of the data detected by the sensors (21,22) and the stored data to obtain an air-fuel ratio estimate to calculate a correction injection amount of fuel by a neural network of a forward neuro-computing unit (210), which has learned beforehand about relations of the injected fuel amount, the detected air-fuel ratio, parameters, and dead time. dated 1997-08-19"
5659666,device for the autonomous generation of useful information,"a device for simulating human creativity employing a neural network trained to produce input-output maps within some predetermined knowledge domain, an apparatus for subjecting the neural network to perturbations that produce changes in the predetermined knowledge domain, the neural network having an optional output for feeding the outputs of the neural network to a second neural network that evaluates and selects outputs based on training within the second neural network. the device may also include a reciprocal feed back connection from the output of the second neural network to the first neural network to further influence and change what takes place in the aforesaid neural network.",1997-08-19,"The title of the patent is device for the autonomous generation of useful information and its abstract is a device for simulating human creativity employing a neural network trained to produce input-output maps within some predetermined knowledge domain, an apparatus for subjecting the neural network to perturbations that produce changes in the predetermined knowledge domain, the neural network having an optional output for feeding the outputs of the neural network to a second neural network that evaluates and selects outputs based on training within the second neural network. the device may also include a reciprocal feed back connection from the output of the second neural network to the first neural network to further influence and change what takes place in the aforesaid neural network. dated 1997-08-19"
5659667,adaptive model predictive process control using neural networks,"a control system for controlling the output of at least one plant process output parameter is implemented by adaptive model predictive control using a neural network. an improved method and apparatus provides for sampling plant output and control input at a first sampling rate to provide control inputs at the fast rate. the mpc system is, however, provided with a network state vector that is constructed at a second, slower rate so that the input control values used by the mpc system are averaged over a gapped time period. another improvement is a provision for on-line training that may include difference training, curvature training, and basis center adjustment to maintain the weights and basis centers of the neural in an updated state that can follow changes in the plant operation apart from initial off-line training data.",1997-08-19,"The title of the patent is adaptive model predictive process control using neural networks and its abstract is a control system for controlling the output of at least one plant process output parameter is implemented by adaptive model predictive control using a neural network. an improved method and apparatus provides for sampling plant output and control input at a first sampling rate to provide control inputs at the fast rate. the mpc system is, however, provided with a network state vector that is constructed at a second, slower rate so that the input control values used by the mpc system are averaged over a gapped time period. another improvement is a provision for on-line training that may include difference training, curvature training, and basis center adjustment to maintain the weights and basis centers of the neural in an updated state that can follow changes in the plant operation apart from initial off-line training data. dated 1997-08-19"
5660181,hybrid neural network and multiple fiber probe for in-depth 3-d mapping,"an apparatus for in-depth three dimensional tumor mapping including (a) a light source; (b) a multi-fiber bundle including at least one illumination fiber and at least two receiving fibers, the at least one illumination fiber being connected to the light source; (c) a spectrometer connected to the at least two receiving fibers; and (d) a hybrid neural network connected to the spectrometer, said hybrid neural network including a principle component analysis processor and a neural network classifier.",1997-08-26,"The title of the patent is hybrid neural network and multiple fiber probe for in-depth 3-d mapping and its abstract is an apparatus for in-depth three dimensional tumor mapping including (a) a light source; (b) a multi-fiber bundle including at least one illumination fiber and at least two receiving fibers, the at least one illumination fiber being connected to the light source; (c) a spectrometer connected to the at least two receiving fibers; and (d) a hybrid neural network connected to the spectrometer, said hybrid neural network including a principle component analysis processor and a neural network classifier. dated 1997-08-26"
5664065,pulse-coupled automatic object recognition system dedicatory clause,the pulse-coupled automatic recognition system processes information utilizing parallel neural network and transmits information to and from the processing sites as time signals which encode in their pulse phase structure the geometrical content of the spatial distributions of light. a locating device then measures the coordinates of the spatial distributions (pulsating segment) and displays the coordinates of the pulsating segment on a screen for observation. the pulsating segment indicates the presence and location of the component of the input scene that corresponds to the segment.,1997-09-02,The title of the patent is pulse-coupled automatic object recognition system dedicatory clause and its abstract is the pulse-coupled automatic recognition system processes information utilizing parallel neural network and transmits information to and from the processing sites as time signals which encode in their pulse phase structure the geometrical content of the spatial distributions of light. a locating device then measures the coordinates of the spatial distributions (pulsating segment) and displays the coordinates of the pulsating segment on a screen for observation. the pulsating segment indicates the presence and location of the component of the input scene that corresponds to the segment. dated 1997-09-02
5664066,intelligent system for automatic feature detection and selection or identification,"a neural network uses a fuzzy membership function, the parameters of which are adaptive during the training process, to parameterize the interconnection weights between an (n-1)'th layer and an n'th layer of the network. each j'th node in each k'th layer of the network except the input layer produces its output value y.sub.k,j according to the function ##equ1## where n.sub.k-1 is the number of nodes in layer k-1, i indexes the nodes of layer k-1 and all the w.sub.k,i,j are interconnection weights. the interconnection weights to all nodes j in the n'th layer are given by w.sub.n,i,j =w.sub.n,j (i, p.sub.n,j,1, . . . , p.sub.n,j,p.sbsb.n). the apparatus is trained by setting values for at least one of the parameters p.sub.n,j,1, . . . , p.sub.n,j,pn. preferably the number of parameters p.sub.n is less than the number of nodes n.sub.n-1 in layer n-1. w.sub.n,j (i,p.sub.n,j,1, . . . , p.sub.n,j,pn) can be convex in i, and it can be bell-shaped. sample functions for w.sub.n,j (i, p.sub.n,j,1, . . . , p.sub.n,j,pn) include ##equ2##",1997-09-02,"The title of the patent is intelligent system for automatic feature detection and selection or identification and its abstract is a neural network uses a fuzzy membership function, the parameters of which are adaptive during the training process, to parameterize the interconnection weights between an (n-1)'th layer and an n'th layer of the network. each j'th node in each k'th layer of the network except the input layer produces its output value y.sub.k,j according to the function ##equ1## where n.sub.k-1 is the number of nodes in layer k-1, i indexes the nodes of layer k-1 and all the w.sub.k,i,j are interconnection weights. the interconnection weights to all nodes j in the n'th layer are given by w.sub.n,i,j =w.sub.n,j (i, p.sub.n,j,1, . . . , p.sub.n,j,p.sbsb.n). the apparatus is trained by setting values for at least one of the parameters p.sub.n,j,1, . . . , p.sub.n,j,pn. preferably the number of parameters p.sub.n is less than the number of nodes n.sub.n-1 in layer n-1. w.sub.n,j (i,p.sub.n,j,1, . . . , p.sub.n,j,pn) can be convex in i, and it can be bell-shaped. sample functions for w.sub.n,j (i, p.sub.n,j,1, . . . , p.sub.n,j,pn) include ##equ2## dated 1997-09-02"
5664067,method and apparatus for training a neural network,"a plurality of training inputs are selected, wherein each training input corresponds to a first possible output. the quality of each of the plurality of training inputs is characterized. a first training input is selected from the training inputs, where the first training input is of higher quality than a second training input of the training inputs. the neural network is trained with the higher-quality first training input prior to training with the second training input. a neuron may be added to the neural network in accordance with the first training input, wherein the neuron is associated with the first possible output.",1997-09-02,"The title of the patent is method and apparatus for training a neural network and its abstract is a plurality of training inputs are selected, wherein each training input corresponds to a first possible output. the quality of each of the plurality of training inputs is characterized. a first training input is selected from the training inputs, where the first training input is of higher quality than a second training input of the training inputs. the neural network is trained with the higher-quality first training input prior to training with the second training input. a neuron may be added to the neural network in accordance with the first training input, wherein the neuron is associated with the first possible output. dated 1997-09-02"
5664068,method and apparatus for pattern classification using distributed adaptive fuzzy windows,"a method for pattern classification and, in particular, a method which distributes the classification criteria across a neural network. the classification criteria for a pattern class is stored distributively in the neural network in two aspects. first, it manifests itself as one or more levels of templates, each of which represents a fuzzily unique perspective of the pattern class. second, the template at each level is represented by a set of fuzzy windows, each of which defines a classification criterion of a corresponding feature of the pattern class.",1997-09-02,"The title of the patent is method and apparatus for pattern classification using distributed adaptive fuzzy windows and its abstract is a method for pattern classification and, in particular, a method which distributes the classification criteria across a neural network. the classification criteria for a pattern class is stored distributively in the neural network in two aspects. first, it manifests itself as one or more levels of templates, each of which represents a fuzzily unique perspective of the pattern class. second, the template at each level is represented by a set of fuzzy windows, each of which defines a classification criterion of a corresponding feature of the pattern class. dated 1997-09-02"
5666467,neural network using inhomogeneities in a medium as neurons and transmitting input signals as an unchannelled wave pattern through the medium,an information processing system comprises a neural net with fully distributed neuron and synapse functionalities in a spatially inhomogeneous medium to propagate a response field from an input to an output. the response field is a reaction of the medium to a plurality of input signals and depends non-linearly on the input signals. the response field is also determined by the inhomogeneities. the value of the field at one or more particular locations is indicative of one or more output signals of the neural net.,1997-09-09,The title of the patent is neural network using inhomogeneities in a medium as neurons and transmitting input signals as an unchannelled wave pattern through the medium and its abstract is an information processing system comprises a neural net with fully distributed neuron and synapse functionalities in a spatially inhomogeneous medium to propagate a response field from an input to an output. the response field is a reaction of the medium to a plurality of input signals and depends non-linearly on the input signals. the response field is also determined by the inhomogeneities. the value of the field at one or more particular locations is indicative of one or more output signals of the neural net. dated 1997-09-09
5666468,neural network binary code recognizer,"a neural network binary code recognizer for decoding n-bit binary code words. this apparatus includes inputs for inputting n signals into the recognizer, each of the n signals representing a bit value of an n-bit binary code word, which may or may not be corrupted. the apparatus also includes n amplifiers, each having an input for receiving a respective input signal. the amplifiers condition the respective input signals to generate respective output signals at one or more bit values of one or more corresponding predetermined valid n-bit binary code words. the apparatus further includes a device for approximating an image product for each of the one or more predetermined valid n-bit binary code words, each approximated image product comprising the product of an output signal from each amplifier in accordance with a respective predetermined valid n-bit binary coded word. each approximated image product is fed back to an input of each amplifier in accordance with a bit value of one or more predetermined valid n-bit binary code words to enable an output signal of a respective amplifier to dynamically approach each bit value of the input n-bit binary code word.",1997-09-09,"The title of the patent is neural network binary code recognizer and its abstract is a neural network binary code recognizer for decoding n-bit binary code words. this apparatus includes inputs for inputting n signals into the recognizer, each of the n signals representing a bit value of an n-bit binary code word, which may or may not be corrupted. the apparatus also includes n amplifiers, each having an input for receiving a respective input signal. the amplifiers condition the respective input signals to generate respective output signals at one or more bit values of one or more corresponding predetermined valid n-bit binary code words. the apparatus further includes a device for approximating an image product for each of the one or more predetermined valid n-bit binary code words, each approximated image product comprising the product of an output signal from each amplifier in accordance with a respective predetermined valid n-bit binary coded word. each approximated image product is fed back to an input of each amplifier in accordance with a bit value of one or more predetermined valid n-bit binary code words to enable an output signal of a respective amplifier to dynamically approach each bit value of the input n-bit binary code word. dated 1997-09-09"
5666481,method and apparatus for resolving faults in communications networks,"an improved method and apparatus of resolving faults in a communications network. the preferred system uses a trouble ticket data structure to describe communications network faults. completed trouble tickets are stored in a library and when an outstanding trouble ticket is received, the system uses at least one determinator to correlate the outstanding communications network fault to data fields in the set of data fields of the trouble ticket data structure to determine which completed trouble tickets in the library are relevant to the outstanding communications network fault. the system retrieves a set of completed trouble tickets from the library that are similar to the outstanding trouble ticket and uses at least a portion of the resolution from at least one completed trouble ticket to provide a resolution of the outstanding trouble ticket. the determinators may be macros, rules, a decision tree derived from an information theoretic induction algorithm and/or a neural network memory derived from a neural network learning algorithm. the system may adapt the resolution from a retrieved trouble ticket to provide the resolution using null adaptation, parameterized adaptation, abstraction/respecialization adaptation, or critic-based adaptation techniques.",1997-09-09,"The title of the patent is method and apparatus for resolving faults in communications networks and its abstract is an improved method and apparatus of resolving faults in a communications network. the preferred system uses a trouble ticket data structure to describe communications network faults. completed trouble tickets are stored in a library and when an outstanding trouble ticket is received, the system uses at least one determinator to correlate the outstanding communications network fault to data fields in the set of data fields of the trouble ticket data structure to determine which completed trouble tickets in the library are relevant to the outstanding communications network fault. the system retrieves a set of completed trouble tickets from the library that are similar to the outstanding trouble ticket and uses at least a portion of the resolution from at least one completed trouble ticket to provide a resolution of the outstanding trouble ticket. the determinators may be macros, rules, a decision tree derived from an information theoretic induction algorithm and/or a neural network memory derived from a neural network learning algorithm. the system may adapt the resolution from a retrieved trouble ticket to provide the resolution using null adaptation, parameterized adaptation, abstraction/respecialization adaptation, or critic-based adaptation techniques. dated 1997-09-09"
5666518,pattern recognition by simulated neural-like networks,apparatus and method for replacing the traditional amplifications by rime delays in a neural network that can be trained to analyze temporally-related patterns. time delays comprise the synapses between feeder and stimulus cells in the network. the result is a multi-temporal trainable delay neural network.,1997-09-09,The title of the patent is pattern recognition by simulated neural-like networks and its abstract is apparatus and method for replacing the traditional amplifications by rime delays in a neural network that can be trained to analyze temporally-related patterns. time delays comprise the synapses between feeder and stimulus cells in the network. the result is a multi-temporal trainable delay neural network. dated 1997-09-09
5668717,method and apparatus for model-free optimal signal timing for system-wide traffic control,"a method and apparatus for model-free, real-time, system-wide signal timing for a complex road network is provided. it provides timings in response to instantaneous flow conditions while accounting for the inherent stochastic variations in traffic flow through the use of a simultaneous perturbation stochastic approximation (spsa) algorithm. this is achieved by setting up several (m) parallel neural networks, each of which produces optimal controls (signal timings) for any time instant (within one of the m time periods) based on observed traffic conditions. the spsa optimization technique is critical to the feasibility of the approach since it provides the values of weight parameters in each of the neural networks without the need for a model of the traffic flow dynamics.",1997-09-16,"The title of the patent is method and apparatus for model-free optimal signal timing for system-wide traffic control and its abstract is a method and apparatus for model-free, real-time, system-wide signal timing for a complex road network is provided. it provides timings in response to instantaneous flow conditions while accounting for the inherent stochastic variations in traffic flow through the use of a simultaneous perturbation stochastic approximation (spsa) algorithm. this is achieved by setting up several (m) parallel neural networks, each of which produces optimal controls (signal timings) for any time instant (within one of the m time periods) based on observed traffic conditions. the spsa optimization technique is critical to the feasibility of the approach since it provides the values of weight parameters in each of the neural networks without the need for a model of the traffic flow dynamics. dated 1997-09-16"
5668926,method and apparatus for converting text into audible signals using a neural network,"text may be converted to audible signals, such as speech, by first training a neural network 106 using recorded audio messages 204. to begin the training, the recorded audio messages are converted into a series of audio frames 205 having a fixed duration 213. then, each audio frame is assigned a phonetic representation 203 and a target acoustic representation 208, where the phonetic representation 203 is a binary word that represents the phone and articulation characteristics of the audio frame, while the target acoustic representation 208 is a vector of audio information such as pitch and energy. after training, the neural network 106 is used in conversion of text into speech. first, text that is to be convened is translated to a series of phonetic frames 401 of the same form as the phonetic representations 208 and having the fixed duration 213. then the neural network produces acoustic representations in response to context descriptions 207 that include some of the phonetic frames 401. the acoustic representations are then converted into a speech wave form by a synthesizer 107.",1997-09-16,"The title of the patent is method and apparatus for converting text into audible signals using a neural network and its abstract is text may be converted to audible signals, such as speech, by first training a neural network 106 using recorded audio messages 204. to begin the training, the recorded audio messages are converted into a series of audio frames 205 having a fixed duration 213. then, each audio frame is assigned a phonetic representation 203 and a target acoustic representation 208, where the phonetic representation 203 is a binary word that represents the phone and articulation characteristics of the audio frame, while the target acoustic representation 208 is a vector of audio information such as pitch and energy. after training, the neural network 106 is used in conversion of text into speech. first, text that is to be convened is translated to a series of phonetic frames 401 of the same form as the phonetic representations 208 and having the fixed duration 213. then the neural network produces acoustic representations in response to context descriptions 207 that include some of the phonetic frames 401. the acoustic representations are then converted into a speech wave form by a synthesizer 107. dated 1997-09-16"
5671334,neural network model for reaching a goal state,"an object, such as a robot, is located at an initial state in a finite state space area and moves under the control of the unsupervised neural network model of the invention. the network instructs the object to move in one of several directions from the initial state. upon reaching another state, the model again instructs the object to move in one of several directions. these instructions continue until either: a) the object has completed a cycle by ending up back at a state it has been to previously during this cycle, or b) the object has completed a cycle by reaching the goal state. upon reaching a state, the neural network model calculates a level of satisfaction with its progress towards reaching the goal state. if the level of satisfaction is low, the neural network model is more likely to override what has been learned thus far and deviate from a path known to lead to the goal state to experiment with new and possibly better paths. if the level of satisfaction is high, the neural network model is much less likely to experiment with new paths. the object is guaranteed to eventually find the best path to the goal state from any starting location, assuming that the level of satisfaction does not exceed a threshold point where learning ceases.",1997-09-23,"The title of the patent is neural network model for reaching a goal state and its abstract is an object, such as a robot, is located at an initial state in a finite state space area and moves under the control of the unsupervised neural network model of the invention. the network instructs the object to move in one of several directions from the initial state. upon reaching another state, the model again instructs the object to move in one of several directions. these instructions continue until either: a) the object has completed a cycle by ending up back at a state it has been to previously during this cycle, or b) the object has completed a cycle by reaching the goal state. upon reaching a state, the neural network model calculates a level of satisfaction with its progress towards reaching the goal state. if the level of satisfaction is low, the neural network model is more likely to override what has been learned thus far and deviate from a path known to lead to the goal state to experiment with new and possibly better paths. if the level of satisfaction is high, the neural network model is much less likely to experiment with new paths. the object is guaranteed to eventually find the best path to the goal state from any starting location, assuming that the level of satisfaction does not exceed a threshold point where learning ceases. dated 1997-09-23"
5671335,process optimization using a neural network,"an input to a complex multi-input process, such as injection molding, is optimized to produce a target output from that process through the use of a neural network trained to that process. a trial input is forward-propagated through the neural network and the output of the network compared to the target output. the difference is back-propagated through the network to determine an input error value in the network. this error value is used to correct the trial input. this correction process is repeated until the trial input produces the target output to within a predetermined degree of accuracy.",1997-09-23,"The title of the patent is process optimization using a neural network and its abstract is an input to a complex multi-input process, such as injection molding, is optimized to produce a target output from that process through the use of a neural network trained to that process. a trial input is forward-propagated through the neural network and the output of the network compared to the target output. the difference is back-propagated through the network to determine an input error value in the network. this error value is used to correct the trial input. this correction process is repeated until the trial input produces the target output to within a predetermined degree of accuracy. dated 1997-09-23"
5671336,apparatus based on n-variable unlimited recurrent adjustable network,"an apparatus based on an n-variable unlimited recurrent adjustable network (uran.sub.n) comprises: two layers, each layer having the same number (n) of neuron elements; linear neuron elements x.sub.i constituting a first layer; nonlinear artificial neuron elements y.sub.j having respective temperature-dependent parameters tj and constituting a second layer. each linear and nonlinear neuron element of the first and second layers is connected using a feedforward connection part, a recurrent connection part, and an auto connection part. a nonlinear oscillation apparatus having the recurrent neural network is generally operated in accordance with the equation (1) described below: ##equ1##",1997-09-23,"The title of the patent is apparatus based on n-variable unlimited recurrent adjustable network and its abstract is an apparatus based on an n-variable unlimited recurrent adjustable network (uran.sub.n) comprises: two layers, each layer having the same number (n) of neuron elements; linear neuron elements x.sub.i constituting a first layer; nonlinear artificial neuron elements y.sub.j having respective temperature-dependent parameters tj and constituting a second layer. each linear and nonlinear neuron element of the first and second layers is connected using a feedforward connection part, a recurrent connection part, and an auto connection part. a nonlinear oscillation apparatus having the recurrent neural network is generally operated in accordance with the equation (1) described below: ##equ1## dated 1997-09-23"
5672853,elevator control neural network,a remaining response time for an elevator car under consideration for assignment to a newly registered hall call is estimated by using a neural network. the neural network or any other downstream module may be standardized for use in any building by use of an upstream fixed length stop description that summarizes the state of the building at the time of the registration of the new hall call for one or more postulated paths of each and every car under consideration for answering the new hall call.,1997-09-30,The title of the patent is elevator control neural network and its abstract is a remaining response time for an elevator car under consideration for assignment to a newly registered hall call is estimated by using a neural network. the neural network or any other downstream module may be standardized for use in any building by use of an upstream fixed length stop description that summarizes the state of the building at the time of the registration of the new hall call for one or more postulated paths of each and every car under consideration for answering the new hall call. dated 1997-09-30
5673367,method for neural network control of motion using real-time environmental feedback,"a method of motion control for robotics and other automatically controlled machinery using a neural network controller with real-time environmental feedback. the method is illustrated with a two-finger robotic hand having proximity sensors and force sensors that provide environmental feedback signals. the neural network controller is taught to control the robotic hand through training sets using back- propagation methods. the training sets are created by recording the control signals and the feedback signal as the robotic hand or a simulation of the robotic hand is moved through a representative grasping motion. the data recorded is divided into discrete increments of time and the feedback data is shifted out of phase with the control signal data so that the feedback signal data lag one time increment behind the control signal data. the modified data is presented to the neural network controller as a training set. the time lag introduced into the data allows the neural network controller to account for the temporal component of the robotic motion. thus trained, the neural network controlled robotic hand is able to grasp a wide variety of different objects by generalizing from the training sets.",1997-09-30,"The title of the patent is method for neural network control of motion using real-time environmental feedback and its abstract is a method of motion control for robotics and other automatically controlled machinery using a neural network controller with real-time environmental feedback. the method is illustrated with a two-finger robotic hand having proximity sensors and force sensors that provide environmental feedback signals. the neural network controller is taught to control the robotic hand through training sets using back- propagation methods. the training sets are created by recording the control signals and the feedback signal as the robotic hand or a simulation of the robotic hand is moved through a representative grasping motion. the data recorded is divided into discrete increments of time and the feedback data is shifted out of phase with the control signal data so that the feedback signal data lag one time increment behind the control signal data. the modified data is presented to the neural network controller as a training set. the time lag introduced into the data allows the neural network controller to account for the temporal component of the robotic motion. thus trained, the neural network controlled robotic hand is able to grasp a wide variety of different objects by generalizing from the training sets. dated 1997-09-30"
5673368,method and device for conducting a process in a controlled system with at least one precomputed process parameter determined using a mathematical model having variable model parameters adjusted based on a network response of a neural network,"in known processes for conducting a process in an automatically controlled system, the system is preset at the beginning of each process run based on at least one process parameter. the at least one process parameter is precomputed with a model of the process, containing at least one model parameter and input values supplied to the model. during the process, the input values and the process parameter are measured and used to adaptively improve the precomputed process parameter after the process run. a neural network is used to determine the model parameters whose dependence on the input values is unknown or insufficiently known. network parameters of the neural network are modified after each process run to adapt the model to the actual process events.",1997-09-30,"The title of the patent is method and device for conducting a process in a controlled system with at least one precomputed process parameter determined using a mathematical model having variable model parameters adjusted based on a network response of a neural network and its abstract is in known processes for conducting a process in an automatically controlled system, the system is preset at the beginning of each process run based on at least one process parameter. the at least one process parameter is precomputed with a model of the process, containing at least one model parameter and input values supplied to the model. during the process, the input values and the process parameter are measured and used to adaptively improve the precomputed process parameter after the process run. a neural network is used to determine the model parameters whose dependence on the input values is unknown or insufficiently known. network parameters of the neural network are modified after each process run to adapt the model to the actual process events. dated 1997-09-30"
5675253,partial least square regression techniques in obtaining measurements of one or more polymer properties with an on-line nmr system,"an on-line nuclear magnetic resonance (nmr) system, and related methods, are useful for predicting one or more properties of interest of a polymer. in one embodiment, a neural network is used to develop a model which correlates process variables in addition to manipulated nmr output to predict a polymer property of interest. in another embodiment, a partial least square regression technique is used to develop a model of enhanced accuracy. either the neural network technique or the partial least square regression technique may be used in conjunction with a described multi-model or best-model-selection scheme according to the invention. the polymer can be a plastic such as polyethylene, polypropylene, or polystyrene, or a rubber such as ethylene propylene rubber.",1997-10-07,"The title of the patent is partial least square regression techniques in obtaining measurements of one or more polymer properties with an on-line nmr system and its abstract is an on-line nuclear magnetic resonance (nmr) system, and related methods, are useful for predicting one or more properties of interest of a polymer. in one embodiment, a neural network is used to develop a model which correlates process variables in addition to manipulated nmr output to predict a polymer property of interest. in another embodiment, a partial least square regression technique is used to develop a model of enhanced accuracy. either the neural network technique or the partial least square regression technique may be used in conjunction with a described multi-model or best-model-selection scheme according to the invention. the polymer can be a plastic such as polyethylene, polypropylene, or polystyrene, or a rubber such as ethylene propylene rubber. dated 1997-10-07"
5675497,method for monitoring an electric motor and detecting a departure from normal operation,a method for detecting a departure from normal operation of an electric motor comprises the steps of modelling a set of normal current measurements and a set of operational current measurements for the motor being monitored. the modelling is carried out by a neural network auto-associator which is trained to reproduce its inputs on its output. a potential failure is indicated whenever the set of normal current measurements and the set of operational current measurements differ by more than a predetermined criterion.,1997-10-07,The title of the patent is method for monitoring an electric motor and detecting a departure from normal operation and its abstract is a method for detecting a departure from normal operation of an electric motor comprises the steps of modelling a set of normal current measurements and a set of operational current measurements for the motor being monitored. the modelling is carried out by a neural network auto-associator which is trained to reproduce its inputs on its output. a potential failure is indicated whenever the set of normal current measurements and the set of operational current measurements differ by more than a predetermined criterion. dated 1997-10-07
5675504,method of predicting residual chlorine in water supply systems,"a method of configuring an artificial neural network for predicting residual chlorine concentration in water contained in a storage tank of a water supply system, the storage tank having an inlet for admitting water into the tank and an outlet for discharging water from the tank. the method of the invention comprises the steps of (a) collecting historical data representative of selected operational and water quality parameters associated with chlorine demand in the tank; (b) scaling the data collected in step (a); (c) organizing the data scaled in step (b) in the form of a set of time-lagged data; and (d) processing the data organized in step (c) through an artificial neural network by scanning a window over the set of data, the window having a size corresponding to a sub-set of the data, to associate a predicted value with a respective one of the sub-sets scanned by the window, the predicted value being representative of the residual chlorine concentration at the tank outlet. the artificial neural network is thus configured to recognize a set of data processed therethrough as corresponding substantially to one of the sub-sets of data and to associate with the recognized set of data the optimized predicted value associated with the sub-set of data.",1997-10-07,"The title of the patent is method of predicting residual chlorine in water supply systems and its abstract is a method of configuring an artificial neural network for predicting residual chlorine concentration in water contained in a storage tank of a water supply system, the storage tank having an inlet for admitting water into the tank and an outlet for discharging water from the tank. the method of the invention comprises the steps of (a) collecting historical data representative of selected operational and water quality parameters associated with chlorine demand in the tank; (b) scaling the data collected in step (a); (c) organizing the data scaled in step (b) in the form of a set of time-lagged data; and (d) processing the data organized in step (c) through an artificial neural network by scanning a window over the set of data, the window having a size corresponding to a sub-set of the data, to associate a predicted value with a respective one of the sub-sets scanned by the window, the predicted value being representative of the residual chlorine concentration at the tank outlet. the artificial neural network is thus configured to recognize a set of data processed therethrough as corresponding substantially to one of the sub-sets of data and to associate with the recognized set of data the optimized predicted value associated with the sub-set of data. dated 1997-10-07"
5675712,method and apparatus for using a neural network to extract an optimal number of data objects from an available class of data objects,"method and apparatus for selecting an optimal number of trajectories from an available class of potential trajectories in a velocimetry application. in an exemplary method, a neural network is constructed wherein each neuron in the network represents a trajectory in the overall class. a binary output of each neuron indicates whether the trajectory represented by the neuron is to be selected. in the exemplary method, the neural network is initialized with a starting solution wherein the network is in an initially converged state. the network is then alternately excited and constrained so that it settles to additional converged states. during excitation, correction factors including a set-size maximizing term are applied to neuron input potentials. during constraint, the set-size maximizing terms are interrupted. each time the network converges, the outputs of the neurons in the network are decoded to obtain a subset of trajectories which are to be selected. the excitation and constraint phases are then repeated in an attempt to obtain larger and larger subsets of trajectories to be selected. once the excitation and constraint phases have been repeated a suitable number of times, the trajectories identified in the largest obtained subset of trajectories are selected.",1997-10-07,"The title of the patent is method and apparatus for using a neural network to extract an optimal number of data objects from an available class of data objects and its abstract is method and apparatus for selecting an optimal number of trajectories from an available class of potential trajectories in a velocimetry application. in an exemplary method, a neural network is constructed wherein each neuron in the network represents a trajectory in the overall class. a binary output of each neuron indicates whether the trajectory represented by the neuron is to be selected. in the exemplary method, the neural network is initialized with a starting solution wherein the network is in an initially converged state. the network is then alternately excited and constrained so that it settles to additional converged states. during excitation, correction factors including a set-size maximizing term are applied to neuron input potentials. during constraint, the set-size maximizing terms are interrupted. each time the network converges, the outputs of the neurons in the network are decoded to obtain a subset of trajectories which are to be selected. the excitation and constraint phases are then repeated in an attempt to obtain larger and larger subsets of trajectories to be selected. once the excitation and constraint phases have been repeated a suitable number of times, the trajectories identified in the largest obtained subset of trajectories are selected. dated 1997-10-07"
5675713,sensor for use in a neural network,"an artificial neuron for use in a neural processing network comprises a plurality of input signal lines, an arrangement for computing a nonlinear function of the sum of the inputs multiplied by associated weights, and a saturating delta-sigma modulator which oversamples the computed value and produces an encoded neuron output signal. conversion of signals for use by these neurons preferably is performed by delta-sigma modulators at the inputs to the neurons, which may be incorporated directly into sensors. processing of the output signals from the neuron includes low-pass filtering and decimation. the present invention may be used in many diverse areas. for example, arrays of sensors with delta signal modulators may be coupled with a network of the neurons to form an intelligent vision system. linear signal processing, both conventional and adaptive, can be done by a simple neuronal system that operates linearly.",1997-10-07,"The title of the patent is sensor for use in a neural network and its abstract is an artificial neuron for use in a neural processing network comprises a plurality of input signal lines, an arrangement for computing a nonlinear function of the sum of the inputs multiplied by associated weights, and a saturating delta-sigma modulator which oversamples the computed value and produces an encoded neuron output signal. conversion of signals for use by these neurons preferably is performed by delta-sigma modulators at the inputs to the neurons, which may be incorporated directly into sensors. processing of the output signals from the neuron includes low-pass filtering and decimation. the present invention may be used in many diverse areas. for example, arrays of sensors with delta signal modulators may be coupled with a network of the neurons to form an intelligent vision system. linear signal processing, both conventional and adaptive, can be done by a simple neuronal system that operates linearly. dated 1997-10-07"
5677609,intelligent servomechanism controller,""" a servomechanism controller for controlling the movement and positioning of a servomechanism (such as that of an actuator assembly for the read/write heads of a hard disk drive assembly) in accordance with an improved """"bang-bang"""" seek technique includes at least one neural network. in one embodiment, a single neural network, connected to a plant with servomechanism, receives two input positioning signals and provides an output positioning signal. one input positioning signal represents a desired servomechanism position, while the other represents its present position. the output positioning signal represents a positioning time period within which the servomechanism will reach the desired position plus a deceleration time period within the positioning time period upon the termination of which the servomechanism will reach its desired position. in another embodiment, one neural network, connected to the plant, receives an input positioning signal and provides an output positioning signal and a status signal. the input positioning signal represents a desired servomechanism position. the output positioning signal represents a positioning time period within which the servomechanism will reach its desired position. the status signal represents a control status of the neural network. another neural network, also connected to the plant, receives the status signal plus a feedback signal and provides a correction signal. the feedback signal represents the present servomechanism position, while the correction signal represents a deceleration time period within the positioning time period upon the termination of which the servomechanism reaches its desired position. """,1997-10-14,"The title of the patent is intelligent servomechanism controller and its abstract is "" a servomechanism controller for controlling the movement and positioning of a servomechanism (such as that of an actuator assembly for the read/write heads of a hard disk drive assembly) in accordance with an improved """"bang-bang"""" seek technique includes at least one neural network. in one embodiment, a single neural network, connected to a plant with servomechanism, receives two input positioning signals and provides an output positioning signal. one input positioning signal represents a desired servomechanism position, while the other represents its present position. the output positioning signal represents a positioning time period within which the servomechanism will reach the desired position plus a deceleration time period within the positioning time period upon the termination of which the servomechanism will reach its desired position. in another embodiment, one neural network, connected to the plant, receives an input positioning signal and provides an output positioning signal and a status signal. the input positioning signal represents a desired servomechanism position. the output positioning signal represents a positioning time period within which the servomechanism will reach its desired position. the status signal represents a control status of the neural network. another neural network, also connected to the plant, receives the status signal plus a feedback signal and provides a correction signal. the feedback signal represents the present servomechanism position, while the correction signal represents a deceleration time period within the positioning time period upon the termination of which the servomechanism reaches its desired position. "" dated 1997-10-14"
5677998,apparatus for pulsed electrical circuitry,"apparatus for pulsed electrical power circuitry includes a pulsed power supply circuit, monitor for monitoring a voltage or current characteristic within the supply circuit and an artificial neural network trained to identify whether the detected characteristic is acceptable or harmful and adapted to produce an output accordingly. the circuit may be arranged to deliver its output to a gas discharge load and the output of the trained network may be adapted to be fed to control a parameter in the circuit or in the load whereby a harmful condition in the circuit may be avoided.",1997-10-14,"The title of the patent is apparatus for pulsed electrical circuitry and its abstract is apparatus for pulsed electrical power circuitry includes a pulsed power supply circuit, monitor for monitoring a voltage or current characteristic within the supply circuit and an artificial neural network trained to identify whether the detected characteristic is acceptable or harmful and adapted to produce an output accordingly. the circuit may be arranged to deliver its output to a gas discharge load and the output of the trained network may be adapted to be fed to control a parameter in the circuit or in the load whereby a harmful condition in the circuit may be avoided. dated 1997-10-14"
5678677,method and apparatus for the classification of an article,"in a process for the classification of an article such as a banknote described by of a k-dimensional feature vector (agf) which is prepared by a preliminary processing system (7), a test specimen is either assigned to one of n target classes or classified as a counterfeit. for the n target classes n recognition units (15.1 to 15.n) are used, exactly one of the n target classes being recognisable by one recognition unit (15.j) using a respective feature vector (agfj) prepared for that class. a recognised target class is transmitted by an output unit (14) to a service system (11). there are assigned to a target class in a learning phase several k-dimensional target vectors which are compared with the feature vector during the classification. the recognition unit (15.j) is advantageously a neural network, one neuron comparing the feature vector (agfj) with one of the target vectors.",1997-10-21,"The title of the patent is method and apparatus for the classification of an article and its abstract is in a process for the classification of an article such as a banknote described by of a k-dimensional feature vector (agf) which is prepared by a preliminary processing system (7), a test specimen is either assigned to one of n target classes or classified as a counterfeit. for the n target classes n recognition units (15.1 to 15.n) are used, exactly one of the n target classes being recognisable by one recognition unit (15.j) using a respective feature vector (agfj) prepared for that class. a recognised target class is transmitted by an output unit (14) to a service system (11). there are assigned to a target class in a learning phase several k-dimensional target vectors which are compared with the feature vector during the classification. the recognition unit (15.j) is advantageously a neural network, one neuron comparing the feature vector (agfj) with one of the target vectors. dated 1997-10-21"
5680096,process and apparatus for monitoring a vehicle interior,"a monitoring of a vehicle interior is effected by detecting sound waves in a vehicle interior, either from an incursion source or as reflected as echo waves and decomposing the electrical signals representing those detected sound waves into measurement vectors which are compared with sample vectors in a neural network so that a correlation parameter is generated which triggers an alarm when the correlation parameter indicates incursion. the system can respond first to glass breakage before an echo system is used to then further establish the nature of the incursion.",1997-10-21,"The title of the patent is process and apparatus for monitoring a vehicle interior and its abstract is a monitoring of a vehicle interior is effected by detecting sound waves in a vehicle interior, either from an incursion source or as reflected as echo waves and decomposing the electrical signals representing those detected sound waves into measurement vectors which are compared with sample vectors in a neural network so that a correlation parameter is generated which triggers an alarm when the correlation parameter indicates incursion. the system can respond first to glass breakage before an echo system is used to then further establish the nature of the incursion. dated 1997-10-21"
5680470,method of automated signature verification,"a method of automated signature verification, in which a test signature, e.g., a signature entered by an operator, may be preprocessed and examined for test features. the test features may be compared against features of a set of template signatures, and verified in response to the presence or absence of the test features in the template signatures. the test signature may be preprocessed, so as to normalize it and remove artifacts which are irrelevant to verification. the features of the template signatures may be determined and stored in an associative memory or a data structure with associative memory capabilities, e.g., a discrete hopfield artificial neural network. the method of verification may be adjusted to greater or lesser sensitivity in response to external conditions.",1997-10-21,"The title of the patent is method of automated signature verification and its abstract is a method of automated signature verification, in which a test signature, e.g., a signature entered by an operator, may be preprocessed and examined for test features. the test features may be compared against features of a set of template signatures, and verified in response to the presence or absence of the test features in the template signatures. the test signature may be preprocessed, so as to normalize it and remove artifacts which are irrelevant to verification. the features of the template signatures may be determined and stored in an associative memory or a data structure with associative memory capabilities, e.g., a discrete hopfield artificial neural network. the method of verification may be adjusted to greater or lesser sensitivity in response to external conditions. dated 1997-10-21"
5680475,"system for processing textured images, texture analyser and texture synthesizer","a system (22) for synthesis of textured images, comprising a texture analyser (24) and a texture synthesizer (28). the texture analyser (24) comprises an analyzing neural network (30) which learns to characterize a texture by calculating synaptic coefficients (c.sub.ab), utilizing at least one proximity function which characterizes a neighbourhood around pixels of said texture. the texture synthesizer (28) comprises a synthesizing neural network (40) which receives the synaptic coefficients (c.sub.ab) thus calculated and which, utilizing a relaxation mechanism, synthesizes a replica of the texture learned. the neural networks (30), (40) may have a tree-type structure.",1997-10-21,"The title of the patent is system for processing textured images, texture analyser and texture synthesizer and its abstract is a system (22) for synthesis of textured images, comprising a texture analyser (24) and a texture synthesizer (28). the texture analyser (24) comprises an analyzing neural network (30) which learns to characterize a texture by calculating synaptic coefficients (c.sub.ab), utilizing at least one proximity function which characterizes a neighbourhood around pixels of said texture. the texture synthesizer (28) comprises a synthesizing neural network (40) which receives the synaptic coefficients (c.sub.ab) thus calculated and which, utilizing a relaxation mechanism, synthesizes a replica of the texture learned. the neural networks (30), (40) may have a tree-type structure. dated 1997-10-21"
5680481,facial feature extraction method and apparatus for a neural network acoustic and visual speech recognition system,"a facial feature extraction method and apparatus uses the variation in light intensity (gray-scale) of a frontal view of a speaker's face. the sequence of video images are sampled and quantized into a regular array of 150.times.150 pixels that naturally form a coordinate system of scan lines and pixel position along a scan line. left and right eye areas and a mouth are located by thresholding the pixel gray-scale and finding the centroids of the three areas. the line segment joining the eye area centroids is bisected at right angle to form an axis of symmetry. a straight line through the centroid of the mouth area that is at right angle to the axis of symmetry constitutes the mouth line. pixels along the mouth line and the axis of symmetry in the vicinity of the mouth area form a horizontal and vertical gray-scale profile, respectively. the profiles could be used as feature vectors but it is more efficient to select peaks and valleys (maximas and minimas) of the profile that correspond to the important physiological speech features such as lower and upper lip, mouth corner, and mouth area positions and pixel values and their time derivatives as visual vector components. time derivatives are estimated by pixel position and value changes between video image frames. a speech recognition system uses the visual feature vector in combination with a concomitant acoustic vector as inputs to a time-delay neural network.",1997-10-21,"The title of the patent is facial feature extraction method and apparatus for a neural network acoustic and visual speech recognition system and its abstract is a facial feature extraction method and apparatus uses the variation in light intensity (gray-scale) of a frontal view of a speaker's face. the sequence of video images are sampled and quantized into a regular array of 150.times.150 pixels that naturally form a coordinate system of scan lines and pixel position along a scan line. left and right eye areas and a mouth are located by thresholding the pixel gray-scale and finding the centroids of the three areas. the line segment joining the eye area centroids is bisected at right angle to form an axis of symmetry. a straight line through the centroid of the mouth area that is at right angle to the axis of symmetry constitutes the mouth line. pixels along the mouth line and the axis of symmetry in the vicinity of the mouth area form a horizontal and vertical gray-scale profile, respectively. the profiles could be used as feature vectors but it is more efficient to select peaks and valleys (maximas and minimas) of the profile that correspond to the important physiological speech features such as lower and upper lip, mouth corner, and mouth area positions and pixel values and their time derivatives as visual vector components. time derivatives are estimated by pixel position and value changes between video image frames. a speech recognition system uses the visual feature vector in combination with a concomitant acoustic vector as inputs to a time-delay neural network. dated 1997-10-21"
5680513,series parallel approach to identification of dynamic systems,"a system and method for identifying and controlling an unknown dynamic system in which the system is identified, at least in part from test data, and a control scheme may be adapted on-line. in addition, the system may be used to develop an off-line solution to complex problems related to both dynamic and static systems. the system may use a multiprocessor architecture which may have a variety of configurations but is particularly suited for a neural network. the neural network may be built up of neurons that are either purely one way (forward signal path) or two way. each neuron may be provided with its own synaptic weight, adjusted using only the local and backward signals.",1997-10-21,"The title of the patent is series parallel approach to identification of dynamic systems and its abstract is a system and method for identifying and controlling an unknown dynamic system in which the system is identified, at least in part from test data, and a control scheme may be adapted on-line. in addition, the system may be used to develop an off-line solution to complex problems related to both dynamic and static systems. the system may use a multiprocessor architecture which may have a variety of configurations but is particularly suited for a neural network. the neural network may be built up of neurons that are either purely one way (forward signal path) or two way. each neuron may be provided with its own synaptic weight, adjusted using only the local and backward signals. dated 1997-10-21"
5680515,high precision computing with charge domain devices and a pseudo-spectral method therefor,"the present invention enhances the bit resolution of a ccd/cid mvm processor by storing each bit of each matrix element as a separate ccd charge packet. the bits of each input vector are separately multiplied by each bit of each matrix element in massive parallelism and the resulting products are combined appropriately to synthesize the correct product. in another aspect of the invention, such arrays are employed in a pseudo-spectral method of the invention, in which partial differential equations are solved by expressing each derivative analytically as matrices, and the state function is updated at each computation cycle by multiplying it by the matrices. the matrices are treated as synaptic arrays of a neural network and the state function vector elements are treated as neurons. in a further aspect of the invention, moving target detection is performed by driving the soliton equation with a vector of detector outputs. the neural architecture consists of two synaptic arrays corresponding to the two differential terms of the soliton-equation and an adder connected to the output thereof and to the output of the detector array to drive the soliton equation.",1997-10-21,"The title of the patent is high precision computing with charge domain devices and a pseudo-spectral method therefor and its abstract is the present invention enhances the bit resolution of a ccd/cid mvm processor by storing each bit of each matrix element as a separate ccd charge packet. the bits of each input vector are separately multiplied by each bit of each matrix element in massive parallelism and the resulting products are combined appropriately to synthesize the correct product. in another aspect of the invention, such arrays are employed in a pseudo-spectral method of the invention, in which partial differential equations are solved by expressing each derivative analytically as matrices, and the state function is updated at each computation cycle by multiplying it by the matrices. the matrices are treated as synaptic arrays of a neural network and the state function vector elements are treated as neurons. in a further aspect of the invention, moving target detection is performed by driving the soliton equation with a vector of detector outputs. the neural architecture consists of two synaptic arrays corresponding to the two differential terms of the soliton-equation and an adder connected to the output thereof and to the output of the detector array to drive the soliton equation. dated 1997-10-21"
5680627,method and apparatus for character preprocessing which translates textual description into numeric form for input to a neural network,"a preprocessor translates problem descriptions in text format to a numeric form, usable by a neural network. problem descriptions and associated solutions are used as training sets for the neural network. from a word domain collected from a number of problem descriptions, each character used in the problem domain is assigned a character frequency value. these frequency values are used to translate each problem description to an input vector, where each input value represents a character and each input value is associated with an input node of the neural network. in addition to being associated with a frequency value, each character is scaled and normalized to improve accuracy of the neural network's recall capabilities.",1997-10-21,"The title of the patent is method and apparatus for character preprocessing which translates textual description into numeric form for input to a neural network and its abstract is a preprocessor translates problem descriptions in text format to a numeric form, usable by a neural network. problem descriptions and associated solutions are used as training sets for the neural network. from a word domain collected from a number of problem descriptions, each character used in the problem domain is assigned a character frequency value. these frequency values are used to translate each problem description to an input vector, where each input value represents a character and each input value is associated with an input node of the neural network. in addition to being associated with a frequency value, each character is scaled and normalized to improve accuracy of the neural network's recall capabilities. dated 1997-10-21"
5680866,artificial neural network cardiopulmonary modeling and diagnosis,"the present invention is a method of diagnosing a cardiopulmonary condition in an individual by comparing data from a progressive multi-stage test for the individual to a non-linear multi-variate model, preferably a recurrent artificial neural network having sensor fusion. the present invention relies on a cardiovascular model developed from physiological measurements of an individual. any differences between the modeled parameters and the parameters of an individual at a given time are used for diagnosis.",1997-10-28,"The title of the patent is artificial neural network cardiopulmonary modeling and diagnosis and its abstract is the present invention is a method of diagnosing a cardiopulmonary condition in an individual by comparing data from a progressive multi-stage test for the individual to a non-linear multi-variate model, preferably a recurrent artificial neural network having sensor fusion. the present invention relies on a cardiovascular model developed from physiological measurements of an individual. any differences between the modeled parameters and the parameters of an individual at a given time are used for diagnosis. dated 1997-10-28"
5681496,apparatus for and method of controlling a microwave oven and a microwave oven controlled thereby,a sensor based automated cooking apparatus is provided. a humidity sensor measures the moisture content within a cooking cavity. an output of the sensor is provided to a digital filter to remove noise therefrom before being passed to a feature extractor which performs a data compression step and extracts salient features relating to the shape of the humidity versus time characteristic. the parameters are analyzed by a neural network to estimate a degree of doneness of the food. a controller uses the degree of doneness to estimate the remaining cooking time and appropriate power level. the cooking apparatus then operates in an open loop mode for the remainder of the cooking time using the appropriate power level.,1997-10-28,The title of the patent is apparatus for and method of controlling a microwave oven and a microwave oven controlled thereby and its abstract is a sensor based automated cooking apparatus is provided. a humidity sensor measures the moisture content within a cooking cavity. an output of the sensor is provided to a digital filter to remove noise therefrom before being passed to a feature extractor which performs a data compression step and extracts salient features relating to the shape of the humidity versus time characteristic. the parameters are analyzed by a neural network to estimate a degree of doneness of the food. a controller uses the degree of doneness to estimate the remaining cooking time and appropriate power level. the cooking apparatus then operates in an open loop mode for the remainder of the cooking time using the appropriate power level. dated 1997-10-28
5682503,self-organizing neural network for pattern classification,"a neural network includes a plurality of input nodes for receiving the respective elements of the input vector. a copy of all of the elements of the input vector is sent to the next level of nodes in the neural network denoted as intermediate nodes. the intermediate nodes each encode a separate template pattern. they compare the actual input pattern with the template and generate a signal indicative of the difference between the input pattern and the template pattern. each of the templates encoded in the intermediate nodes has a class associated with it. the difference calculated by the intermediate nodes is passed to an output node for each of the intermediate nodes at a given class. the output node then selects the minimum difference amongst the values sent from the intermediate nodes. this lowest difference for the class represented by the output node is then forwarded to a selector. the selector receives such values from each of the output nodes of all of the classes and then selects that to output value which is a minimum difference. the selector in turn, generates a signal indicative of the class of the intermediate node that sent the smallest difference value.",1997-10-28,"The title of the patent is self-organizing neural network for pattern classification and its abstract is a neural network includes a plurality of input nodes for receiving the respective elements of the input vector. a copy of all of the elements of the input vector is sent to the next level of nodes in the neural network denoted as intermediate nodes. the intermediate nodes each encode a separate template pattern. they compare the actual input pattern with the template and generate a signal indicative of the difference between the input pattern and the template pattern. each of the templates encoded in the intermediate nodes has a class associated with it. the difference calculated by the intermediate nodes is passed to an output node for each of the intermediate nodes at a given class. the output node then selects the minimum difference amongst the values sent from the intermediate nodes. this lowest difference for the class represented by the output node is then forwarded to a selector. the selector receives such values from each of the output nodes of all of the classes and then selects that to output value which is a minimum difference. the selector in turn, generates a signal indicative of the class of the intermediate node that sent the smallest difference value. dated 1997-10-28"
5683605,heater controlling unit using a fuzzy neural network,"a heater controlling unit used in a heat fusing device of an image forming apparatus, such as a copying machine, comprises a temperature detecting unit for outputting a detected temperature, a temperature change rate computing unit for computing the surface temperature of a heat fusing roller using the output of the temperature detecting unit, and an on-time computing and controlling unit for computing a heater on-time within a predetermined period by means of a fuzzy neural network using the above temperature and a temperature change rate found by the temperature-change rate computing unit, a heater controlling circuit for controlling the on/off action of the heater, a predictive computing unit for predicting the surface temperature, and a comparing and adjusting unit for comparing the actual temperature with the predicted temperature, and, when the difference is greater than a predetermined value (e.g., .+-.5.degree. c.), for adjusting the weights of the links within the network. accordingly, the roughly-set parameters can be amended to the ones for an optimal on-time by the sequential learning. as a result, the program can be generated in a simpler manner, and the program can be changed easily for individual units depending on aged distortion and environments thereof.",1997-11-04,"The title of the patent is heater controlling unit using a fuzzy neural network and its abstract is a heater controlling unit used in a heat fusing device of an image forming apparatus, such as a copying machine, comprises a temperature detecting unit for outputting a detected temperature, a temperature change rate computing unit for computing the surface temperature of a heat fusing roller using the output of the temperature detecting unit, and an on-time computing and controlling unit for computing a heater on-time within a predetermined period by means of a fuzzy neural network using the above temperature and a temperature change rate found by the temperature-change rate computing unit, a heater controlling circuit for controlling the on/off action of the heater, a predictive computing unit for predicting the surface temperature, and a comparing and adjusting unit for comparing the actual temperature with the predicted temperature, and, when the difference is greater than a predetermined value (e.g., .+-.5.degree. c.), for adjusting the weights of the links within the network. accordingly, the roughly-set parameters can be amended to the ones for an optimal on-time by the sequential learning. as a result, the program can be generated in a simpler manner, and the program can be changed easily for individual units depending on aged distortion and environments thereof. dated 1997-11-04"
5684699,travel characteristic control system for automotive vehicle,"an electric control system for controlling a travel characteristic of an automotive vehicle, wherein a driving condition, movement conditions and a travel course of the vehicle are detected and a number of parameters indicative of a resultant of learning based on a teacher data are memorized, and wherein a neural network is adapted to estimate a characteristic index indicative of a driving characteristic of the driver on a basis of the memorized parameters and the detected driving condition and travel course of the vehicle. in the control system, a control signal for control of the travel characteristic of the vehicle is produced in accordance with the detected movement conditions of the vehicle and corrected in accordance with the estimated characteristic index.",1997-11-04,"The title of the patent is travel characteristic control system for automotive vehicle and its abstract is an electric control system for controlling a travel characteristic of an automotive vehicle, wherein a driving condition, movement conditions and a travel course of the vehicle are detected and a number of parameters indicative of a resultant of learning based on a teacher data are memorized, and wherein a neural network is adapted to estimate a characteristic index indicative of a driving characteristic of the driver on a basis of the memorized parameters and the detected driving condition and travel course of the vehicle. in the control system, a control signal for control of the travel characteristic of the vehicle is produced in accordance with the detected movement conditions of the vehicle and corrected in accordance with the estimated characteristic index. dated 1997-11-04"
5687286,neural networks with subdivision,"a neural network apparatus, and methods for training the neural network apparatus, for processing input information, supplied as a data array, for a prespecified application to indicate output categories characteristic of the processing for that application. in the invention, an input stage accepts the data array and converts it to a corresponding internal representation, and a data preprocessor analyzes the data array based on a plurality of feature attributes to generate a corresponding plurality of attribute measures. a neural network, comprising a plurality of interconnected neurons, processes the attribute measures to reach a neural state representative of corresponding category attributes; portions of the network are predefined to include a number of neurons and prespecified with a particular correspondence to the feature attributes to accept corresponding attribute measures for the data array, and portions of the network are prespecified with a particular correspondence to the category attributes. a data postprocessor indicates the category attributes by correlating the neural state with predefined category attribute measures, and an output stage combines the category measures in a prespecified manner to generate on output category for the input information.",1997-11-11,"The title of the patent is neural networks with subdivision and its abstract is a neural network apparatus, and methods for training the neural network apparatus, for processing input information, supplied as a data array, for a prespecified application to indicate output categories characteristic of the processing for that application. in the invention, an input stage accepts the data array and converts it to a corresponding internal representation, and a data preprocessor analyzes the data array based on a plurality of feature attributes to generate a corresponding plurality of attribute measures. a neural network, comprising a plurality of interconnected neurons, processes the attribute measures to reach a neural state representative of corresponding category attributes; portions of the network are predefined to include a number of neurons and prespecified with a particular correspondence to the feature attributes to accept corresponding attribute measures for the data array, and portions of the network are prespecified with a particular correspondence to the category attributes. a data postprocessor indicates the category attributes by correlating the neural state with predefined category attribute measures, and an output stage combines the category measures in a prespecified manner to generate on output category for the input information. dated 1997-11-11"
5687291,method and apparatus for estimating a cognitive decision made in response to a known stimulus from the corresponding single-event evoked cerebral potential,"the present invention estimates the cognitive decision made in response to a known stimulus from the corresponding single-event evoked cerebral potential. the present invention uses a unique recursive procedure to identify the decision from a mathematical description of the potential as the output of a cerebrally located, autoregressive, moving average filter with the stimulus as an exogenous input. the procedure employs in a two-step sequence, the least squares algorithm to update the filter coefficients, followed by a taylor's series approximation for updating an internal cerebral source signal which is generated in response to the external stimulus. the recursive procedure computes the attenuation used by the moving average component of the filter to produce the cerebral source signal. this procedure is repeated for all feasible cerebral source signals, computed from the set of possible event evoked average response potentials, to produce a set of attenuator-values. these values are then used as input to a multiple-layered, feed-forward artificial neural network for identifying the decision made from the set of feasible responses. in turn, the power spectrum computed from the autoregressive coefficients is used to track the cognitive state and therefore the reliability of the decision estimate. the present invention may be used for the control by mental thought of computerized visual and aural display functions, by measuring the electroencephalogram in time with the operant orientation of the user onto a displayed stimulus.",1997-11-11,"The title of the patent is method and apparatus for estimating a cognitive decision made in response to a known stimulus from the corresponding single-event evoked cerebral potential and its abstract is the present invention estimates the cognitive decision made in response to a known stimulus from the corresponding single-event evoked cerebral potential. the present invention uses a unique recursive procedure to identify the decision from a mathematical description of the potential as the output of a cerebrally located, autoregressive, moving average filter with the stimulus as an exogenous input. the procedure employs in a two-step sequence, the least squares algorithm to update the filter coefficients, followed by a taylor's series approximation for updating an internal cerebral source signal which is generated in response to the external stimulus. the recursive procedure computes the attenuation used by the moving average component of the filter to produce the cerebral source signal. this procedure is repeated for all feasible cerebral source signals, computed from the set of possible event evoked average response potentials, to produce a set of attenuator-values. these values are then used as input to a multiple-layered, feed-forward artificial neural network for identifying the decision made from the set of feasible responses. in turn, the power spectrum computed from the autoregressive coefficients is used to track the cognitive state and therefore the reliability of the decision estimate. the present invention may be used for the control by mental thought of computerized visual and aural display functions, by measuring the electroencephalogram in time with the operant orientation of the user onto a displayed stimulus. dated 1997-11-11"
5687292,device and method for determining a distribution of resources of a physical network,"a device for distributing resources of a given physical network among logical links by subdividing physical link capacities into logical links using an algorithm. the device comprises a first neural network, in which one part of the algorithm is implemented, and a second neural network, in which a second part of the algorithm is implemented, said two neural networks interworking to compute logical link capacities. furthermore, a method for distributing resources of a given physical network among logical links by subdividing physical link capacities into said logical links is provided. more specifically, the method involves the use of a first neural network, in which one part of an algorithm is implemented, and a second neural network, in which a second part of an algorithm is implemented, said two neural networks interworking to compute logical link capacities so that the operation of the physical network, given an objective function, is generally optimized, according to the objective function.",1997-11-11,"The title of the patent is device and method for determining a distribution of resources of a physical network and its abstract is a device for distributing resources of a given physical network among logical links by subdividing physical link capacities into logical links using an algorithm. the device comprises a first neural network, in which one part of the algorithm is implemented, and a second neural network, in which a second part of the algorithm is implemented, said two neural networks interworking to compute logical link capacities. furthermore, a method for distributing resources of a given physical network among logical links by subdividing physical link capacities into said logical links is provided. more specifically, the method involves the use of a first neural network, in which one part of an algorithm is implemented, and a second neural network, in which a second part of an algorithm is implemented, said two neural networks interworking to compute logical link capacities so that the operation of the physical network, given an objective function, is generally optimized, according to the objective function. dated 1997-11-11"
5687716,selective differentiating diagnostic process based on broad data bases,"a process for analyzing for a variety of physical medical disorders, which comprises: pa1 a) clinical testing by a common analysis, patients free of such disorders and those patients that possess such one or more disorders, which analysis distinquishes any such disorder thereby obtaining the characterization of a collection of such disorders as numerical data; pa1 b) scaling the matrix of said numerical data in a computer; pa1 c) configuring in a computer, an artificial neural network prescribed by the number of variables and the number of disorders, which artificial neural network is electronic data that possesses an input layer, one or more hidden layers and an output layer; pa1 d) fitting the artificial neural network electronic data in a computer to the numerical data according to adjustable parameters of the artificial neural network, including one or more of pa2 i) the number of neurons in a hidden layer; pa2 ii) the number of hidden layers; pa2 iii) the type of transfer functions in the layers; and pa2 iv) the weights connecting the neurons; pa1 e) whereby the artificial neural network is trained in the computer to automatically provide an analytic model; pa1 f) withdrawing from a new patient that has not been diagnosed for such disorder, a sample or samples of a kind taken from said reference patients and subjecting said diagnostic sample to said clinical testing to obtain new numerical data; pa1 g) automatically scaling in the computer the new data to the data derived from the reference patients; pa1 h) feeding the scaled new data to the trained artificial neural network and the analytic model thereof, and pa1 i) automatically obtaining a diagnosis of the new patient with respect to such disorders.",1997-11-18,"The title of the patent is selective differentiating diagnostic process based on broad data bases and its abstract is a process for analyzing for a variety of physical medical disorders, which comprises: pa1 a) clinical testing by a common analysis, patients free of such disorders and those patients that possess such one or more disorders, which analysis distinquishes any such disorder thereby obtaining the characterization of a collection of such disorders as numerical data; pa1 b) scaling the matrix of said numerical data in a computer; pa1 c) configuring in a computer, an artificial neural network prescribed by the number of variables and the number of disorders, which artificial neural network is electronic data that possesses an input layer, one or more hidden layers and an output layer; pa1 d) fitting the artificial neural network electronic data in a computer to the numerical data according to adjustable parameters of the artificial neural network, including one or more of pa2 i) the number of neurons in a hidden layer; pa2 ii) the number of hidden layers; pa2 iii) the type of transfer functions in the layers; and pa2 iv) the weights connecting the neurons; pa1 e) whereby the artificial neural network is trained in the computer to automatically provide an analytic model; pa1 f) withdrawing from a new patient that has not been diagnosed for such disorder, a sample or samples of a kind taken from said reference patients and subjecting said diagnostic sample to said clinical testing to obtain new numerical data; pa1 g) automatically scaling in the computer the new data to the data derived from the reference patients; pa1 h) feeding the scaled new data to the trained artificial neural network and the analytic model thereof, and pa1 i) automatically obtaining a diagnosis of the new patient with respect to such disorders. dated 1997-11-18"
5689581,methods of inspection,"an inspection method for judging whether an object is defective or not uses a neural network comprising an input neuron which quantizes input feature quantities based on present quantization ranges, a plurality of intermediate neuron coupled to the input neuron with coupling coefficients determined from quantized feature quantities and a pair of output neurons coupled to respective intermediate neurons with intermediate coupling coefficients obtained by learning in which the intermediate coupling coefficients are initially set so that any object is judged to be defective and are fitted so that if an object is nondefective then it is judged to be nondefective by performing learning with use of nondefective objects.",1997-11-18,"The title of the patent is methods of inspection and its abstract is an inspection method for judging whether an object is defective or not uses a neural network comprising an input neuron which quantizes input feature quantities based on present quantization ranges, a plurality of intermediate neuron coupled to the input neuron with coupling coefficients determined from quantized feature quantities and a pair of output neurons coupled to respective intermediate neurons with intermediate coupling coefficients obtained by learning in which the intermediate coupling coefficients are initially set so that any object is judged to be defective and are fitted so that if an object is nondefective then it is judged to be nondefective by performing learning with use of nondefective objects. dated 1997-11-18"
5689616,automatic language identification/verification system,"a language identification and verification system is described whereby language identification is determined by finding the closest match of a speech utterance to multiple speaker sets. the language identification and verification system is implemented through use of a speaker identification/verification system as a baseline to find a set of well matched speakers in each of a plurality of languages. a comparison of unknown speech to speech features from such well-matched speakers is then made and a language decision is arrived on based on a closest match between the unknown speech features and speech features for such well matched reference speakers in a particular language. to avoid a problem associated with prior-art language identification systems, wherein speech feature are based on short-term spectral features determined at a system frame rate--thereby seriously limiting the resolution and accuracy of such prior-art systems, the invention uses speech features derived from vocalic or syllabic nuclei, from which related phonetic speech features may then be extracted. detection of such vocalic centers or syllabic nuclei is accomplished using a trained back-error propagation multi-level neural network.",1997-11-18,"The title of the patent is automatic language identification/verification system and its abstract is a language identification and verification system is described whereby language identification is determined by finding the closest match of a speech utterance to multiple speaker sets. the language identification and verification system is implemented through use of a speaker identification/verification system as a baseline to find a set of well matched speakers in each of a plurality of languages. a comparison of unknown speech to speech features from such well-matched speakers is then made and a language decision is arrived on based on a closest match between the unknown speech features and speech features for such well matched reference speakers in a particular language. to avoid a problem associated with prior-art language identification systems, wherein speech feature are based on short-term spectral features determined at a system frame rate--thereby seriously limiting the resolution and accuracy of such prior-art systems, the invention uses speech features derived from vocalic or syllabic nuclei, from which related phonetic speech features may then be extracted. detection of such vocalic centers or syllabic nuclei is accomplished using a trained back-error propagation multi-level neural network. dated 1997-11-18"
5689621,modular feedforward neural network architecture with learning,"a first feedforward network receives an input vector signal and a plurality of weight signals and forms an output vector signal based thereupon. a second feedforward network, substantially identical to the first feedforward network, receives a first learning vector signal and the weight signals and forms a learning output vector based thereupon. a weight updating circuit generates the weight signals in accordance with a back propagation updating rule. the weight signals are updated based upon the first learning vector signal and a second learning vector signal in response to receiving a first predetermined layer signal, and are updated based upon the first learning vector signal, the second learning vector signal, and the learning output vector signal in response to receiving a second predetermined layer signal. a back propagation feedback network receives the second learning vector signal and the weight signals and generates a back propagated error vector signal based thereupon. the feedforward networks, the weight updating circuit, and the back propagation feedback network are implemented as analog circuits on a single integrated circuit.",1997-11-18,"The title of the patent is modular feedforward neural network architecture with learning and its abstract is a first feedforward network receives an input vector signal and a plurality of weight signals and forms an output vector signal based thereupon. a second feedforward network, substantially identical to the first feedforward network, receives a first learning vector signal and the weight signals and forms a learning output vector based thereupon. a weight updating circuit generates the weight signals in accordance with a back propagation updating rule. the weight signals are updated based upon the first learning vector signal and a second learning vector signal in response to receiving a first predetermined layer signal, and are updated based upon the first learning vector signal, the second learning vector signal, and the learning output vector signal in response to receiving a second predetermined layer signal. a back propagation feedback network receives the second learning vector signal and the weight signals and generates a back propagated error vector signal based thereupon. the feedforward networks, the weight updating circuit, and the back propagation feedback network are implemented as analog circuits on a single integrated circuit. dated 1997-11-18"
5690103,detection/exclusion of acute myocardial infarction using neural network analysis of measurements of biochemical markers,"the overall invention categorizes patients with suspected acute myocardial infarction (ami) with regard to a) ami/non-ami; b) infarct size (e.g. major/minor); c) time since onset of infarction; and d) non-ami with/without minor myocardial damage (mmd). generally, the above categorization is based on frequent timed blood sampling and measurement of selected biochemical markers of ami with different rates of appearance in circulating blood. the computations are performed by using specially designed artificial neural networks. according to a first main aspect of the invention, early, i.e. generally within 3 hours from admission of the patient, detection/exclusion of acute myocardial infarction is provided. furthermore, early prediction of the infarct size and early estimation of the time from onset are also provided.",1997-11-25,"The title of the patent is detection/exclusion of acute myocardial infarction using neural network analysis of measurements of biochemical markers and its abstract is the overall invention categorizes patients with suspected acute myocardial infarction (ami) with regard to a) ami/non-ami; b) infarct size (e.g. major/minor); c) time since onset of infarction; and d) non-ami with/without minor myocardial damage (mmd). generally, the above categorization is based on frequent timed blood sampling and measurement of selected biochemical markers of ami with different rates of appearance in circulating blood. the computations are performed by using specially designed artificial neural networks. according to a first main aspect of the invention, early, i.e. generally within 3 hours from admission of the patient, detection/exclusion of acute myocardial infarction is provided. furthermore, early prediction of the infarct size and early estimation of the time from onset are also provided. dated 1997-11-25"
5692098,real-time mozer phase recoding using a neural-network for speech compression,"a system and method for compressing speech using an artificial neural network to calculate the recoded phase vector (mozer code) resulting from the spectral magnitude-to-phase transformation. raw speech is equalized to remove the spectral tilt and segmented into analysis frames. the spectral magnitudes of each frame segment are determined at a plurality of points by a fourier transform, normalized, and applied to a neural net magnitude-to-phase transform calculator to provide a recoded phase vector. an inverse discrete fourier transform is used to calculate the new recoded speech waveform in which the two quarters with minimum power are zeroed to produce the compressed speech output signal.",1997-11-25,"The title of the patent is real-time mozer phase recoding using a neural-network for speech compression and its abstract is a system and method for compressing speech using an artificial neural network to calculate the recoded phase vector (mozer code) resulting from the spectral magnitude-to-phase transformation. raw speech is equalized to remove the spectral tilt and segmented into analysis frames. the spectral magnitudes of each frame segment are determined at a plurality of points by a fourier transform, normalized, and applied to a neural net magnitude-to-phase transform calculator to provide a recoded phase vector. an inverse discrete fourier transform is used to calculate the new recoded speech waveform in which the two quarters with minimum power are zeroed to produce the compressed speech output signal. dated 1997-11-25"
5696838,pattern searching method using neural networks and correlation,"a pattern searching method using neural networks and correlation. this method combines the quickness and adaptiveness of neural networks with the accuracy of the mathematical correlation approach. images are divided into small sub-images which are presented to the trained neural network. sub-images that may contain the pattern or partial pattern are selected by the neural network. the neural network also provides the approximate location of the pattern, therefore the selected sub-images can be adjusted to contain the complete pattern. desired patterns can be located by measuring the new sub-images' correlation values against the reference models in a small area. experiments show that this superior method is able to find the desired patterns. moreover, this method is much faster than traditional pattern searching methods which use only correlation.",1997-12-09,"The title of the patent is pattern searching method using neural networks and correlation and its abstract is a pattern searching method using neural networks and correlation. this method combines the quickness and adaptiveness of neural networks with the accuracy of the mathematical correlation approach. images are divided into small sub-images which are presented to the trained neural network. sub-images that may contain the pattern or partial pattern are selected by the neural network. the neural network also provides the approximate location of the pattern, therefore the selected sub-images can be adjusted to contain the complete pattern. desired patterns can be located by measuring the new sub-images' correlation values against the reference models in a small area. experiments show that this superior method is able to find the desired patterns. moreover, this method is much faster than traditional pattern searching methods which use only correlation. dated 1997-12-09"
5696877,pattern recognition using a predictive neural network,"input feature vectors (a(t)) is considered a pattern selected from a plurality of reference patterns which represent categories of recognition objects. each reference pattern is defined by a sequence of state models, successively supplied with the time sequence of the input feature vectors and with a sequence of preceding state vectors (h(t, s, n)). the sequence of the state models produces a time sequence of predicted feature vectors (a(t+1, s, n) and a sequence of new state vectors (h(t+1, s, n)). the recognized pattern is selected from one of the reference patterns that minimizes a prediction error between the time sequence of the input feature vectors and the time sequence of the predicted feature vectors. the prediction error is calculated by using a dynamic programming algorithm. training of the reference pattern is carried out by a gradient descent method such as back-propagation technique.",1997-12-09,"The title of the patent is pattern recognition using a predictive neural network and its abstract is input feature vectors (a(t)) is considered a pattern selected from a plurality of reference patterns which represent categories of recognition objects. each reference pattern is defined by a sequence of state models, successively supplied with the time sequence of the input feature vectors and with a sequence of preceding state vectors (h(t, s, n)). the sequence of the state models produces a time sequence of predicted feature vectors (a(t+1, s, n) and a sequence of new state vectors (h(t+1, s, n)). the recognized pattern is selected from one of the reference patterns that minimizes a prediction error between the time sequence of the input feature vectors and the time sequence of the predicted feature vectors. the prediction error is calculated by using a dynamic programming algorithm. training of the reference pattern is carried out by a gradient descent method such as back-propagation technique. dated 1997-12-09"
5696881,method for use in a processing system having a network that generates a network output signal as a polynomial function of a plurality of network input signals,"a continuous logic system using a neural network is characterized by defining input and output variables that do not use a membership function, by employing production rules (if/then rules) that relate the output variables to the input variables, and by using the neural network to compute or interpolate the outputs. the neural network first learns the given production rules and then produces the outputs in real time. the neural network is constructed of artificial neurons each having only one significant processing element in the form of a multiplier. the neural network utilizes a training algorithm which does not require repetitive training and which yields a global minimum to each given set of input vectors.",1997-12-09,"The title of the patent is method for use in a processing system having a network that generates a network output signal as a polynomial function of a plurality of network input signals and its abstract is a continuous logic system using a neural network is characterized by defining input and output variables that do not use a membership function, by employing production rules (if/then rules) that relate the output variables to the input variables, and by using the neural network to compute or interpolate the outputs. the neural network first learns the given production rules and then produces the outputs in real time. the neural network is constructed of artificial neurons each having only one significant processing element in the form of a multiplier. the neural network utilizes a training algorithm which does not require repetitive training and which yields a global minimum to each given set of input vectors. dated 1997-12-09"
5696882,method and apparatus for using a neural network to extract an optimal subset of data objects from an available class of data objects,"method and apparatus for selecting an optimized subset of trajectories from an available class of potential trajectories in a velocimetry application. in an exemplary method, a neural network is constructed wherein each neuron in the network represents a trajectory in the overall class. a binary output of each neuron indicates whether the object represented by the neuron is to be selected. in the exemplary method, the neural network is in an initially converged state. the network is then alternately excited and constrained so that it settles to additional converged states. during excitation, correction factors including a parameter-optimizing term are applied to neuron input potentials. during constraint, the parameter-optimizing terms are interrupted. each time the network converges, the outputs of the neurons in the network are decoded to establish a subset of trajectories to be selected, and a value for an optimization parameter associated with the established subset is computed. the excitation and constraint phases are repeated in an attempt to establish subsets of trajectories having superior optimization parameter values. once the excitation and constraint phases have been repeated a suitable number of times, the trajectories identified in a subset of objects having a superior optimization parameter value are selected. in exemplary embodiments, neurons are updated in a novel block-sequential fashion.",1997-12-09,"The title of the patent is method and apparatus for using a neural network to extract an optimal subset of data objects from an available class of data objects and its abstract is method and apparatus for selecting an optimized subset of trajectories from an available class of potential trajectories in a velocimetry application. in an exemplary method, a neural network is constructed wherein each neuron in the network represents a trajectory in the overall class. a binary output of each neuron indicates whether the object represented by the neuron is to be selected. in the exemplary method, the neural network is in an initially converged state. the network is then alternately excited and constrained so that it settles to additional converged states. during excitation, correction factors including a parameter-optimizing term are applied to neuron input potentials. during constraint, the parameter-optimizing terms are interrupted. each time the network converges, the outputs of the neurons in the network are decoded to establish a subset of trajectories to be selected, and a value for an optimization parameter associated with the established subset is computed. the excitation and constraint phases are repeated in an attempt to establish subsets of trajectories having superior optimization parameter values. once the excitation and constraint phases have been repeated a suitable number of times, the trajectories identified in a subset of objects having a superior optimization parameter value are selected. in exemplary embodiments, neurons are updated in a novel block-sequential fashion. dated 1997-12-09"
5696883,neural network expressing apparatus including refresh of stored synapse load value information,"a self-organizable neural network expressing unit includes a plurality of neuron units electronically expressing nerve cell bodies, and a plurality of synapse expressing units electronically expressing synapses for coupling neuron units through programmed coupling strengths represented by synapse load values, and a control circuit for supplying a pattern of random number data as an educator data. when the pattern of random number data is generated, the neural network expressing unit carries out correction of synapse load values as in a learning mode of operation using the pattern of random number data as an educator data. the memorized internal states in the neural network expressing unit is reinforced based on a faded memory thereof, and the synapse load values are precisely maintained for a long time, resulting in a reliable neural network expressing unit.",1997-12-09,"The title of the patent is neural network expressing apparatus including refresh of stored synapse load value information and its abstract is a self-organizable neural network expressing unit includes a plurality of neuron units electronically expressing nerve cell bodies, and a plurality of synapse expressing units electronically expressing synapses for coupling neuron units through programmed coupling strengths represented by synapse load values, and a control circuit for supplying a pattern of random number data as an educator data. when the pattern of random number data is generated, the neural network expressing unit carries out correction of synapse load values as in a learning mode of operation using the pattern of random number data as an educator data. the memorized internal states in the neural network expressing unit is reinforced based on a faded memory thereof, and the synapse load values are precisely maintained for a long time, resulting in a reliable neural network expressing unit. dated 1997-12-09"
5696907,system and method for performing risk and credit analysis of financial service applications,"the present invention discloses a method and system for performing risk and credit analysis of financial service applications with a neural network. the neural network imitates and perfects a credit manager's evaluation and decision process to control loss and guide business expansion. in particular, the neural network screens applications to control loss and to find directions where business volume can increase with a minimum increase in loss. initially, data variables are pre-processed and applied to the neural network. the neural network in the present invention is optimized by a non-iterative regression process, as opposed to the computationally intensive back propagation algorithm.",1997-12-09,"The title of the patent is system and method for performing risk and credit analysis of financial service applications and its abstract is the present invention discloses a method and system for performing risk and credit analysis of financial service applications with a neural network. the neural network imitates and perfects a credit manager's evaluation and decision process to control loss and guide business expansion. in particular, the neural network screens applications to control loss and to find directions where business volume can increase with a minimum increase in loss. initially, data variables are pre-processed and applied to the neural network. the neural network in the present invention is optimized by a non-iterative regression process, as opposed to the computationally intensive back propagation algorithm. dated 1997-12-09"
5697369,"method and apparatus for disease, injury and bodily condition screening or sensing","a method and apparatus for measuring biopotentials at a test site on a human or animal subject and determining therefrom the existence or status of a condition at such test site. the efficacy of treatment of a condition is monitored by comparing values obtained from biopotentials taken before treatment with those obtained during or after treatment. for disease diagnosis or screening, values obtained from biopotential measurements taken during a test period are processed in a neural network which has been programmed to recognize potential value patterns indicative of a disease condition.",1997-12-16,"The title of the patent is method and apparatus for disease, injury and bodily condition screening or sensing and its abstract is a method and apparatus for measuring biopotentials at a test site on a human or animal subject and determining therefrom the existence or status of a condition at such test site. the efficacy of treatment of a condition is monitored by comparing values obtained from biopotentials taken before treatment with those obtained during or after treatment. for disease diagnosis or screening, values obtained from biopotential measurements taken during a test period are processed in a neural network which has been programmed to recognize potential value patterns indicative of a disease condition. dated 1997-12-16"
5699096,potential estimating apparatus using a plurality of neural networks for carrying out an electrophotographic process,"a potential estimation apparatus estimates a potential of a photosensitive body of an image forming apparatus that carries out an electro-photography process using the photosensitive body. the potential estimation apparatus includes a sensor group for sensing and outputting data related to information which affects the electro-photography process, a storage unit for at least storing the data output from the sensor group and information related to charge of the photosensitive body, and an estimation circuit including a neural network for estimating a charged portion potential of the photosensitive body based on a charge retentivity of the photosensitive body learned by the neural network. the neural network in a learning mode receives at least one of the data output from the sensor group and time-sequentially sampled, and parameters which affect the charge retentivity of the photosensitive body as an input, and receives as a teaching value a charged portion potential which is obtained in advance with respect to at least an amount of charge and the charge retentivity of the photosensitive body.",1997-12-16,"The title of the patent is potential estimating apparatus using a plurality of neural networks for carrying out an electrophotographic process and its abstract is a potential estimation apparatus estimates a potential of a photosensitive body of an image forming apparatus that carries out an electro-photography process using the photosensitive body. the potential estimation apparatus includes a sensor group for sensing and outputting data related to information which affects the electro-photography process, a storage unit for at least storing the data output from the sensor group and information related to charge of the photosensitive body, and an estimation circuit including a neural network for estimating a charged portion potential of the photosensitive body based on a charge retentivity of the photosensitive body learned by the neural network. the neural network in a learning mode receives at least one of the data output from the sensor group and time-sequentially sampled, and parameters which affect the charge retentivity of the photosensitive body as an input, and receives as a teaching value a charged portion potential which is obtained in advance with respect to at least an amount of charge and the charge retentivity of the photosensitive body. dated 1997-12-16"
5699253,nonlinear dynamic transform for correction of crankshaft acceleration having torsional oscillations,"irregularities in crankshaft velocity introduced when measuring crankshaft rotation at a section of a crankshaft in an internal combustion engine that is less damped to torsional oscillations than is another more accessible crankshaft section are corrected by performing a nonlinear transformation via a neural network to predict rotation measurements that would have been obtained at the inaccessible section from data actually collected at the accessible crankshaft section. thus, the effects of torsional oscillations in the crankshaft are substantially filtered away, resulting in crankshaft acceleration values that form the basis of a misfire detector having nearly maximum signal-to-noise performance.",1997-12-16,"The title of the patent is nonlinear dynamic transform for correction of crankshaft acceleration having torsional oscillations and its abstract is irregularities in crankshaft velocity introduced when measuring crankshaft rotation at a section of a crankshaft in an internal combustion engine that is less damped to torsional oscillations than is another more accessible crankshaft section are corrected by performing a nonlinear transformation via a neural network to predict rotation measurements that would have been obtained at the inaccessible section from data actually collected at the accessible crankshaft section. thus, the effects of torsional oscillations in the crankshaft are substantially filtered away, resulting in crankshaft acceleration values that form the basis of a misfire detector having nearly maximum signal-to-noise performance. dated 1997-12-16"
5699449,method and apparatus for implementation of neural networks for face recognition,"a method and apparatus for implementation of neural networks for face recognition is presented. a nonlinear filter or a nonlinear joint transform correlator (jtc) employs a supervised perceptron learning algorithm in a two-layer neural network for real-time face recognition. the nonlinear filter is generally implemented electronically, while the nonlinear joint transform correlator is generally implemented optically. the system implements perception learning to train with a sequence of facial images and then classifies a distorted input image in real-time. computer simulations and optical experimental results show that the system can identify the input with the probability of error less than 3%. by using time multiplexing of the input image under investigation, that is, using more than one input image, the probability of error for classification can be reduced to zero.",1997-12-16,"The title of the patent is method and apparatus for implementation of neural networks for face recognition and its abstract is a method and apparatus for implementation of neural networks for face recognition is presented. a nonlinear filter or a nonlinear joint transform correlator (jtc) employs a supervised perceptron learning algorithm in a two-layer neural network for real-time face recognition. the nonlinear filter is generally implemented electronically, while the nonlinear joint transform correlator is generally implemented optically. the system implements perception learning to train with a sequence of facial images and then classifies a distorted input image in real-time. computer simulations and optical experimental results show that the system can identify the input with the probability of error less than 3%. by using time multiplexing of the input image under investigation, that is, using more than one input image, the probability of error for classification can be reduced to zero. dated 1997-12-16"
5699450,detector array method and apparatus for real time in situ color control in printers and copiers,"a detector array method and apparatus for real time in-situ color control in printers and copiers includes an array for performing a linear matrix transformation on color patch information generated by a printer or copier. a light sensor array detects color components from a series of color test patches and performs a predetermined matrix transformation on the color information to produce a set of control signals for feedback to the printer or copier. in another embodiment, the light sensor array is extended to produce a fully analog neural network processor which is capable of arbitrary mappings of the color information into control signals for use by the printer or copier apparatus. the system is fully programmable, adaptive and, in one embodiment, trainable using backpropagation or other techniques.",1997-12-16,"The title of the patent is detector array method and apparatus for real time in situ color control in printers and copiers and its abstract is a detector array method and apparatus for real time in-situ color control in printers and copiers includes an array for performing a linear matrix transformation on color patch information generated by a printer or copier. a light sensor array detects color components from a series of color test patches and performs a predetermined matrix transformation on the color information to produce a set of control signals for feedback to the printer or copier. in another embodiment, the light sensor array is extended to produce a fully analog neural network processor which is capable of arbitrary mappings of the color information into control signals for use by the printer or copier apparatus. the system is fully programmable, adaptive and, in one embodiment, trainable using backpropagation or other techniques. dated 1997-12-16"
5699487,artificial neural network read channel,"a magnetic read channel employs an artificial neural network for reconstruction of a recorded magnetic signal and its corresponding synchronization signal. a magnetic read head receives magnetic signals from a magnetic recording media such as a magnetic tape or disk and converts it to an electronic signal. a preamplifier receives and amplifies the electronic signal from the magnetic read head to produce an amplified electronic signal. a delay line receives the amplified electronic signal from the preamplifier, storing delayed successive representations of the received signal. an artificial neural network receives the delayed successive representations from the delay line for reconstruction of the originally recorded data signal. prior to use in an application, the artificial neural network is trained via a training method with a known training set of corresponding simultaneously generated data and clock pairs. training the network with data having such known clock (synchronization) signal enables extraction of the synchronization signal from its nonlinear properties hidden within its corresponding data.",1997-12-16,"The title of the patent is artificial neural network read channel and its abstract is a magnetic read channel employs an artificial neural network for reconstruction of a recorded magnetic signal and its corresponding synchronization signal. a magnetic read head receives magnetic signals from a magnetic recording media such as a magnetic tape or disk and converts it to an electronic signal. a preamplifier receives and amplifies the electronic signal from the magnetic read head to produce an amplified electronic signal. a delay line receives the amplified electronic signal from the preamplifier, storing delayed successive representations of the received signal. an artificial neural network receives the delayed successive representations from the delay line for reconstruction of the originally recorded data signal. prior to use in an application, the artificial neural network is trained via a training method with a known training set of corresponding simultaneously generated data and clock pairs. training the network with data having such known clock (synchronization) signal enables extraction of the synchronization signal from its nonlinear properties hidden within its corresponding data. dated 1997-12-16"
5701394,information processing apparatus having a neural network and an expert system,"disclosed is an information processing system having a neural processing unit, an expert system for executing an inference using a rule, and an information selection unit. the neural processing unit and the expert system being arranged in parallel to each other. the information selection unit selects and outputs necessary information from the information outputted from the neural processing unit and the expert system. since necessary information is selected and outputted from the information obtained from the neural processing unit and the expert system as a result of processing executed by the neural processing unit and the expert system in the parallelly different points of view, the certainty of the information to be selected is improved.",1997-12-23,"The title of the patent is information processing apparatus having a neural network and an expert system and its abstract is disclosed is an information processing system having a neural processing unit, an expert system for executing an inference using a rule, and an information selection unit. the neural processing unit and the expert system being arranged in parallel to each other. the information selection unit selects and outputs necessary information from the information outputted from the neural processing unit and the expert system. since necessary information is selected and outputted from the information obtained from the neural processing unit and the expert system as a result of processing executed by the neural processing unit and the expert system in the parallelly different points of view, the certainty of the information to be selected is improved. dated 1997-12-23"
5701397,circuit for pre-charging a free neuron circuit,"in each neuron in a neural network of a plurality of neuron circuits either in an engaged or a free state, a pre-charge circuit, that allows loading the components of an input vector (a) only into a determined free neuron circuit during a recognition phase as a potential prototype vector (b) attached to the determined neuron circuit. the pre-charge circuit is a weight memory (251) controlled by a memory control signal (rs) and the circuit generating the memory control signal. the memory control signal identifies the determined free neuron circuit. during the recognition phase, the memory control signal is active only for the determined free neuron circuit. when the neural network is a chain of neuron circuits, the determined free neuron circuit is the first free neuron in the chain. the input vector components on an input data bus (data-bus) are connected to the weight memory of all neuron circuits. the data therefrom are available in each neuron on an output data bus (ram-bus). the pre-charge circuit may further include an address counter (252) for addressing the weight memory and a register (253) to latch the data output on the output data bus. after the determined neuron circuit has been engaged, the contents of its weight memory cannot be modified. pre-charging the input vector during the recognition phase makes the engagement process more efficient and significantly reduces learning time in learning the input vector.",1997-12-23,"The title of the patent is circuit for pre-charging a free neuron circuit and its abstract is in each neuron in a neural network of a plurality of neuron circuits either in an engaged or a free state, a pre-charge circuit, that allows loading the components of an input vector (a) only into a determined free neuron circuit during a recognition phase as a potential prototype vector (b) attached to the determined neuron circuit. the pre-charge circuit is a weight memory (251) controlled by a memory control signal (rs) and the circuit generating the memory control signal. the memory control signal identifies the determined free neuron circuit. during the recognition phase, the memory control signal is active only for the determined free neuron circuit. when the neural network is a chain of neuron circuits, the determined free neuron circuit is the first free neuron in the chain. the input vector components on an input data bus (data-bus) are connected to the weight memory of all neuron circuits. the data therefrom are available in each neuron on an output data bus (ram-bus). the pre-charge circuit may further include an address counter (252) for addressing the weight memory and a register (253) to latch the data output on the output data bus. after the determined neuron circuit has been engaged, the contents of its weight memory cannot be modified. pre-charging the input vector during the recognition phase makes the engagement process more efficient and significantly reduces learning time in learning the input vector. dated 1997-12-23"
5701398,adaptive classifier having multiple subnetworks,"a neural network including a plurality of sub-nets stored in a memory array and a method of operation. each of the sub-nets includes a corresponding plurality of weights and is individually operable to classify an input vector. a computation unit, including a distance calculation unit and a math unit, is responsive to an input vector comprising input training features for determining a distance between the weights of each sub-net to each of the input features of the input vector and for determining whether the distance is within a particular region of influence. also described is a parallel process for training the sub-nets.",1997-12-23,"The title of the patent is adaptive classifier having multiple subnetworks and its abstract is a neural network including a plurality of sub-nets stored in a memory array and a method of operation. each of the sub-nets includes a corresponding plurality of weights and is individually operable to classify an input vector. a computation unit, including a distance calculation unit and a math unit, is responsive to an input vector comprising input training features for determining a distance between the weights of each sub-net to each of the input features of the input vector and for determining whether the distance is within a particular region of influence. also described is a parallel process for training the sub-nets. dated 1997-12-23"
5703959,method and device for analyzing particles,"when a mixture of different kinds of particles, such as blood, is passed through a particle detector such as a flow cytometer to measure various characteristics of each particle and the measurements of two characteristics are plotted oh a two-dimensional rectangular co-ordinate, for example, the resultant dots tend to part into groups (clusters) corresponding to the kinds of the particles. this invention relates to a method and a device for presuming informations of the position of center of gravity, variances, number of dots and likes of each cluster by learning and, based upon these informations, presuming a specific cluster or category to which each measured particle should belong from the corresponding measurements of the particle, and it intends to execute such information processing by a fuzzy clustering technique using a neural network.",1997-12-30,"The title of the patent is method and device for analyzing particles and its abstract is when a mixture of different kinds of particles, such as blood, is passed through a particle detector such as a flow cytometer to measure various characteristics of each particle and the measurements of two characteristics are plotted oh a two-dimensional rectangular co-ordinate, for example, the resultant dots tend to part into groups (clusters) corresponding to the kinds of the particles. this invention relates to a method and a device for presuming informations of the position of center of gravity, variances, number of dots and likes of each cluster by learning and, based upon these informations, presuming a specific cluster or category to which each measured particle should belong from the corresponding measurements of the particle, and it intends to execute such information processing by a fuzzy clustering technique using a neural network. dated 1997-12-30"
5704011,method and apparatus for providing multivariable nonlinear control,"a method and apparatus for training and optimizing a neural network for use in controlling multivariable nonlinear processes. the neural network can be used as a controller generating manipulated variables for directly controlling the process or as part of a controller structure generating predicted process outputs. the neural network is trained and optimized off-line with historical values of the process inputs, outputs, and their rates of change. the determination of the manipulated variables or the predicted process outputs are based on an optimum prediction time which represents the effective response time of the process output to the setpoint such that the greatest change to the process output occurs as a result of a small change made to its paired manipulated variable.",1997-12-30,"The title of the patent is method and apparatus for providing multivariable nonlinear control and its abstract is a method and apparatus for training and optimizing a neural network for use in controlling multivariable nonlinear processes. the neural network can be used as a controller generating manipulated variables for directly controlling the process or as part of a controller structure generating predicted process outputs. the neural network is trained and optimized off-line with historical values of the process inputs, outputs, and their rates of change. the determination of the manipulated variables or the predicted process outputs are based on an optimum prediction time which represents the effective response time of the process output to the setpoint such that the greatest change to the process output occurs as a result of a small change made to its paired manipulated variable. dated 1997-12-30"
5704012,adaptive resource allocation using neural networks,"in a system comprising a plurality of resources for performing useful work, a resource allocation controller function, which is customized to the particular system's available resources and configuration, dynamically allocates resources and/or alters configuration to accommodate a changing workload. preferably, the resource allocation controller is part of the computer's operating system which allocates resources of the computer system. the resource allocation controller uses a controller neural network for control, and a separate system model neural network for modelling the system and training the controller neural network. performance data is collected by the system and used to train the system model neural network. a system administrator specifies computer system performance targets which indicate the desired performance of the system. deviations in actual performance from desired performance are propagated back through the system model and ultimately to the controller neural network to create a closed loop system for resource allocation.",1997-12-30,"The title of the patent is adaptive resource allocation using neural networks and its abstract is in a system comprising a plurality of resources for performing useful work, a resource allocation controller function, which is customized to the particular system's available resources and configuration, dynamically allocates resources and/or alters configuration to accommodate a changing workload. preferably, the resource allocation controller is part of the computer's operating system which allocates resources of the computer system. the resource allocation controller uses a controller neural network for control, and a separate system model neural network for modelling the system and training the controller neural network. performance data is collected by the system and used to train the system model neural network. a system administrator specifies computer system performance targets which indicate the desired performance of the system. deviations in actual performance from desired performance are propagated back through the system model and ultimately to the controller neural network to create a closed loop system for resource allocation. dated 1997-12-30"
5704014,voltage-current conversion circuit employing mos transistor cells as synapses of neural network,"a cell of mos transistors for converting a voltage into a current for forming synapses of neural nets, in particular for converting the difference between an input voltage (v.sub.in) and a voltage (v.sub.w) for weighting the synapse into a current, realized by means of a differential stage comprising a first transistor (m1) operating as a current generator, in which a first and a second branch in parallel end, which branches respectively comprise a second (m2) and a third (m3) push-pull connected transistor, to the gate regions of which the input voltage (v.sub.in) and the voltage (v.sub.w) for weighting the synapse, and to which a fourth (m4) and a fifth (m5) transistor are respectively connected in series, in which the fourth (m4) and the fifth (m5) transistor are p-mos transistors having their gate regions short-circuited and said fourth (m4) p-mos transistor is connected as a diode, and in which the output current (i.sub.out) is drawn from the node (n) that connects said third (m3) and said fifth (m5) transistors inserted in series in said second branch of the circuit and a capacitor (c) is connected to the gate region of said third (m3) transistor to store the voltage (v.sub.w) for weighting the synapse applied to the circuit.",1997-12-30,"The title of the patent is voltage-current conversion circuit employing mos transistor cells as synapses of neural network and its abstract is a cell of mos transistors for converting a voltage into a current for forming synapses of neural nets, in particular for converting the difference between an input voltage (v.sub.in) and a voltage (v.sub.w) for weighting the synapse into a current, realized by means of a differential stage comprising a first transistor (m1) operating as a current generator, in which a first and a second branch in parallel end, which branches respectively comprise a second (m2) and a third (m3) push-pull connected transistor, to the gate regions of which the input voltage (v.sub.in) and the voltage (v.sub.w) for weighting the synapse, and to which a fourth (m4) and a fifth (m5) transistor are respectively connected in series, in which the fourth (m4) and the fifth (m5) transistor are p-mos transistors having their gate regions short-circuited and said fourth (m4) p-mos transistor is connected as a diode, and in which the output current (i.sub.out) is drawn from the node (n) that connects said third (m3) and said fifth (m5) transistors inserted in series in said second branch of the circuit and a capacitor (c) is connected to the gate region of said third (m3) transistor to store the voltage (v.sub.w) for weighting the synapse applied to the circuit. dated 1997-12-30"
5704015,"optical operation element, optical data processing circuit and photoelectric operation element","a modulating optical operation element is composed of a light receiving separation element having a pair of light receiving elements, a photoelectric operation element made of an operation element for calculating electric signals outputted from the light receiving separation element and outputting an electric signal in response to a result of this calculation, a transmission type light modulation element made of a transmission type liquid crystal sandwiched between transparent electrodes, and polarizers disposed on both sides of this light modulation element. an optical neuron yet for outputting an information beam consisting of light outputted from each optical neuron element is constituted by arranging a plurality of these elements a two-dimensional pattern. a constituent unit of an optical neural network is made by the combination of this neuron layer with a weighted information beam output means for weighting a modulated information beam outputted from the neuron layer and outputting the modulated information beam. an optical neural network which acts as optical data processing circuit is constituted by connecting a plurality of these constituent units together.",1997-12-30,"The title of the patent is optical operation element, optical data processing circuit and photoelectric operation element and its abstract is a modulating optical operation element is composed of a light receiving separation element having a pair of light receiving elements, a photoelectric operation element made of an operation element for calculating electric signals outputted from the light receiving separation element and outputting an electric signal in response to a result of this calculation, a transmission type light modulation element made of a transmission type liquid crystal sandwiched between transparent electrodes, and polarizers disposed on both sides of this light modulation element. an optical neuron yet for outputting an information beam consisting of light outputted from each optical neuron element is constituted by arranging a plurality of these elements a two-dimensional pattern. a constituent unit of an optical neural network is made by the combination of this neuron layer with a weighted information beam output means for weighting a modulated information beam outputted from the neuron layer and outputting the modulated information beam. an optical neural network which acts as optical data processing circuit is constituted by connecting a plurality of these constituent units together. dated 1997-12-30"
5704016,temporal learning neural network,"a temporal learning neural network includes a plurality of temporal learning neural processing elements and an input/output control section. each element includes a calculation device and a learning device. the calculation device includes an input memory section and a response calculation circuit. the learning device includes a learning processing circuit and a history evaluation circuit. the calculation circuit calculates a sum of a total summation value of a product of input values and connection efficacies, and an internal potential, compares the sum with a predetermined threshold value, outputs a 1 or 0 signal depending on the comparison and substitutes internal potential of a next time for the sum. the processing circuit receives an input history evaluation value when the calculation circuit has produced an output 1 signal which strengthens, weakens or leaves unchanged the connection efficacies depending on the comparison. the evaluation circuit obtains an input history value, compares the obtained input history value with the learning threshold value, generates an evaluation signal and distributes the evaluation signal to the input memory section. the input/output control section is provided with input terminals and output terminals, sends signals input from the calculation circuit and evaluation circuit to the input memory section, receives signals output from the calculation circuit and evaluation circuit, and effects communication with each of the processing elements. this process is an input temporal associative learning process.",1997-12-30,"The title of the patent is temporal learning neural network and its abstract is a temporal learning neural network includes a plurality of temporal learning neural processing elements and an input/output control section. each element includes a calculation device and a learning device. the calculation device includes an input memory section and a response calculation circuit. the learning device includes a learning processing circuit and a history evaluation circuit. the calculation circuit calculates a sum of a total summation value of a product of input values and connection efficacies, and an internal potential, compares the sum with a predetermined threshold value, outputs a 1 or 0 signal depending on the comparison and substitutes internal potential of a next time for the sum. the processing circuit receives an input history evaluation value when the calculation circuit has produced an output 1 signal which strengthens, weakens or leaves unchanged the connection efficacies depending on the comparison. the evaluation circuit obtains an input history value, compares the obtained input history value with the learning threshold value, generates an evaluation signal and distributes the evaluation signal to the input memory section. the input/output control section is provided with input terminals and output terminals, sends signals input from the calculation circuit and evaluation circuit to the input memory section, receives signals output from the calculation circuit and evaluation circuit, and effects communication with each of the processing elements. this process is an input temporal associative learning process. dated 1997-12-30"
5705956,neural network based pll,a neural network controlled phase lock loop (pll) with an artificial neural network for comparing the phase of the feedback signal from the voltage controlled oscillator (vco) with the phase of the pll reference signal. the vco feedback signal and pll reference signal are inputted to and processed by a multiple layer perception which has been trained with a model of the vco to generate the appropriate control voltage necessary to establish and maintain the desired phase and/or frequency of the vco output signal.,1998-01-06,The title of the patent is neural network based pll and its abstract is a neural network controlled phase lock loop (pll) with an artificial neural network for comparing the phase of the feedback signal from the voltage controlled oscillator (vco) with the phase of the pll reference signal. the vco feedback signal and pll reference signal are inputted to and processed by a multiple layer perception which has been trained with a model of the vco to generate the appropriate control voltage necessary to establish and maintain the desired phase and/or frequency of the vco output signal. dated 1998-01-06
5706400,fault-tolerant implementation of finite-state automata in recurrent neural networks,""" any deterministic finite-state automata (dfa) can be implemented in a sparse recurrent neural network (rnn) with second-order weights and sigmoidal discriminant functions. construction algorithms can be extended to fault-tolerant dfa implementations such that faults in an analog implementation of neurons or weights do not affect the desired network performance. the weights are replicated k times for k-1 fault tolerance. alternatively, the independent network is replicated 2k+1 times and the majority of the outputs is used for a k fault tolerance. in a further alternative solution, a single network with k.eta. neurons uses a """"n choose k""""encoding algorithm for k fault tolerance. """,1998-01-06,"The title of the patent is fault-tolerant implementation of finite-state automata in recurrent neural networks and its abstract is "" any deterministic finite-state automata (dfa) can be implemented in a sparse recurrent neural network (rnn) with second-order weights and sigmoidal discriminant functions. construction algorithms can be extended to fault-tolerant dfa implementations such that faults in an analog implementation of neurons or weights do not affect the desired network performance. the weights are replicated k times for k-1 fault tolerance. alternatively, the independent network is replicated 2k+1 times and the majority of the outputs is used for a k fault tolerance. in a further alternative solution, a single network with k.eta. neurons uses a """"n choose k""""encoding algorithm for k fault tolerance. "" dated 1998-01-06"
5706401,method for editing an input quantity for a neural network,"in a method for supplementing missing data in a time series used as an input to a neural network or for improving noise-infested data supplied to a neural network, error distribution densities for the missing values are calculated on the basis of the known measured values from the time series and their known or predetermined error distribution density, and samples are taken from this error distribution density according to the monte carlo method. these each lead to an estimated or predicted value whose average is introduced for the value to be predicted. the method can be employed for the operation as well as for the training of the neural network, and is suitable for use in all known fields of utilization of neural networks.",1998-01-06,"The title of the patent is method for editing an input quantity for a neural network and its abstract is in a method for supplementing missing data in a time series used as an input to a neural network or for improving noise-infested data supplied to a neural network, error distribution densities for the missing values are calculated on the basis of the known measured values from the time series and their known or predetermined error distribution density, and samples are taken from this error distribution density according to the monte carlo method. these each lead to an estimated or predicted value whose average is introduced for the value to be predicted. the method can be employed for the operation as well as for the training of the neural network, and is suitable for use in all known fields of utilization of neural networks. dated 1998-01-06"
5706402,blind signal processing system employing information maximization to recover unknown signals through unsupervised minimization of output redundancy,a neural network system and unsupervised learning process for separating unknown source signals from their received mixtures by solving the independent components analysis (ica) problem. the unsupervised learning procedure solves the general blind signal processing problem by maximizing joint output entropy through gradient ascent to minimize mutual information in the outputs. the neural network system can separate a multiplicity of unknown source signals from measured mixture signals where the mixture characteristics and the original source signals are both unknown. the system can be easily adapted to solve the related blind deconvolution problem that extracts an unknown source signal from the output of an unknown reverberating channel.,1998-01-06,The title of the patent is blind signal processing system employing information maximization to recover unknown signals through unsupervised minimization of output redundancy and its abstract is a neural network system and unsupervised learning process for separating unknown source signals from their received mixtures by solving the independent components analysis (ica) problem. the unsupervised learning procedure solves the general blind signal processing problem by maximizing joint output entropy through gradient ascent to minimize mutual information in the outputs. the neural network system can separate a multiplicity of unknown source signals from measured mixture signals where the mixture characteristics and the original source signals are both unknown. the system can be easily adapted to solve the related blind deconvolution problem that extracts an unknown source signal from the output of an unknown reverberating channel. dated 1998-01-06
5706403,semiconductor neural circuit device,"a semiconductor neural circuit device having a very simple circuit and a self-teaching function, by which a neural network is allowed to learn. the device comprises synapse circuits which output weighted values, and neuron circuits which execute linear addition of the output signals from the synapse circuits, and output the signal voltages of high and low levels with respect to a given threshold value v.sub.th. in the case of learning of increasing the total value z, only when v.sub.th -.epsilon.<z<v.sub.th +.alpha. with respect to two positive parameters .epsilon. and .alpha., the weighted value of predetermined synapse circuits which input signals to the neuron circuit is increased by a given positive value. conversely, in the case of learning of the decreasing the total value z, only when the v.sub.th -.alpha.<z<v.sub.th +.epsilon., the weighted value of predetermined synapse circuit which input signals to the neuron circuit is decreased by a given positive value.",1998-01-06,"The title of the patent is semiconductor neural circuit device and its abstract is a semiconductor neural circuit device having a very simple circuit and a self-teaching function, by which a neural network is allowed to learn. the device comprises synapse circuits which output weighted values, and neuron circuits which execute linear addition of the output signals from the synapse circuits, and output the signal voltages of high and low levels with respect to a given threshold value v.sub.th. in the case of learning of increasing the total value z, only when v.sub.th -.epsilon.<z<v.sub.th +.alpha. with respect to two positive parameters .epsilon. and .alpha., the weighted value of predetermined synapse circuits which input signals to the neuron circuit is increased by a given positive value. conversely, in the case of learning of the decreasing the total value z, only when the v.sub.th -.alpha.<z<v.sub.th +.epsilon., the weighted value of predetermined synapse circuit which input signals to the neuron circuit is decreased by a given positive value. dated 1998-01-06"
5706404,neural network using inhomogeneities in a medium as neurons and transmitting input signals in an unchannelled wave pattern through the medium,neural net with spatially distributed functionalities. an information processing system comprises a neural net with fully distributed neuron and synapse functionalities in a spatially inhomogeneous medium to propagate a response field from an input to an output. the response field is a reaction of the medium to a plurality of input signals and depends non-linearly on the input signals. the response field is also determined by the inhomogeneities. the value of the field at one or more particular locations is indicative of one or more output signals of the neural net.,1998-01-06,The title of the patent is neural network using inhomogeneities in a medium as neurons and transmitting input signals in an unchannelled wave pattern through the medium and its abstract is neural net with spatially distributed functionalities. an information processing system comprises a neural net with fully distributed neuron and synapse functionalities in a spatially inhomogeneous medium to propagate a response field from an input to an output. the response field is a reaction of the medium to a plurality of input signals and depends non-linearly on the input signals. the response field is also determined by the inhomogeneities. the value of the field at one or more particular locations is indicative of one or more output signals of the neural net. dated 1998-01-06
5708727,neuroprocessing service,"a neuroprocessing center executes a neuroprocessing using a neurocomputer. the neuroprocessing center is a public facility available for a user having a user terminal and executes the neuroprocessing as requested by the user. the user is given a result of the neuroprocessing. it is unnecessary for a user to have a computer implementing the neural network and anyone can participate in the profits of the neuroprocessing service. examples of neuroprocessing service achieved by the neural network are graphic pattern generating services, character recognition services, sound synthesizing services, etc. a result of the neuroprocessing is effectively used by the user.",1998-01-13,"The title of the patent is neuroprocessing service and its abstract is a neuroprocessing center executes a neuroprocessing using a neurocomputer. the neuroprocessing center is a public facility available for a user having a user terminal and executes the neuroprocessing as requested by the user. the user is given a result of the neuroprocessing. it is unnecessary for a user to have a computer implementing the neural network and anyone can participate in the profits of the neuroprocessing service. examples of neuroprocessing service achieved by the neural network are graphic pattern generating services, character recognition services, sound synthesizing services, etc. a result of the neuroprocessing is effectively used by the user. dated 1998-01-13"
5710830,method of and apparatus for segmenting foreground and background information for optical character recognition of labels employing single layer recurrent neural network,"a method and apparatus for processing a greyscale input of an image, particularly of a shipping label, into a binary output image in which foreground information is segmented from the background information and contrasts between adjacent regions having different background densities are obliterated is described. a neuron employing a 5.times.5 input neighborhood having a unique neuron activation function is shown. no explicit line process is employed. output is biased toward a particular one of the output values by employing non-linear feedback as a function of both the grey scale value for the pixel corresponding to the label site being updated and the most recent value of the label site. the otherwise strong contribution from a gradient term in the energy function is suppressed by a shunting inhibition when the shunting inhibition function detects that the pixel lies on or near a boundary between adjacent regions of differing background intensities.",1998-01-20,"The title of the patent is method of and apparatus for segmenting foreground and background information for optical character recognition of labels employing single layer recurrent neural network and its abstract is a method and apparatus for processing a greyscale input of an image, particularly of a shipping label, into a binary output image in which foreground information is segmented from the background information and contrasts between adjacent regions having different background densities are obliterated is described. a neuron employing a 5.times.5 input neighborhood having a unique neuron activation function is shown. no explicit line process is employed. output is biased toward a particular one of the output values by employing non-linear feedback as a function of both the grey scale value for the pixel corresponding to the label site being updated and the most recent value of the label site. the otherwise strong contribution from a gradient term in the energy function is suppressed by a shunting inhibition when the shunting inhibition function detects that the pixel lies on or near a boundary between adjacent regions of differing background intensities. dated 1998-01-20"
5710869,daisy chain circuit for serial connection of neuron circuits,""" each daisy chain circuit is serially connected to the two adjacent neuron circuits, so that all the neuron circuits form a chain. the daisy chain circuit distinguishes between the two possible states of the neuron circuit (engaged or free) and identifies the first free """"or ready to learn"""" neuron circuit in the chain, based on the respective values of the input (dci) and output (dco) signals of the daisy chain circuit. the ready to learn neuron circuit is the only neuron circuit of the neural network having daisy chain input and output signals complementary to each other. the daisy chain circuit includes a 1-bit register (601) controlled by a store enable signal (st) which is active at initialization or, during the learning phase when a new neuron circuit is engaged. at initialization, all the daisy registers of the chain are forced to a first logic value. the dci input of the first daisy chain circuit in the chain is connected to a second logic value, such that after initialization, it is the ready to learn neuron circuit. in the learning phase, the ready to learn neuron's 1-bit daisy register contents are set to the second logic value by the store enable signal, it is said """"engaged"""". as neurons are engaged, each subsequent neuron circuit in the chain then becomes the next ready to learn neuron circuit. """,1998-01-20,"The title of the patent is daisy chain circuit for serial connection of neuron circuits and its abstract is "" each daisy chain circuit is serially connected to the two adjacent neuron circuits, so that all the neuron circuits form a chain. the daisy chain circuit distinguishes between the two possible states of the neuron circuit (engaged or free) and identifies the first free """"or ready to learn"""" neuron circuit in the chain, based on the respective values of the input (dci) and output (dco) signals of the daisy chain circuit. the ready to learn neuron circuit is the only neuron circuit of the neural network having daisy chain input and output signals complementary to each other. the daisy chain circuit includes a 1-bit register (601) controlled by a store enable signal (st) which is active at initialization or, during the learning phase when a new neuron circuit is engaged. at initialization, all the daisy registers of the chain are forced to a first logic value. the dci input of the first daisy chain circuit in the chain is connected to a second logic value, such that after initialization, it is the ready to learn neuron circuit. in the learning phase, the ready to learn neuron's 1-bit daisy register contents are set to the second logic value by the store enable signal, it is said """"engaged"""". as neurons are engaged, each subsequent neuron circuit in the chain then becomes the next ready to learn neuron circuit. "" dated 1998-01-20"
5711843,system for indirectly monitoring and controlling a process with particular application to plasma processes,"the invention enables real-time control of a process using information regarding process properties that are indirectly related to the state of the process. a set of properties that characterize the process environment (fingerprint) is measured and used by a process results estimator to infer information regarding the state of the process and by a process condition monitor for monitoring the process to ascertain whether a particular type of condition exists. in one embodiment, optical emission spectra (oes) are used as the fingerprint. the process results estimator is sufficiently powerful to enable the process state to be inferred even when the relationship between the process environmental properties and the process state is complicated and difficult to describe with traditional mathematical models. in one embodiment, the process results estimator is embodied by a neural network. the process condition monitor can also be embodied by a neural network. because the invention does not directly measure the process state, the invention is particularly useful in situations in which it is difficult or undesirable to directly determine the state of the process during the process such as monitoring and control of a plasma process such as plasma etching.",1998-01-27,"The title of the patent is system for indirectly monitoring and controlling a process with particular application to plasma processes and its abstract is the invention enables real-time control of a process using information regarding process properties that are indirectly related to the state of the process. a set of properties that characterize the process environment (fingerprint) is measured and used by a process results estimator to infer information regarding the state of the process and by a process condition monitor for monitoring the process to ascertain whether a particular type of condition exists. in one embodiment, optical emission spectra (oes) are used as the fingerprint. the process results estimator is sufficiently powerful to enable the process state to be inferred even when the relationship between the process environmental properties and the process state is complicated and difficult to describe with traditional mathematical models. in one embodiment, the process results estimator is embodied by a neural network. the process condition monitor can also be embodied by a neural network. because the invention does not directly measure the process state, the invention is particularly useful in situations in which it is difficult or undesirable to directly determine the state of the process during the process such as monitoring and control of a plasma process such as plasma etching. dated 1998-01-27"
5712729,"artificial retina cell, artificial retina and artificial visual apparatus","an artificial retina cell effectively used to recognize a plurality of objects from an image containing them with ease and at high speed. also disclosed are an artificial retina and an artificial visual apparatus employing the same. the artificial visual apparatus includes an artificial eyeball (3) having a focusing means (2) and an artificial retina (1) including a first artificial retina cell disposed in a central visual field (1a) to detect a bright-dark boundary by optical filtering and a second artificial retina cell disposed in a peripheral visual field (1b) to detect an object position by optical filtering, and a neural network (4) for executing pattern recognition of an object on the basis of information detected by the first artificial retina cell. the apparatus further includes a means for determining an object to be recognized subsequently from information detected by the second artificial retina cell of the artificial retina (1), and a means (5) for moving the artificial eyeball (3) toward the object to be recognized. thus, a specific object in an image containing a plurality of objects of recognition is selectively recognized with ease and at high speed.",1998-01-27,"The title of the patent is artificial retina cell, artificial retina and artificial visual apparatus and its abstract is an artificial retina cell effectively used to recognize a plurality of objects from an image containing them with ease and at high speed. also disclosed are an artificial retina and an artificial visual apparatus employing the same. the artificial visual apparatus includes an artificial eyeball (3) having a focusing means (2) and an artificial retina (1) including a first artificial retina cell disposed in a central visual field (1a) to detect a bright-dark boundary by optical filtering and a second artificial retina cell disposed in a peripheral visual field (1b) to detect an object position by optical filtering, and a neural network (4) for executing pattern recognition of an object on the basis of information detected by the first artificial retina cell. the apparatus further includes a means for determining an object to be recognized subsequently from information detected by the second artificial retina cell of the artificial retina (1), and a means (5) for moving the artificial eyeball (3) toward the object to be recognized. thus, a specific object in an image containing a plurality of objects of recognition is selectively recognized with ease and at high speed. dated 1998-01-27"
5712796,method for evaluating the faulted sections and states in a power transmission line,"the faulted sections are evaluated by calculating the measuring information resulted from the faults in various positions by the previous fault simulative calculation, introducing the resulting fault simulative measuring information into a self-organizing neural network having output elements of which number being more than that of input elements to permit the self-organizing neural network to learn the classification of the simulative measuring information, preparing an evaluation rule representing the correspondent relation between the output from the classification of the stimulative measuring information and the faulted position, thereafter introducing actual measured information into the self-organizing neural network to permit the self-organizing neural network to classify the introduced actual measured information, and evaluating the faulted section from an output classified by the self-organized neural network on the basis of the evaluation rule.",1998-01-27,"The title of the patent is method for evaluating the faulted sections and states in a power transmission line and its abstract is the faulted sections are evaluated by calculating the measuring information resulted from the faults in various positions by the previous fault simulative calculation, introducing the resulting fault simulative measuring information into a self-organizing neural network having output elements of which number being more than that of input elements to permit the self-organizing neural network to learn the classification of the simulative measuring information, preparing an evaluation rule representing the correspondent relation between the output from the classification of the stimulative measuring information and the faulted position, thereafter introducing actual measured information into the self-organizing neural network to permit the self-organizing neural network to classify the introduced actual measured information, and evaluating the faulted section from an output classified by the self-organized neural network on the basis of the evaluation rule. dated 1998-01-27"
5712922,neural network optical character recognition system and method for classifying characters in a moving web,"a neural network based optical character recognition technique is presented for identifying characters in a moving web. image acquisition means defines an imaging window through which the moving web passes such that the characters printed thereon can be imaged. classification data is extracted and accumulated for each printed web character passing through the imaging window. a light source provides transmissive illumination of the web as it is being imaged. a neural network accelerator is coupled to the image acquisition means for intelligent processing of the accumulated classification data to produce therefrom printed character classification information indicative of each corresponding character imaged. a processor is coupled to the accelerator for converting the classification information into the appropriate ascii character code. the technique is particularly useful for reading dot-matrix-type characters on a noisy, semi-transparent background at fast real-time rates. a neural network algorithm based recognition method is also described.",1998-01-27,"The title of the patent is neural network optical character recognition system and method for classifying characters in a moving web and its abstract is a neural network based optical character recognition technique is presented for identifying characters in a moving web. image acquisition means defines an imaging window through which the moving web passes such that the characters printed thereon can be imaged. classification data is extracted and accumulated for each printed web character passing through the imaging window. a light source provides transmissive illumination of the web as it is being imaged. a neural network accelerator is coupled to the image acquisition means for intelligent processing of the accumulated classification data to produce therefrom printed character classification information indicative of each corresponding character imaged. a processor is coupled to the accelerator for converting the classification information into the appropriate ascii character code. the technique is particularly useful for reading dot-matrix-type characters on a noisy, semi-transparent background at fast real-time rates. a neural network algorithm based recognition method is also described. dated 1998-01-27"
5712959,neural network architecture for non-gaussian components of a mixture density function,"a neural network for classifying input vectors to an outcome class, under the assumption that the outcome classes are characterized by mixtures of component populations, with each component population having a multivariate non-gaussian likelihood distribution. the neural network comprising an input layer for receiving the components of the input vector, two hidden layers for generating an number of outcome class component values, and an output layer that identifies an outcome class in response to the outcome class component values. the first hidden layer includes a number of first layer nodes each connected receive input vector components from the input layer and generating in response a first layer output value representing a selected first layer power of the absolute value of the sum of the difference between a function of each input vector component and a threshold value. the second hidden layer includes a plurality of second layer nodes each being connected to predetermined ones of the first layer nodes and generating in response to the first layer output values an outcome class component value representing a function related to the exponential of the negative square of the sum of first layer output values. the output layer includes a plurality of output layer nodes each associated with an outcome class. each output layer node receives the output class component values from all of the second layer nodes and uses them, in combination with respective weighting values, to generate a value that represents the likelihood that the input vector is properly classified to the output layer node's outcome class.",1998-01-27,"The title of the patent is neural network architecture for non-gaussian components of a mixture density function and its abstract is a neural network for classifying input vectors to an outcome class, under the assumption that the outcome classes are characterized by mixtures of component populations, with each component population having a multivariate non-gaussian likelihood distribution. the neural network comprising an input layer for receiving the components of the input vector, two hidden layers for generating an number of outcome class component values, and an output layer that identifies an outcome class in response to the outcome class component values. the first hidden layer includes a number of first layer nodes each connected receive input vector components from the input layer and generating in response a first layer output value representing a selected first layer power of the absolute value of the sum of the difference between a function of each input vector component and a threshold value. the second hidden layer includes a plurality of second layer nodes each being connected to predetermined ones of the first layer nodes and generating in response to the first layer output values an outcome class component value representing a function related to the exponential of the negative square of the sum of first layer output values. the output layer includes a plurality of output layer nodes each associated with an outcome class. each output layer node receives the output class component values from all of the second layer nodes and uses them, in combination with respective weighting values, to generate a value that represents the likelihood that the input vector is properly classified to the output layer node's outcome class. dated 1998-01-27"
5714866,method and apparatus for fast battery charging using neural network fuzzy logic based control,a method and apparatus for providing fast charging of secondary cells in an electronic device. the charging process is under the control of a microcontroller which contains a read-only-memory (rom) in which is embedded code which determines the charging method. the charge method controls the charge provided to a battery back by a variable current source. an intelligent control scheme based on a neural network fuzzy logic methodology is used to optimize the charging current in response to measured characteristics of the battery.,1998-02-03,The title of the patent is method and apparatus for fast battery charging using neural network fuzzy logic based control and its abstract is a method and apparatus for providing fast charging of secondary cells in an electronic device. the charging process is under the control of a microcontroller which contains a read-only-memory (rom) in which is embedded code which determines the charging method. the charge method controls the charge provided to a battery back by a variable current source. an intelligent control scheme based on a neural network fuzzy logic methodology is used to optimize the charging current in response to measured characteristics of the battery. dated 1998-02-03
5714886,method of calibrating the trip point of an overload relay,a method of calibrating the selectable trip points of an overload relay wherein a neural network is used in the calibration process. the calibration method involves precisely positioning an indicia comprised of a number of graduation marks or symbols each representing a particular trip point and having a range and spacing unique to the characteristics of an overload detection circuit of a particular overload relay. the indicia is calibrated with respect to predetermined positions of a trip point indicator and fixed with respect to the predetermined positions of the trip point indicator. the range and spacing of the indicia graduations is derived from the neural networks learned trip point values and from trip point values obtained from the particular overload relay being calibrated.,1998-02-03,The title of the patent is method of calibrating the trip point of an overload relay and its abstract is a method of calibrating the selectable trip points of an overload relay wherein a neural network is used in the calibration process. the calibration method involves precisely positioning an indicia comprised of a number of graduation marks or symbols each representing a particular trip point and having a range and spacing unique to the characteristics of an overload detection circuit of a particular overload relay. the indicia is calibrated with respect to predetermined positions of a trip point indicator and fixed with respect to the predetermined positions of the trip point indicator. the range and spacing of the indicia graduations is derived from the neural networks learned trip point values and from trip point values obtained from the particular overload relay being calibrated. dated 1998-02-03
5715182,device for the classification and examination of particles in fluid,"a particle image in a sample is formed at an imaging position by an objective lens of a microscope, projected on the image picking up plane of a tv camera via a projection lens and is subjected to photo-electric conversion. image signals from the tv camera are supplied to an image memory via an a/d converter as well as to an image processing control unit. image signals outputted from the image memory are supplied to a characteristic picking out unit and there a plurality of characteristics of the particle concerned are picked out. the picked-out characteristics are supplied to the classification unit and there classification of the sediment components is perfumed via a neural network with a learning capability. accordingly, the classification unit performs provisionally an automatic classification of the objective sediment components by making use of the inputted characteristic parameters. the device allows accurate and fast automatic component particle analysis even for patient specimens containing a variety of components in high concentration.",1998-02-03,"The title of the patent is device for the classification and examination of particles in fluid and its abstract is a particle image in a sample is formed at an imaging position by an objective lens of a microscope, projected on the image picking up plane of a tv camera via a projection lens and is subjected to photo-electric conversion. image signals from the tv camera are supplied to an image memory via an a/d converter as well as to an image processing control unit. image signals outputted from the image memory are supplied to a characteristic picking out unit and there a plurality of characteristics of the particle concerned are picked out. the picked-out characteristics are supplied to the classification unit and there classification of the sediment components is perfumed via a neural network with a learning capability. accordingly, the classification unit performs provisionally an automatic classification of the objective sediment components by making use of the inputted characteristic parameters. the device allows accurate and fast automatic component particle analysis even for patient specimens containing a variety of components in high concentration. dated 1998-02-03"
5715372,method and apparatus for characterizing an input signal,"the present invention provides a method and apparatus for measuring at least one signal characteristic. initially, a set of features is selected which characterize a signal. an intelligent system, such as a neural network, is trained in the relationship between feature sets and signal characteristics. the selected feature set is then extracted from a first input signal. the extracted feature set from the first signal is input to the trained intelligent system. the intelligent system creates an output signal based on the feature set extracted from the first input signal. this output signal is then used to characterize the input signal. in one embodiment, the invention assesses voice quality, typically as expressed in mos scores, in a manner which accurately corresponds to the analysis of human evaluators. for voice signals processed by voice coders, the present invention provides a measurement technique which is independent of various voice coding algorithms and consistent for any given algorithm.",1998-02-03,"The title of the patent is method and apparatus for characterizing an input signal and its abstract is the present invention provides a method and apparatus for measuring at least one signal characteristic. initially, a set of features is selected which characterize a signal. an intelligent system, such as a neural network, is trained in the relationship between feature sets and signal characteristics. the selected feature set is then extracted from a first input signal. the extracted feature set from the first signal is input to the trained intelligent system. the intelligent system creates an output signal based on the feature set extracted from the first input signal. this output signal is then used to characterize the input signal. in one embodiment, the invention assesses voice quality, typically as expressed in mos scores, in a manner which accurately corresponds to the analysis of human evaluators. for voice signals processed by voice coders, the present invention provides a measurement technique which is independent of various voice coding algorithms and consistent for any given algorithm. dated 1998-02-03"
5715821,"neural network method and apparatus for disease, injury and bodily condition screening or sensing","a method and apparatus for disease, injury or condition screening or sensing wherein biopotentials are received from a plurality of measuring sensors located in the area of a suspected disease, injury or condition change site. these potentials are then processed and the processed values are provided to a particular type of neural network or a combination of neural networks uniquely adapted to receive and analyze data of an identifiable type to provide an indication of specific conditions.",1998-02-10,"The title of the patent is neural network method and apparatus for disease, injury and bodily condition screening or sensing and its abstract is a method and apparatus for disease, injury or condition screening or sensing wherein biopotentials are received from a plurality of measuring sensors located in the area of a suspected disease, injury or condition change site. these potentials are then processed and the processed values are provided to a particular type of neural network or a combination of neural networks uniquely adapted to receive and analyze data of an identifiable type to provide an indication of specific conditions. dated 1998-02-10"
5717832,neural semiconductor chip and neural networks incorporated therein,""" a base neural semiconductor chip (10) including a neural network or unit (11(#)). the neural network (11(#)) has a plurality of neuron circuits fed by different buses transporting data such as the input vector data, set-up parameters, and control signals. each neuron circuit (11) includes logic for generating local result signals of the """"fire"""" type (f) and a local output signal (nout) of the distance or category type on respective buses (nr-bus, nout-bus). an or circuit (12) performs an or function for all corresponding local result and output signals to generate respective first global result (r*) and output (out*) signals on respective buses (r*-bus, out*-bus) that are merged in an on-chip common communication bus (com*-bus) shared by all neuron circuits of the chip. in a multi-chip network, an additional or function is performed between all corresponding first global result and output signals (which are intermediate signals) to generate second global result (r**) and output (out**) signals, preferably by dotting onto an off-chip common communication bus (com**-bus) in the chip's driver block (19). this latter bus is shared by all the base neural network chips that are connected to it in order to incorporate a neural network of the desired size. in the chip, a multiplexer (21) may select either the intermediate output or the global output signal to be fed back to all neuron circuits of the neural network, depending on whether the chip is used in a single or multi-chip environment via a feed-back bus (or-bus). the feedback signal is the result of a collective processing of all the local output signals. """,1998-02-10,"The title of the patent is neural semiconductor chip and neural networks incorporated therein and its abstract is "" a base neural semiconductor chip (10) including a neural network or unit (11(#)). the neural network (11(#)) has a plurality of neuron circuits fed by different buses transporting data such as the input vector data, set-up parameters, and control signals. each neuron circuit (11) includes logic for generating local result signals of the """"fire"""" type (f) and a local output signal (nout) of the distance or category type on respective buses (nr-bus, nout-bus). an or circuit (12) performs an or function for all corresponding local result and output signals to generate respective first global result (r*) and output (out*) signals on respective buses (r*-bus, out*-bus) that are merged in an on-chip common communication bus (com*-bus) shared by all neuron circuits of the chip. in a multi-chip network, an additional or function is performed between all corresponding first global result and output signals (which are intermediate signals) to generate second global result (r**) and output (out**) signals, preferably by dotting onto an off-chip common communication bus (com**-bus) in the chip's driver block (19). this latter bus is shared by all the base neural network chips that are connected to it in order to incorporate a neural network of the desired size. in the chip, a multiplexer (21) may select either the intermediate output or the global output signal to be fed back to all neuron circuits of the neural network, depending on whether the chip is used in a single or multi-chip environment via a feed-back bus (or-bus). the feedback signal is the result of a collective processing of all the local output signals. "" dated 1998-02-10"
5717833,system and method for designing fixed weight analog neural networks,"a system and method for designing a fixed weight analog neural network to perform analog signal processing allows the neural network to be designed with off-line training and implemented with low precision components. a global system error is iteratively computed in accordance with initialized neural functions and weights corresponding to a desired analog neural network configuration for analog signal processing. the neural weights are selectively modified during training and then expected values of weight implementation errors are added thereto. the error adjusted neural weights are used to recompute the global system error and the result thereof is compared to a desired global system error. these steps are repeated as long as the recomputed global system error is greater than the desired global system error. following that, mosfet parameters representing mosfet channel widths and lengths are computed which correspond to the neural functions and weights. such mosfet device parameters are then used to implement the desired analog neural network configuration.",1998-02-10,"The title of the patent is system and method for designing fixed weight analog neural networks and its abstract is a system and method for designing a fixed weight analog neural network to perform analog signal processing allows the neural network to be designed with off-line training and implemented with low precision components. a global system error is iteratively computed in accordance with initialized neural functions and weights corresponding to a desired analog neural network configuration for analog signal processing. the neural weights are selectively modified during training and then expected values of weight implementation errors are added thereto. the error adjusted neural weights are used to recompute the global system error and the result thereof is compared to a desired global system error. these steps are repeated as long as the recomputed global system error is greater than the desired global system error. following that, mosfet parameters representing mosfet channel widths and lengths are computed which correspond to the neural functions and weights. such mosfet device parameters are then used to implement the desired analog neural network configuration. dated 1998-02-10"
5717834,cnn programamble topographic sensory device,""" the main design components underlying the implementation of physiologically faithful retina and other topographic sensory organ models on cellular neural network (cnn) universal chips is discussed. if the various retinas are implemented on a cnn universal chip, in a programmable way, it can be called a """"cnn bionic eye"""", a device capable of performing a broad range of image processing functions similar to those performed by biological retinas. the cnn universal machine has the special properties that it is 1) programmable and 2) includes local memory. programming is stored in analog and logical form (the analogic program) generated by an analogic programming and control unit, so the functions of the cnn universal machine can be modified as a function of complex internal and external constraints. further, several cnn bionic eyes and other topographic sensory modalities can be combined on a single cnn universal chip, and, for more complex sensory tasks, the necessary physical microsensors to provide the input signals can be implemented on the chip, in most instances. """,1998-02-10,"The title of the patent is cnn programamble topographic sensory device and its abstract is "" the main design components underlying the implementation of physiologically faithful retina and other topographic sensory organ models on cellular neural network (cnn) universal chips is discussed. if the various retinas are implemented on a cnn universal chip, in a programmable way, it can be called a """"cnn bionic eye"""", a device capable of performing a broad range of image processing functions similar to those performed by biological retinas. the cnn universal machine has the special properties that it is 1) programmable and 2) includes local memory. programming is stored in analog and logical form (the analogic program) generated by an analogic programming and control unit, so the functions of the cnn universal machine can be modified as a function of complex internal and external constraints. further, several cnn bionic eyes and other topographic sensory modalities can be combined on a single cnn universal chip, and, for more complex sensory tasks, the necessary physical microsensors to provide the input signals can be implemented on the chip, in most instances. "" dated 1998-02-10"
5719480,parametric control device,"an adaptive control system for mechanical and dynamic systems being transferred from an initial to a desired final state by control devices. the control system comprises an adaptive controller (e.g. in the form of an artificial neural network) for providing scaling parameters p, from inputs to the controller of coordinates for desired final and initial states, to a function generator which provides prototypical time functions scaled by the above parameters, those signals being provided directly via connections to the control devices which generate the required control signals. the adaptive control system may be used for a robotic manipulator having a number of rigid links interconnected by joints, the links being moveable by actuators. the manipulator's kinematics and dynamics, including joint interaction effects, are taken into account by this adaptive control system and it controls a manipulator's movements without the need for any real-time feedback circuity or any explicit calculations of inverse kinematics or inverse dynamics.",1998-02-17,"The title of the patent is parametric control device and its abstract is an adaptive control system for mechanical and dynamic systems being transferred from an initial to a desired final state by control devices. the control system comprises an adaptive controller (e.g. in the form of an artificial neural network) for providing scaling parameters p, from inputs to the controller of coordinates for desired final and initial states, to a function generator which provides prototypical time functions scaled by the above parameters, those signals being provided directly via connections to the control devices which generate the required control signals. the adaptive control system may be used for a robotic manipulator having a number of rigid links interconnected by joints, the links being moveable by actuators. the manipulator's kinematics and dynamics, including joint interaction effects, are taken into account by this adaptive control system and it controls a manipulator's movements without the need for any real-time feedback circuity or any explicit calculations of inverse kinematics or inverse dynamics. dated 1998-02-17"
5719955,data processing using neural networks having conversion tables in an intermediate layer,"in a neural network which includes one input layer, one or more intermediate layers and one output layer, neural elements in the input layer and neural elements in the intermediate layer are divided into groups. arithmetic operations representing the coupling between the neural elements of the input layer and the neural elements of the intermediate layer are put into table form.",1998-02-17,"The title of the patent is data processing using neural networks having conversion tables in an intermediate layer and its abstract is in a neural network which includes one input layer, one or more intermediate layers and one output layer, neural elements in the input layer and neural elements in the intermediate layer are divided into groups. arithmetic operations representing the coupling between the neural elements of the input layer and the neural elements of the intermediate layer are put into table form. dated 1998-02-17"
5720002,neural network and method of using same,"a neural network, which can be implemented either in hardware or software, is constructed of neurons or neuron circuits each having only one significant processing element in the form of a multiplier. the number of training examples is compared to the number of neurons in the neural network to effectuate training. the neural network utilizes a training algorithm which does not require repetitive training and which yields a global minimum to each given set of input vectors.",1998-02-17,"The title of the patent is neural network and method of using same and its abstract is a neural network, which can be implemented either in hardware or software, is constructed of neurons or neuron circuits each having only one significant processing element in the form of a multiplier. the number of training examples is compared to the number of neurons in the neural network to effectuate training. the neural network utilizes a training algorithm which does not require repetitive training and which yields a global minimum to each given set of input vectors. dated 1998-02-17"
5720004,current-mode hamming neural network,"a current-mode hamming neural network is provided with n binary inputs, and has a template matching calculation subnet and a winner-take-all subnet. the template matching calculation subnet includes m first neurons in which m exemplar templates are stored respectively. each first neuron is consisted of current mirrors connected to and controlled by the n binary inputs respectively, to generate a template matching current signal which is substantially proportional to the number of matched bits between the n binary inputs and the corresponding stored exemplar template. the winner-take-all subnet includes m second neurons, each including m transistors with their gate electrodes connected together to form a template competition node, their source electrodes connected to ground, and their drain electrodes connected to the template competition nodes respectively. the template competition nodes are coupled to and receive the template matching current signals respectively, so that the template competition node connecting with the largest template matching current signal is eventually at a relatively high voltage level, and the other template competition nodes are at a relatively low voltage level, after competition.",1998-02-17,"The title of the patent is current-mode hamming neural network and its abstract is a current-mode hamming neural network is provided with n binary inputs, and has a template matching calculation subnet and a winner-take-all subnet. the template matching calculation subnet includes m first neurons in which m exemplar templates are stored respectively. each first neuron is consisted of current mirrors connected to and controlled by the n binary inputs respectively, to generate a template matching current signal which is substantially proportional to the number of matched bits between the n binary inputs and the corresponding stored exemplar template. the winner-take-all subnet includes m second neurons, each including m transistors with their gate electrodes connected together to form a template competition node, their source electrodes connected to ground, and their drain electrodes connected to the template competition nodes respectively. the template competition nodes are coupled to and receive the template matching current signals respectively, so that the template competition node connecting with the largest template matching current signal is eventually at a relatively high voltage level, and the other template competition nodes are at a relatively low voltage level, after competition. dated 1998-02-17"
5721807,method and neural network for speech recognition using a correlogram as input,"a method and device for recognizing individual words of spoken speech can be used to control technical processes. the method proposed by the invention is based on feature extraction which is particularly efficient in terms of computing capacity and recognition rate, plus subsequent classification of the individual words using a neural network.",1998-02-24,"The title of the patent is method and neural network for speech recognition using a correlogram as input and its abstract is a method and device for recognizing individual words of spoken speech can be used to control technical processes. the method proposed by the invention is based on feature extraction which is particularly efficient in terms of computing capacity and recognition rate, plus subsequent classification of the individual words using a neural network. dated 1998-02-24"
5722109,vacuum cleaner with floor type detection means and motor power control as a function of the detected floor type,"a vacuum cleaner has: an air outlet, an air inlet, a dust chamber in communication with the air inlet, a fan driven by a main motor provided in a housing in communication with the air outlet and the dust chamber, a pressure detector which provides a signal showing variations characteristic of the floor type being cleaned; and circuitry, including a neural network for recognizing and classifying the floor type as a function of these variations, a setpoint generator for determining a pressure setpoint as a function of the variations, and a control circuit for controlling the power of the main motor so as to maintain the pressure setpoint.",1998-03-03,"The title of the patent is vacuum cleaner with floor type detection means and motor power control as a function of the detected floor type and its abstract is a vacuum cleaner has: an air outlet, an air inlet, a dust chamber in communication with the air inlet, a fan driven by a main motor provided in a housing in communication with the air outlet and the dust chamber, a pressure detector which provides a signal showing variations characteristic of the floor type being cleaned; and circuitry, including a neural network for recognizing and classifying the floor type as a function of these variations, a setpoint generator for determining a pressure setpoint as a function of the variations, and a control circuit for controlling the power of the main motor so as to maintain the pressure setpoint. dated 1998-03-03"
5722893,card dispensing shoe with scanner,""" the present invention is directed to a shoe of the type described wherein the shoe has a card scanner which scans indicia on a playing card as the card moves along and out of a chute by manual direction by the dealer in the normal fashion. the scanner can be one of several different types of devices which will sense each card as it is moved downwardly and out of the shoe. a feed forward neural-network which is trained using error back-propagation to recognize all possible card suits and card values sensed by the scanner. such a neural-network becomes a part of a scanning system which provides a proper reading of the cards to determine the progress of the play of the game including how the game might suffer if the game players are allowed to count cards using a card count system and perform other acts which would limit the profit margin of the casino. the shoe of the present invention is also provided with additional devices which make it simple and easy to record data relevant to the play of the game. for instance, the shoe has means for accommodating a """"customer-tracking-card"""" or preferred customer card which reads the personal information of a card holder from a magnetic stripe on the card and this information travels with the preferred customer from game to game, throughout a casino, which the customer likes to play. an lcd display can also be part of the shoe and this display can be used to enter and retrieve vital player information as deemed necessary or desirable to the customer file opened when the magnetic stripe reader reads the preferred customer card with the customer name and account number embedded within the cards magnetic stripe. """,1998-03-03,"The title of the patent is card dispensing shoe with scanner and its abstract is "" the present invention is directed to a shoe of the type described wherein the shoe has a card scanner which scans indicia on a playing card as the card moves along and out of a chute by manual direction by the dealer in the normal fashion. the scanner can be one of several different types of devices which will sense each card as it is moved downwardly and out of the shoe. a feed forward neural-network which is trained using error back-propagation to recognize all possible card suits and card values sensed by the scanner. such a neural-network becomes a part of a scanning system which provides a proper reading of the cards to determine the progress of the play of the game including how the game might suffer if the game players are allowed to count cards using a card count system and perform other acts which would limit the profit margin of the casino. the shoe of the present invention is also provided with additional devices which make it simple and easy to record data relevant to the play of the game. for instance, the shoe has means for accommodating a """"customer-tracking-card"""" or preferred customer card which reads the personal information of a card holder from a magnetic stripe on the card and this information travels with the preferred customer from game to game, throughout a casino, which the customer likes to play. an lcd display can also be part of the shoe and this display can be used to enter and retrieve vital player information as deemed necessary or desirable to the customer file opened when the magnetic stripe reader reads the preferred customer card with the customer name and account number embedded within the cards magnetic stripe. "" dated 1998-03-03"
5723794,photoelastic neural torque sensor,"an opto-mechanical torque sensing device suitable for use with rotary machinery integrates a photoelastic polymer detector, a light source, a photoelastic image sensor and an artificial intelligence neural network and algorithm. the photoelastic polymer is formed into a hollow cylinder and bonded to metal collars located at each end of the cylinder. the collars serve to readily place the cylinder detector over a machine shaft and affix it to the shaft using a keyway, setscrew, or spring pin. alternatively, split collars and a split sleeve can also be used to clamp the detector to the shaft. in the presence of polarized light, the photoelastic polymer detector generates an optical fringe pattern that varies as a function of torque applied to the shaft or other machine part on which the detector is mounted. the artificial intelligence neural network learns the mapping relationship between the observed optical fringe pattern and the applied torque using a training procedure. once trained, the neural network generates a signal representative of torque in the shaft based on observed fringe pattern.",1998-03-03,"The title of the patent is photoelastic neural torque sensor and its abstract is an opto-mechanical torque sensing device suitable for use with rotary machinery integrates a photoelastic polymer detector, a light source, a photoelastic image sensor and an artificial intelligence neural network and algorithm. the photoelastic polymer is formed into a hollow cylinder and bonded to metal collars located at each end of the cylinder. the collars serve to readily place the cylinder detector over a machine shaft and affix it to the shaft using a keyway, setscrew, or spring pin. alternatively, split collars and a split sleeve can also be used to clamp the detector to the shaft. in the presence of polarized light, the photoelastic polymer detector generates an optical fringe pattern that varies as a function of torque applied to the shaft or other machine part on which the detector is mounted. the artificial intelligence neural network learns the mapping relationship between the observed optical fringe pattern and the applied torque using a training procedure. once trained, the neural network generates a signal representative of torque in the shaft based on observed fringe pattern. dated 1998-03-03"
5724247,method of generating a signal indicating the direction of a short-circuit,"a process for generating a direction signal indicating the direction of a short-circuit current in a power transmission line to be monitored. to reliably produce a direction signal even in the event of a short-circuit occurring in the proximity of the control point, each derived current and voltage signal (j.sub.r (t)) is separately sampled and normalized to obtain different normalized sample values (j.sub.rn). taking the triggering reference values (s.sub.n) into account, difference values (.delta.i.sub.r (t), .delta.i.sub.r (t-1), .delta.i.sub.r (t-2), .delta.i.sub.r (t-3)) are formed. the difference values (.delta.i.sub.r (t), .delta.i.sub.r (t-1), .delta.i.sub.r (t-2), .delta.i.sub.r (t-3)) of a series (s.sub.ir) are supplied to different input neurons of a suitably trained neural network successively and simultaneously with the difference values corresponding to the same times of the other series (s.sub.is, s.sub.it, s.sub.ur, s.sub.ut). in the event of a short-circuit in one direction (forward direction), the output signal (s.sub.m) of the output neuron exceeds a predefined upper threshold value, while, in the event of a short-circuit in the reverse direction, the output signal (s.sub.m) remains below a predefined lower threshold value. the different magnitudes of the output signal (s.sub.m) are used for generating the corresponding direction signals.",1998-03-03,"The title of the patent is method of generating a signal indicating the direction of a short-circuit and its abstract is a process for generating a direction signal indicating the direction of a short-circuit current in a power transmission line to be monitored. to reliably produce a direction signal even in the event of a short-circuit occurring in the proximity of the control point, each derived current and voltage signal (j.sub.r (t)) is separately sampled and normalized to obtain different normalized sample values (j.sub.rn). taking the triggering reference values (s.sub.n) into account, difference values (.delta.i.sub.r (t), .delta.i.sub.r (t-1), .delta.i.sub.r (t-2), .delta.i.sub.r (t-3)) are formed. the difference values (.delta.i.sub.r (t), .delta.i.sub.r (t-1), .delta.i.sub.r (t-2), .delta.i.sub.r (t-3)) of a series (s.sub.ir) are supplied to different input neurons of a suitably trained neural network successively and simultaneously with the difference values corresponding to the same times of the other series (s.sub.is, s.sub.it, s.sub.ur, s.sub.ut). in the event of a short-circuit in one direction (forward direction), the output signal (s.sub.m) of the output neuron exceeds a predefined upper threshold value, while, in the event of a short-circuit in the reverse direction, the output signal (s.sub.m) remains below a predefined lower threshold value. the different magnitudes of the output signal (s.sub.m) are used for generating the corresponding direction signals. dated 1998-03-03"
5724258,neural network analysis for multifocal contact lens design,"the present invention discloses a method for optimizing multifocal lens designs using neural network analysis. more specifically, a neural network is trained using data collected in clinical evaluations of various multifocal lens designs. the trained neural network is then used to predict optimal lens designs for large populations of patients.",1998-03-03,"The title of the patent is neural network analysis for multifocal contact lens design and its abstract is the present invention discloses a method for optimizing multifocal lens designs using neural network analysis. more specifically, a neural network is trained using data collected in clinical evaluations of various multifocal lens designs. the trained neural network is then used to predict optimal lens designs for large populations of patients. dated 1998-03-03"
5724487,neural network for maximum likelihood classification with supervised and unsupervised training capability,"a neural network comprising an input layer, two hidden layers for generating an number of outcome class component values, and an output layer for classifying input vectors to an outcome class, under the assumption that the outcome classes are characterized by mixtures of component populations with each component population having a multivariate gaussian likelihood distribution. the first hidden layer includes a number of first layer nodes each connected receive input vector components from the input layer and generates in response a first layer output value representing the absolute value of the sum of a function of the difference between each input vector component and a threshold value. the second hidden layer includes a plurality of second layer nodes each for generating an outcome class component value, each second layer node being connected to predetermined ones of the first layer nodes and generating in response to the first layer output values an outcome class component value representing a function related to the exponential of the negative square of the sum of first layer output values connected thereto. the output layer includes a plurality of output layer nodes each associated with an outcome class. each output layer node uses the output class component values from the second layer nodes in combination with weighting values to generate the likelihood that the input vector is properly classified to the output layer node's outcome class.",1998-03-03,"The title of the patent is neural network for maximum likelihood classification with supervised and unsupervised training capability and its abstract is a neural network comprising an input layer, two hidden layers for generating an number of outcome class component values, and an output layer for classifying input vectors to an outcome class, under the assumption that the outcome classes are characterized by mixtures of component populations with each component population having a multivariate gaussian likelihood distribution. the first hidden layer includes a number of first layer nodes each connected receive input vector components from the input layer and generates in response a first layer output value representing the absolute value of the sum of a function of the difference between each input vector component and a threshold value. the second hidden layer includes a plurality of second layer nodes each for generating an outcome class component value, each second layer node being connected to predetermined ones of the first layer nodes and generating in response to the first layer output values an outcome class component value representing a function related to the exponential of the negative square of the sum of first layer output values connected thereto. the output layer includes a plurality of output layer nodes each associated with an outcome class. each output layer node uses the output class component values from the second layer nodes in combination with weighting values to generate the likelihood that the input vector is properly classified to the output layer node's outcome class. dated 1998-03-03"
5726847,method of generating a protection-triggering signal,"a method of generating a protection-triggering signal using a triggering device of a selective-protective arrangement for an electrical power network to be monitored. in order to produce a protection-triggering signal in a relatively short time after the occurrence of a fault in the power network, a triggering device is provided in which a neural network is associated with each phase conductor in the power network. successively sampled normalized values of the current in the power network are applied at the same time to the various neurons in the input layer of each neural network and a subsequently sampled normalized value (comparison value) of the current compared with the signal from the output neuron. if the normalized comparison value exceeds the output from the output neuron, a protection-triggering signal is generated.",1998-03-10,"The title of the patent is method of generating a protection-triggering signal and its abstract is a method of generating a protection-triggering signal using a triggering device of a selective-protective arrangement for an electrical power network to be monitored. in order to produce a protection-triggering signal in a relatively short time after the occurrence of a fault in the power network, a triggering device is provided in which a neural network is associated with each phase conductor in the power network. successively sampled normalized values of the current in the power network are applied at the same time to the various neurons in the input layer of each neural network and a subsequently sampled normalized value (comparison value) of the current compared with the signal from the output neuron. if the normalized comparison value exceeds the output from the output neuron, a protection-triggering signal is generated. dated 1998-03-10"
5727128,system and method for automatically determining a set of variables for use in creating a process model,"a process modeling system and method develop a set of process model inputs for a process model, such as a neural network, from values for a number of process input variables and at least one process output variable. the system and method first determine a correlation measurement between each of the process input variables and the process output variable and select a set of potential model input variables based on the correlation measurements. the system and method then iteratively determine a succession of sets of potential model input variables by performing a regression analysis on the selected set of potential model input variables and the model output variable and by then refining the set of potential model input variables based on the result of the regression analysis and on the correlation measurements. after a number of iterations, the system and method choose a set of potential model input variables as the set of model inputs and develop a process model from the chosen set of model inputs.",1998-03-10,"The title of the patent is system and method for automatically determining a set of variables for use in creating a process model and its abstract is a process modeling system and method develop a set of process model inputs for a process model, such as a neural network, from values for a number of process input variables and at least one process output variable. the system and method first determine a correlation measurement between each of the process input variables and the process output variable and select a set of potential model input variables based on the correlation measurements. the system and method then iteratively determine a succession of sets of potential model input variables by performing a regression analysis on the selected set of potential model input variables and the model output variable and by then refining the set of potential model input variables based on the result of the regression analysis and on the correlation measurements. after a number of iterations, the system and method choose a set of potential model input variables as the set of model inputs and develop a process model from the chosen set of model inputs. dated 1998-03-10"
5727131,neural network learning device,"a learning device that affects only the input-output relationships that should be additionally learned. a learning nn unit 8 capable of executing additional learning is provided separately from a learned nn unit 4 which is a basic control unit. the learned nn unit 4 produces a basic output in response to an input signal from a signal input unit 14, the learning nn unit 8 produces a correction amount desired by an individual person, and a desired control is performed based on the total value. when the output is changed, a difference is calculated between the changed output value and the basic output value from a first output unit 15, and the learning nn unit 8 executes the additional learning based upon the difference and the input value at this moment in compliance with a back-propagation method.",1998-03-10,"The title of the patent is neural network learning device and its abstract is a learning device that affects only the input-output relationships that should be additionally learned. a learning nn unit 8 capable of executing additional learning is provided separately from a learned nn unit 4 which is a basic control unit. the learned nn unit 4 produces a basic output in response to an input signal from a signal input unit 14, the learning nn unit 8 produces a correction amount desired by an individual person, and a desired control is performed based on the total value. when the output is changed, a difference is calculated between the changed output value and the basic output value from a first output unit 15, and the learning nn unit 8 executes the additional learning based upon the difference and the input value at this moment in compliance with a back-propagation method. dated 1998-03-10"
5729623,pattern recognition apparatus and method of optimizing mask for pattern recognition according to genetic algorithm,"a bill-recognition apparatus includes a neural network having a learning capability and performs high-efficiency pattern recognition of seven kinds of u.s. dollar bills. pattern image data optically inputted through a sensor is compressed using plurality of column masks, and then a plurality of values representative of images (slab values) are determined. the image data is divided into a large number of strip-shaped segments, and some of theses segments are masked with column areas of masks. the values representative of images compressed through column masks are not influenced by a slight inclination of the pattern image during the reading operation. these values representative of images are inputted to a separation processing unit (neural network). from these values, the separation processing unit calculates separation values corresponding to respective decision patterns associated with pattern images, using weights which have been adjusted to optimum values for respective decision patterns. a correct pattern image is determined from the maximum value of the separation values. the above arrangement allows for a reduction in scale of the neural network and control system. furthermore, bill recognition may also be achieved by separation processing using a plurality of small-scaled neural networks connected in cascade, or replacing weight functions in the same neural network and performing separation processing a plurality of times for the same slab values (cascade processing). in this way, it is possible to reduce the scale of the neural network and the control system.",1998-03-17,"The title of the patent is pattern recognition apparatus and method of optimizing mask for pattern recognition according to genetic algorithm and its abstract is a bill-recognition apparatus includes a neural network having a learning capability and performs high-efficiency pattern recognition of seven kinds of u.s. dollar bills. pattern image data optically inputted through a sensor is compressed using plurality of column masks, and then a plurality of values representative of images (slab values) are determined. the image data is divided into a large number of strip-shaped segments, and some of theses segments are masked with column areas of masks. the values representative of images compressed through column masks are not influenced by a slight inclination of the pattern image during the reading operation. these values representative of images are inputted to a separation processing unit (neural network). from these values, the separation processing unit calculates separation values corresponding to respective decision patterns associated with pattern images, using weights which have been adjusted to optimum values for respective decision patterns. a correct pattern image is determined from the maximum value of the separation values. the above arrangement allows for a reduction in scale of the neural network and control system. furthermore, bill recognition may also be achieved by separation processing using a plurality of small-scaled neural networks connected in cascade, or replacing weight functions in the same neural network and performing separation processing a plurality of times for the same slab values (cascade processing). in this way, it is possible to reduce the scale of the neural network and the control system. dated 1998-03-17"
5729660,3-d inverse scattering by artificial intelligence : apparatus and method,""" an unknown object is non-destructively and quantitatively evaluated for three-dimensional spatial distribution of a set of material constitutive parameters of the unknown object, using a multi-element array-source transducer and a multi-element array-detector transducer located near the unknown object. the array-source transducer exposes the array-detector transducer to a set of source-field patterns pursuant to a set of electrical input signals. an unknown object located near these transducers will be the cause of scattering, thus presenting a scattered-field pattern to the array detector transducer, for each pattern of the set of source-field patterns. in a related computation, a set of training signals is determined by evaluating on a computer the scattered field from a set of computer simulated training objects. a computer, a signal processor and a neural network operate from detector response to the computer simulated and unknown object scattered-field patterns, in each of two modes. in an initial mode, the neural network is """"trained"""" or configured to process a set of transfer functions involved in array-detector response to scattered-field patterns evaluated by computer simulations for the known computer simulated objects; in another mode, the neural network utilizes its """"trained"""" configuration in application to a set of transfer functions involved in array-detector response to scattered-field patterns produced by an unknown object, to generate estimates of the three-dimensional spatial distribution of the material constitutive parameters of the unknown object. in another embodiment, a set of the biot poro-elastic material parameters of an unknown object is estimated. """,1998-03-17,"The title of the patent is 3-d inverse scattering by artificial intelligence : apparatus and method and its abstract is "" an unknown object is non-destructively and quantitatively evaluated for three-dimensional spatial distribution of a set of material constitutive parameters of the unknown object, using a multi-element array-source transducer and a multi-element array-detector transducer located near the unknown object. the array-source transducer exposes the array-detector transducer to a set of source-field patterns pursuant to a set of electrical input signals. an unknown object located near these transducers will be the cause of scattering, thus presenting a scattered-field pattern to the array detector transducer, for each pattern of the set of source-field patterns. in a related computation, a set of training signals is determined by evaluating on a computer the scattered field from a set of computer simulated training objects. a computer, a signal processor and a neural network operate from detector response to the computer simulated and unknown object scattered-field patterns, in each of two modes. in an initial mode, the neural network is """"trained"""" or configured to process a set of transfer functions involved in array-detector response to scattered-field patterns evaluated by computer simulations for the known computer simulated objects; in another mode, the neural network utilizes its """"trained"""" configuration in application to a set of transfer functions involved in array-detector response to scattered-field patterns produced by an unknown object, to generate estimates of the three-dimensional spatial distribution of the material constitutive parameters of the unknown object. in another embodiment, a set of the biot poro-elastic material parameters of an unknown object is estimated. "" dated 1998-03-17"
5729661,method and apparatus for preprocessing input data to a neural network,"a preprocessing system for preprocessing input data to a neural network includes a training system for training a model (20) on data from a data file (10). the data is first preprocessed in a preprocessor (12) to fill in bad or missing data and merge all the time values on a common time scale. the preprocess operation utilizes preprocessing algorithms and time merging algorithms which are stored in a storage area (14). the output of the preprocessor (12) is then delayed in a delay block (16) in accordance with delay settings in storage area (18). these delayed outputs are then utilized to train the model (20), the model parameter is then stored in a storage area (22) during run time, a distributed control system (24) outputs the data to a preprocess block (34) and then preprocesses data in accordance with the algorithms in storage area (14). these outputs are then delayed in accordance with a delay block (36) with the delay settings (18). the output of the delay block (36) comprises inputs to a run time system model (26) which is built to provide a representation of the system in accordance with the model parameters in the storage area (22). a predicted control output or predicted control inputs are then generated. the control input is input back to the dcs (24).",1998-03-17,"The title of the patent is method and apparatus for preprocessing input data to a neural network and its abstract is a preprocessing system for preprocessing input data to a neural network includes a training system for training a model (20) on data from a data file (10). the data is first preprocessed in a preprocessor (12) to fill in bad or missing data and merge all the time values on a common time scale. the preprocess operation utilizes preprocessing algorithms and time merging algorithms which are stored in a storage area (14). the output of the preprocessor (12) is then delayed in a delay block (16) in accordance with delay settings in storage area (18). these delayed outputs are then utilized to train the model (20), the model parameter is then stored in a storage area (22) during run time, a distributed control system (24) outputs the data to a preprocess block (34) and then preprocesses data in accordance with the algorithms in storage area (14). these outputs are then delayed in accordance with a delay block (36) with the delay settings (18). the output of the delay block (36) comprises inputs to a run time system model (26) which is built to provide a representation of the system in accordance with the model parameters in the storage area (22). a predicted control output or predicted control inputs are then generated. the control input is input back to the dcs (24). dated 1998-03-17"
5729662,neural network for classification of patterns with improved method and apparatus for ordering vectors,"a type of neural network called a self-organizing map (som) is useful in pattern classification. the ability of the som to map the density of the input distribution is improved with two techniques. in the first technique, the som is improved by monitoring the frequency for which each node is the winning node, and splitting frequently winning nodes into two nodes, while eliminating infrequently winning nodes. topological order is preserved by inserting a link between the preceding and following nodes so that such preceding and following nodes are now adjacent in the output index space. in the second technique, the som is trained by applying a weight correction to each node based on the frequencies of that node and its neighbors. if any of the adjacent nodes have a frequency greater than the frequency of the present node, then the weight vector of the present node is adjusted towards the highest-frequency neighboring node. the topological order of the nodes is preserved because the weight vector is moved along a line of connection from the present node to the highest-frequency adjacent node. this second technique is suitable for mapping to an index space of any dimension, while the first technique is practical only for a one-dimensional output space.",1998-03-17,"The title of the patent is neural network for classification of patterns with improved method and apparatus for ordering vectors and its abstract is a type of neural network called a self-organizing map (som) is useful in pattern classification. the ability of the som to map the density of the input distribution is improved with two techniques. in the first technique, the som is improved by monitoring the frequency for which each node is the winning node, and splitting frequently winning nodes into two nodes, while eliminating infrequently winning nodes. topological order is preserved by inserting a link between the preceding and following nodes so that such preceding and following nodes are now adjacent in the output index space. in the second technique, the som is trained by applying a weight correction to each node based on the frequencies of that node and its neighbors. if any of the adjacent nodes have a frequency greater than the frequency of the present node, then the weight vector of the present node is adjusted towards the highest-frequency neighboring node. the topological order of the nodes is preserved because the weight vector is moved along a line of connection from the present node to the highest-frequency adjacent node. this second technique is suitable for mapping to an index space of any dimension, while the first technique is practical only for a one-dimensional output space. dated 1998-03-17"
5732288,auto-focusing device for camera,"an auto-focusing device for a camera according to the present invention comprises many different systems. included is a focus detection section which intermittently calculates focus detection information corresponding to the distance to the photographic subject. also included is a photographic subject position prediction section which predicts a future position of the photographic subject based on the focus detection information. finally, a lens driving section then drives a photographic lens based on a predicted result of the photographic subject position predicting section. the photographic subject position predicting section includes a neural network that predicts the future position of the photographic subject with an input parameter that has values regarding focusing positions of the photographic lens. these values correspond to focus detection data calculated by the focus detection section. the neural network makes it possible to predict the future focusing position of the photographic lens accurately.",1998-03-24,"The title of the patent is auto-focusing device for camera and its abstract is an auto-focusing device for a camera according to the present invention comprises many different systems. included is a focus detection section which intermittently calculates focus detection information corresponding to the distance to the photographic subject. also included is a photographic subject position prediction section which predicts a future position of the photographic subject based on the focus detection information. finally, a lens driving section then drives a photographic lens based on a predicted result of the photographic subject position predicting section. the photographic subject position predicting section includes a neural network that predicts the future position of the photographic subject with an input parameter that has values regarding focusing positions of the photographic lens. these values correspond to focus detection data calculated by the focus detection section. the neural network makes it possible to predict the future focusing position of the photographic lens accurately. dated 1998-03-24"
5732382,method for identifying misfire events of an internal combustion engine,"a method for identifying engine combustion failure of an internal combustion engine having a plurality of cylinders, a crankshaft and a crankshaft position sensor includes the steps of operating the internal combustion engine to rotate the crankshaft, measuring rotational quantities of the crankshaft corresponding to events created by each of the plurality of cylinders during operation of the internal combustion engine, correcting the rotational quantities measured to remove periodic position irregularities to generate a corrected temporal signal, generating an acceleration signal of the crankshaft using the corrected temporal signals, and identifying combustion failures as a function of the acceleration signal. a time-lagged recurrent neural network utilizes the acceleration signal, along with other engine parameters to identify the cylinder-specific misfire events.",1998-03-24,"The title of the patent is method for identifying misfire events of an internal combustion engine and its abstract is a method for identifying engine combustion failure of an internal combustion engine having a plurality of cylinders, a crankshaft and a crankshaft position sensor includes the steps of operating the internal combustion engine to rotate the crankshaft, measuring rotational quantities of the crankshaft corresponding to events created by each of the plurality of cylinders during operation of the internal combustion engine, correcting the rotational quantities measured to remove periodic position irregularities to generate a corrected temporal signal, generating an acceleration signal of the crankshaft using the corrected temporal signals, and identifying combustion failures as a function of the acceleration signal. a time-lagged recurrent neural network utilizes the acceleration signal, along with other engine parameters to identify the cylinder-specific misfire events. dated 1998-03-24"
5732697,shift-invariant artificial neural network for computerized detection of clustered microcalcifications in mammography,a computerized method and system using a shift-invariant artificial neural network (siann) for the quantitative analysis of image data. a series of digitized medical images are used to train an artificial neural network to differentiate between diseased and normal tissue. the sum of the weights in groups between layers is constrained to be substantially zero so as to avoid saturation of layers which would otherwise be saturated by low frequency background noise. the method and system also include utilizing training-free zones to exclude from training the center portions of microcalcifications in the digitized images. the method and system further include rule-based selection criteria for providing a more accurate diagnosis.,1998-03-31,The title of the patent is shift-invariant artificial neural network for computerized detection of clustered microcalcifications in mammography and its abstract is a computerized method and system using a shift-invariant artificial neural network (siann) for the quantitative analysis of image data. a series of digitized medical images are used to train an artificial neural network to differentiate between diseased and normal tissue. the sum of the weights in groups between layers is constrained to be substantially zero so as to avoid saturation of layers which would otherwise be saturated by low frequency background noise. the method and system also include utilizing training-free zones to exclude from training the center portions of microcalcifications in the digitized images. the method and system further include rule-based selection criteria for providing a more accurate diagnosis. dated 1998-03-31
5734319,method of determining the inflation pressure of a tire on a moving vehicle,"a method of determining the inflation pressure of one or more pneumatic tires on a moving vehicle having a plurality of n wheels fitted with tires by performing a set-up procedure comprising, for a range of tire inflation set-up procedure comprising, for a range of tire inflation pressures including all the vehicle tires at their scheduled inflation pressure and combinations of one or more tires at a range of pressures below their scheduled and for a range of vehicle speeds and driving conditions such as accelerating, braking, straight-ahead driving and cornering, deriving for each of the wheels a speed value cn proportion to the wheel angular velocity, determining and saving the relationship of the set of set-up speed values to each related tire pressure, and subsequently in normal driving monitoring at intervals of time the wheel speed signals of the wheels on the vehicle, deriving an equivalent set of normal driving speed values based on these monitored wheel speed signals, obtaining the associated tire inflation pressures. the method may utilize a trainable information processing system such as a neural network to map the speed values onto the set of tire inflation pressures. the method may particularly be applied to detecting a deflated tire on a vehicle.",1998-03-31,"The title of the patent is method of determining the inflation pressure of a tire on a moving vehicle and its abstract is a method of determining the inflation pressure of one or more pneumatic tires on a moving vehicle having a plurality of n wheels fitted with tires by performing a set-up procedure comprising, for a range of tire inflation set-up procedure comprising, for a range of tire inflation pressures including all the vehicle tires at their scheduled inflation pressure and combinations of one or more tires at a range of pressures below their scheduled and for a range of vehicle speeds and driving conditions such as accelerating, braking, straight-ahead driving and cornering, deriving for each of the wheels a speed value cn proportion to the wheel angular velocity, determining and saving the relationship of the set of set-up speed values to each related tire pressure, and subsequently in normal driving monitoring at intervals of time the wheel speed signals of the wheels on the vehicle, deriving an equivalent set of normal driving speed values based on these monitored wheel speed signals, obtaining the associated tire inflation pressures. the method may utilize a trainable information processing system such as a neural network to map the speed values onto the set of tire inflation pressures. the method may particularly be applied to detecting a deflated tire on a vehicle. dated 1998-03-31"
5734575,method and apparatus for detecting high-impedance faults in electrical power systems,"the present invention features a method and apparatus for detecting and enabling the clearance of high impedance faults (hifs) in an electrical transmission or distribution system. current in at least one phase in a distribution system is monitored in real time by sensors. analog current signature information is then digitized for processing by a digital computer. zero crossings are identified and current maxima and minima located. the first derivatives of the maxima and minima are computed and a modified fast fourier transform (fft) is then performed to convert time domain to frequency domain information. the transformed data is formatted and normalized and then applied to a trained neural network, which provides an output trigger signal when an hif condition is probable. the trigger signal is made available to either a network administrator for manual intervention, or directly to switchgear to deactivate an affected portion of the network. the inventive method may be practiced using either conventional computer hardware and software or dedicated custom hardware such as a vlsi chip.",1998-03-31,"The title of the patent is method and apparatus for detecting high-impedance faults in electrical power systems and its abstract is the present invention features a method and apparatus for detecting and enabling the clearance of high impedance faults (hifs) in an electrical transmission or distribution system. current in at least one phase in a distribution system is monitored in real time by sensors. analog current signature information is then digitized for processing by a digital computer. zero crossings are identified and current maxima and minima located. the first derivatives of the maxima and minima are computed and a modified fast fourier transform (fft) is then performed to convert time domain to frequency domain information. the transformed data is formatted and normalized and then applied to a trained neural network, which provides an output trigger signal when an hif condition is probable. the trigger signal is made available to either a network administrator for manual intervention, or directly to switchgear to deactivate an affected portion of the network. the inventive method may be practiced using either conventional computer hardware and software or dedicated custom hardware such as a vlsi chip. dated 1998-03-31"
5734797,system and method for determining class discrimination features,a system for generating a minimal artificial neural network (ann) architecture having as inputs the minimal number of features necessary to discriminate between event classes. a network generator generates an initial ann architecture. a training processor generates a trained ann by training the initial ann to a desired degree of accuracy. a pruning processor prunes the trained ann by removing interconnections and nodes from the trained ann. the training processor and pruning processor continue to train and prune the ann until a minimal network architecture having the class discrimination features as its only inputs is obtained.,1998-03-31,The title of the patent is system and method for determining class discrimination features and its abstract is a system for generating a minimal artificial neural network (ann) architecture having as inputs the minimal number of features necessary to discriminate between event classes. a network generator generates an initial ann architecture. a training processor generates a trained ann by training the initial ann to a desired degree of accuracy. a pruning processor prunes the trained ann by removing interconnections and nodes from the trained ann. the training processor and pruning processor continue to train and prune the ann until a minimal network architecture having the class discrimination features as its only inputs is obtained. dated 1998-03-31
5737485,method and apparatus including microphone arrays and neural networks for speech/speaker recognition systems,"a neural network is trained to transform distant-talking cepstrum coefficients, derived from a microphone array receiving speech from a speaker distant therefrom, into a form substantially similar to close-talking cepstrum coefficients that would be derived from a microphone close to the speaker, for providing robust hands-free speech and speaker recognition in adverse practical environments with existing speech and speaker recognition systems which have been trained on close-talking speech.",1998-04-07,"The title of the patent is method and apparatus including microphone arrays and neural networks for speech/speaker recognition systems and its abstract is a neural network is trained to transform distant-talking cepstrum coefficients, derived from a microphone array receiving speech from a speaker distant therefrom, into a form substantially similar to close-talking cepstrum coefficients that would be derived from a microphone close to the speaker, for providing robust hands-free speech and speaker recognition in adverse practical environments with existing speech and speaker recognition systems which have been trained on close-talking speech. dated 1998-04-07"
5737496,active neural network control of wafer attributes in a plasma etch process,"the present invention is predicated upon the fact that an emission trace from a plasma glow used in fabricating integrated circuits contains information about phenoma which cause variations in the fabrication process such as age of the plasma reactor, densities of the wafers exposed to the plasma, chemistry of the plasma, and concentration of the remaining material. in accordance with the present invention, a method for using neural networks to determine plasma etch end-point times in an integrated circuit fabrication process is disclosed. the end-point time is based on in-situ monitoring of the optical emission trace. the back-propagation method is used to train the network. more generally, a neural network can be used to regulate control variables and materials in a manufacturing process to yield an output product with desired quality attributes. an identified process signature which reflects the relation between the quality attribute and the process may be used to train the neural network.",1998-04-07,"The title of the patent is active neural network control of wafer attributes in a plasma etch process and its abstract is the present invention is predicated upon the fact that an emission trace from a plasma glow used in fabricating integrated circuits contains information about phenoma which cause variations in the fabrication process such as age of the plasma reactor, densities of the wafers exposed to the plasma, chemistry of the plasma, and concentration of the remaining material. in accordance with the present invention, a method for using neural networks to determine plasma etch end-point times in an integrated circuit fabrication process is disclosed. the end-point time is based on in-situ monitoring of the optical emission trace. the back-propagation method is used to train the network. more generally, a neural network can be used to regulate control variables and materials in a manufacturing process to yield an output product with desired quality attributes. an identified process signature which reflects the relation between the quality attribute and the process may be used to train the neural network. dated 1998-04-07"
5737716,method and apparatus for encoding speech using neural network technology for speech classification,"a low-rate voice coding method and apparatus uses vocoder-embedded neural network techniques. a neural network controlled speech analysis processor includes a neural network which manages speech characterization, encoding , decoding, and reconstruction methodologies. the voice coding method and apparatus uses multi-layer perceptron (mlp) based neural network structures in single or multi-stage arrangements.",1998-04-07,"The title of the patent is method and apparatus for encoding speech using neural network technology for speech classification and its abstract is a low-rate voice coding method and apparatus uses vocoder-embedded neural network techniques. a neural network controlled speech analysis processor includes a neural network which manages speech characterization, encoding , decoding, and reconstruction methodologies. the voice coding method and apparatus uses multi-layer perceptron (mlp) based neural network structures in single or multi-stage arrangements. dated 1998-04-07"
5740274,method for recognizing object images and learning method for neural networks,"a method for recognizing an object image comprises the steps of extracting a candidate for mineda predetermined object image from an overall image, and making a judgment as to whether the extracted candidate for the predetermined object image is or is not the predetermined object image. the candidate for the predetermined object image is extracted by causing the center point of a view window, which has a predetermined size, to travel to the position of the candidate for the predetermined object image, and determining an extraction area in accordance with the size and/or the shape of the candidate for the predetermined object image, the center point of the view window being taken as a reference during the determination of the extraction area. a learning method for a neural network comprises the steps of extracting a target object image, for which learning operations are to be carried out, from an image, feeding a signal, which represents the extracted target object image, into a neural network, and carrying out the learning operations of the neural network in accordance with the input target object image.",1998-04-14,"The title of the patent is method for recognizing object images and learning method for neural networks and its abstract is a method for recognizing an object image comprises the steps of extracting a candidate for mineda predetermined object image from an overall image, and making a judgment as to whether the extracted candidate for the predetermined object image is or is not the predetermined object image. the candidate for the predetermined object image is extracted by causing the center point of a view window, which has a predetermined size, to travel to the position of the candidate for the predetermined object image, and determining an extraction area in accordance with the size and/or the shape of the candidate for the predetermined object image, the center point of the view window being taken as a reference during the determination of the extraction area. a learning method for a neural network comprises the steps of extracting a target object image, for which learning operations are to be carried out, from an image, feeding a signal, which represents the extracted target object image, into a neural network, and carrying out the learning operations of the neural network in accordance with the input target object image. dated 1998-04-14"
5740322,fuzzy-neural network system,"a fuzzy neural network system includes a learning function. the learning is performed by determining degrees of coincidence of rules from combinations of membership functions for realizing fuzzy rules, constructing a network based on the number of input and output items, in such a manner as to produce output in conformity with the degrees of coincidence, to thereby properly simulate relations between input and output as to sample data represented by a subject input pattern and an output pattern corresponding thereto. the system further includes a fuzzy rule setup portion for programming fuzzy rules created by engineers into the fuzzy neural network, a fuzzy rule extracting portion for extracting each fuzzy rule from the network after a learning period, and a degree-of-importance extracting portion which extracts, for each input, a respective contribution ratio of each input on each output in the network, after a learning period. in a preferred embodiment, the network structure includes an input layer, a membership layer having front and rear halves, a rule layer, and an output layer.",1998-04-14,"The title of the patent is fuzzy-neural network system and its abstract is a fuzzy neural network system includes a learning function. the learning is performed by determining degrees of coincidence of rules from combinations of membership functions for realizing fuzzy rules, constructing a network based on the number of input and output items, in such a manner as to produce output in conformity with the degrees of coincidence, to thereby properly simulate relations between input and output as to sample data represented by a subject input pattern and an output pattern corresponding thereto. the system further includes a fuzzy rule setup portion for programming fuzzy rules created by engineers into the fuzzy neural network, a fuzzy rule extracting portion for extracting each fuzzy rule from the network after a learning period, and a degree-of-importance extracting portion which extracts, for each input, a respective contribution ratio of each input on each output in the network, after a learning period. in a preferred embodiment, the network structure includes an input layer, a membership layer having front and rear halves, a rule layer, and an output layer. dated 1998-04-14"
5740324,method for process system identification using neural network,"the method of making the tool, for process system identification that is based on the general purpose learning capabilities of neural networks. the method can be used for a wide variety of system identification problems with little or no analytic effort. a neural network is trained using a process model to approximate a function which relates process input and output data to process parameter values. once trained, the network can be used as a system identification tool. in principle, this approach can be used for linear or nonlinear processes, for open or closed loop identification, and for identifying any or all process parameters.",1998-04-14,"The title of the patent is method for process system identification using neural network and its abstract is the method of making the tool, for process system identification that is based on the general purpose learning capabilities of neural networks. the method can be used for a wide variety of system identification problems with little or no analytic effort. a neural network is trained using a process model to approximate a function which relates process input and output data to process parameter values. once trained, the network can be used as a system identification tool. in principle, this approach can be used for linear or nonlinear processes, for open or closed loop identification, and for identifying any or all process parameters. dated 1998-04-14"
5740325,computer system having a polynomial co-processor,"a computing device, which may be implemented as an integrated circuit, is constructed of a microprocessor and one or more neural network co-processors. the microprocessor normally executes programs which transfer data to the neural network co-processors, which are used to compute complicated mathematical functions. direct memory access (dma) is also used to transfer data. each neural network co-processor interfaces to the microprocessor in a manner substantially similar to that of a conventional memory device. the co-processor does not require any instructions and is configured to execute mathematical operations simply by being pre-loaded with gating functions and weight values. in addition, the co-processor executes a plurality of arithmetic operations in parallel, and the results of such operations are simply read from the co-processor.",1998-04-14,"The title of the patent is computer system having a polynomial co-processor and its abstract is a computing device, which may be implemented as an integrated circuit, is constructed of a microprocessor and one or more neural network co-processors. the microprocessor normally executes programs which transfer data to the neural network co-processors, which are used to compute complicated mathematical functions. direct memory access (dma) is also used to transfer data. each neural network co-processor interfaces to the microprocessor in a manner substantially similar to that of a conventional memory device. the co-processor does not require any instructions and is configured to execute mathematical operations simply by being pre-loaded with gating functions and weight values. in addition, the co-processor executes a plurality of arithmetic operations in parallel, and the results of such operations are simply read from the co-processor. dated 1998-04-14"
5740326,circuit for searching/sorting data in neural networks,"in a neural network of n neuron circuits, having an engaged neuron's calculated p bit wide distance between an input vector and a prototype vector and stored in the weight memory thereof, an aggregate search/sort circuit (517) of n engaged neurons' search/sort circuits. the aggregate search/sort circuit determines the minimum distance among the calculated distances. each search/sort circuit (502-1) has p elementary search/sort units connected in series to form a column, such that the aggregate circuit is a matrix of elementary search/sort units. the distance bit signals of the same bit rank are applied to search/sort units in each row. a feedback signal is generated by oring in an or gate (12.1) all local search/sort output signals from the elementary search/sort units of the same row. the search process is based on identifying zeroes in the distance bit signals, from the msb's to the lsb's. as a zero is found in a row, all the columns with a one in that row are excluded from the subsequent row search. the search process continues until only one distance, the minimum distance, remains and is available at the output of the or circuit. the above described search/sort circuit may further include a latch allowing the aggregate circuit to sort remaining distances in increasing order.",1998-04-14,"The title of the patent is circuit for searching/sorting data in neural networks and its abstract is in a neural network of n neuron circuits, having an engaged neuron's calculated p bit wide distance between an input vector and a prototype vector and stored in the weight memory thereof, an aggregate search/sort circuit (517) of n engaged neurons' search/sort circuits. the aggregate search/sort circuit determines the minimum distance among the calculated distances. each search/sort circuit (502-1) has p elementary search/sort units connected in series to form a column, such that the aggregate circuit is a matrix of elementary search/sort units. the distance bit signals of the same bit rank are applied to search/sort units in each row. a feedback signal is generated by oring in an or gate (12.1) all local search/sort output signals from the elementary search/sort units of the same row. the search process is based on identifying zeroes in the distance bit signals, from the msb's to the lsb's. as a zero is found in a row, all the columns with a one in that row are excluded from the subsequent row search. the search process continues until only one distance, the minimum distance, remains and is available at the output of the or circuit. the above described search/sort circuit may further include a latch allowing the aggregate circuit to sort remaining distances in increasing order. dated 1998-04-14"
5740686,method and apparatus for rolling a metal strip,"in rolling a metal strip in a roughing line and a finishing line, the rolling process in the roughing line is adjusted as a function of a predicted value for the change in width of the metal strip in the finishing line such that the metal strip has a given specified finished strip width on leaving the finishing line. in order to permit a reliable prediction of the change in width despite the lack of accurate information regarding the dependence of the change in width on influencing parameters that affect the process, this dependence is simulated in a neural network whose network parameters are adapted after each passage of a metal strip through the finishing line as a function of the influencing parameters measured or calculated during the passage and the measured actual change in width.",1998-04-21,"The title of the patent is method and apparatus for rolling a metal strip and its abstract is in rolling a metal strip in a roughing line and a finishing line, the rolling process in the roughing line is adjusted as a function of a predicted value for the change in width of the metal strip in the finishing line such that the metal strip has a given specified finished strip width on leaving the finishing line. in order to permit a reliable prediction of the change in width despite the lack of accurate information regarding the dependence of the change in width on influencing parameters that affect the process, this dependence is simulated in a neural network whose network parameters are adapted after each passage of a metal strip through the finishing line as a function of the influencing parameters measured or calculated during the passage and the measured actual change in width. dated 1998-04-21"
5741980,flow analysis system and method,"a non-invasive flow analysis system and method wherein a sensor, such as an acoustic sensor, is coupled to a conduit for transmitting a signal which varies depending on the characteristics of the flow in the conduit. the signal is amplified and there is a filter, responsive to the sensor signal, and tuned to pass a narrow band of frequencies proximate the resonant frequency of the sensor. a demodulator generates an amplitude envelope of the filtered signal and a number of flow indicator quantities are calculated based on variations in amplitude of the amplitude envelope. a neural network, or its equivalent, is then used to determine the flow rate of the flow in the conduit based on the flow indicator quantities.",1998-04-21,"The title of the patent is flow analysis system and method and its abstract is a non-invasive flow analysis system and method wherein a sensor, such as an acoustic sensor, is coupled to a conduit for transmitting a signal which varies depending on the characteristics of the flow in the conduit. the signal is amplified and there is a filter, responsive to the sensor signal, and tuned to pass a narrow band of frequencies proximate the resonant frequency of the sensor. a demodulator generates an amplitude envelope of the filtered signal and a number of flow indicator quantities are calculated based on variations in amplitude of the amplitude envelope. a neural network, or its equivalent, is then used to determine the flow rate of the flow in the conduit based on the flow indicator quantities. dated 1998-04-21"
5742018,device for and method of setting machining conditions for electrical discharge machining,"a machining condition setting apparatus for electrical discharge machining that sets the machining conditions, including the machining pulse energy, the feed of the electrode, etc. based on set data regarding specifications including the material of the workpiece and the desired surface roughness of the product. plural sets of basic data showing the relationship between the machining conditions and the specifications resulting from the machining conditions are selected. the relationship between the specifications and machining conditions are stored in an inference unit which includes a neural network type computation unit, by using selected basic data. the inference unit determines the most appropriate machining conditions based on the set data and the specifications. thus, the storage unit only need store a very small amount of basic data.",1998-04-21,"The title of the patent is device for and method of setting machining conditions for electrical discharge machining and its abstract is a machining condition setting apparatus for electrical discharge machining that sets the machining conditions, including the machining pulse energy, the feed of the electrode, etc. based on set data regarding specifications including the material of the workpiece and the desired surface roughness of the product. plural sets of basic data showing the relationship between the machining conditions and the specifications resulting from the machining conditions are selected. the relationship between the specifications and machining conditions are stored in an inference unit which includes a neural network type computation unit, by using selected basic data. the inference unit determines the most appropriate machining conditions based on the set data and the specifications. thus, the storage unit only need store a very small amount of basic data. dated 1998-04-21"
5742517,method for randomly accessing stored video and a field inspection system employing the same,"the present invention relates to a field inspection system for compressing video signals received from a field inspection video camera into compressed video data and for burning the compressed video data on a compact disc along with an electronic logsheet. the electronic logsheet includes a listing of suspected defects or anomalies and associated pointers to reference frames in the compressed video data. the electronic logsheet may be displayed and an operator may access a portion of the field inspection video showing a listed defect by clicking a mouse button when a pointer icon is posited on the listed defect. to perform this task, the present invention utilizes a technique for randomly accessing the compressed video data in which reference frames included therein are used as access points to the video footage. the field inspection system of the present invention may also utilize a neural network and an artificial intelligence system to detect, identify, and log defects without human intervention. the artificial intelligence system may also automatically generate a report making recommendations for repairing each detected defect.",1998-04-21,"The title of the patent is method for randomly accessing stored video and a field inspection system employing the same and its abstract is the present invention relates to a field inspection system for compressing video signals received from a field inspection video camera into compressed video data and for burning the compressed video data on a compact disc along with an electronic logsheet. the electronic logsheet includes a listing of suspected defects or anomalies and associated pointers to reference frames in the compressed video data. the electronic logsheet may be displayed and an operator may access a portion of the field inspection video showing a listed defect by clicking a mouse button when a pointer icon is posited on the listed defect. to perform this task, the present invention utilizes a technique for randomly accessing the compressed video data in which reference frames included therein are used as access points to the video footage. the field inspection system of the present invention may also utilize a neural network and an artificial intelligence system to detect, identify, and log defects without human intervention. the artificial intelligence system may also automatically generate a report making recommendations for repairing each detected defect. dated 1998-04-21"
5742700,quantitative dental caries detection system and method,a caries detection system and method for quantifying a probability of lesions existing in tissues are presented. digital x-ray images are segmented and further processed to generate feature statistics inputs for a neural network. the feature statistics include colinearity measurements of candidate lesions in different tissue segments. the neural network is trained by back propagation with an extensive data set of radiographs and histologic examinations and processes the statistics to determine the probability of lesions existing in the tissues.,1998-04-21,The title of the patent is quantitative dental caries detection system and method and its abstract is a caries detection system and method for quantifying a probability of lesions existing in tissues are presented. digital x-ray images are segmented and further processed to generate feature statistics inputs for a neural network. the feature statistics include colinearity measurements of candidate lesions in different tissue segments. the neural network is trained by back propagation with an extensive data set of radiographs and histologic examinations and processes the statistics to determine the probability of lesions existing in the tissues. dated 1998-04-21
5742702,neural network for character recognition and verification,""" a neural network is used to recognize characters from a character set. based upon the character recognized, a smaller neural network is used for verification of the character recognized. the smaller neural network is trained to recognize only a single character of the set and provides a """"yes"""" or """"no"""" type verification of correct identification of the character. """,1998-04-21,"The title of the patent is neural network for character recognition and verification and its abstract is "" a neural network is used to recognize characters from a character set. based upon the character recognized, a smaller neural network is used for verification of the character recognized. the smaller neural network is trained to recognize only a single character of the set and provides a """"yes"""" or """"no"""" type verification of correct identification of the character. "" dated 1998-04-21"
5742739,method of speeding up the execution of neural networks for correlated signal processing,"a method of speeding up the execution of a wide class of neural networks for processing input signals evolving slowly through time, such as, for instance, voice, radar, sonar, video signals, and which requires no specialized, costly or hard-to-find hardware. the method requires storing, for the neurons in at least one level of the network, the activation value at a certain instant and comparing it with the one computed at the subsequent instant. if the activation is equal, the neuron carries out no activity, otherwise it propagates the difference in activation, multiplied by the interconnection weights, to the neurons it is connected to.",1998-04-21,"The title of the patent is method of speeding up the execution of neural networks for correlated signal processing and its abstract is a method of speeding up the execution of a wide class of neural networks for processing input signals evolving slowly through time, such as, for instance, voice, radar, sonar, video signals, and which requires no specialized, costly or hard-to-find hardware. the method requires storing, for the neurons in at least one level of the network, the activation value at a certain instant and comparing it with the one computed at the subsequent instant. if the activation is equal, the neuron carries out no activity, otherwise it propagates the difference in activation, multiplied by the interconnection weights, to the neurons it is connected to. dated 1998-04-21"
5742740,adaptive network for automated first break picking of seismic refraction events and method of operating the same,"a method of operating an adaptive, or neural, network is disclosed for performing first break analysis for seismic shot records. the adaptive network is first trained according to the generalized delta rule. the disclosed training method includes selection of the seismic trace with the highest error, where the backpropagation is performed according to the error of this worst trace. the learning and momentum factors in the generalized delta rule are adjusted according to the value of the worst error, so that the learning and momentum factors increase as the error decreases. the training method further includes detection of slow convergence regions, and methods for escaping such regions including restoration of previously trimmed dormant links, renormalization of the weighting factor values, and the addition of new layers to the network. the network, after the addition of a new layer, includes links between nodes which skip the hidden layer. the error value used in the backpropagation is reduced from that actually calculated, by adjusting the desired output value, in order to reduce the growth of the weighting factors. after the training of the network, data corresponding to an average of the graphical display of a portion of the shot record, including multiple traces over a period of time, is provided to the network. the time of interest of the data is incremented until such time as the network indicates that the time of interest equals the first break time. the analysis may be repeated for all of the traces in the shot record.",1998-04-21,"The title of the patent is adaptive network for automated first break picking of seismic refraction events and method of operating the same and its abstract is a method of operating an adaptive, or neural, network is disclosed for performing first break analysis for seismic shot records. the adaptive network is first trained according to the generalized delta rule. the disclosed training method includes selection of the seismic trace with the highest error, where the backpropagation is performed according to the error of this worst trace. the learning and momentum factors in the generalized delta rule are adjusted according to the value of the worst error, so that the learning and momentum factors increase as the error decreases. the training method further includes detection of slow convergence regions, and methods for escaping such regions including restoration of previously trimmed dormant links, renormalization of the weighting factor values, and the addition of new layers to the network. the network, after the addition of a new layer, includes links between nodes which skip the hidden layer. the error value used in the backpropagation is reduced from that actually calculated, by adjusting the desired output value, in order to reduce the growth of the weighting factors. after the training of the network, data corresponding to an average of the graphical display of a portion of the shot record, including multiple traces over a period of time, is provided to the network. the time of interest of the data is incremented until such time as the network indicates that the time of interest equals the first break time. the analysis may be repeated for all of the traces in the shot record. dated 1998-04-21"
5742741,reconfigurable neural network,"a reconfigurable neural network is disclosed. the neural network includes a plurality of switches each having at least two conductive leads, wherein data flow direction of the conductive leads of the switches is programmed to select one of the conductive leads as input switch lead and select another one of the conductive leads as an output switch lead. a plurality of processing elements each having a plurality of leads connected to the switches, wherein the processing elements and the switches are interconnected in one-dimension manner. each of the processing elements comprising: (a) a serial-in-parallel-out accumulator having a first input coupled to one of the interconnected switches and generating a first output; (b) an activation function for transforming the first output of the serial-in-parallel-out accumulator and generating a second output; and (c) a parallel-in-serial-out shift register for shifting out the second output of the activation function serially to one of the interconnected switches.",1998-04-21,"The title of the patent is reconfigurable neural network and its abstract is a reconfigurable neural network is disclosed. the neural network includes a plurality of switches each having at least two conductive leads, wherein data flow direction of the conductive leads of the switches is programmed to select one of the conductive leads as input switch lead and select another one of the conductive leads as an output switch lead. a plurality of processing elements each having a plurality of leads connected to the switches, wherein the processing elements and the switches are interconnected in one-dimension manner. each of the processing elements comprising: (a) a serial-in-parallel-out accumulator having a first input coupled to one of the interconnected switches and generating a first output; (b) an activation function for transforming the first output of the serial-in-parallel-out accumulator and generating a second output; and (c) a parallel-in-serial-out shift register for shifting out the second output of the activation function serially to one of the interconnected switches. dated 1998-04-21"
5744967,apparatus for detecting intermittent and continuous faults in multiple conductor wiring and terminations for electronic systems,"a latching tester for testing continuity of wires in a system has an array of pin electronics cells, where each pin cell couples a signal to a row sense line and a column sense line when a change in current flow through a pin occurs; apparatus for detecting the signal on the row sense line; and apparatus for detecting the signal on the column sense line. the array of pin electronics cells may also operate as a capacitively coupled neural network, where a signal coupled onto the row and column lines from a stimulus line varies with the load on each pin of the pin electronics cells. an alternate mode of operation permits stimulus of the network and attached loads, and generation of a signature based upon the response of the network to the stimulus, as observed on the row and column sense lines. in yet another mode of operation, the tester may serve to recognize particular signals.",1998-04-28,"The title of the patent is apparatus for detecting intermittent and continuous faults in multiple conductor wiring and terminations for electronic systems and its abstract is a latching tester for testing continuity of wires in a system has an array of pin electronics cells, where each pin cell couples a signal to a row sense line and a column sense line when a change in current flow through a pin occurs; apparatus for detecting the signal on the row sense line; and apparatus for detecting the signal on the column sense line. the array of pin electronics cells may also operate as a capacitively coupled neural network, where a signal coupled onto the row and column lines from a stimulus line varies with the load on each pin of the pin electronics cells. an alternate mode of operation permits stimulus of the network and attached loads, and generation of a signature based upon the response of the network to the stimulus, as observed on the row and column sense lines. in yet another mode of operation, the tester may serve to recognize particular signals. dated 1998-04-28"
5745034,providing an alarm in response to a determination that a person may have suddenly experienced fear,"a computer system processes physiological data signals provided by a physiological-condition monitoring system to determine whether the person may have suddenly experienced fear by comparing the monitored physiological data with stored stress profile data for the person that is based upon measurements of the monitored physiological conditions of the person during situations of stress and/or upon statistical classification in view of a combination of predetermined characteristics of the person. when such processing by the computer system determines that the person may have suddenly experienced fear, the computer system activates an alarm indicator. the computer system includes a neural network for modifying the stored stress profile data in response to an input signal indicating that the computer system provided a false alarm indication. the surrounding conditions that may have caused the determination of probable fear are recorded and transmitted to a remote location.",1998-04-28,"The title of the patent is providing an alarm in response to a determination that a person may have suddenly experienced fear and its abstract is a computer system processes physiological data signals provided by a physiological-condition monitoring system to determine whether the person may have suddenly experienced fear by comparing the monitored physiological data with stored stress profile data for the person that is based upon measurements of the monitored physiological conditions of the person during situations of stress and/or upon statistical classification in view of a combination of predetermined characteristics of the person. when such processing by the computer system determines that the person may have suddenly experienced fear, the computer system activates an alarm indicator. the computer system includes a neural network for modifying the stored stress profile data in response to an input signal indicating that the computer system provided a false alarm indication. the surrounding conditions that may have caused the determination of probable fear are recorded and transmitted to a remote location. dated 1998-04-28"
5745382,neural network based system for equipment surveillance,"a method and system for performing surveillance of transient signals of an industrial device to ascertain the operating state. the method and system involves the steps of reading into a memory training data, determining neural network weighting values until achieving target outputs close to the neural network output. if the target outputs are inadequate, wavelet parameters are determined to yield neural network outputs close to the desired set of target outputs and then providing signals characteristic of an industrial process and comparing the neural network output to the industrial process signals to evaluate the operating state of the industrial process.",1998-04-28,"The title of the patent is neural network based system for equipment surveillance and its abstract is a method and system for performing surveillance of transient signals of an industrial device to ascertain the operating state. the method and system involves the steps of reading into a memory training data, determining neural network weighting values until achieving target outputs close to the neural network output. if the target outputs are inadequate, wavelet parameters are determined to yield neural network outputs close to the desired set of target outputs and then providing signals characteristic of an industrial process and comparing the neural network output to the industrial process signals to evaluate the operating state of the industrial process. dated 1998-04-28"
5745652,adaptive resource allocation using neural networks,"in a system comprising a plurality of resources for performing useful work, a resource allocation controller function, which is customized to the particular system's available resources and configuration, dynamically allocates resources and/or alters configuration to accommodate a changing workload. preferably, the resource allocation controller is part of the computer's operating system which allocates resources of the computer system. the resource allocation controller uses a controller neural network for control, and a separate system model neural network for modelling the system and training the controller neural network. performance data is collected by the system and used to train the system model neural network. a system administrator specifies computer system performance targets which indicate the desired performance of the system. deviations in actual performance from desired performance are propagated back through the system model and ultimately to the controller neural network to create a closed loop system for resource allocation.",1998-04-28,"The title of the patent is adaptive resource allocation using neural networks and its abstract is in a system comprising a plurality of resources for performing useful work, a resource allocation controller function, which is customized to the particular system's available resources and configuration, dynamically allocates resources and/or alters configuration to accommodate a changing workload. preferably, the resource allocation controller is part of the computer's operating system which allocates resources of the computer system. the resource allocation controller uses a controller neural network for control, and a separate system model neural network for modelling the system and training the controller neural network. performance data is collected by the system and used to train the system model neural network. a system administrator specifies computer system performance targets which indicate the desired performance of the system. deviations in actual performance from desired performance are propagated back through the system model and ultimately to the controller neural network to create a closed loop system for resource allocation. dated 1998-04-28"
5745653,generic neural network training and processing system,"a electronic engine control (eec) module executes a generic neural network processing program to perform one or more neural network control funtions. each neural network funtion is defined by a unitary data structure which defines the network architecture, including the number of node layers, the number of nodes per layer, and the interconnections between nodes. in addition, the data structure holds weight values which determine the manner in which network signals are combined. the network definition data structures are created by a network training system which utilizes an external training processor which employs gradient methods to derive network weight values in accordance with a cost function which quantitatively defines system objectives and an identification network which is pretrained to provide gradient signals representative the behavior of the physical plant. the training processor executes training cycles asynchronously with the operation of the eec module in a representative test vehicle.",1998-04-28,"The title of the patent is generic neural network training and processing system and its abstract is a electronic engine control (eec) module executes a generic neural network processing program to perform one or more neural network control funtions. each neural network funtion is defined by a unitary data structure which defines the network architecture, including the number of node layers, the number of nodes per layer, and the interconnections between nodes. in addition, the data structure holds weight values which determine the manner in which network signals are combined. the network definition data structures are created by a network training system which utilizes an external training processor which employs gradient methods to derive network weight values in accordance with a cost function which quantitatively defines system objectives and an identification network which is pretrained to provide gradient signals representative the behavior of the physical plant. the training processor executes training cycles asynchronously with the operation of the eec module in a representative test vehicle. dated 1998-04-28"
5745654,fast explanations of scored observations,"a system, method, and product provide rapid explanations for the scores determined by a neural network for new observations input into the neural network. the explanations are associated with a table of percentile bins for each of the input variables used to define the observation. the table contains for each input variable a number of percentile bins. each percentile bin contains an expected score for values of the input variable containing with the percentile bin. the expected score in each percentile bin is determined from historical observation data. preferably each percentile bin is associated with an explanation that describes the meaning of the value of the input variable falling within the percentile bin. during observation processing, a new observation is scored. the value of each input variable in the new observation is compared with the percentile bins for the input variable in the table. the expected score in the percentile bin that contains the value of the input variable is retrieved, and this is repeated for all input variables in the new observation. the explanation associated with the percentile bin that has an expected score closest to the actual score is retrieved and provided as the explanation of the most significant input variable accounting for score. other explanations from the next closest expected scores may also be retrieved.",1998-04-28,"The title of the patent is fast explanations of scored observations and its abstract is a system, method, and product provide rapid explanations for the scores determined by a neural network for new observations input into the neural network. the explanations are associated with a table of percentile bins for each of the input variables used to define the observation. the table contains for each input variable a number of percentile bins. each percentile bin contains an expected score for values of the input variable containing with the percentile bin. the expected score in each percentile bin is determined from historical observation data. preferably each percentile bin is associated with an explanation that describes the meaning of the value of the input variable falling within the percentile bin. during observation processing, a new observation is scored. the value of each input variable in the new observation is compared with the percentile bins for the input variable in the table. the expected score in the percentile bin that contains the value of the input variable is retrieved, and this is repeated for all input variables in the new observation. the explanation associated with the percentile bin that has an expected score closest to the actual score is retrieved and provided as the explanation of the most significant input variable accounting for score. other explanations from the next closest expected scores may also be retrieved. dated 1998-04-28"
5745655,chaotic neural circuit and chaotic neural network using the same,"a mapping circuit includes a linear circuit for outputting a signal which is linearly changed with respect to its input, a non-linear circuit for outputting a signal which is non-linearly changed with respect to its input, and an adder for summing the output signals of the linear and non-linear circuits and an external input signal. a chaotic neuron circuit using the mapping circuit has a simple structure and more precise chaos characteristics. a chaotic neural network can thus be formed by the serial and/or parallel interconnection of a plurality of chaotic neuron circuits, wherein the weight of each neuron is controlled.",1998-04-28,"The title of the patent is chaotic neural circuit and chaotic neural network using the same and its abstract is a mapping circuit includes a linear circuit for outputting a signal which is linearly changed with respect to its input, a non-linear circuit for outputting a signal which is non-linearly changed with respect to its input, and an adder for summing the output signals of the linear and non-linear circuits and an external input signal. a chaotic neuron circuit using the mapping circuit has a simple structure and more precise chaos characteristics. a chaotic neural network can thus be formed by the serial and/or parallel interconnection of a plurality of chaotic neuron circuits, wherein the weight of each neuron is controlled. dated 1998-04-28"
5745668,example-based image analysis and synthesis using pixelwise correspondence,"synthesis of novel images from example images is achieved by determining a pixelwise optical flow among example images, computing a parameter set for a new image, and synthesizing the new image based on the parameter vector and the example images. the parameter set may describe characteristics of the image, in which case the characteristics are applied to a neural network trained with the example images in order to synthesize the novel image. the parameter set may also be an estimate of relative contributions of each of the example images to the new image, in which case the new image may be synthesized by taking a linear combination of the example images, weighted by the parameter set. in one embodiment, both the example and the new images are of the same object. in another embodiment, the example image represents a first object while the new image represents a second object. in yet another embodiment, a set of target images is synthesized from a single target image and a set of source images. analysis of existing images to determine image parameters is achieved by determining parameters for the existing image based on comparison of the existing image with an image set for which parameters are known.",1998-04-28,"The title of the patent is example-based image analysis and synthesis using pixelwise correspondence and its abstract is synthesis of novel images from example images is achieved by determining a pixelwise optical flow among example images, computing a parameter set for a new image, and synthesizing the new image based on the parameter vector and the example images. the parameter set may describe characteristics of the image, in which case the characteristics are applied to a neural network trained with the example images in order to synthesize the novel image. the parameter set may also be an estimate of relative contributions of each of the example images to the new image, in which case the new image may be synthesized by taking a linear combination of the example images, weighted by the parameter set. in one embodiment, both the example and the new images are of the same object. in another embodiment, the example image represents a first object while the new image represents a second object. in yet another embodiment, a set of target images is synthesized from a single target image and a set of source images. analysis of existing images to determine image parameters is achieved by determining parameters for the existing image based on comparison of the existing image with an image set for which parameters are known. dated 1998-04-28"
5745874,preprocessor for automatic speech recognition system,a preprocessor for automatic speech recognition based upon auditory modeling includes a tapped delay line and a neural network in the form of a multilayer perceptron. the tapped delay line receives an analog speech signal and provides multiple time delayed samples thereof in parallel as inputs for the neural network. the single analog output of the neural network is suitable for interfacing with a signal processor for further processing of the speech information using spectral signal analysis so as to provide a speech representation with desirable characteristics of an auditory based spectral analysis model while simultaneously maintaining a standard analog signal interface.,1998-04-28,The title of the patent is preprocessor for automatic speech recognition system and its abstract is a preprocessor for automatic speech recognition based upon auditory modeling includes a tapped delay line and a neural network in the form of a multilayer perceptron. the tapped delay line receives an analog speech signal and provides multiple time delayed samples thereof in parallel as inputs for the neural network. the single analog output of the neural network is suitable for interfacing with a signal processor for further processing of the speech information using spectral signal analysis so as to provide a speech representation with desirable characteristics of an auditory based spectral analysis model while simultaneously maintaining a standard analog signal interface. dated 1998-04-28
5746698,method and device for determining brachial arterial pressure wave on the basis of nonivasively measured finger blood pressure wave,"method and device for determining a proximal arterial blood pressure waveform in a person, starting from a distally measured arterial pressure waveform, by first applying age-dependent waveform filtering to the distal pressure waveform, in order to obtain the proximal pressure waveform with mutually correct systolic, diastolic and mean pressure levels, and by then shifting the filtered pressure waveform by means of calibration to the correct proximal pressure level, for example by calibration of one level of the filtered systolic, diastolic or mean pressure levels with the corresponding proximal pressure level. this can be a single noninvasively measured systolic or diastolic or mean pressure level. the age for the purpose of the age-dependent waveform filtering is derived from the distally measured pressure waveform, for example by means of a trained neural network. the level shift in the filtered pressure waveform can be obtained by means of a regression equation which has entered in it only the filtered pressure waveform with corresponding systolic and diastolic pressure levels, or the above combined with a noninvasively measured single brachial pressure level.",1998-05-05,"The title of the patent is method and device for determining brachial arterial pressure wave on the basis of nonivasively measured finger blood pressure wave and its abstract is method and device for determining a proximal arterial blood pressure waveform in a person, starting from a distally measured arterial pressure waveform, by first applying age-dependent waveform filtering to the distal pressure waveform, in order to obtain the proximal pressure waveform with mutually correct systolic, diastolic and mean pressure levels, and by then shifting the filtered pressure waveform by means of calibration to the correct proximal pressure level, for example by calibration of one level of the filtered systolic, diastolic or mean pressure levels with the corresponding proximal pressure level. this can be a single noninvasively measured systolic or diastolic or mean pressure level. the age for the purpose of the age-dependent waveform filtering is derived from the distally measured pressure waveform, for example by means of a trained neural network. the level shift in the filtered pressure waveform can be obtained by means of a regression equation which has entered in it only the filtered pressure waveform with corresponding systolic and diastolic pressure levels, or the above combined with a noninvasively measured single brachial pressure level. dated 1998-05-05"
5747777,heater control device,"a heater on-time computing unit, provided in a heater control device, has a first fuzzy neural network for computing a heater on-time in accordance with a surface temperature of a heat fixing roller and a surface temperature change obtained from surface temperatures. a roller surface temperature predicting unit has the second fuzzy neural network for computing a predicted temperature of the heater in accordance with a surface temperature, a surface temperature change obtained from surface temperatures, and a heater on-time computed by the heater on-time computing unit. thereby only roughly setting of parameters is required because the parameters are adjusted by sequential learning so that the optimal heater on-time is obtained. therefore, the programming is simplified and it is possible to comply with differences in such as models, individuals, deterioration due to aging, and changes in environments.",1998-05-05,"The title of the patent is heater control device and its abstract is a heater on-time computing unit, provided in a heater control device, has a first fuzzy neural network for computing a heater on-time in accordance with a surface temperature of a heat fixing roller and a surface temperature change obtained from surface temperatures. a roller surface temperature predicting unit has the second fuzzy neural network for computing a predicted temperature of the heater in accordance with a surface temperature, a surface temperature change obtained from surface temperatures, and a heater on-time computed by the heater on-time computing unit. thereby only roughly setting of parameters is required because the parameters are adjusted by sequential learning so that the optimal heater on-time is obtained. therefore, the programming is simplified and it is possible to comply with differences in such as models, individuals, deterioration due to aging, and changes in environments. dated 1998-05-05"
5748329,method and apparatus for adaptive color scanning/printing data correction employing neural networks,"a method and apparatus for adaptive color scanning/printing data correction to effectively adjust to an arbitrary combination of color image scanning input and printing output devices for reproducing color image at optimized resemblance. an improved back propagation algorithm is employed to reduce the learning error and accelerate the process of learning procedure by increasing the rate of convergence. a method of characteristics extracting functionalization is utilized to urge the successful convergence of the learning procedure of the neural network, and reduce the color discrepancy and accelerate the process of learning convergence. an enhanced grey-scale balancing scheme is also utilized to extract the grey component of a learning sample under a predetermined condition and re fetch again to the neural network for accelerated convergence of the learning behavior. an apparatus for performing the color data correction includes an integrated neural network color processor that utilizes a small number of neural elements to reduce the complexity of the color data processing and computational complication.",1998-05-05,"The title of the patent is method and apparatus for adaptive color scanning/printing data correction employing neural networks and its abstract is a method and apparatus for adaptive color scanning/printing data correction to effectively adjust to an arbitrary combination of color image scanning input and printing output devices for reproducing color image at optimized resemblance. an improved back propagation algorithm is employed to reduce the learning error and accelerate the process of learning procedure by increasing the rate of convergence. a method of characteristics extracting functionalization is utilized to urge the successful convergence of the learning procedure of the neural network, and reduce the color discrepancy and accelerate the process of learning convergence. an enhanced grey-scale balancing scheme is also utilized to extract the grey component of a learning sample under a predetermined condition and re fetch again to the neural network for accelerated convergence of the learning behavior. an apparatus for performing the color data correction includes an integrated neural network color processor that utilizes a small number of neural elements to reduce the complexity of the color data processing and computational complication. dated 1998-05-05"
5748847,nonadaptively trained adaptive neural systems,"an adaptive neural system (ans) disclosed herein comprises a processor and an adaptor. the processor includes mainly a neural network whose adjustable weights are divided into nonadaptively and adaptively adjustable weights. the nonadaptively adjustable weights are determined through minimizing or reducing a nonadaptive training criterion in an off-line nonadaptive training. being constructed with a priori training data, the nonadaptive training criterion is a function of the nonadaptively adjustable weights and the diversity variables associated with typical values of the environmental parameter. during an operation of the adaptive neural system, only the adaptively adjustable weights are adjusted on-line to adapt to the unknown environmental parameter. this adaptive training is achieved by minimizing or reducing an adaptive training criterion. the nonadaptive training allows the ans to make full advantage of a priori information about the ans's operating environment and helps the ans focus on learning about and adapting to the unknown environmental parameter during the adaptive training. in many applications, the adaptively adjustable weights can be selected, without adversely affecting the ans's performance, such that they appear quadratically in the adaptive training criterion. in this case, the adaptive training criterion has no undesirable local minima and the existing fast algorithms for adaptive linear filters are applicable to the adaptive training.",1998-05-05,"The title of the patent is nonadaptively trained adaptive neural systems and its abstract is an adaptive neural system (ans) disclosed herein comprises a processor and an adaptor. the processor includes mainly a neural network whose adjustable weights are divided into nonadaptively and adaptively adjustable weights. the nonadaptively adjustable weights are determined through minimizing or reducing a nonadaptive training criterion in an off-line nonadaptive training. being constructed with a priori training data, the nonadaptive training criterion is a function of the nonadaptively adjustable weights and the diversity variables associated with typical values of the environmental parameter. during an operation of the adaptive neural system, only the adaptively adjustable weights are adjusted on-line to adapt to the unknown environmental parameter. this adaptive training is achieved by minimizing or reducing an adaptive training criterion. the nonadaptive training allows the ans to make full advantage of a priori information about the ans's operating environment and helps the ans focus on learning about and adapting to the unknown environmental parameter during the adaptive training. in many applications, the adaptively adjustable weights can be selected, without adversely affecting the ans's performance, such that they appear quadratically in the adaptive training criterion. in this case, the adaptive training criterion has no undesirable local minima and the existing fast algorithms for adaptive linear filters are applicable to the adaptive training. dated 1998-05-05"
5748848,learning method for a neural network,"in a learning method for training a recurrent neural network having a number of inputs and a number of outputs with at least one output being connected via a return line to an input, the return line is separated during training of the neural network, thereby freeing the input connected to the return line for use as an additional input during training, together with the other inputs. the additional input values, which must be estimated or predicted for supply to the thus-produced additional training inputs, are generated by treating each additional input value to be generated as a missing value in the time series of input quantities. error distribution densities for the additional input values are calculated on the basis of the known values from the time series and their known or predetermined error distribution density, and samples are taken from this error distribution density according to the monte carlo method. these each lead to an estimated or predicted value whose average is introduced for the additional input value to be predicted. the method can be employed for the operation as well as for the training of the neural network, and is suitable for use in all known fields of utilization of neural networks.",1998-05-05,"The title of the patent is learning method for a neural network and its abstract is in a learning method for training a recurrent neural network having a number of inputs and a number of outputs with at least one output being connected via a return line to an input, the return line is separated during training of the neural network, thereby freeing the input connected to the return line for use as an additional input during training, together with the other inputs. the additional input values, which must be estimated or predicted for supply to the thus-produced additional training inputs, are generated by treating each additional input value to be generated as a missing value in the time series of input quantities. error distribution densities for the additional input values are calculated on the basis of the known values from the time series and their known or predetermined error distribution density, and samples are taken from this error distribution density according to the monte carlo method. these each lead to an estimated or predicted value whose average is introduced for the additional input value to be predicted. the method can be employed for the operation as well as for the training of the neural network, and is suitable for use in all known fields of utilization of neural networks. dated 1998-05-05"
5749066,method and apparatus for developing a neural network for phoneme recognition,"an automated speech recognition system converts a speech signal into a compact, coded representation that correlates to a speech phoneme set. a number of different neural network pattern matching schemes may be used to perform the necessary speech coding. an integrated user interface guides a user unfamiliar with the details of speech recognition or neural networks to quickly develop and test a neural network for phoneme recognition. to train the neural network, digitized voice data containing known phonemes that the user wants the neural network to ultimately recognize are processed by the integrated user interface. the digitized speech is segmented into phonemes with each segment being labelled with a corresponding phoneme code. based on a user selected transformation method and transformation parameters, each segment is transformed into a series of multiple dimension vectors representative of the speech characteristics of that segment. these vectors are iteratively presented to a neural network to train/adapt that neural network to consistently distinguish and recognize these vectors and assign an appropriate phoneme code to each vector. simultaneous display of the digitized speech, segments, vector sets, and a representation of the trained neural network assist the user in visually confirming the acceptability of the phoneme training set. a user may also selectively audibly confirm the acceptability of the digitization scheme, the segments, and the transform vectors so that satisfactory training data are presented to the neural network. if the user finds a particular step or parameter produces an unacceptable result, the user may modify one or more of the parameters and verify whether the modification effected an improvement in performance. the trained neural network is also automatically tested by presenting a test speech signal to the integrated user interface and observing both audibly and visually automatic segmentation of the speech, transformation into multidimensional vectors, and the resulting neural network assigned phoneme codes. a method of decoding such phoneme codes using the neural network is also disclosed.",1998-05-05,"The title of the patent is method and apparatus for developing a neural network for phoneme recognition and its abstract is an automated speech recognition system converts a speech signal into a compact, coded representation that correlates to a speech phoneme set. a number of different neural network pattern matching schemes may be used to perform the necessary speech coding. an integrated user interface guides a user unfamiliar with the details of speech recognition or neural networks to quickly develop and test a neural network for phoneme recognition. to train the neural network, digitized voice data containing known phonemes that the user wants the neural network to ultimately recognize are processed by the integrated user interface. the digitized speech is segmented into phonemes with each segment being labelled with a corresponding phoneme code. based on a user selected transformation method and transformation parameters, each segment is transformed into a series of multiple dimension vectors representative of the speech characteristics of that segment. these vectors are iteratively presented to a neural network to train/adapt that neural network to consistently distinguish and recognize these vectors and assign an appropriate phoneme code to each vector. simultaneous display of the digitized speech, segments, vector sets, and a representation of the trained neural network assist the user in visually confirming the acceptability of the phoneme training set. a user may also selectively audibly confirm the acceptability of the digitization scheme, the segments, and the transform vectors so that satisfactory training data are presented to the neural network. if the user finds a particular step or parameter produces an unacceptable result, the user may modify one or more of the parameters and verify whether the modification effected an improvement in performance. the trained neural network is also automatically tested by presenting a test speech signal to the integrated user interface and observing both audibly and visually automatic segmentation of the speech, transformation into multidimensional vectors, and the resulting neural network assigned phoneme codes. a method of decoding such phoneme codes using the neural network is also disclosed. dated 1998-05-05"
5749367,heart monitoring apparatus and method,a heart monitoring apparatus and method is disclosed wherein an electrocardiograph signal is obtained from a patient and processed to enhance the salient features and to suppress noise. a plurality n of values representative of the features of the electrocardiograph signal are generated and used in a kohonen neural network to generate an n dimensional vector. this vector is compared with a stored plurality m of n dimensional reference vectors defining an n dimensional kohonen feature map to determine the proximity of the vector to the reference vectors. if it is determined by the kohonen neural network that the vector is within or beyond a threshold range of the reference vectors a signal is output which can be used to initiate an event such as the generation of an alarm or the storage of ecg data.,1998-05-12,The title of the patent is heart monitoring apparatus and method and its abstract is a heart monitoring apparatus and method is disclosed wherein an electrocardiograph signal is obtained from a patient and processed to enhance the salient features and to suppress noise. a plurality n of values representative of the features of the electrocardiograph signal are generated and used in a kohonen neural network to generate an n dimensional vector. this vector is compared with a stored plurality m of n dimensional reference vectors defining an n dimensional kohonen feature map to determine the proximity of the vector to the reference vectors. if it is determined by the kohonen neural network that the vector is within or beyond a threshold range of the reference vectors a signal is output which can be used to initiate an event such as the generation of an alarm or the storage of ecg data. dated 1998-05-12
5751209,system for the early detection of fires,"a fire detection system contains several detectors connected to a control center, some of which are fitted with at least two sensors for monitoring different fire parameters. preferably, one sensor is a thermal sensor and another is an optical sensor. an arrangement for processing the sensor signals is located within each of the detectors. it contains a microcontroller for conditioning the sensor signals and for signal processing, with the aim of generating alarm signals. the alarm signals are obtained in a neural network.",1998-05-12,"The title of the patent is system for the early detection of fires and its abstract is a fire detection system contains several detectors connected to a control center, some of which are fitted with at least two sensors for monitoring different fire parameters. preferably, one sensor is a thermal sensor and another is an optical sensor. an arrangement for processing the sensor signals is located within each of the detectors. it contains a microcontroller for conditioning the sensor signals and for signal processing, with the aim of generating alarm signals. the alarm signals are obtained in a neural network. dated 1998-05-12"
5751609,neural network based method for estimating helicopter low airspeed,"the invention is directed to a method, utilizing a neural network, for estimating helicopter airspeed in the low airspeed flight range of below about 50 knots using only fixed system parameters as inputs to the neural network. the method includes the steps of: (a) defining input parameters derivable from variable state parameters generated during flight of the helicopter and measurable in a nonrotating reference frame associated with the helicopter; (b) determining the input parameters and a corresponding helicopter airspeed at a plurality of flight conditions representing a predetermined low airspeed flight domain of the helicopter; (c) establishing a learned relationship between the determined input parameters and the corresponding helicopter airspeed wherein the relationship is represented by at least one nonlinear equation; (d) storing the at least one nonlinear equation in a memory onboard the helicopter; (e) measuring real time values of the variable state parameters during low airspeed flight of the helicopter; (f) calculating real time values of the input parameters; (g) storing the real time values of the input parameters in the memory; (h) processing the real time values of the input parameters in accordance with the at least one nonlinear equation to determine real time airspeed; and (i) displaying the real time airspeed.",1998-05-12,"The title of the patent is neural network based method for estimating helicopter low airspeed and its abstract is the invention is directed to a method, utilizing a neural network, for estimating helicopter airspeed in the low airspeed flight range of below about 50 knots using only fixed system parameters as inputs to the neural network. the method includes the steps of: (a) defining input parameters derivable from variable state parameters generated during flight of the helicopter and measurable in a nonrotating reference frame associated with the helicopter; (b) determining the input parameters and a corresponding helicopter airspeed at a plurality of flight conditions representing a predetermined low airspeed flight domain of the helicopter; (c) establishing a learned relationship between the determined input parameters and the corresponding helicopter airspeed wherein the relationship is represented by at least one nonlinear equation; (d) storing the at least one nonlinear equation in a memory onboard the helicopter; (e) measuring real time values of the variable state parameters during low airspeed flight of the helicopter; (f) calculating real time values of the input parameters; (g) storing the real time values of the input parameters in the memory; (h) processing the real time values of the input parameters in accordance with the at least one nonlinear equation to determine real time airspeed; and (i) displaying the real time airspeed. dated 1998-05-12"
5751831,method for extracting object images and method for detecting movements thereof,"in a method for extracting an object image, an extraction area for extraction of a candidate for a predetermined object image from an image is determined. the center point of a view window, which has a predetermined size, is caused to travel to the position of the candidate for the predetermined object image. the extraction area is determined in accordance with the size and/or the shape of the candidate for the predetermined object image, the center point of the view window being taken as a reference during the determination of the extraction area. the extraction of the candidate for the predetermined object image is carried out by using a neural network. even if a plurality of object images, which are to be extracted, are embedded in a given image, the object images are extracted efficiently such that an object image, which has already been extracted, may not be extracted again.",1998-05-12,"The title of the patent is method for extracting object images and method for detecting movements thereof and its abstract is in a method for extracting an object image, an extraction area for extraction of a candidate for a predetermined object image from an image is determined. the center point of a view window, which has a predetermined size, is caused to travel to the position of the candidate for the predetermined object image. the extraction area is determined in accordance with the size and/or the shape of the candidate for the predetermined object image, the center point of the view window being taken as a reference during the determination of the extraction area. the extraction of the candidate for the predetermined object image is carried out by using a neural network. even if a plurality of object images, which are to be extracted, are embedded in a given image, the object images are extracted efficiently such that an object image, which has already been extracted, may not be extracted again. dated 1998-05-12"
5751844,method and apparatus for image acquisition with adaptive compensation for image exposure variation,"in an image acquisition system, which produces at least one image, each of which is scorable with reference to at least one image quality criterion, a control system is provided for optimizing the image quality criterion. typically, the image quality criterion is at least in-part established by an exposure parameter. when considered as an apparatus, the present invention includes a number of components which cooperate together to automatically and continually adjust the value of the exposure parameter to optimize the image quality criterion. an image acquisition means is provided to obtain one or more acquired images under selected exposure characteristics. a transform system is provided for receiving the one or more acquired images and developing an energy distribution map, or histogram, of at least a portion of the one or more acquired images. a neural network means is provided for maintaining a learned relationship between the image quality criterion and the exposure parameter, and for receiving the energy distribution map of the one or more acquired images, and for automatically providing a corrected exposure parameter, so that subsequent acquired images will be obtained under optimal settings of the exposure parameter. a controller member is provided for supplying corrected exposure parameters, which are the output of the neural network means, to the image acquisition means.",1998-05-12,"The title of the patent is method and apparatus for image acquisition with adaptive compensation for image exposure variation and its abstract is in an image acquisition system, which produces at least one image, each of which is scorable with reference to at least one image quality criterion, a control system is provided for optimizing the image quality criterion. typically, the image quality criterion is at least in-part established by an exposure parameter. when considered as an apparatus, the present invention includes a number of components which cooperate together to automatically and continually adjust the value of the exposure parameter to optimize the image quality criterion. an image acquisition means is provided to obtain one or more acquired images under selected exposure characteristics. a transform system is provided for receiving the one or more acquired images and developing an energy distribution map, or histogram, of at least a portion of the one or more acquired images. a neural network means is provided for maintaining a learned relationship between the image quality criterion and the exposure parameter, and for receiving the energy distribution map of the one or more acquired images, and for automatically providing a corrected exposure parameter, so that subsequent acquired images will be obtained under optimal settings of the exposure parameter. a controller member is provided for supplying corrected exposure parameters, which are the output of the neural network means, to the image acquisition means. dated 1998-05-12"
5751904,speech recognition system using neural networks,a speech recognition system can recognize a plurality of voice data having different patterns. the speech recognition system has a voice recognizing and processing device including a plurality of speech recognition neural networks that have previously learned different voice patterns to recognize given voice data. each of the speech recognition neutral networks is adapted to judge whether or not input voice data coincides with one of the voice data to be recognized. each neural network then outputs adaptation judgment data representing the adaptation in speech recognition. a selector responsive to the adaptation judgment data from each of the speech recognition neural networks selects one of the neural networks that has the highest adaptation in speech recognition. an output control device outputs the result of speech recognition from the speech recognition neural network selected by the selector.,1998-05-12,The title of the patent is speech recognition system using neural networks and its abstract is a speech recognition system can recognize a plurality of voice data having different patterns. the speech recognition system has a voice recognizing and processing device including a plurality of speech recognition neural networks that have previously learned different voice patterns to recognize given voice data. each of the speech recognition neutral networks is adapted to judge whether or not input voice data coincides with one of the voice data to be recognized. each neural network then outputs adaptation judgment data representing the adaptation in speech recognition. a selector responsive to the adaptation judgment data from each of the speech recognition neural networks selects one of the neural networks that has the highest adaptation in speech recognition. an output control device outputs the result of speech recognition from the speech recognition neural network selected by the selector. dated 1998-05-12
5751910,neural network solder paste inspection system,a solder paste brick inspection and physical quality scoring system 10 employs a neural network 70 trained with a fuzzified output vector. an image of solder paste bricks 64 on a printed circuit board 12 is acquired by a ccd camera 30. values of a predetermined set of brick metrics are extracted from the image by a computer 28 and used as a crisp input vector to trained neural network 70. a defuzzifier 76 converts a fuzzy output vector from neural network 70 into a crisp quality score output which can be used for monitoring and process control.,1998-05-12,The title of the patent is neural network solder paste inspection system and its abstract is a solder paste brick inspection and physical quality scoring system 10 employs a neural network 70 trained with a fuzzified output vector. an image of solder paste bricks 64 on a printed circuit board 12 is acquired by a ccd camera 30. values of a predetermined set of brick metrics are extracted from the image by a computer 28 and used as a crisp input vector to trained neural network 70. a defuzzifier 76 converts a fuzzy output vector from neural network 70 into a crisp quality score output which can be used for monitoring and process control. dated 1998-05-12
5751911,real-time waveform analysis using artificial neural networks,"a real-time waveform analysis system utilizes neural networks to perform various stages of the analysis. the signal containing the waveform is first stored in a buffer and the buffer contents transmitted to a first and second neural network which have been previously trained to recognize the start point and the end point of the waveform respectively. a third neural network receives the signal occurring between the start and end points and classifies that waveform as comprising either an incomplete waveform, a normal waveform or one of a variety of predetermined characteristic classifications. ambiguities in the output of the third neural network are arbitrated by a fourth neural network which may be given additional information which serves to resolve these ambiguities. in accordance with the preferred embodiment, the present invention is applied to a system analyzing respiratory waveforms of a patient undergoing anesthesia and the classifications of the waveform correspond to normal or various categories of abnormal features functioning in the respiratory signal. the system performs the analysis rapidly enough to be used in real-time systems and can be operated with relatively low cost hardware and with minimal software development required.",1998-05-12,"The title of the patent is real-time waveform analysis using artificial neural networks and its abstract is a real-time waveform analysis system utilizes neural networks to perform various stages of the analysis. the signal containing the waveform is first stored in a buffer and the buffer contents transmitted to a first and second neural network which have been previously trained to recognize the start point and the end point of the waveform respectively. a third neural network receives the signal occurring between the start and end points and classifies that waveform as comprising either an incomplete waveform, a normal waveform or one of a variety of predetermined characteristic classifications. ambiguities in the output of the third neural network are arbitrated by a fourth neural network which may be given additional information which serves to resolve these ambiguities. in accordance with the preferred embodiment, the present invention is applied to a system analyzing respiratory waveforms of a patient undergoing anesthesia and the classifications of the waveform correspond to normal or various categories of abnormal features functioning in the respiratory signal. the system performs the analysis rapidly enough to be used in real-time systems and can be operated with relatively low cost hardware and with minimal software development required. dated 1998-05-12"
5751913,reconfigurable neural network and difference-square neuron,"a reconfigurable neural network includes several switches each having at least two conductive leads, data flow direction of the conductive leads is programmed to select one of the conductive leads as input switch lead and select another one of the conductive leads as an output switch lead. several processing elements each having leads connected to the switches, where the processing elements and the switches are interconnected in one-dimension manner. the neural network of interconnected switches and processing elements has a bit-serial input and a bit-serial output. each of the processing elements comprising: (a) a serial-in-parallel-out difference-square accumulator having a first input coupled to one of the interconnected switches and generating a first output; (b) an activation function for transforming the first output of the serial-in-parallel-out difference-square accumulator and generating a second output; and (c) a parallel-in-serial-out shift register for shifting out the second output of the activation function serially to one of the interconnected switches.",1998-05-12,"The title of the patent is reconfigurable neural network and difference-square neuron and its abstract is a reconfigurable neural network includes several switches each having at least two conductive leads, data flow direction of the conductive leads is programmed to select one of the conductive leads as input switch lead and select another one of the conductive leads as an output switch lead. several processing elements each having leads connected to the switches, where the processing elements and the switches are interconnected in one-dimension manner. the neural network of interconnected switches and processing elements has a bit-serial input and a bit-serial output. each of the processing elements comprising: (a) a serial-in-parallel-out difference-square accumulator having a first input coupled to one of the interconnected switches and generating a first output; (b) an activation function for transforming the first output of the serial-in-parallel-out difference-square accumulator and generating a second output; and (c) a parallel-in-serial-out shift register for shifting out the second output of the activation function serially to one of the interconnected switches. dated 1998-05-12"
5751915,elastic fuzzy logic system,""" an artificial intelligence system is provided which makes use of a dual subroutine to adapt weights. elastic fuzzy logic (""""elf"""") system is provided in which classical neural network learning techniques are combined with fuzzy logic techniques in order to accomplish artificial intelligence tasks such as pattern recognition, expert cloning and trajectory control. the system may be implemented in a computer provided with multiplier means and storage means for storing a vector of weights to be used as multiplier factors in an apparatus for fuzzy control. """,1998-05-12,"The title of the patent is elastic fuzzy logic system and its abstract is "" an artificial intelligence system is provided which makes use of a dual subroutine to adapt weights. elastic fuzzy logic (""""elf"""") system is provided in which classical neural network learning techniques are combined with fuzzy logic techniques in order to accomplish artificial intelligence tasks such as pattern recognition, expert cloning and trajectory control. the system may be implemented in a computer provided with multiplier means and storage means for storing a vector of weights to be used as multiplier factors in an apparatus for fuzzy control. "" dated 1998-05-12"
5751916,building management system having set offset value learning and set bias value determining system for controlling thermal environment,"a building management set value decision support apparatus includes a thermal environment index calculation/display section, and a pseudo-thermal environment index calculation/display section. the thermal environment index calculation/display section calculates/displays the current value of a thermal environment index on the basis of the measured value of an air-conditioning control target and the measured or preset values of other predetermined parameters. the pseudo-thermal environment index calculation/display section calculates/displays a pseudo-thermal environment index on the basis of a supplied building management set value and the measured or preset values of the other predetermined parameters. a set value learning apparatus, a set value determining apparatus, and a neural network operation apparatus are also disclosed.",1998-05-12,"The title of the patent is building management system having set offset value learning and set bias value determining system for controlling thermal environment and its abstract is a building management set value decision support apparatus includes a thermal environment index calculation/display section, and a pseudo-thermal environment index calculation/display section. the thermal environment index calculation/display section calculates/displays the current value of a thermal environment index on the basis of the measured value of an air-conditioning control target and the measured or preset values of other predetermined parameters. the pseudo-thermal environment index calculation/display section calculates/displays a pseudo-thermal environment index on the basis of a supplied building management set value and the measured or preset values of the other predetermined parameters. a set value learning apparatus, a set value determining apparatus, and a neural network operation apparatus are also disclosed. dated 1998-05-12"
5754661,programmable hearing aid,"a programmable hearing aid has improved signal processing, particularly improved separation of the useful signals from unwanted noise, by virtue of signals of the signal path from at least one microphone to the earphone being conducted through a neural network and being processed therein.",1998-05-19,"The title of the patent is programmable hearing aid and its abstract is a programmable hearing aid has improved signal processing, particularly improved separation of the useful signals from unwanted noise, by virtue of signals of the signal path from at least one microphone to the earphone being conducted through a neural network and being processed therein. dated 1998-05-19"
5755212,air-fuel ratio control system for internal combustion engine,"an air-fuel ratio control system with a high accuracy for an internal combustion engine which is capable of particularly improving the transient response characteristic irrespective of the occurrence of an air-fuel ratio sensor delay and a fuel attachment. an in-cylinder air-fuel ratio is calculated on the basis of engine data obtained in advance so that an neural network (nn) receiving a fuel injection quantity involving the past value and air quantity estimating information such as an intake pressure and outputting a calculated in-cylinder air-fuel ratio undergoes learning. in the actual control, a difference between the in-cylinder air-fuel ratio estimated in the nn and the target air-fuel ratio is taken on the basis of information such as a fuel injection quantity varying with the time and the output of the nn is partially differentiated with respect to the fuel injection quantity, so that the difference therebetween is divided by the resultant partial differential coefficient to obtain a fuel correction amount whereby the in-cylinder air-fuel ratio coincides with the target air-fuel ratio. the fuel injection quantity is corrected with this correction amount to calculate a final fuel injection quantity. that is, the in-cylinder air-fuel ratio is controlled to approach the target air-fuel ratio so that the exhaust gas air-fuel ratio equals the target air-fuel ratio.",1998-05-26,"The title of the patent is air-fuel ratio control system for internal combustion engine and its abstract is an air-fuel ratio control system with a high accuracy for an internal combustion engine which is capable of particularly improving the transient response characteristic irrespective of the occurrence of an air-fuel ratio sensor delay and a fuel attachment. an in-cylinder air-fuel ratio is calculated on the basis of engine data obtained in advance so that an neural network (nn) receiving a fuel injection quantity involving the past value and air quantity estimating information such as an intake pressure and outputting a calculated in-cylinder air-fuel ratio undergoes learning. in the actual control, a difference between the in-cylinder air-fuel ratio estimated in the nn and the target air-fuel ratio is taken on the basis of information such as a fuel injection quantity varying with the time and the output of the nn is partially differentiated with respect to the fuel injection quantity, so that the difference therebetween is divided by the resultant partial differential coefficient to obtain a fuel correction amount whereby the in-cylinder air-fuel ratio coincides with the target air-fuel ratio. the fuel injection quantity is corrected with this correction amount to calculate a final fuel injection quantity. that is, the in-cylinder air-fuel ratio is controlled to approach the target air-fuel ratio so that the exhaust gas air-fuel ratio equals the target air-fuel ratio. dated 1998-05-26"
5758021,speech recognition combining dynamic programming and neural network techniques,"recognition of speech with successive expansion of a reference vocabulary, can be used for automatic telephone dialing by voice input. neural and conventional recognition methods are performed in parallel so that during training and configuration of the neural network, a conventional recognizer operating according to the dynamic programming principle has available newly added word patterns as references for immediate use in recognition. upon completion of the training and configuration, the neural network takes over the recognition of the now expanded vocabulary.",1998-05-26,"The title of the patent is speech recognition combining dynamic programming and neural network techniques and its abstract is recognition of speech with successive expansion of a reference vocabulary, can be used for automatic telephone dialing by voice input. neural and conventional recognition methods are performed in parallel so that during training and configuration of the neural network, a conventional recognizer operating according to the dynamic programming principle has available newly added word patterns as references for immediate use in recognition. upon completion of the training and configuration, the neural network takes over the recognition of the now expanded vocabulary. dated 1998-05-26"
5758022,method and apparatus for improved speech recognition from stress-induced pronunciation variations with a neural network utilizing non-linear imaging characteristics,"speech recognition of lombard-induced speech at a high rate of recognition is shown using a neural network (nn) utilizing nonlinear imaging characteristics. in a training phase, systematic parameter changes of lombard-induced speech are trained to a neural network. in a speech recognition phase, imaging of lombard-induced speech patterns to lombard-free speech patterns takes place through the trained parameter changes.",1998-05-26,"The title of the patent is method and apparatus for improved speech recognition from stress-induced pronunciation variations with a neural network utilizing non-linear imaging characteristics and its abstract is speech recognition of lombard-induced speech at a high rate of recognition is shown using a neural network (nn) utilizing nonlinear imaging characteristics. in a training phase, systematic parameter changes of lombard-induced speech are trained to a neural network. in a speech recognition phase, imaging of lombard-induced speech patterns to lombard-free speech patterns takes place through the trained parameter changes. dated 1998-05-26"
5761326,method and apparatus for machine vision classification and tracking,"a method and apparatus for classification and tracking objects in three-dimensional space is described. a machine vision system acquires images from roadway scenes and processes the images by analyzing the intensities of edge elements within the image. the system then applies fuzzy set theory to the location and angles of each pixel after the pixel intensities have been characterized by vectors. a neural network interprets the data created by the fuzzy set operators and classifies objects within the roadway scene. the system can also track objects within the roadway scene, such as vehicles, by forecasting potential track regions and then calculating match scores for each potential track region based on how well the edge elements from the target track regions match those from the source region as weighted by the extent the edge elements have moved.",1998-06-02,"The title of the patent is method and apparatus for machine vision classification and tracking and its abstract is a method and apparatus for classification and tracking objects in three-dimensional space is described. a machine vision system acquires images from roadway scenes and processes the images by analyzing the intensities of edge elements within the image. the system then applies fuzzy set theory to the location and angles of each pixel after the pixel intensities have been characterized by vectors. a neural network interprets the data created by the fuzzy set operators and classifies objects within the roadway scene. the system can also track objects within the roadway scene, such as vehicles, by forecasting potential track regions and then calculating match scores for each potential track region based on how well the edge elements from the target track regions match those from the source region as weighted by the extent the edge elements have moved. dated 1998-06-02"
5761383,adaptive filtering neural network classifier,"an adaptive filtering neural network classifier for classifying input signals, includes a neural network and one or more adaptive filters for receiving input analog signals to be classified and generates inputs for the classifier. each adaptive filter is characterized as having a predetermined number of operating parameters. an analog to digital converter converts each input signal into a digital signal before input to the neural network. the neural network processes each digital signal to generate therefrom a plurality of weighted output signals in accordance with the type of network implemented. one of the weighted output signals represents a class for the input signal, and an error signal representing a difference between the weighted output signals and a predetermined desired output is also generated by the network. a control device responsive to the error signal generates a further set of operating filter parameters for input to each of the adaptive filters to change the operating response thereof and minimize the error signal.",1998-06-02,"The title of the patent is adaptive filtering neural network classifier and its abstract is an adaptive filtering neural network classifier for classifying input signals, includes a neural network and one or more adaptive filters for receiving input analog signals to be classified and generates inputs for the classifier. each adaptive filter is characterized as having a predetermined number of operating parameters. an analog to digital converter converts each input signal into a digital signal before input to the neural network. the neural network processes each digital signal to generate therefrom a plurality of weighted output signals in accordance with the type of network implemented. one of the weighted output signals represents a class for the input signal, and an error signal representing a difference between the weighted output signals and a predetermined desired output is also generated by the network. a control device responsive to the error signal generates a further set of operating filter parameters for input to each of the adaptive filters to change the operating response thereof and minimize the error signal. dated 1998-06-02"
5761384,fuzzy neural network system,"a fuzzy neural network system which is provided with an input layer, a membership layer front half section, a membership layer back half section, a rule layer, and an output layer; and constructs a network from a plurality of input/output items. the input layer and the membership layers are structured so as to divide each input value into three regions of fuzzy sets, respectively. the rule layer selects one each respectively from the above mentioned items divided into three regions, and is structured so as to make these the and rules for the two input items.",1998-06-02,"The title of the patent is fuzzy neural network system and its abstract is a fuzzy neural network system which is provided with an input layer, a membership layer front half section, a membership layer back half section, a rule layer, and an output layer; and constructs a network from a plurality of input/output items. the input layer and the membership layers are structured so as to divide each input value into three regions of fuzzy sets, respectively. the rule layer selects one each respectively from the above mentioned items divided into three regions, and is structured so as to make these the and rules for the two input items. dated 1998-06-02"
5761385,product and method for extracting image data,"a system and associated method for extracting image data from a representation of a geographic area to generate a digital geographic database that can be accessed by computer programs to analyze various features contained within the database. in particular, the method includes scanning an image of a geographic area, manipulating the image into a processable format of a plurality of pixels, and submitting the format to a neural network with user-guided rules. the neural network classifies the format into a plurality of categories which comprise the geographic database, thus foregoing the need to manually identify and classify each pixel of the scanned image as having a particular feature. to enhance the processing and structure of the database, an optimizer and a combiner are employed. accordingly, the method provides a digital geographic database for use by government agencies, urban planners and any user desiring to obtain information from a geographic representation.",1998-06-02,"The title of the patent is product and method for extracting image data and its abstract is a system and associated method for extracting image data from a representation of a geographic area to generate a digital geographic database that can be accessed by computer programs to analyze various features contained within the database. in particular, the method includes scanning an image of a geographic area, manipulating the image into a processable format of a plurality of pixels, and submitting the format to a neural network with user-guided rules. the neural network classifies the format into a plurality of categories which comprise the geographic database, thus foregoing the need to manually identify and classify each pixel of the scanned image as having a particular feature. to enhance the processing and structure of the database, an optimizer and a combiner are employed. accordingly, the method provides a digital geographic database for use by government agencies, urban planners and any user desiring to obtain information from a geographic representation. dated 1998-06-02"
5761386,method and apparatus for foreign exchange rate time series prediction and classification,"a method and apparatus for the prediction of time series data, specifically, the prediction of a foreign currency exchange rate. the method disclosed transforms the time series data into a difference of a series, compresses the transformed data using a log transformation, converts the compressed data into symbols, and subsequently trains one or more neural networks on the symbols such that a prediction is generated. alternative embodiments demonstrate the conversion by a self-organizing map and training by a recurrent neural network.",1998-06-02,"The title of the patent is method and apparatus for foreign exchange rate time series prediction and classification and its abstract is a method and apparatus for the prediction of time series data, specifically, the prediction of a foreign currency exchange rate. the method disclosed transforms the time series data into a difference of a series, compresses the transformed data using a log transformation, converts the compressed data into symbols, and subsequently trains one or more neural networks on the symbols such that a prediction is generated. alternative embodiments demonstrate the conversion by a self-organizing map and training by a recurrent neural network. dated 1998-06-02"
5761442,predictive neural network means and method for selecting a portfolio of securities wherein each network has been trained using data relating to a corresponding security,"a data processing system and method for selecting securities and constructing an investment portfolio is based on a set of artificial neural networks which are designed to model and track the performance of each security in a given capital market and output a parameter which is related to the expected risk adjusted return for the security. each artificial neural network is trained using a number of fundamental and price and volume history input parameters about the security and the underlying index. the system combines the expected return/appreciation potential data for each security via an optimization process to construct an investment portfolio which satisfies predetermined aggregate statistics. the data processing system receives input from the capital market and periodically evaluates the performance of the investment portfolio, rebalancing it whenever necessary to correct performance degradations.",1998-06-02,"The title of the patent is predictive neural network means and method for selecting a portfolio of securities wherein each network has been trained using data relating to a corresponding security and its abstract is a data processing system and method for selecting securities and constructing an investment portfolio is based on a set of artificial neural networks which are designed to model and track the performance of each security in a given capital market and output a parameter which is related to the expected risk adjusted return for the security. each artificial neural network is trained using a number of fundamental and price and volume history input parameters about the security and the underlying index. the system combines the expected return/appreciation potential data for each security via an optimization process to construct an investment portfolio which satisfies predetermined aggregate statistics. the data processing system receives input from the capital market and periodically evaluates the performance of the investment portfolio, rebalancing it whenever necessary to correct performance degradations. dated 1998-06-02"
5761626,system and method for distinguishing and characterizing motor vehicles for control of automatic drivers,"a motor vehicle related operating signal, such as a torque signal produced by operation of an engine of the motor vehicle, is monitored to produce a frequency signature for the vehicle. the frequency signature is filtered by a fuzzy spectral filter to extract a frequency membership for the vehicle which is utilized to generate characteristic signals representative of the vehicle. signals representative of vehicle inertia (j), vehicle horsepower (hp) and relative vehicle temperature (rvt) are extracted by a radial basis function (rbf) neural network. these characteristic signals are then used to control the vehicle directly via a powertrain control module (pcm) or via a robot driver for vehicle test purposes. for robot control, an anticipated throttle lag and an anticipated brake lag are generated and used to more accurately and repeatedly control the robot to simulate a human driver in following acceleration curves for vehicle testing purposes.",1998-06-02,"The title of the patent is system and method for distinguishing and characterizing motor vehicles for control of automatic drivers and its abstract is a motor vehicle related operating signal, such as a torque signal produced by operation of an engine of the motor vehicle, is monitored to produce a frequency signature for the vehicle. the frequency signature is filtered by a fuzzy spectral filter to extract a frequency membership for the vehicle which is utilized to generate characteristic signals representative of the vehicle. signals representative of vehicle inertia (j), vehicle horsepower (hp) and relative vehicle temperature (rvt) are extracted by a radial basis function (rbf) neural network. these characteristic signals are then used to control the vehicle directly via a powertrain control module (pcm) or via a robot driver for vehicle test purposes. for robot control, an anticipated throttle lag and an anticipated brake lag are generated and used to more accurately and repeatedly control the robot to simulate a human driver in following acceleration curves for vehicle testing purposes. dated 1998-06-02"
5764856,parallel neural networks having one neural network providing evaluated data to another neural network,"a data processing system is provided that consists of a connection of a first neural network (n.sub.1) with at least one other neural network (n.sub.21, n.sub.22, . . . , n.sub.2n). the first neural network (n.sub.1) and the at least one other neural network (n.sub.21, n.sub.22, . . . , n.sub.2n) is an associative memory. first input data (e.sub.0) are supplied to both the first neural network (n.sub.1) and also to at least the one other neural network (n.sub.21, n.sub.22, . . . , n.sub.2n) data (e.sub.11, e.sub.12, . . . , e.sub.1n) which are evaluated by at least the one other neural network (n.sub.21, n.sub.22, . . . , n.sub.2n), are supplied as further input data (e.sub.1) to the first neural network (n.sub.1).",1998-06-09,"The title of the patent is parallel neural networks having one neural network providing evaluated data to another neural network and its abstract is a data processing system is provided that consists of a connection of a first neural network (n.sub.1) with at least one other neural network (n.sub.21, n.sub.22, . . . , n.sub.2n). the first neural network (n.sub.1) and the at least one other neural network (n.sub.21, n.sub.22, . . . , n.sub.2n) is an associative memory. first input data (e.sub.0) are supplied to both the first neural network (n.sub.1) and also to at least the one other neural network (n.sub.21, n.sub.22, . . . , n.sub.2n) data (e.sub.11, e.sub.12, . . . , e.sub.1n) which are evaluated by at least the one other neural network (n.sub.21, n.sub.22, . . . , n.sub.2n), are supplied as further input data (e.sub.1) to the first neural network (n.sub.1). dated 1998-06-09"
5764858,one-dimensional signal processor with optimized solution capability,"an architecture and design of compact neural networks is presented for the maximum-likelihood sequence estimation (mlse) of one-dimensional signals, such as sound, in digital communications. optimization of a concave lyapunov function associated with a compact neural network performs a combinatorial minimization of the detection cost, and truly paralleled operations in the analog domain are achievable via the collective computational behaviors. in addition, the mlse performance can be improved by paralleled hardware annealing, a technique for obtaining optimal or near-optimal solutions in high-speed, real-time applications. for a sequence of length n, the network of complexity and throughput rate are o(l) and n/t.sub.c, respectively, where l is the number of symbols the inference spans and t.sub.c is the convergence time. the hardware architecture as well as network models, neuron models, and methods of feeding the input to the network are addressed in terms of the probability of error.",1998-06-09,"The title of the patent is one-dimensional signal processor with optimized solution capability and its abstract is an architecture and design of compact neural networks is presented for the maximum-likelihood sequence estimation (mlse) of one-dimensional signals, such as sound, in digital communications. optimization of a concave lyapunov function associated with a compact neural network performs a combinatorial minimization of the detection cost, and truly paralleled operations in the analog domain are achievable via the collective computational behaviors. in addition, the mlse performance can be improved by paralleled hardware annealing, a technique for obtaining optimal or near-optimal solutions in high-speed, real-time applications. for a sequence of length n, the network of complexity and throughput rate are o(l) and n/t.sub.c, respectively, where l is the number of symbols the inference spans and t.sub.c is the convergence time. the hardware architecture as well as network models, neuron models, and methods of feeding the input to the network are addressed in terms of the probability of error. dated 1998-06-09"
5764859,apparatus for nondestructive on-line inspection of electric resistance welding states and a method thereof,"apparatus for inspecting an electric resistance welding state including a first electrode connected to a power source, a second electrode connected to another terminal of the power source, and a welding object interposed between the first and the second electrodes. a voltage waveform measuring system includes a first analog-to-digital converter for detecting voltage applied, during a welding process, to both ends of the welding object. an electrode movement measuring system includes a sensor for detecting a change of a gap between the first and the second electrodes during the welding process, and a second analog-to-digital converter for receiving an output of the sensor. a computer system which includes a neural network inspection system for receiving outputs from the voltage waveform measuring system and from the electrode movement measuring system is provided.",1998-06-09,"The title of the patent is apparatus for nondestructive on-line inspection of electric resistance welding states and a method thereof and its abstract is apparatus for inspecting an electric resistance welding state including a first electrode connected to a power source, a second electrode connected to another terminal of the power source, and a welding object interposed between the first and the second electrodes. a voltage waveform measuring system includes a first analog-to-digital converter for detecting voltage applied, during a welding process, to both ends of the welding object. an electrode movement measuring system includes a sensor for detecting a change of a gap between the first and the second electrodes during the welding process, and a second analog-to-digital converter for receiving an output of the sensor. a computer system which includes a neural network inspection system for receiving outputs from the voltage waveform measuring system and from the electrode movement measuring system is provided. dated 1998-06-09"
5764860,learning method for multi-level neural network,"a learning method supervised by a binary teacher signal for a binary neural network comprises at least an error signal generator 10 for weighting factor updating, which generates an error signal for weighting factor updating having an opposite polarity to that of a difference signal between an output unit signal of the binary neural network and the binary teacher signal on an output unit whereat a binary output unit signal coincides with the binary teacher signal, and an amplitude which decreases by increase of distance from the binary teacher signal, when an absolute value of the difference signal is smaller than a threshold, generates an error signal which has the same polarity as that of the difference signal and an amplitude smaller than that of the difference signal, when the absolute value of the difference signal is larger than the threshold, or generates an error signal which has an amplitude equal to or smaller than that of the difference signal on an output unit providing a wrong binary output unit signal which is different from the binary teacher signal. updating the weighting factors by the error signal which is optimally generated according to discrimination between the correct binary output unit signal and the wrong one, can provide a binary neural network which converges very quickly and reliably to obtain a desired binary output and also realizes a high generalization ability.",1998-06-09,"The title of the patent is learning method for multi-level neural network and its abstract is a learning method supervised by a binary teacher signal for a binary neural network comprises at least an error signal generator 10 for weighting factor updating, which generates an error signal for weighting factor updating having an opposite polarity to that of a difference signal between an output unit signal of the binary neural network and the binary teacher signal on an output unit whereat a binary output unit signal coincides with the binary teacher signal, and an amplitude which decreases by increase of distance from the binary teacher signal, when an absolute value of the difference signal is smaller than a threshold, generates an error signal which has the same polarity as that of the difference signal and an amplitude smaller than that of the difference signal, when the absolute value of the difference signal is larger than the threshold, or generates an error signal which has an amplitude equal to or smaller than that of the difference signal on an output unit providing a wrong binary output unit signal which is different from the binary teacher signal. updating the weighting factors by the error signal which is optimally generated according to discrimination between the correct binary output unit signal and the wrong one, can provide a binary neural network which converges very quickly and reliably to obtain a desired binary output and also realizes a high generalization ability. dated 1998-06-09"
5765028,method and apparatus for providing neural intelligence to a mail query agent in an online analytical processing system,"a neural intelligent mail query agent. the neural intelligent mail query agent includes an online analytical processing system for accessing and analyzing data in at least one database, a query-by-mail system coupled to the online analytical processing system for receiving and processing queries from users for information derived from databases, and a neural network coupled to the remote query-by-mail system for providing learning capabilities in response to the remote mail queries. an expert system provides inference functions and the neural network is trained using a data stream from the databases as it is generated by the received mail queries. the neural network reports intelligence abstracts to the query-by-mail system as well as reports and organizes new rules constructed from the intelligence abstracts and existing rules.",1998-06-09,"The title of the patent is method and apparatus for providing neural intelligence to a mail query agent in an online analytical processing system and its abstract is a neural intelligent mail query agent. the neural intelligent mail query agent includes an online analytical processing system for accessing and analyzing data in at least one database, a query-by-mail system coupled to the online analytical processing system for receiving and processing queries from users for information derived from databases, and a neural network coupled to the remote query-by-mail system for providing learning capabilities in response to the remote mail queries. an expert system provides inference functions and the neural network is trained using a data stream from the databases as it is generated by the received mail queries. the neural network reports intelligence abstracts to the query-by-mail system as well as reports and organizes new rules constructed from the intelligence abstracts and existing rules. dated 1998-06-09"
5768476,parallel multi-value neural networks,"in a parallel multi-value neural network having a main neural network 16 and a sub neural network 18 coupled with the main neural network 16 in parallel for an input signal, the main neural network 16 is trained with a training input signal by using a main multi-value teacher signal, and the sub neural network is successively trained with the training input signal by using multi-value errors between a multi-value output signal of the main neural network 16 derived through a multi-value threshold means 17 and the main multi-value teacher signal, so as to compensate the multi-value errors involved in the multi-value output signal of the main neural network 16 by the multi-value output signal of the sub neural network 18 derived through a multi-value threshold means 19. a desired multi-value output signal of the parallel multi-value neural network 15 is obtained by adding in modulo the multi-value output signals of both the neural networks through a multi-value modulo adder 20.",1998-06-16,"The title of the patent is parallel multi-value neural networks and its abstract is in a parallel multi-value neural network having a main neural network 16 and a sub neural network 18 coupled with the main neural network 16 in parallel for an input signal, the main neural network 16 is trained with a training input signal by using a main multi-value teacher signal, and the sub neural network is successively trained with the training input signal by using multi-value errors between a multi-value output signal of the main neural network 16 derived through a multi-value threshold means 17 and the main multi-value teacher signal, so as to compensate the multi-value errors involved in the multi-value output signal of the main neural network 16 by the multi-value output signal of the sub neural network 18 derived through a multi-value threshold means 19. a desired multi-value output signal of the parallel multi-value neural network 15 is obtained by adding in modulo the multi-value output signals of both the neural networks through a multi-value modulo adder 20. dated 1998-06-16"
5768477,radio direction finding system for narrowband multiple signals,"a method of training a neural network to determine directions of arrival when there are multiple input signals within a narrow frequency band. for a single input signal having a given angle of arrival, an antenna array is used to sample signal values, from which spatial covariance matrix values are obtained. each value is applied at an input node of the neural network. the neural network is adjusted so that an output node associated with that signal's angle of arrival fires in response to these inputs. this process is repeated for different angles of arrival, such that for l different angles of arrival, l output nodes are trained with l different training sets of data. once trained, the neural network can be used in a direction finding system that detects whatever number of signals are present in a received wavefield having any number of signals.",1998-06-16,"The title of the patent is radio direction finding system for narrowband multiple signals and its abstract is a method of training a neural network to determine directions of arrival when there are multiple input signals within a narrow frequency band. for a single input signal having a given angle of arrival, an antenna array is used to sample signal values, from which spatial covariance matrix values are obtained. each value is applied at an input node of the neural network. the neural network is adjusted so that an output node associated with that signal's angle of arrival fires in response to these inputs. this process is repeated for different angles of arrival, such that for l different angles of arrival, l output nodes are trained with l different training sets of data. once trained, the neural network can be used in a direction finding system that detects whatever number of signals are present in a received wavefield having any number of signals. dated 1998-06-16"
5768478,artificial neurons using delta-sigma modulation,"an artificial neuron for use in a neural processing network comprises a plurality of input signal lines, an arrangement for computing a nonlinear function of the sum of the inputs multiplied by associated weights, and a saturating delta-sigma modulator which oversamples the computed value and produces an encoded neuron output signal. conversion of signals for use by these neurons preferably is performed by delta-sigma modulators at the inputs to the neurons, which may be incorporated directly into sensors. processing of the output signals from the neuron includes low-pass filtering and decimation. the present invention may be used in many diverse areas. for example, arrays of sensors with delta signal modulators may be coupled with a network of the neurons to form an intelligent vision system. linear signal processing, both conventional and adaptive, can be done by a simple neuronal system that operates linearly.",1998-06-16,"The title of the patent is artificial neurons using delta-sigma modulation and its abstract is an artificial neuron for use in a neural processing network comprises a plurality of input signal lines, an arrangement for computing a nonlinear function of the sum of the inputs multiplied by associated weights, and a saturating delta-sigma modulator which oversamples the computed value and produces an encoded neuron output signal. conversion of signals for use by these neurons preferably is performed by delta-sigma modulators at the inputs to the neurons, which may be incorporated directly into sensors. processing of the output signals from the neuron includes low-pass filtering and decimation. the present invention may be used in many diverse areas. for example, arrays of sensors with delta signal modulators may be coupled with a network of the neurons to form an intelligent vision system. linear signal processing, both conventional and adaptive, can be done by a simple neuronal system that operates linearly. dated 1998-06-16"
5769074,computer assisted methods for diagnosing diseases,"the simultaneous multi access reasoning technology system of the present invention utilizes both existing knowledge and implicit information that can be numerically extracted from training data to provide a method and apparatus for diagnosing disease and treating a patient. this technology further comprises a system for receiving patient data from another location, analyzing the data in a trained neural network, producing a diagnostic value, and optionally transmitting the diagnostic value to another location.",1998-06-23,"The title of the patent is computer assisted methods for diagnosing diseases and its abstract is the simultaneous multi access reasoning technology system of the present invention utilizes both existing knowledge and implicit information that can be numerically extracted from training data to provide a method and apparatus for diagnosing disease and treating a patient. this technology further comprises a system for receiving patient data from another location, analyzing the data in a trained neural network, producing a diagnostic value, and optionally transmitting the diagnostic value to another location. dated 1998-06-23"
5771306,method and apparatus for extracting speech related facial features for use in speech recognition systems,"the apparatus for the recognition of speech comprises an acoustic preprocessor, a visual preprocessor, and a speech classifier that operates the acoustic and visual preprocessed data. the acoustic preprocessor comprises a log mel spectrum analyzer that produces an equal mel bandwidth log power spectrum. the visual processor detects the motion of a set of fiducial markers on the speaker's face and extracts a set of normalized distance vectors describing lip and mouth movement. the speech classifier uses a multilevel time-delay neural network operating on the preprocessed acoustic and visual data to form an output probability distribution that indicates the probability of each candidate utterance having been spoken, based on the acoustic and visual data.",1998-06-23,"The title of the patent is method and apparatus for extracting speech related facial features for use in speech recognition systems and its abstract is the apparatus for the recognition of speech comprises an acoustic preprocessor, a visual preprocessor, and a speech classifier that operates the acoustic and visual preprocessed data. the acoustic preprocessor comprises a log mel spectrum analyzer that produces an equal mel bandwidth log power spectrum. the visual processor detects the motion of a set of fiducial markers on the speaker's face and extracts a set of normalized distance vectors describing lip and mouth movement. the speech classifier uses a multilevel time-delay neural network operating on the preprocessed acoustic and visual data to form an output probability distribution that indicates the probability of each candidate utterance having been spoken, based on the acoustic and visual data. dated 1998-06-23"
5771311,method and apparatus for correction of color shifts due to illuminant changes,"color separation values such as cmyk values from an input device are transformed into characteristic parameter values by a first transforming device having a neural network which has undergone learning in advance in such a manner as to output characteristic parameter values obtained by multivariate analysis of spectral distributions which correspond to the color separation values and are illuminant-independent. these characteristic parameter values are subjected to a linear transformation by a second transforming device, the linear transformation having a constraint that calorimetric values under a specified illuminant are to be equal or another similar constraint, by using mean vectors and eigenvectors (principal component vectors) of predetermined spectral reflectances stored in a storage device. the transformed characteristic parameter values are transformed into color separation values of a target color reproduction device by a third transforming device having a neural network which has undergone learning in advance.",1998-06-23,"The title of the patent is method and apparatus for correction of color shifts due to illuminant changes and its abstract is color separation values such as cmyk values from an input device are transformed into characteristic parameter values by a first transforming device having a neural network which has undergone learning in advance in such a manner as to output characteristic parameter values obtained by multivariate analysis of spectral distributions which correspond to the color separation values and are illuminant-independent. these characteristic parameter values are subjected to a linear transformation by a second transforming device, the linear transformation having a constraint that calorimetric values under a specified illuminant are to be equal or another similar constraint, by using mean vectors and eigenvectors (principal component vectors) of predetermined spectral reflectances stored in a storage device. the transformed characteristic parameter values are transformed into color separation values of a target color reproduction device by a third transforming device having a neural network which has undergone learning in advance. dated 1998-06-23"
5774230,color image processing apparatus for color-correcting input image data using neural network,"a color image processing apparatus including a read device for reading an original image, a color correcting device for color-correcting input image data using a color-correcting neural network, and an output device for outputting color-corrected image data. the neural network with teacher data, which are in effect learning color data reflecting the visual characteristics of human eyes, using learning data produced by reading a copy image produced based on the learning color data. accordingly, the neural network can correct colors accurately in a manner to fully reflect the visual characteristics of human eyes, thereby making the color difference between the original image and a copy of the same less noticeable.",1998-06-30,"The title of the patent is color image processing apparatus for color-correcting input image data using neural network and its abstract is a color image processing apparatus including a read device for reading an original image, a color correcting device for color-correcting input image data using a color-correcting neural network, and an output device for outputting color-corrected image data. the neural network with teacher data, which are in effect learning color data reflecting the visual characteristics of human eyes, using learning data produced by reading a copy image produced based on the learning color data. accordingly, the neural network can correct colors accurately in a manner to fully reflect the visual characteristics of human eyes, thereby making the color difference between the original image and a copy of the same less noticeable. dated 1998-06-30"
5774376,structural health monitoring using active members and neural networks,"a system for monitoring the structural integrity of a mechanical structure. the system utilizes a trainable adaptive interpreter such as a neural network to analyze data from the structure to characterize the structure's health. an actuator is attached to the mechanical structure for generating vibrations in response to an input signal. a sensor, also attached to the mechanical structure, senses the vibrations and generates an output signal in response thereto. the sensor output signal is then coupled to a pre-trained adaptive interpreter for generating an output which characterizes the structural integrity of the mechanical structure. the system can provide continual health monitoring of a structural system to detect structural damage and pinpoint probable location of the damage. the system can operate while the structural system is in service there by significantly reducing structural inspection costs.",1998-06-30,"The title of the patent is structural health monitoring using active members and neural networks and its abstract is a system for monitoring the structural integrity of a mechanical structure. the system utilizes a trainable adaptive interpreter such as a neural network to analyze data from the structure to characterize the structure's health. an actuator is attached to the mechanical structure for generating vibrations in response to an input signal. a sensor, also attached to the mechanical structure, senses the vibrations and generates an output signal in response thereto. the sensor output signal is then coupled to a pre-trained adaptive interpreter for generating an output which characterizes the structural integrity of the mechanical structure. the system can provide continual health monitoring of a structural system to detect structural damage and pinpoint probable location of the damage. the system can operate while the structural system is in service there by significantly reducing structural inspection costs. dated 1998-06-30"
5774631,3-d reconstruction of objects by artificial intelligence: apparatus and method,""" an unknown object is non-destructively and quantitatively evaluated for three-dimensional spatial distribution of a set of material constitutive parameters, using a multi-element array-source transducer and a multi-element array-detector transducer in spaced, mutually facing relation. the array-source transducer exposes the array-detector transducer to a set of source-field patterns pursuant to a set of electrical input signals. either a known object or an unknown object positioned between these transducers will be the cause of scattering, thus presenting a scattered-field pattern to the array detector transducer, for each pattern of the set of source-field patterns. a computer, a signal processor and a neural network operate from detector response to each set of scattered-field patterns, in each of two modes. in an initial mode, the neural network is """"trained"""" or configured to process a set of transfer functions involved in array-detector response to scattered-field patterns produced by the known object; in another mode, the neural network utilizes its """"trained"""" configuration in application to a set of transfer functions involved in array-detector response to scattered-field patterns produced by an unknown object, to generate estimates of the three-dimensional spatial distribution of the material constitutive parameters of the unknown object. """,1998-06-30,"The title of the patent is 3-d reconstruction of objects by artificial intelligence: apparatus and method and its abstract is "" an unknown object is non-destructively and quantitatively evaluated for three-dimensional spatial distribution of a set of material constitutive parameters, using a multi-element array-source transducer and a multi-element array-detector transducer in spaced, mutually facing relation. the array-source transducer exposes the array-detector transducer to a set of source-field patterns pursuant to a set of electrical input signals. either a known object or an unknown object positioned between these transducers will be the cause of scattering, thus presenting a scattered-field pattern to the array detector transducer, for each pattern of the set of source-field patterns. a computer, a signal processor and a neural network operate from detector response to each set of scattered-field patterns, in each of two modes. in an initial mode, the neural network is """"trained"""" or configured to process a set of transfer functions involved in array-detector response to scattered-field patterns produced by the known object; in another mode, the neural network utilizes its """"trained"""" configuration in application to a set of transfer functions involved in array-detector response to scattered-field patterns produced by an unknown object, to generate estimates of the three-dimensional spatial distribution of the material constitutive parameters of the unknown object. "" dated 1998-06-30"
5774633,supporting neural network method for process operation,"a method and a system for causing a neural circuit model to learn typical past control results of a process and using the neural circuit model for supporting an operation of the process. the neural circuit model is caused to learn by using, as input signals, a typical pattern of values of input variables at different points in time and, as a teacher signal, its corresponding values of the control variable. an unlearned pattern of input variables is inputted to the thus-learned neuron circuit model, whereby a corresponding value of the control variable is determined. preferably, plural patterns at given time intervals can be simultaneously used as patterns to be learned.",1998-06-30,"The title of the patent is supporting neural network method for process operation and its abstract is a method and a system for causing a neural circuit model to learn typical past control results of a process and using the neural circuit model for supporting an operation of the process. the neural circuit model is caused to learn by using, as input signals, a typical pattern of values of input variables at different points in time and, as a teacher signal, its corresponding values of the control variable. an unlearned pattern of input variables is inputted to the thus-learned neuron circuit model, whereby a corresponding value of the control variable is determined. preferably, plural patterns at given time intervals can be simultaneously used as patterns to be learned. dated 1998-06-30"
5774823,method of generation correction tables for misfire detection using neural networks,"a method of automating the calibration of lookup tables containing correction values to be used in an on-board vehicle system is disclosed. the method includes training a neural network to model engine behavior by outputting cylinder specific crankshaft acceleration correction values in response to any engine speed and load input conditions. the correction values generated are stored in a memory device. the training takes place off-board the vehicle, using a data set previously obtained from operating a representative engine under normal operating conditions.",1998-06-30,"The title of the patent is method of generation correction tables for misfire detection using neural networks and its abstract is a method of automating the calibration of lookup tables containing correction values to be used in an on-board vehicle system is disclosed. the method includes training a neural network to model engine behavior by outputting cylinder specific crankshaft acceleration correction values in response to any engine speed and load input conditions. the correction values generated are stored in a memory device. the training takes place off-board the vehicle, using a data set previously obtained from operating a representative engine under normal operating conditions. dated 1998-06-30"
5774831,system for improving average accuracy of signals from global positioning system by using a neural network to obtain signal correction values,"a neural network is used to process raw, uncorrected signals received by a global positioning system (gps) receiver to obtain signal corrections which are used to correct the raw signals and obtain highly accurate position coordinate data. the neural network is trained with a particular gps receiver. once the neural network is trained, the weight matrices used for the particular gps receiver may be used in different gps receivers without requiring training of the different gps receivers.",1998-06-30,"The title of the patent is system for improving average accuracy of signals from global positioning system by using a neural network to obtain signal correction values and its abstract is a neural network is used to process raw, uncorrected signals received by a global positioning system (gps) receiver to obtain signal corrections which are used to correct the raw signals and obtain highly accurate position coordinate data. the neural network is trained with a particular gps receiver. once the neural network is trained, the weight matrices used for the particular gps receiver may be used in different gps receivers without requiring training of the different gps receivers. dated 1998-06-30"
5774868,automatic sales promotion selection system and method,"an automated sales promotion selection system uses neural networks to identify promising sales promotions based on recent customer purchases. the system includes a customer information device that receives customer data relating to customer purchases of items from an inventory of items, a central processing unit having a sales promotion neural network and a storage unit containing a plurality of item identifiers comprising potential customer purchases of additional items from the inventory, wherein the sales opportunity neural network responds to customer data received from the customer information device by determining if one or more of the item identifiers in the storage unit corresponds to an item likely to be purchased by one of the customers, and an output device that receives the item identifiers of the likely purchases determined by the sales promotion neural network and produces a sales promotion relating to at least one of the item identifiers.",1998-06-30,"The title of the patent is automatic sales promotion selection system and method and its abstract is an automated sales promotion selection system uses neural networks to identify promising sales promotions based on recent customer purchases. the system includes a customer information device that receives customer data relating to customer purchases of items from an inventory of items, a central processing unit having a sales promotion neural network and a storage unit containing a plurality of item identifiers comprising potential customer purchases of additional items from the inventory, wherein the sales opportunity neural network responds to customer data received from the customer information device by determining if one or more of the item identifiers in the storage unit corresponds to an item likely to be purchased by one of the customers, and an output device that receives the item identifiers of the likely purchases determined by the sales promotion neural network and produces a sales promotion relating to at least one of the item identifiers. dated 1998-06-30"
5775330,neurometric assessment of intraoperative anesthetic,"the present invention is a method and apparatus for collecting eeg data, reducing the eeg data into coefficients, and correlating those coefficients with a depth of unconsciousness or anesthetic depth, and which obtains a bounded first derivative of anesthetic depth to indicate trends. the present invention provides a developed artificial neural network based method capable of continuously analyzing eeg data to discriminate between awake and anesthetized states in an individual and continuously monitoring anesthetic depth trends in real-time. the present invention enables an anesthesiologist to respond immediately to changes in anesthetic depth of the patient during surgery and to administer the correct amount of anesthetic.",1998-07-07,"The title of the patent is neurometric assessment of intraoperative anesthetic and its abstract is the present invention is a method and apparatus for collecting eeg data, reducing the eeg data into coefficients, and correlating those coefficients with a depth of unconsciousness or anesthetic depth, and which obtains a bounded first derivative of anesthetic depth to indicate trends. the present invention provides a developed artificial neural network based method capable of continuously analyzing eeg data to discriminate between awake and anesthetized states in an individual and continuously monitoring anesthetic depth trends in real-time. the present invention enables an anesthesiologist to respond immediately to changes in anesthetic depth of the patient during surgery and to administer the correct amount of anesthetic. dated 1998-07-07"
5775806,infrared assessment system,"a process and system for determining the integrity of an object by analyzing its dynamic heat properties is disclosed. a properly functioning reference object is heated and an infrared camera is positioned above the object. a digital computer collects the infrared images of the object and analyzes its dynamic heat properties. only heat changes within predefined regions of the object are analyzed and the data is reduced to a peak temperature across each region for a given time. this data is stored and used for future reference. next, a test object is sampled in the same manner. the reference and test object data is then compared by employing a complex neural network. the neural network uses confidence estimates and historical data on similar reference and test objects to determine the integrity of the object. thermal images of objects under test are graphically displayed on a video screen. out-of-profile regions are indicated on the thermal image by displaying the region's actual heating rate and a range of acceptable heating rates.",1998-07-07,"The title of the patent is infrared assessment system and its abstract is a process and system for determining the integrity of an object by analyzing its dynamic heat properties is disclosed. a properly functioning reference object is heated and an infrared camera is positioned above the object. a digital computer collects the infrared images of the object and analyzes its dynamic heat properties. only heat changes within predefined regions of the object are analyzed and the data is reduced to a peak temperature across each region for a given time. this data is stored and used for future reference. next, a test object is sampled in the same manner. the reference and test object data is then compared by employing a complex neural network. the neural network uses confidence estimates and historical data on similar reference and test objects to determine the integrity of the object. thermal images of objects under test are graphically displayed on a video screen. out-of-profile regions are indicated on the thermal image by displaying the region's actual heating rate and a range of acceptable heating rates. dated 1998-07-07"
5776063,analysis of ultrasound images in the presence of contrast agent,""" a method and apparatus for directly identifying and characterizing input data derived from regions of interest in ultrasound images of organs in the presence of attenuation from interposed contrast agent, for the purpose of diagnosing abnormalities. the input data is classified into one of a number of classes depending upon the characteristics of that data, in order to distinguish normal conditions from abnormal conditions. the invention is based on the recognition that significant information relating to the health of tissue exists in regions of interest in ultrasound images in the presence of attenuation from interposed contrast agent. this information is in the form of backscatter speckle patterns that have """"texture"""" characteristics that are distinguishable in healthy versus diseased tissue. the invention classifies such patterns as probably normal or abnormal by means of an analysis system that may include a neural network system. the preferred embodiment of the present invention includes: (1) a data acquisition system for acquiring ultrasound image data indicative of a region of interest in the presence of attenuation from interposed contrast agent; (2) an optional signal conditioning stage to remove signals (e.g., noise) from the input data; and (3) an analysis system designed to detect """"texture"""" characteristics that distinguish healthy tissue from diseased tissue even in the presence of the contrast agent. the output classifies the input data in a uniform, unambiguous manner. the invention is preferably implemented as a computer program executing on a programmable computer. """,1998-07-07,"The title of the patent is analysis of ultrasound images in the presence of contrast agent and its abstract is "" a method and apparatus for directly identifying and characterizing input data derived from regions of interest in ultrasound images of organs in the presence of attenuation from interposed contrast agent, for the purpose of diagnosing abnormalities. the input data is classified into one of a number of classes depending upon the characteristics of that data, in order to distinguish normal conditions from abnormal conditions. the invention is based on the recognition that significant information relating to the health of tissue exists in regions of interest in ultrasound images in the presence of attenuation from interposed contrast agent. this information is in the form of backscatter speckle patterns that have """"texture"""" characteristics that are distinguishable in healthy versus diseased tissue. the invention classifies such patterns as probably normal or abnormal by means of an analysis system that may include a neural network system. the preferred embodiment of the present invention includes: (1) a data acquisition system for acquiring ultrasound image data indicative of a region of interest in the presence of attenuation from interposed contrast agent; (2) an optional signal conditioning stage to remove signals (e.g., noise) from the input data; and (3) an analysis system designed to detect """"texture"""" characteristics that distinguish healthy tissue from diseased tissue even in the presence of the contrast agent. the output classifies the input data in a uniform, unambiguous manner. the invention is preferably implemented as a computer program executing on a programmable computer. "" dated 1998-07-07"
5778151,method and control device for controlling a material-processing process,"in the control of a material-processing process in a regulated system, a preliminary adjustment of the system takes place at the beginning of each process cycle as a function of a precalculated process parameter. a material characteristic which is relevant for the process and which in turn is dependent on state variables (such as the composition of the material and its temperature), is included in an advance calculation of the process parameter. the relationship between the state variables and the material property is modelled in a neural network which forms a prediction value for the material property on its output side. as a function of the deviation between the prediction value and an actual value for the material property which is determined based on measuring the process parameter during the process cycle, an adaptive change of the network parameters takes place in the sense of reducing this deviation.",1998-07-07,"The title of the patent is method and control device for controlling a material-processing process and its abstract is in the control of a material-processing process in a regulated system, a preliminary adjustment of the system takes place at the beginning of each process cycle as a function of a precalculated process parameter. a material characteristic which is relevant for the process and which in turn is dependent on state variables (such as the composition of the material and its temperature), is included in an advance calculation of the process parameter. the relationship between the state variables and the material property is modelled in a neural network which forms a prediction value for the material property on its output side. as a function of the deviation between the prediction value and an actual value for the material property which is determined based on measuring the process parameter during the process cycle, an adaptive change of the network parameters takes place in the sense of reducing this deviation. dated 1998-07-07"
5778152,training method for neural network,"a neural network designed to recognize a particular character is supplied with initial tap weights for a first hidden node which are an image of the character to be recognized. the additive inverse of this set of weights is used as the tap weights for a second hidden node. a third node, if used, is initialized with random noise. the network is then trained with back propagation.",1998-07-07,"The title of the patent is training method for neural network and its abstract is a neural network designed to recognize a particular character is supplied with initial tap weights for a first hidden node which are an image of the character to be recognized. the additive inverse of this set of weights is used as the tap weights for a second hidden node. a third node, if used, is initialized with random noise. the network is then trained with back propagation. dated 1998-07-07"
5778153,neural network utilizing logarithmic function and method of using same,"a neural network, which may be implemented either in hardware or software, is constructed of neurons or neuron circuits each having only one significant processing element in the form of an adder. each neural network further includes circuits for applying a logarithmic function to its inputs and for applying an inverse-logarithmic function to the outputs of its neurons. the neural network utilizes a training algorithm which does not require repetitive training and which yields a global minimum to each given set of input vectors.",1998-07-07,"The title of the patent is neural network utilizing logarithmic function and method of using same and its abstract is a neural network, which may be implemented either in hardware or software, is constructed of neurons or neuron circuits each having only one significant processing element in the form of an adder. each neural network further includes circuits for applying a logarithmic function to its inputs and for applying an inverse-logarithmic function to the outputs of its neurons. the neural network utilizes a training algorithm which does not require repetitive training and which yields a global minimum to each given set of input vectors. dated 1998-07-07"
5778279,image forming apparatus estimating a consumable life of a component using fuzzy logic,"a copying machine equipped with a pc counter for counting cumulative rotation time of a photoconductive drum, a development counter for counting cumulative drive time of a developing unit, a sensor for detecting deposition amount of toner onto the photoconductive drum, a sensor for detecting toner concentration in the developer, and a humidity sensor. based on output values from the counters and sensors, a control section of the copying machine estimates the degree of consumption of a consumable article by using fuzzy inference method or neural network, and thereby decides whether the consumable article have reached an end of life.",1998-07-07,"The title of the patent is image forming apparatus estimating a consumable life of a component using fuzzy logic and its abstract is a copying machine equipped with a pc counter for counting cumulative rotation time of a photoconductive drum, a development counter for counting cumulative drive time of a developing unit, a sensor for detecting deposition amount of toner onto the photoconductive drum, a sensor for detecting toner concentration in the developer, and a humidity sensor. based on output values from the counters and sensors, a control section of the copying machine estimates the degree of consumption of a consumable article by using fuzzy inference method or neural network, and thereby decides whether the consumable article have reached an end of life. dated 1998-07-07"
5781128,data compression system and method,"a method and system for encoding an input signal for storage or transmission over a communications channel by transforming successive vectors of parameters of the input signal into a corresponding succession of index signals, each index signal being associated with a quantized vector that corresponds to an ordered set of values of the input signal parameters. supplying the input vector and a set of distance parameters to an artificial neural network for causing the artificial neural network to produce at least one control output signal; providing a vector quantization system composed of: at least one table having a first plurality of storage locations, each storage location storing a representative vector and having an address represented by a respective index signal; and search means for comparing each representative vector with each input signal parameter value; applying the at least one control output signal to the vector quantization system for identifying a second plurality of the storage locations, which second plurality is a subset of the first plurality of storage locations; searching, in said search means, over the second plurality of storage locations to locate that storage location of the second plurality which most closely approximates the extracted features vector according to a criterion utilizing the set of distance parameters; and outputting the index of the storage location which is found to most closely approximate the input vector.",1998-07-14,"The title of the patent is data compression system and method and its abstract is a method and system for encoding an input signal for storage or transmission over a communications channel by transforming successive vectors of parameters of the input signal into a corresponding succession of index signals, each index signal being associated with a quantized vector that corresponds to an ordered set of values of the input signal parameters. supplying the input vector and a set of distance parameters to an artificial neural network for causing the artificial neural network to produce at least one control output signal; providing a vector quantization system composed of: at least one table having a first plurality of storage locations, each storage location storing a representative vector and having an address represented by a respective index signal; and search means for comparing each representative vector with each input signal parameter value; applying the at least one control output signal to the vector quantization system for identifying a second plurality of the storage locations, which second plurality is a subset of the first plurality of storage locations; searching, in said search means, over the second plurality of storage locations to locate that storage location of the second plurality which most closely approximates the extracted features vector according to a criterion utilizing the set of distance parameters; and outputting the index of the storage location which is found to most closely approximate the input vector. dated 1998-07-14"
5781432,method and apparatus for analyzing a neural network within desired operating parameter constraints,"a distributed control system (14) receives on the input thereof the control inputs and then outputs control signals to a plant (10) for the operation thereof. the measured variables of the plant and the control inputs are input to a predictive model (34) that operates in conjunction with an inverse model (36) to generate predicted control inputs. the predicted control inputs are processed through a filter (46) to apply hard constraints, the values of which are received from a control parameter block (22). during operation, predetermined criterion stored in the control parameter block (22) are utilized by a cost minimization block (42) to generate an error control signal which is minimized by the inverse model (36) to generate the control signals. the system works in two modes, an analyze mode and a runtime mode. in the analyze mode, the predictive model (34) and the inverse model (36) are connected to either training data or simulated data from the analyzer (30) and the operation of the plant (10) evaluated. the values of the hard constraints in filter (46) and the criterion utilized for the cost minimization (42) can then be varied to change the constraints on the control signals input to the control network, the predicted output of the predictive model (34) and the hard constraints stored in the filter (46). cost coefficients can be utilized as the criterion to set the input values in accordance with predetermined cost constraints.",1998-07-14,"The title of the patent is method and apparatus for analyzing a neural network within desired operating parameter constraints and its abstract is a distributed control system (14) receives on the input thereof the control inputs and then outputs control signals to a plant (10) for the operation thereof. the measured variables of the plant and the control inputs are input to a predictive model (34) that operates in conjunction with an inverse model (36) to generate predicted control inputs. the predicted control inputs are processed through a filter (46) to apply hard constraints, the values of which are received from a control parameter block (22). during operation, predetermined criterion stored in the control parameter block (22) are utilized by a cost minimization block (42) to generate an error control signal which is minimized by the inverse model (36) to generate the control signals. the system works in two modes, an analyze mode and a runtime mode. in the analyze mode, the predictive model (34) and the inverse model (36) are connected to either training data or simulated data from the analyzer (30) and the operation of the plant (10) evaluated. the values of the hard constraints in filter (46) and the criterion utilized for the cost minimization (42) can then be varied to change the constraints on the control signals input to the control network, the predicted output of the predictive model (34) and the hard constraints stored in the filter (46). cost coefficients can be utilized as the criterion to set the input values in accordance with predetermined cost constraints. dated 1998-07-14"
5781659,ocr classification based on transition ground data,"an ocr system 10 classifies an input image vector of an unclassified symbol with respect to a library 14t of template image vectors of pre-classified characters. each template vector is in the form of a sequence of elements representing the image intensity level of a pixel within the character defined by that template vector. each template element is part of the image background, foreground, or transition ground between the background and foreground. each input vector, like the template vectors, is also in the form of a sequence of elements. however, in the input vector, each element represents the sum or an image intensity level signal component defining the symbol within the image of the unclassified symbol plus a greyscale noise component. each input element is also part of the background, foreground, or transition ground. the input vector and at least one of the template vectors are entered into a classifier device 18. the input vector is classified based on the backgrounds, foregrounds, and transition grounds. the presence of transition ground in the input vector and the template vector produces a robust classification response with a more uniform correlation coefficient between repeated classifications of the same input symbol. the classifier device may be a distance function classifier or a neural network classifier.",1998-07-14,"The title of the patent is ocr classification based on transition ground data and its abstract is an ocr system 10 classifies an input image vector of an unclassified symbol with respect to a library 14t of template image vectors of pre-classified characters. each template vector is in the form of a sequence of elements representing the image intensity level of a pixel within the character defined by that template vector. each template element is part of the image background, foreground, or transition ground between the background and foreground. each input vector, like the template vectors, is also in the form of a sequence of elements. however, in the input vector, each element represents the sum or an image intensity level signal component defining the symbol within the image of the unclassified symbol plus a greyscale noise component. each input element is also part of the background, foreground, or transition ground. the input vector and at least one of the template vectors are entered into a classifier device 18. the input vector is classified based on the backgrounds, foregrounds, and transition grounds. the presence of transition ground in the input vector and the template vector produces a robust classification response with a more uniform correlation coefficient between repeated classifications of the same input symbol. the classifier device may be a distance function classifier or a neural network classifier. dated 1998-07-14"
5781700,trained neural network air/fuel control system,"an electronic engine control (eec) module executes both open loop and closed loop neural network processes to control the air/fuel mixture ratio of a vehicle engine to hold the fuel mixture at stoichiometry. the open loop neural network provides transient air/fuel control to provide a base stoichiometric air/fuel mixture ratio signal in response to throttle position under current engine speed and load conditions. the base air/fuel mixture ratio signal from the open loop network is additively combined with a closed loop trimming signal which varies the air/fuel mixture ratio in response to variations in the sensed exhaust gas oxygen level. each neural network function is defined by a unitary data structure which defines the network architecture, including the number of node layers, the number of nodes per layer, and the interconnections between nodes. in addition, the data structure holds weight values which determine the manner in which network signals are combined. the network definition data structures are created by a network training system which utilizes an external training processor which employs gradient methods to derive network weight values in accordance with a cost function which quantitatively defines system objectives and an identification network which is pretrained to provide gradient signals representative of the behavior of the physical plant. the training processor executes training cycles asynchronously with the operation of the eec module in a representative test vehicle.",1998-07-14,"The title of the patent is trained neural network air/fuel control system and its abstract is an electronic engine control (eec) module executes both open loop and closed loop neural network processes to control the air/fuel mixture ratio of a vehicle engine to hold the fuel mixture at stoichiometry. the open loop neural network provides transient air/fuel control to provide a base stoichiometric air/fuel mixture ratio signal in response to throttle position under current engine speed and load conditions. the base air/fuel mixture ratio signal from the open loop network is additively combined with a closed loop trimming signal which varies the air/fuel mixture ratio in response to variations in the sensed exhaust gas oxygen level. each neural network function is defined by a unitary data structure which defines the network architecture, including the number of node layers, the number of nodes per layer, and the interconnections between nodes. in addition, the data structure holds weight values which determine the manner in which network signals are combined. the network definition data structures are created by a network training system which utilizes an external training processor which employs gradient methods to derive network weight values in accordance with a cost function which quantitatively defines system objectives and an identification network which is pretrained to provide gradient signals representative of the behavior of the physical plant. the training processor executes training cycles asynchronously with the operation of the eec module in a representative test vehicle. dated 1998-07-14"
5781701,neural network and method of using same,"a method of operating a neural network and a neural network, which is implemented either in hardware or software, is constructed of neurons or neuron circuits each having only one significant processing element in the form of a multiplier. the neural network utilizes a training algorithm which does not require repetitive training and which yields a global minimum to each given set of input vectors.",1998-07-14,"The title of the patent is neural network and method of using same and its abstract is a method of operating a neural network and a neural network, which is implemented either in hardware or software, is constructed of neurons or neuron circuits each having only one significant processing element in the form of a multiplier. the neural network utilizes a training algorithm which does not require repetitive training and which yields a global minimum to each given set of input vectors. dated 1998-07-14"
5781702,hybrid chip-set architecture for artificial neural network system,"a self-contained chip set architecture for ann systems, based on back-propagation model with full-connectivity topology, and on-chip learning and refreshing, based on analog chip set technology providing self-contained synapse and neuron modules with fault tolerant neural computing, capable of growing to any arbitrary size as a result of embedded electronic addressing. direct analog and digital i/o ports allow real-time computation and interface communication with other systems including digital host of any bus bandwidth. scalability is provided, allowing accommodation of all input/output data sizes and different host platform.",1998-07-14,"The title of the patent is hybrid chip-set architecture for artificial neural network system and its abstract is a self-contained chip set architecture for ann systems, based on back-propagation model with full-connectivity topology, and on-chip learning and refreshing, based on analog chip set technology providing self-contained synapse and neuron modules with fault tolerant neural computing, capable of growing to any arbitrary size as a result of embedded electronic addressing. direct analog and digital i/o ports allow real-time computation and interface communication with other systems including digital host of any bus bandwidth. scalability is provided, allowing accommodation of all input/output data sizes and different host platform. dated 1998-07-14"
5782763,electromagnetic bone-assessment apparatus and method,"non-invasive quantitative in-vivo electromagnetic evaluation of bone is performed by subjecting bone to an electrical excitation wave-form supplied to a pair of electrodes on opposite sides of a bony member, and involving a repetitive finite duration signal consisting of plural frequencies that are in the range 0 hz-200 mhz. signal-processing of a bone-current response signal and a bone-voltage response signal is operative to sequentially average the most recently received given number of successive bone-current and bone-voltage response signals to obtain an averaged per-pulse bone-current signal and an averaged per-pulse bone-voltage signal, and to produce their associated fourier transforms. these fourier transforms are further processed to obtain the frequency-dependent bone-admittance function. in a separate operation, the same electrodes respond to the same excitation signal via a medium of known electromagnetic properties and path length to establish a reference-voltage signal and reference-current signal, which are processed to produce their associated fourier transforms. these two fourier transforms are further processed to produce a frequency-dependent reference-admittance function, which together with the bone-admittance function are processed to derive the frequency-dependent bone-conductivity real function, .sigma.'.sub.b (f), and frequency-dependent dielectric bone-permittivity real function, .di-elect cons.'.sub.b (f). the function .sigma.'.sub.b (f) is related to the energy loss in the bony member, and the function .di-elect cons.'.sub.b (f) is related to the energy storage in the bony member. a neural network, configured to generate an estimate of one or more of the desired bone-related quantities, is connected for response to the functions .sigma.'.sub.b (f) and .di-elect cons.'.sub.b (f), whereby to generate the indicated estimates of bone status, namely, bone-density, bone-architecture, bone-strength and bone-fracture risk.",1998-07-21,"The title of the patent is electromagnetic bone-assessment apparatus and method and its abstract is non-invasive quantitative in-vivo electromagnetic evaluation of bone is performed by subjecting bone to an electrical excitation wave-form supplied to a pair of electrodes on opposite sides of a bony member, and involving a repetitive finite duration signal consisting of plural frequencies that are in the range 0 hz-200 mhz. signal-processing of a bone-current response signal and a bone-voltage response signal is operative to sequentially average the most recently received given number of successive bone-current and bone-voltage response signals to obtain an averaged per-pulse bone-current signal and an averaged per-pulse bone-voltage signal, and to produce their associated fourier transforms. these fourier transforms are further processed to obtain the frequency-dependent bone-admittance function. in a separate operation, the same electrodes respond to the same excitation signal via a medium of known electromagnetic properties and path length to establish a reference-voltage signal and reference-current signal, which are processed to produce their associated fourier transforms. these two fourier transforms are further processed to produce a frequency-dependent reference-admittance function, which together with the bone-admittance function are processed to derive the frequency-dependent bone-conductivity real function, .sigma.'.sub.b (f), and frequency-dependent dielectric bone-permittivity real function, .di-elect cons.'.sub.b (f). the function .sigma.'.sub.b (f) is related to the energy loss in the bony member, and the function .di-elect cons.'.sub.b (f) is related to the energy storage in the bony member. a neural network, configured to generate an estimate of one or more of the desired bone-related quantities, is connected for response to the functions .sigma.'.sub.b (f) and .di-elect cons.'.sub.b (f), whereby to generate the indicated estimates of bone status, namely, bone-density, bone-architecture, bone-strength and bone-fracture risk. dated 1998-07-21"
5782885,rate responsive heart stimulation device using neural network and iegm classifier,"in a method for classifying iegm waveforms in dependence of the workload of a patient, a predetermined number of iegm-signals, each signal extending over at least one segment of one heart beat cycle, are registered, whereafter the iegm-signals are fed to a neural network and an encoded form of the signals is formed using said neural network. this encoded form is stored in a memory, for use in classifying further registered iegm-signals. in an active cardiac implant connectable to an implantable electrode arrangement adapted for in vivo delivery of stimulation pulses to a heart, iegm signals present are obtained from one or more of the electrodes. a pulse generator connected to the electrode arrangement, generates and emits stimulation pulses with a variable stimulation interval between successive stimulation pulses. the implant also contains a classifying device for classification of a predetermined number of iegm-signals registered during predetermined time intervals at predetermined points of time according to a predetermined classification stored in the classification device, this classification being related to preregistered waveforms of measured iegm-signals. a control unit in the implant supplies a control signal to a control input of the pulse generator dependent on the classification of each of the registered iegm signals. the control signal causes the pulse generator to adjust the stimulation rate dependent on each of the registered iegm-signals.",1998-07-21,"The title of the patent is rate responsive heart stimulation device using neural network and iegm classifier and its abstract is in a method for classifying iegm waveforms in dependence of the workload of a patient, a predetermined number of iegm-signals, each signal extending over at least one segment of one heart beat cycle, are registered, whereafter the iegm-signals are fed to a neural network and an encoded form of the signals is formed using said neural network. this encoded form is stored in a memory, for use in classifying further registered iegm-signals. in an active cardiac implant connectable to an implantable electrode arrangement adapted for in vivo delivery of stimulation pulses to a heart, iegm signals present are obtained from one or more of the electrodes. a pulse generator connected to the electrode arrangement, generates and emits stimulation pulses with a variable stimulation interval between successive stimulation pulses. the implant also contains a classifying device for classification of a predetermined number of iegm-signals registered during predetermined time intervals at predetermined points of time according to a predetermined classification stored in the classification device, this classification being related to preregistered waveforms of measured iegm-signals. a control unit in the implant supplies a control signal to a control input of the pulse generator dependent on the classification of each of the registered iegm signals. the control signal causes the pulse generator to adjust the stimulation rate dependent on each of the registered iegm-signals. dated 1998-07-21"
5783829,energy and position sensitive radiation detectors,"a device and a method for determining the interaction position and energy measurement of a radiation detector. a plurality of waveshifting optical fiber are positioned on the surface of a scintillator which is subjected to incident radiation causing the release of photons of a first wavelength from the scintillator which are absorbed by the optical fiber with a portion of the released photons being reemitted from one end of the fibers. the reemitted photons are measured and the centroid of the distribution of these measured photons is determined in order to provide an indication of the interaction position of incident radiation. a measurement of photons generated by the incident radiation which pass through the fibers but which are not transported down said fibers provide an indication of the energy of the incident radiation. in a further embodiment, a crossed wire anode photomultiplier is used in conjunction with a plurality of scintillating fibers embedded in grooves on both sides of the scintillator in order to provide an accurate position for the centroid of a more extended distribution of incident photons. furthermore a designed neural network extracts, from the crossed wire anode photomultiplier output pulses, a linear position to provide real time imaging.",1998-07-21,"The title of the patent is energy and position sensitive radiation detectors and its abstract is a device and a method for determining the interaction position and energy measurement of a radiation detector. a plurality of waveshifting optical fiber are positioned on the surface of a scintillator which is subjected to incident radiation causing the release of photons of a first wavelength from the scintillator which are absorbed by the optical fiber with a portion of the released photons being reemitted from one end of the fibers. the reemitted photons are measured and the centroid of the distribution of these measured photons is determined in order to provide an indication of the interaction position of incident radiation. a measurement of photons generated by the incident radiation which pass through the fibers but which are not transported down said fibers provide an indication of the energy of the incident radiation. in a further embodiment, a crossed wire anode photomultiplier is used in conjunction with a plurality of scintillating fibers embedded in grooves on both sides of the scintillator in order to provide an accurate position for the centroid of a more extended distribution of incident photons. furthermore a designed neural network extracts, from the crossed wire anode photomultiplier output pulses, a linear position to provide real time imaging. dated 1998-07-21"
5784233,differential protection device of a power transformer,"a preprocessing circuit receives signals representative of a current circulating in a primary winding and of a current circulating in a secondary winding of a transformer. the signals representative of currents are used to calculate the values of a through current and a differential current. the preprocessing circuit performs a spectral analysis and provides a neural network with signals representative of the fundamental component of the through current, of the fundamental component of the differential current, of the second harmonic and of the fifth harmonic of the differential current. the neural network identifies fault conditions and normal operation states, and supplies a triggering and/or alarm signal to an output when a fault condition is detected.",1998-07-21,"The title of the patent is differential protection device of a power transformer and its abstract is a preprocessing circuit receives signals representative of a current circulating in a primary winding and of a current circulating in a secondary winding of a transformer. the signals representative of currents are used to calculate the values of a through current and a differential current. the preprocessing circuit performs a spectral analysis and provides a neural network with signals representative of the fundamental component of the through current, of the fundamental component of the differential current, of the second harmonic and of the fifth harmonic of the differential current. the neural network identifies fault conditions and normal operation states, and supplies a triggering and/or alarm signal to an output when a fault condition is detected. dated 1998-07-21"
5784485,method and apparatus for automated pattern recognition,"the present invention relates to a new and useful automated pattern recognition device comprising a neural-network system, implemented on a general purpose computer, and capable of recognizing not only printed characters but also handwritten characters and other patterns in n-dimensions. the system incorporates novel feature extraction which generates an additional dimension from an n-dimensional input pattern, for example, a three-dimensional feature pattern from a two dimensional input pattern, resulting in shift-invariance, scale-invariance, and invariance to slight rotation.",1998-07-21,"The title of the patent is method and apparatus for automated pattern recognition and its abstract is the present invention relates to a new and useful automated pattern recognition device comprising a neural-network system, implemented on a general purpose computer, and capable of recognizing not only printed characters but also handwritten characters and other patterns in n-dimensions. the system incorporates novel feature extraction which generates an additional dimension from an n-dimensional input pattern, for example, a three-dimensional feature pattern from a two dimensional input pattern, resulting in shift-invariance, scale-invariance, and invariance to slight rotation. dated 1998-07-21"
5785653,system and method for predicting internal condition of live body,"a live body internal condition predicting and expressing system is capable of predicting the internal condition of a live body on the basis of an electromagnetic field distribution at significantly reduced period. the system comprises an electromagnetic field distribution measuring portion, a training data generating portion generating a plurality of training data on the basis of an electromagnetic field distribution derived from a head model data designating a head model and a predetermined dipole parameter and an internal condition category data describing a relationship between the active region of a brain and the internal condition of the live body, an inference portion deriving a numeric value representative of the active region of the brain from the electromagnetic field distribution in the training data with employing a neural network having predetermined coupling coefficients representative of coupling condition between each unit forming each layer and transforming portion transforming the numeric value representative of the active region in the brain output from said inferencing means into an expression indicative of the internal condition of the live body.",1998-07-28,"The title of the patent is system and method for predicting internal condition of live body and its abstract is a live body internal condition predicting and expressing system is capable of predicting the internal condition of a live body on the basis of an electromagnetic field distribution at significantly reduced period. the system comprises an electromagnetic field distribution measuring portion, a training data generating portion generating a plurality of training data on the basis of an electromagnetic field distribution derived from a head model data designating a head model and a predetermined dipole parameter and an internal condition category data describing a relationship between the active region of a brain and the internal condition of the live body, an inference portion deriving a numeric value representative of the active region of the brain from the electromagnetic field distribution in the training data with employing a neural network having predetermined coupling coefficients representative of coupling condition between each unit forming each layer and transforming portion transforming the numeric value representative of the active region in the brain output from said inferencing means into an expression indicative of the internal condition of the live body. dated 1998-07-28"
5787138,supervision of a neutron detector in a nuclear reactor,"a method of monitoring neutron flex detectors in a nuclear reactor in a neural network is provided. the network comprises an input layer which receives a number of input signals corresponding to values of a measured signal at different times and an output layer which is adapted to deliver a number of state signals. the network is taught to identify a number of different trends of the measured signal such that the state signals indicate the trends, by supplying the network with a plurality of input signals which have known state signals. the network is then supplied with input signals which correspond to values of a measured signal taken at different times from a supervised detector. also, input signals corresponding to values of measured signals taken at different times from a reference detector are supplied to the network. next, state signals for the supervised detector and for the reference detectors are calculated and whether the supervised detector is defective is determined based on the state signals.",1998-07-28,"The title of the patent is supervision of a neutron detector in a nuclear reactor and its abstract is a method of monitoring neutron flex detectors in a nuclear reactor in a neural network is provided. the network comprises an input layer which receives a number of input signals corresponding to values of a measured signal at different times and an output layer which is adapted to deliver a number of state signals. the network is taught to identify a number of different trends of the measured signal such that the state signals indicate the trends, by supplying the network with a plurality of input signals which have known state signals. the network is then supplied with input signals which correspond to values of a measured signal taken at different times from a supervised detector. also, input signals corresponding to values of measured signals taken at different times from a reference detector are supplied to the network. next, state signals for the supervised detector and for the reference detectors are calculated and whether the supervised detector is defective is determined based on the state signals. dated 1998-07-28"
5787190,method and apparatus for pattern recognition of wafer test bins,"an automated system and procedure processes wafer test bin data of semiconductor wafers to formulate a fault pattern at statistically significant levels. a processor such as a neural engine or neural network collects wafer test bin results to generate a n/n wafer map to be correlated with wafer maps produced from a wafer electrical test, a wafer level reliability test, and an in-line defect analysis. a n/n wafer map generated by the processor is cross-checked with a wafer map generated from another semiconductor tester to formulate possible overlap fault patterns. the confirmed fault patterns are further analyzed by performing failure analysis to find the root cause of fault patterns. a report containing fault patterns and the root cause for fault patterns is sent back to a fab for making adjustment to the fabrication process to increase the overall yield of the future batch of semiconductor wafers. the report is also stored in a pattern database to serve as a library for future reference of previously recognized fault patterns, thereby to bypass the need to perform a failure analysis for matching fault patterns.",1998-07-28,"The title of the patent is method and apparatus for pattern recognition of wafer test bins and its abstract is an automated system and procedure processes wafer test bin data of semiconductor wafers to formulate a fault pattern at statistically significant levels. a processor such as a neural engine or neural network collects wafer test bin results to generate a n/n wafer map to be correlated with wafer maps produced from a wafer electrical test, a wafer level reliability test, and an in-line defect analysis. a n/n wafer map generated by the processor is cross-checked with a wafer map generated from another semiconductor tester to formulate possible overlap fault patterns. the confirmed fault patterns are further analyzed by performing failure analysis to find the root cause of fault patterns. a report containing fault patterns and the root cause for fault patterns is sent back to a fab for making adjustment to the fabrication process to increase the overall yield of the future batch of semiconductor wafers. the report is also stored in a pattern database to serve as a library for future reference of previously recognized fault patterns, thereby to bypass the need to perform a failure analysis for matching fault patterns. dated 1998-07-28"
5787194,system and method for image processing using segmentation of images and classification and merging of image segments using a cost function,"image processing apparatus for segmenting an input image into image portions each containing a single character, the apparatus comprising identification logic for identifying connected components in the input image; classification logic, including a neural network, for determining into which of a number of predefined classes a connected component falls, at least one of said classes indicating that the connected component is most likely to be a single character; merging logic and splitting logic for merging and splitting the connected components. the merging and splitting logic and the classification logic are arranged to operate so that the connected components are iteratively merged and/or split and the resulting split and/or merged connected components reclassified by the classification logic until an image segmentation is achieved which meets a predefined criterion.",1998-07-28,"The title of the patent is system and method for image processing using segmentation of images and classification and merging of image segments using a cost function and its abstract is image processing apparatus for segmenting an input image into image portions each containing a single character, the apparatus comprising identification logic for identifying connected components in the input image; classification logic, including a neural network, for determining into which of a number of predefined classes a connected component falls, at least one of said classes indicating that the connected component is most likely to be a single character; merging logic and splitting logic for merging and splitting the connected components. the merging and splitting logic and the classification logic are arranged to operate so that the connected components are iteratively merged and/or split and the resulting split and/or merged connected components reclassified by the classification logic until an image segmentation is achieved which meets a predefined criterion. dated 1998-07-28"
5787393,"speech recognition apparatus using neural network, and learning method therefor","a speech recognition apparatus using a neural network. a neuron-like element according to the present invention has a means for storing a value of the inner condition thereof, a means for updating a value of internal status on the basis of an output from the neuron-like element itself, outputs from other neuron-like elements and an external input, and an output value generating means for converting a value of internal status into an external output. accordingly, the neuron-like element itself can retain the history of input data. this enables the time series data, such as speech to be processed without providing any special means in the neural network.",1998-07-28,"The title of the patent is speech recognition apparatus using neural network, and learning method therefor and its abstract is a speech recognition apparatus using a neural network. a neuron-like element according to the present invention has a means for storing a value of the inner condition thereof, a means for updating a value of internal status on the basis of an output from the neuron-like element itself, outputs from other neuron-like elements and an external input, and an output value generating means for converting a value of internal status into an external output. accordingly, the neuron-like element itself can retain the history of input data. this enables the time series data, such as speech to be processed without providing any special means in the neural network. dated 1998-07-28"
5787408,system and method for determining node functionality in artificial neural networks,"a system for unwrapping an artificial neural network (ann) to determine the tilization and functionality of the nodes uses a network generator for generating an initial ann architecture. training and pruning processors operate to generate minimal ann architectures having increasingly lower levels of classification accuracy. a network analyzer uses an analysis controller to receive minimal ann architectures from the pruning processor. a connection analyzer operates on the minimal ann architectures to identify the inputs to the minimal ann architecture and determine the information represented by and contained in the network inputs. a node analyzer, coupled to the connection analyzer, then defines the utilization and functionality of each node in the minimal ann architecture in terms of known functions.",1998-07-28,"The title of the patent is system and method for determining node functionality in artificial neural networks and its abstract is a system for unwrapping an artificial neural network (ann) to determine the tilization and functionality of the nodes uses a network generator for generating an initial ann architecture. training and pruning processors operate to generate minimal ann architectures having increasingly lower levels of classification accuracy. a network analyzer uses an analysis controller to receive minimal ann architectures from the pruning processor. a connection analyzer operates on the minimal ann architectures to identify the inputs to the minimal ann architecture and determine the information represented by and contained in the network inputs. a node analyzer, coupled to the connection analyzer, then defines the utilization and functionality of each node in the minimal ann architecture in terms of known functions. dated 1998-07-28"
5789676,settling process analysis device and method,"a system and method for monitoring the dynamics of particle motion in a liquid-solid media including the rate of settling of particles, the identification of unsettled particle clouds, and the identification and control of the bed level of settled particles in a slurry within a settler. the system includes an ultrasound transducer and a receiver for detecting echoes from particles in the slurry. the echoes are processed to determine the bed level of the settled particles, the position of unsettled particle clouds, and the rate of settling of the particle clouds, employing a neural network preferably with a fuzzy logic controller.",1998-08-04,"The title of the patent is settling process analysis device and method and its abstract is a system and method for monitoring the dynamics of particle motion in a liquid-solid media including the rate of settling of particles, the identification of unsettled particle clouds, and the identification and control of the bed level of settled particles in a slurry within a settler. the system includes an ultrasound transducer and a receiver for detecting echoes from particles in the slurry. the echoes are processed to determine the bed level of the settled particles, the position of unsettled particle clouds, and the rate of settling of the particle clouds, employing a neural network preferably with a fuzzy logic controller. dated 1998-08-04"
5790690,computer-aided method for automated image feature analysis and diagnosis of medical images,"a computerized method for the detection and characterization of disease in an image derived from a chest radiograph, wherein an image in the chest radiograph is processed to determine the ribcage boundary, including lung top edges, right and left ribcage edges, and right and left hemidiaphragm edges. texture measures including rms variations of pixel values within regions of interest are converted to relative exposures and corrected for system noise existing in the system used to produce the image. texture and/or geometric pattern indices are produced. a histogram(s) of the produced index (indices) is produced and values of the histogram(s) are applied as inputs to a trained artificial neural network, which classifies the image as normal or abnormal. in one embodiment, obviously normal and obviously abnormal images are determined based on the ratio of abnormal regions of interest to the total number of regions of interest in a rule-based method, so that only difficult cases to diagnose are applied to the artificial neural network.",1998-08-04,"The title of the patent is computer-aided method for automated image feature analysis and diagnosis of medical images and its abstract is a computerized method for the detection and characterization of disease in an image derived from a chest radiograph, wherein an image in the chest radiograph is processed to determine the ribcage boundary, including lung top edges, right and left ribcage edges, and right and left hemidiaphragm edges. texture measures including rms variations of pixel values within regions of interest are converted to relative exposures and corrected for system noise existing in the system used to produce the image. texture and/or geometric pattern indices are produced. a histogram(s) of the produced index (indices) is produced and values of the histogram(s) are applied as inputs to a trained artificial neural network, which classifies the image as normal or abnormal. in one embodiment, obviously normal and obviously abnormal images are determined based on the ratio of abnormal regions of interest to the total number of regions of interest in a rule-based method, so that only difficult cases to diagnose are applied to the artificial neural network. dated 1998-08-04"
5790754,speech recognition apparatus for consumer electronic applications,"a spoken word or phrase recognition device. the device does not require a digital signal processor, large ram, or extensive analog circuitry. the input audio signal is digitized and passed recursively through a digital difference filter to produce a multiplicity of filtered output waveforms. these waveforms are processed in real time by a microprocessor to generate a pattern that is recognized by a neural network pattern classifier that operates in software in the microprocessor. by application of additional techniques, this device has been shown to recognize an unknown speaker saying a digit from zero through nine with an accuracy greater than 99%. because of the recognition accuracy and cost-effective design, the device may be used in cost sensitive applications such as toys, electronic learning aids, and consumer electronic products.",1998-08-04,"The title of the patent is speech recognition apparatus for consumer electronic applications and its abstract is a spoken word or phrase recognition device. the device does not require a digital signal processor, large ram, or extensive analog circuitry. the input audio signal is digitized and passed recursively through a digital difference filter to produce a multiplicity of filtered output waveforms. these waveforms are processed in real time by a microprocessor to generate a pattern that is recognized by a neural network pattern classifier that operates in software in the microprocessor. by application of additional techniques, this device has been shown to recognize an unknown speaker saying a digit from zero through nine with an accuracy greater than 99%. because of the recognition accuracy and cost-effective design, the device may be used in cost sensitive applications such as toys, electronic learning aids, and consumer electronic products. dated 1998-08-04"
5790758,neural network architecture for gaussian components of a mixture density function,"a neural network for classifying input vectors to an outcome class under the assumption that the classes are characterized by mixtures of component populations having a multivariate gaussian likelihood distribution. the neural network comprises an input layer for receiving components of an input vector, two hidden layers for generating a number of outcome class component values, and an output layer. the first hidden layer includes a number of first layer nodes each connected receive input vector components and generate a first layer output value representing the absolute value of the sum of a function of the difference between each input vector component and a threshold value. the second hidden layer includes a plurality of second layer nodes, each second layer node being connected to the first layer nodes and generating an outcome class component value representing a function related to the exponential of the negative square of a function of the sum of the first layer output values times a weighting value. the output layer includes a plurality of output nodes, each associated with an outcome class, for generating a value that represents the likelihood that the input vector belongs to that outcome class.",1998-08-04,"The title of the patent is neural network architecture for gaussian components of a mixture density function and its abstract is a neural network for classifying input vectors to an outcome class under the assumption that the classes are characterized by mixtures of component populations having a multivariate gaussian likelihood distribution. the neural network comprises an input layer for receiving components of an input vector, two hidden layers for generating a number of outcome class component values, and an output layer. the first hidden layer includes a number of first layer nodes each connected receive input vector components and generate a first layer output value representing the absolute value of the sum of a function of the difference between each input vector component and a threshold value. the second hidden layer includes a plurality of second layer nodes, each second layer node being connected to the first layer nodes and generating an outcome class component value representing a function related to the exponential of the negative square of a function of the sum of the first layer output values times a weighting value. the output layer includes a plurality of output nodes, each associated with an outcome class, for generating a value that represents the likelihood that the input vector belongs to that outcome class. dated 1998-08-04"
5790761,method and apparatus for the diagnosis of colorectal cancer,"a process is set forth in which cancer of the colon is assessed in a patient. the probabilities of developing cancer involves the initial step of extracting a set of sample body fluids from the patient. fluids can be evaluated to determine certain marker constituents in the body fluids. fluids which are extracted have some relationship to me development of cancer, precancer or tendency toward cancerous conditions. the body fluid markers are measured and other quantified. the marker data then is evaluated using a nonlinear technique exemplified through the use of a multiple input and multiple output neural network having a variable learning rate and training rate. the neural network is provided with data from other patients for the same or similar markers. data from other patients who did and did not have cancer is used in the leaning of the neural network which thereby processes the data and provides a determination that the patient has a cancerous condition, precancer cells or a tendency towards cancer.",1998-08-04,"The title of the patent is method and apparatus for the diagnosis of colorectal cancer and its abstract is a process is set forth in which cancer of the colon is assessed in a patient. the probabilities of developing cancer involves the initial step of extracting a set of sample body fluids from the patient. fluids can be evaluated to determine certain marker constituents in the body fluids. fluids which are extracted have some relationship to me development of cancer, precancer or tendency toward cancerous conditions. the body fluid markers are measured and other quantified. the marker data then is evaluated using a nonlinear technique exemplified through the use of a multiple input and multiple output neural network having a variable learning rate and training rate. the neural network is provided with data from other patients for the same or similar markers. data from other patients who did and did not have cancer is used in the leaning of the neural network which thereby processes the data and provides a determination that the patient has a cancerous condition, precancer cells or a tendency towards cancer. dated 1998-08-04"
5791155,system for monitoring expansion valve,a monitoring system for a heating or cooling system includes a neural network for computing the status of one or more expansion valves within the system. the neural network is trained to learn certain characteristics of the heating or cooling system during a development mode of operation. the thus trained neural network timely computes the status of the one or more expansion valves during a run time mode of operation. information as to the status of the one or more expansion valves is made available for real time assessment during the run time mode of operation.,1998-08-11,The title of the patent is system for monitoring expansion valve and its abstract is a monitoring system for a heating or cooling system includes a neural network for computing the status of one or more expansion valves within the system. the neural network is trained to learn certain characteristics of the heating or cooling system during a development mode of operation. the thus trained neural network timely computes the status of the one or more expansion valves during a run time mode of operation. information as to the status of the one or more expansion valves is made available for real time assessment during the run time mode of operation. dated 1998-08-11
5794178,visualization of information using graphical representations of context vector based relationships and attributes,""" a system and method for generating context vectors for use in storage and retrieval of documents and other information items. context vectors represent conceptual relationships among information items by quantitative means. a neural network operates on a training corpus of records to develop relationship-based context vectors based on word proximity and co-importance using a technique of """"windowed co-occurrence"""". relationships among context vectors are deterministic, so that a context vector set has one logical solution, although it may have a plurality of physical solutions. no human knowledge, thesaurus, synonym list, knowledge base, or conceptual hierarchy, is required. summary vectors of records may be clustered to reduce searching time, by forming a tree of clustered nodes. once the context vectors are determined, records may be retrieved using a query interface that allows a user to specify content terms, boolean terms, and/or document feedback. the present invention further facilitates visualization of textual information by translating context vectors into visual and graphical representations. thus, a user can explore visual representations of meaning, and can apply human visual pattern recognition skills to document searches. """,1998-08-11,"The title of the patent is visualization of information using graphical representations of context vector based relationships and attributes and its abstract is "" a system and method for generating context vectors for use in storage and retrieval of documents and other information items. context vectors represent conceptual relationships among information items by quantitative means. a neural network operates on a training corpus of records to develop relationship-based context vectors based on word proximity and co-importance using a technique of """"windowed co-occurrence"""". relationships among context vectors are deterministic, so that a context vector set has one logical solution, although it may have a plurality of physical solutions. no human knowledge, thesaurus, synonym list, knowledge base, or conceptual hierarchy, is required. summary vectors of records may be clustered to reduce searching time, by forming a tree of clustered nodes. once the context vectors are determined, records may be retrieved using a query interface that allows a user to specify content terms, boolean terms, and/or document feedback. the present invention further facilitates visualization of textual information by translating context vectors into visual and graphical representations. thus, a user can explore visual representations of meaning, and can apply human visual pattern recognition skills to document searches. "" dated 1998-08-11"
5794191,neural network based speech recognition method utilizing spectrum-dependent and time-dependent coefficients,"an improved artificial neural network for use in speech recognition is disclosed. it comprises an input layer, a hidden layer, and an output layer, each of these layers consisting of a plurality of nodal points. a set of first weighting coefficients are used between the input layer and the hidden layer which are functions of at least one of the nodal points in the hidden layer and at least one of the nodal points in the input layer; whereas, a set of second weighting coefficients, which are functions of time and at least one of the nodal points in the output, are used to correlate between the hidden layer and output layer. in a preferred embodiment, the first weighting coefficients are calculated using the following formula: ##equ1## i is the index for nodal point in the input layer and a.sub.j, b.sub.j, and c.sub.j are all training coefficients associated with nodal pointj in the hidden layer; and the second weighting coefficients are calculated using the following formula: ##equ2## n is the timeframe number, r is the order of an orthogonal polynomial series (.psi., 60 .sub.jkm is the m-th order training coefficient between nodal points j and k, in the hidden and output layers, respectively. the use of the two different sets of weighting coefficients allows a timeframe-based division of the speech signals, resulting in a substantial reduction of parameters required for accurate speech recognition.",1998-08-11,"The title of the patent is neural network based speech recognition method utilizing spectrum-dependent and time-dependent coefficients and its abstract is an improved artificial neural network for use in speech recognition is disclosed. it comprises an input layer, a hidden layer, and an output layer, each of these layers consisting of a plurality of nodal points. a set of first weighting coefficients are used between the input layer and the hidden layer which are functions of at least one of the nodal points in the hidden layer and at least one of the nodal points in the input layer; whereas, a set of second weighting coefficients, which are functions of time and at least one of the nodal points in the output, are used to correlate between the hidden layer and output layer. in a preferred embodiment, the first weighting coefficients are calculated using the following formula: ##equ1## i is the index for nodal point in the input layer and a.sub.j, b.sub.j, and c.sub.j are all training coefficients associated with nodal pointj in the hidden layer; and the second weighting coefficients are calculated using the following formula: ##equ2## n is the timeframe number, r is the order of an orthogonal polynomial series (.psi., 60 .sub.jkm is the m-th order training coefficient between nodal points j and k, in the hidden and output layers, respectively. the use of the two different sets of weighting coefficients allows a timeframe-based division of the speech signals, resulting in a substantial reduction of parameters required for accurate speech recognition. dated 1998-08-11"
5796920,multiprocessor system and method for identification and adaptive control of dynamic systems,"a system and method for identifying and adapting a control system for a dynamic system includes in one embodiment massively parallel, decentralized signal processing equipment which can (1) identify a dynamic system; (2) adapt an on-line control system for the dynamic system; and/or, (3) solve off-line complex, nonlinear problems related to either static or dynamic systems. the disclosed system may include a neural network in which the neurons are two-way devices (forward and backward signal paths), each of which has a synaptic weight which is independently adjusted using only the forward and backward signals.",1998-08-18,"The title of the patent is multiprocessor system and method for identification and adaptive control of dynamic systems and its abstract is a system and method for identifying and adapting a control system for a dynamic system includes in one embodiment massively parallel, decentralized signal processing equipment which can (1) identify a dynamic system; (2) adapt an on-line control system for the dynamic system; and/or, (3) solve off-line complex, nonlinear problems related to either static or dynamic systems. the disclosed system may include a neural network in which the neurons are two-way devices (forward and backward signal paths), each of which has a synaptic weight which is independently adjusted using only the forward and backward signals. dated 1998-08-18"
5796922,"trainable, state-sampled, network controller","a trainable, state-sampled, network controller (tssnc) or state-sampled controller (ssc) requires little information regarding a plant (as with neural networks), but can use what information is available (as in classical controllers), and provides a linear network (as for cmac) improving calculation speeds. a form of a governing differential equation characterizing a plant may include parameters and their derivatives of various orders as variables combined in linear and nonlinear terms. classical control theory, and a method such as a fourier transform of governing equations, may provide 8a form of a control law, linear in certain weights or coefficients. knowledge of coefficients is not required for either the form of the governing equations or the form of the control law. an optimization method may be used to train the ssc, defining a table of weights (contributions to coefficients) to be used in the matrix equation representing the control law the solution yielding a control output to the plant. sampling plant outputs, during training, may be done at a selected spatial frequency in state space (each dimension a variable from the control law). sampling is used to provide ideal interpolation of the weights over the entire range of interest. minimum memory is used with maximum accuracy of interpolation, and any control/output value may be calculated as needed in real time by a minimal processor.",1998-08-18,"The title of the patent is trainable, state-sampled, network controller and its abstract is a trainable, state-sampled, network controller (tssnc) or state-sampled controller (ssc) requires little information regarding a plant (as with neural networks), but can use what information is available (as in classical controllers), and provides a linear network (as for cmac) improving calculation speeds. a form of a governing differential equation characterizing a plant may include parameters and their derivatives of various orders as variables combined in linear and nonlinear terms. classical control theory, and a method such as a fourier transform of governing equations, may provide 8a form of a control law, linear in certain weights or coefficients. knowledge of coefficients is not required for either the form of the governing equations or the form of the control law. an optimization method may be used to train the ssc, defining a table of weights (contributions to coefficients) to be used in the matrix equation representing the control law the solution yielding a control output to the plant. sampling plant outputs, during training, may be done at a selected spatial frequency in state space (each dimension a variable from the control law). sampling is used to provide ideal interpolation of the weights over the entire range of interest. minimum memory is used with maximum accuracy of interpolation, and any control/output value may be calculated as needed in real time by a minimal processor. dated 1998-08-18"
5799102,bank note scanner utilizing olfactory characteristics for authentication,"the invention concerns a bank note scanner (2) for assessing the authenticity of a bank note, which includes a vacuum pump (4), an olfactory sensor (8), an authentication means (28) for producing an electrical output indicative of the authenticity of a bank note, and suction means (10) connected to the vacuum pump (4) via the olfactory sensor (8). in operation, a bank note is fed through an entry slot (50) in the scanner (2) into co-operative relationship with the suction means (10) such that the bank note covers, and is sucked against, the suction means (10), thus enabling the sensor (8) to test the note. the authentication means (28) comprises a neural network (26) which can be taught the olfactory characteristics of an authentic bank note. the authentication means (28) is arranged to make a determination of the authenticity of the bank note based on the comparison of the electrical output of the olfactory sensor (8) and the olfactory characteristics of one or more authentic bank notes.",1998-08-25,"The title of the patent is bank note scanner utilizing olfactory characteristics for authentication and its abstract is the invention concerns a bank note scanner (2) for assessing the authenticity of a bank note, which includes a vacuum pump (4), an olfactory sensor (8), an authentication means (28) for producing an electrical output indicative of the authenticity of a bank note, and suction means (10) connected to the vacuum pump (4) via the olfactory sensor (8). in operation, a bank note is fed through an entry slot (50) in the scanner (2) into co-operative relationship with the suction means (10) such that the bank note covers, and is sucked against, the suction means (10), thus enabling the sensor (8) to test the note. the authentication means (28) comprises a neural network (26) which can be taught the olfactory characteristics of an authentic bank note. the authentication means (28) is arranged to make a determination of the authenticity of the bank note based on the comparison of the electrical output of the olfactory sensor (8) and the olfactory characteristics of one or more authentic bank notes. dated 1998-08-25"
5799133,training process,"a training apparatus for establishing the network definition function of a trainable processing apparatus, such as, for example, a neural network, is disclosed. the training apparatus analyzes a signal and includes means for providing a training sequence, the training sequence including a first signal and a distorted version of the first signal. the training sequence is transmitted to an analysis means which generates a distortion perception measure that indicates the extent to which the distortion would be perceptible to a human observer. the network definition function is determined by applying the distortion perception measure to the trainable processing apparatus.",1998-08-25,"The title of the patent is training process and its abstract is a training apparatus for establishing the network definition function of a trainable processing apparatus, such as, for example, a neural network, is disclosed. the training apparatus analyzes a signal and includes means for providing a training sequence, the training sequence including a first signal and a distorted version of the first signal. the training sequence is transmitted to an analysis means which generates a distortion perception measure that indicates the extent to which the distortion would be perceptible to a human observer. the network definition function is determined by applying the distortion perception measure to the trainable processing apparatus. dated 1998-08-25"
5799134,one dimensional systolic array architecture for neural network,"a circuit for implementing a neural network comprises a one dimensional systolic array of processing elements controlled by a microprocessor. the one dimensional systolic array can implement weighted sum and radial based type networks including neurons with a variety of different activation functions. pipelined processing and partitioning is used to optimize data flows in the systolic array. accordingly, the inventive circuit can implement a variety of neural networks in a very efficient manner.",1998-08-25,"The title of the patent is one dimensional systolic array architecture for neural network and its abstract is a circuit for implementing a neural network comprises a one dimensional systolic array of processing elements controlled by a microprocessor. the one dimensional systolic array can implement weighted sum and radial based type networks including neurons with a variety of different activation functions. pipelined processing and partitioning is used to optimize data flows in the systolic array. accordingly, the inventive circuit can implement a variety of neural networks in a very efficient manner. dated 1998-08-25"
5799296,system for continuous logic computation and method of using same,"a continuous logic system using a neural network is characterized by defining input and output variables that do not use a membership function, by employing production rules (if/then rules) that relate the output variables to the input variables, and by using the neural network to compute or interpolate the outputs. the neural network first learns the given production rules and then produces the outputs in real time. the neural network is constructed of artificial neurons each having only one significant processing element in the form of a multiplier. the neural network utilizes a training algorithm which does not require repetitive training and which yields a global minimum to each given set of input vectors.",1998-08-25,"The title of the patent is system for continuous logic computation and method of using same and its abstract is a continuous logic system using a neural network is characterized by defining input and output variables that do not use a membership function, by employing production rules (if/then rules) that relate the output variables to the input variables, and by using the neural network to compute or interpolate the outputs. the neural network first learns the given production rules and then produces the outputs in real time. the neural network is constructed of artificial neurons each having only one significant processing element in the form of a multiplier. the neural network utilizes a training algorithm which does not require repetitive training and which yields a global minimum to each given set of input vectors. dated 1998-08-25"
5799496,temperature controlling method and apparatus for refrigerator using velocity control of ventilation fan and direction control of rotary blade,"a temperature controlling method and apparatus for a refrigerator are provided, in which a temperature-equilibrating position into which cool air is to be discharged is calculated based on a fuzzy model and learning of a neural network, using the change in temperatures measured by only a small number of temperature sensors at a predetermined number of portions within a refrigeration compartment, and then the rotation velocity of a ventilation fan and a stop angle of a rotary blade are controlled according to the calculated temperature-equilibrating position. as a result, the cool air is appropriately discharged into each portion according to the distance between the rotary blade and a target position, so that the optimal temperature equilibrium is obtained in the refrigeration compartment.",1998-09-01,"The title of the patent is temperature controlling method and apparatus for refrigerator using velocity control of ventilation fan and direction control of rotary blade and its abstract is a temperature controlling method and apparatus for a refrigerator are provided, in which a temperature-equilibrating position into which cool air is to be discharged is calculated based on a fuzzy model and learning of a neural network, using the change in temperatures measured by only a small number of temperature sensors at a predetermined number of portions within a refrigeration compartment, and then the rotation velocity of a ventilation fan and a stop angle of a rotary blade are controlled according to the calculated temperature-equilibrating position. as a result, the cool air is appropriately discharged into each portion according to the distance between the rotary blade and a target position, so that the optimal temperature equilibrium is obtained in the refrigeration compartment. dated 1998-09-01"
5804940,device designed to compensate for non-linearity of machine shafts,"a method for the numerical control of machines with several axes, in particular machine tools and robots, to compensate for the inaccuracies occurring when the axes are reversed, wherein varying friction conditions, as well as slackness and torsional effects are compensated using a friction precontrol. rotation speed reference values are corrected by injecting a correction pulse with an acceleration-dependent injection amplitude and a constant decay time for each axis at the time of passage from one quadrant to another, with the associated change in direction. the injection amplitude and constant decay time are determined for each machine manually or learned in an additional embodiment automatically in a self-learning system in the form of a neural network.",1998-09-08,"The title of the patent is device designed to compensate for non-linearity of machine shafts and its abstract is a method for the numerical control of machines with several axes, in particular machine tools and robots, to compensate for the inaccuracies occurring when the axes are reversed, wherein varying friction conditions, as well as slackness and torsional effects are compensated using a friction precontrol. rotation speed reference values are corrected by injecting a correction pulse with an acceleration-dependent injection amplitude and a constant decay time for each axis at the time of passage from one quadrant to another, with the associated change in direction. the injection amplitude and constant decay time are determined for each machine manually or learned in an additional embodiment automatically in a self-learning system in the form of a neural network. dated 1998-09-08"
5805453,system identifier for paper machine,a system identifier for a paper machine which allows an operator having no special skills to easily implement step responses in a profile control and which comprises a mechanism for storing theoretical positional correspondence determined by paper contraction; calculator device for calculating profiles for slices associated with the operated ends using measured values at the detected end; another calculating device for calculating the conformity of each operated end by multiplying the profile for the slice by an interference coefficient indicating the effect of an operation on an operated end and on the detected end; an output slice selection device to which data of the conformity of each operated end is inputted and which outputs an output slice selection function by performing a neural network calculation wherein self feedback is performed for the operated end and lateral inhibition is performed to inhibit mutual interference between neighboring operated ends within a predetermined range; and a step response output device for performing step response at the operated ends which have survived within the range in which the mutual interference is inhibited using the output slice selection function.,1998-09-08,The title of the patent is system identifier for paper machine and its abstract is a system identifier for a paper machine which allows an operator having no special skills to easily implement step responses in a profile control and which comprises a mechanism for storing theoretical positional correspondence determined by paper contraction; calculator device for calculating profiles for slices associated with the operated ends using measured values at the detected end; another calculating device for calculating the conformity of each operated end by multiplying the profile for the slice by an interference coefficient indicating the effect of an operation on an operated end and on the detected end; an output slice selection device to which data of the conformity of each operated end is inputted and which outputs an output slice selection function by performing a neural network calculation wherein self feedback is performed for the operated end and lateral inhibition is performed to inhibit mutual interference between neighboring operated ends within a predetermined range; and a step response output device for performing step response at the operated ends which have survived within the range in which the mutual interference is inhibited using the output slice selection function. dated 1998-09-08
5805731,adaptive statistical classifier which provides reliable estimates or output classes having low probabilities,"a statistical classifier for pattern recognition, such as a neural network, produces a plurality of output signals corresponding to the probabilities that a given input pattern belongs in respective classes. the classifier is trained in a manner such that low probabilities which pertain to classes of interest are not suppressed too greatly. this is achieved by modifying the amount by which error signals, corresponding to classes which are incorrectly identified, are employed in the training process, relative to error signals corresponding to the correct class. as a result, output probabilities for incorrect classes are not forced to a low value as much as probabilities for correct classes are raised.",1998-09-08,"The title of the patent is adaptive statistical classifier which provides reliable estimates or output classes having low probabilities and its abstract is a statistical classifier for pattern recognition, such as a neural network, produces a plurality of output signals corresponding to the probabilities that a given input pattern belongs in respective classes. the classifier is trained in a manner such that low probabilities which pertain to classes of interest are not suppressed too greatly. this is achieved by modifying the amount by which error signals, corresponding to classes which are incorrectly identified, are employed in the training process, relative to error signals corresponding to the correct class. as a result, output probabilities for incorrect classes are not forced to a low value as much as probabilities for correct classes are raised. dated 1998-09-08"
5805771,automatic language identification method and system,"this invention consists of three enhancements to hmm-based automatic language identification systems. the three enhancements are: (i) language-discriminant acoustic model training and recognition, (ii) an acoustic model pruning procedure that picks only those phonetic models which are considered useful for language identification, and (iii) a neural network-based language classification method that uses knowledge-based features derived from phone sequences output by the hmm phonetic recognizers.",1998-09-08,"The title of the patent is automatic language identification method and system and its abstract is this invention consists of three enhancements to hmm-based automatic language identification systems. the three enhancements are: (i) language-discriminant acoustic model training and recognition, (ii) an acoustic model pruning procedure that picks only those phonetic models which are considered useful for language identification, and (iii) a neural network-based language classification method that uses knowledge-based features derived from phone sequences output by the hmm phonetic recognizers. dated 1998-09-08"
5806013,control of engine fuel delivery using an artificial neural network in parallel with a feed-forward controller,"a system for controlling operation of engine fuel injectors that includes a feed-forward control unit responsive to signals from sensors on the engine for supplying a basic electronic control signal for the injectors. a neural network is connected in parallel with the feed-forward control unit for receiving the sensor signals and multiplying the sensor signals by associated weighting factors. the sensor signals multiplied by the weighting factors are combined to produce a network output signal, which in turn is combined with the basic control signal from the feed-forward control unit to control operation of the fuel injectors. the weighting factors in the neural network are modified as a function of inputs from the engine sensors so as to reduce any errors in the sensor output signals as compared with desired values.",1998-09-08,"The title of the patent is control of engine fuel delivery using an artificial neural network in parallel with a feed-forward controller and its abstract is a system for controlling operation of engine fuel injectors that includes a feed-forward control unit responsive to signals from sensors on the engine for supplying a basic electronic control signal for the injectors. a neural network is connected in parallel with the feed-forward control unit for receiving the sensor signals and multiplying the sensor signals by associated weighting factors. the sensor signals multiplied by the weighting factors are combined to produce a network output signal, which in turn is combined with the basic control signal from the feed-forward control unit to control operation of the fuel injectors. the weighting factors in the neural network are modified as a function of inputs from the engine sensors so as to reduce any errors in the sensor output signals as compared with desired values. dated 1998-09-08"
5806053,method for training a neural network with the non-deterministic behavior of a technical system,"in a method for tranining a neural network with the non-deterministic behavior of a technical system, weightings for the neurons of the neural network are set during the training using a cost function. the cost function evaluates a beneficial system behavior of the technical system to be modeled, and thereby intensifies or increases the weighting settings which contribute to the beneficial system behavior, and attenuates or minimizes weightings which produce a non-beneficial behavior. arbitrary or random disturbances are generated by disturbing the manipulated variable with noise having a known noise distribution, these random disturbances significantly faciliating the mathematical processing of the weightings which are set, because the terms required for that purpose are simplified. the correct weighting setting for the neural network is thus found on the basis of a statistical method and the application of a cost function to the values emitted by the technical system or its model.",1998-09-08,"The title of the patent is method for training a neural network with the non-deterministic behavior of a technical system and its abstract is in a method for tranining a neural network with the non-deterministic behavior of a technical system, weightings for the neurons of the neural network are set during the training using a cost function. the cost function evaluates a beneficial system behavior of the technical system to be modeled, and thereby intensifies or increases the weighting settings which contribute to the beneficial system behavior, and attenuates or minimizes weightings which produce a non-beneficial behavior. arbitrary or random disturbances are generated by disturbing the manipulated variable with noise having a known noise distribution, these random disturbances significantly faciliating the mathematical processing of the weightings which are set, because the terms required for that purpose are simplified. the correct weighting setting for the neural network is thus found on the basis of a statistical method and the application of a cost function to the values emitted by the technical system or its model. dated 1998-09-08"
5808621,system and method for selecting a color space using a neural network,"a system and method for automatically selecting a color space for use in compressing and decompressing a texture image that automatically determines a compression color space for each texture image. the invention selects a compression color space manually, or preferable, using a neural network algorithm. the invention initializes the neural network that includes an input layer of neurons and a hidden layer of neurons. each input layer neuron has an associated weight that is equal to the combination of the weights of a y neuron, an a neuron, and a b neuron that is associated with each input layer neuron. the texel image is reduced into a representative sample of colors and input vectors from the texel image are randomly selected. for each input vector, the invention determines the two input layer neurons that most closely match the input vector and modifies the weights of the two input layer neurons to more closely match the value of the input vector. the weights of the hidden layer neurons represent the y, a, and b-channel values of the optimal compression color space.",1998-09-15,"The title of the patent is system and method for selecting a color space using a neural network and its abstract is a system and method for automatically selecting a color space for use in compressing and decompressing a texture image that automatically determines a compression color space for each texture image. the invention selects a compression color space manually, or preferable, using a neural network algorithm. the invention initializes the neural network that includes an input layer of neurons and a hidden layer of neurons. each input layer neuron has an associated weight that is equal to the combination of the weights of a y neuron, an a neuron, and a b neuron that is associated with each input layer neuron. the texel image is reduced into a representative sample of colors and input vectors from the texel image are randomly selected. for each input vector, the invention determines the two input layer neurons that most closely match the input vector and modifies the weights of the two input layer neurons to more closely match the value of the input vector. the weights of the hidden layer neurons represent the y, a, and b-channel values of the optimal compression color space. dated 1998-09-15"
5809461,speech recognition apparatus using neural network and learning method therefor,"a speech recognition apparatus using a neural network is provided. a neuron-like element stores a value of its inner conditions. the neuron-like element also updates a value of its internal status on the basis of an output from the neuron-like element itself, outputs from other neuron-like elements and an external input outside. the neuron-like element also converts a value of its internal status into an external output. accordingly, the neuron-like element itself can retain the history of input data. this enables time series data, such as speech, to be processed without providing any special devices in the neural network.",1998-09-15,"The title of the patent is speech recognition apparatus using neural network and learning method therefor and its abstract is a speech recognition apparatus using a neural network is provided. a neuron-like element stores a value of its inner conditions. the neuron-like element also updates a value of its internal status on the basis of an output from the neuron-like element itself, outputs from other neuron-like elements and an external input outside. the neuron-like element also converts a value of its internal status into an external output. accordingly, the neuron-like element itself can retain the history of input data. this enables time series data, such as speech, to be processed without providing any special devices in the neural network. dated 1998-09-15"
5809462,method and apparatus for interfacing and training a neural network for phoneme recognition,"an automated speech recognition system converts a speech signal into a compact, coded representation that correlates to a speech phoneme set. a number of different neural network pattern matching schemes may be used to perform the necessary speech coding. an integrated user interface guides a user unfamiliar with the details of speech recognition or neural networks to quickly develop and test a neural network for phoneme recognition. to train the neural network, digitized voice data containing known phonemes that the user wants the neural network to ultimately recognize are processed by the integrated user interface. the digitized speech is segmented into phonemes with each segment being labelled with a corresponding phoneme code. based on a user selected transformation method and transformation parameters, each segment is transformed into a series of multiple dimension vectors representative of the speech characteristics of that segment. these vectors are iteratively presented to a neural network to train/adapt that neural network to consistently distinguish and recognize these vectors and assign an appropriate phoneme code to each vector. simultaneous display of the digitized speech, segments, vector sets, and a representation of the trained neural network assist the user in visually confirming the acceptability of the phoneme training set. a user may also selectively audibly confirm the acceptability of the digitization scheme, the segments, and the transform vectors so that satisfactory training data are presented to the neural network. if the user finds a particular step or parameter produces an unacceptable result, the user may modify one or more of the parameters and verify whether the modification effected an improvement in performance. the trained neural network is also automatically tested by presenting a test speech signal to the integrated user interface and observing both audibly and visually automatic segmentation of the speech, transformation into multidimensional vectors, and the resulting neural network assigned phoneme codes. a method of decoding such phoneme codes using the neural network is also disclosed.",1998-09-15,"The title of the patent is method and apparatus for interfacing and training a neural network for phoneme recognition and its abstract is an automated speech recognition system converts a speech signal into a compact, coded representation that correlates to a speech phoneme set. a number of different neural network pattern matching schemes may be used to perform the necessary speech coding. an integrated user interface guides a user unfamiliar with the details of speech recognition or neural networks to quickly develop and test a neural network for phoneme recognition. to train the neural network, digitized voice data containing known phonemes that the user wants the neural network to ultimately recognize are processed by the integrated user interface. the digitized speech is segmented into phonemes with each segment being labelled with a corresponding phoneme code. based on a user selected transformation method and transformation parameters, each segment is transformed into a series of multiple dimension vectors representative of the speech characteristics of that segment. these vectors are iteratively presented to a neural network to train/adapt that neural network to consistently distinguish and recognize these vectors and assign an appropriate phoneme code to each vector. simultaneous display of the digitized speech, segments, vector sets, and a representation of the trained neural network assist the user in visually confirming the acceptability of the phoneme training set. a user may also selectively audibly confirm the acceptability of the digitization scheme, the segments, and the transform vectors so that satisfactory training data are presented to the neural network. if the user finds a particular step or parameter produces an unacceptable result, the user may modify one or more of the parameters and verify whether the modification effected an improvement in performance. the trained neural network is also automatically tested by presenting a test speech signal to the integrated user interface and observing both audibly and visually automatic segmentation of the speech, transformation into multidimensional vectors, and the resulting neural network assigned phoneme codes. a method of decoding such phoneme codes using the neural network is also disclosed. dated 1998-09-15"
5809487,arrangement for modeling a non-linear process,"an arrangement for modeling a non-linear process including at least one input variable (x1, x2) and at least one output variable (y) can include a neural network and a device for specifying functional (or operational) values. a function of the neural network is determined in a first part of the domain of input variables (x1, x2) by learning from measuring data, which are obtained from the process by acquiring measured values. an empirically based device for specifying functional (or operational) values, preferably a fuzzy system, is provided in a second part of the domain of input variables (x1, x2) in which there are no measuring data for training the neural network. this arrangement is particularly useful when it is implemented in controllers.",1998-09-15,"The title of the patent is arrangement for modeling a non-linear process and its abstract is an arrangement for modeling a non-linear process including at least one input variable (x1, x2) and at least one output variable (y) can include a neural network and a device for specifying functional (or operational) values. a function of the neural network is determined in a first part of the domain of input variables (x1, x2) by learning from measuring data, which are obtained from the process by acquiring measured values. an empirically based device for specifying functional (or operational) values, preferably a fuzzy system, is provided in a second part of the domain of input variables (x1, x2) in which there are no measuring data for training the neural network. this arrangement is particularly useful when it is implemented in controllers. dated 1998-09-15"
5809488,management system for a power station installation,"in the control of a power station installation having a number of power station blocks, in which each power station block is controlled by using at least one reference variable, it is intended to permit reliable determination of especially favorable reference variables while also taking the current installation condition into account. to this end, a management system for the power station installation includes a computer unit which determines reference variables for the power station block or for each of the power station blocks through the use of a genetic algorithm, and an optimization module which is connected to the computer unit. the optimization module is connected to a number of neural networks and one neural network is assigned to each power station block.",1998-09-15,"The title of the patent is management system for a power station installation and its abstract is in the control of a power station installation having a number of power station blocks, in which each power station block is controlled by using at least one reference variable, it is intended to permit reliable determination of especially favorable reference variables while also taking the current installation condition into account. to this end, a management system for the power station installation includes a computer unit which determines reference variables for the power station block or for each of the power station blocks through the use of a genetic algorithm, and an optimization module which is connected to the computer unit. the optimization module is connected to a number of neural networks and one neural network is assigned to each power station block. dated 1998-09-15"
5809490,apparatus and method for selecting a working data set for model development,"the present invention provides a data selection apparatus which augments a set of training examples with the desired output data. the resulting augmented data set is normalized such that the augmented data values range between -1 and +1 and such that the mean of the augmented data set is zero. the data selection apparatus then groups the augmented and normalized data set into related clusters using a clusterizer. preferably, the clusterizer is a neural network such as a kohonen self-organizing map (som). the data selection apparatus further applies an extractor to cull a working set of data from the clusterized data set. the present invention thus picks, or filters, a set of data which is more nearly uniformly distributed across the portion of the input space of interest to minimize the maximum absolute error over the entire input space. the output of the data selection apparatus is provided to train the analyzer with important sub-sets of the training data rather than with all available training data. a smaller training data set significantly reduces the complexity of the model building or analyzer construction process.",1998-09-15,"The title of the patent is apparatus and method for selecting a working data set for model development and its abstract is the present invention provides a data selection apparatus which augments a set of training examples with the desired output data. the resulting augmented data set is normalized such that the augmented data values range between -1 and +1 and such that the mean of the augmented data set is zero. the data selection apparatus then groups the augmented and normalized data set into related clusters using a clusterizer. preferably, the clusterizer is a neural network such as a kohonen self-organizing map (som). the data selection apparatus further applies an extractor to cull a working set of data from the clusterized data set. the present invention thus picks, or filters, a set of data which is more nearly uniformly distributed across the portion of the input space of interest to minimize the maximum absolute error over the entire input space. the output of the data selection apparatus is provided to train the analyzer with important sub-sets of the training data rather than with all available training data. a smaller training data set significantly reduces the complexity of the model building or analyzer construction process. dated 1998-09-15"
5810747,remote site medical intervention system,"an interactive intervention training system used for monitoring a patient suffering from neurological disorders of movement or a subject seeking to improve skill performance and assisting their training. a patient (or trainee) station is used in interactive training. the patient (or trainee) station includes a computer. a supervisor station is used by, for example, a medical or other professional. the patient (or trainee) station and the supervisor station can communicate with each other, for example, over the internet or over a lan. the patient (or trainee) station may be located remotely or locally with respect to the supervisor station. sensors collect physiologic information and physical information from the patient or subject while the patient or subject is undergoing training. this information is provided to the supervisor station. it may be summarized and displayed to the patient/subject and/or the supervisor. the patient/subject and the supervisor can communicate with each other, for example, via video, in real time. an expert system and neural network determine a goal to be achieved during training. there may be more than one patient (or trainee) station, thus allowing the supervisor to supervise a number of patients/subjects concurrently.",1998-09-22,"The title of the patent is remote site medical intervention system and its abstract is an interactive intervention training system used for monitoring a patient suffering from neurological disorders of movement or a subject seeking to improve skill performance and assisting their training. a patient (or trainee) station is used in interactive training. the patient (or trainee) station includes a computer. a supervisor station is used by, for example, a medical or other professional. the patient (or trainee) station and the supervisor station can communicate with each other, for example, over the internet or over a lan. the patient (or trainee) station may be located remotely or locally with respect to the supervisor station. sensors collect physiologic information and physical information from the patient or subject while the patient or subject is undergoing training. this information is provided to the supervisor station. it may be summarized and displayed to the patient/subject and/or the supervisor. the patient/subject and the supervisor can communicate with each other, for example, via video, in real time. an expert system and neural network determine a goal to be achieved during training. there may be more than one patient (or trainee) station, thus allowing the supervisor to supervise a number of patients/subjects concurrently. dated 1998-09-22"
5812083,non-cooperative target identification using antenna pattern shape,"a method of classification of a device by recognizing distorted signals, generally a rf antenna pattern, emanating therefrom wherein a trainable recognition system, preferably a neural network, is provided. known distorted signals are applied to the trainable recognition system to train the recognition system to recognize individually each of a plurality of different signals having distortion therein. unknown distorted signals are then provided emanating from a remote device and the unknown distorted signals are then used to classify the remote device by analyzing the received unknown distorted signals in the trained recognition system. the steps of training and classifying each include the step of converting the signals from the time domain to the frequency domain. the steps of training and classifying also can each include the step of down sampling the frequency domain signals to compress the signature content.",1998-09-22,"The title of the patent is non-cooperative target identification using antenna pattern shape and its abstract is a method of classification of a device by recognizing distorted signals, generally a rf antenna pattern, emanating therefrom wherein a trainable recognition system, preferably a neural network, is provided. known distorted signals are applied to the trainable recognition system to train the recognition system to recognize individually each of a plurality of different signals having distortion therein. unknown distorted signals are then provided emanating from a remote device and the unknown distorted signals are then used to classify the remote device by analyzing the received unknown distorted signals in the trained recognition system. the steps of training and classifying each include the step of converting the signals from the time domain to the frequency domain. the steps of training and classifying also can each include the step of down sampling the frequency domain signals to compress the signature content. dated 1998-09-22"
5812698,handwriting recognition system and method,"a system for recognizing handwritten characters, including pre-processing apparatus for generating a set of features for each handwritten character, a neural network disposed for operating on sparse data structures of those features and generating a set of confidence values for each possible character symbol which might correspond to the handwritten character, and post-processing apparatus for adjusting those confidence values and for selecting a character symbol consistent with external knowledge about handwritten characters and the language they are written in. the pre-processing apparatus scales and re-parameterizes the handwritten strokes, encodes the scaled and re-parameterizd strokes into fuzzy membership vectors and binary pointwise data, and combines the vectors and data into a sparse data structure of features. the (nonconvolutional) neural network performs a matrix-vector multiply on the sparse data structure, using only the data for nonzero features collected in that structure, and, for a first layer of that neural network, using only successive chunks of the neural weights. the post-processing apparatus adjusts the confidence values for character symbols using a set of expert rules embodying common-sense knowledge, from which it generates a set of character probabilities for each character position; these character probabilities are combined with a markov model of character sequence transitions and a dictionary of known words, to produce a final work output for a sequence of handwritten characters.",1998-09-22,"The title of the patent is handwriting recognition system and method and its abstract is a system for recognizing handwritten characters, including pre-processing apparatus for generating a set of features for each handwritten character, a neural network disposed for operating on sparse data structures of those features and generating a set of confidence values for each possible character symbol which might correspond to the handwritten character, and post-processing apparatus for adjusting those confidence values and for selecting a character symbol consistent with external knowledge about handwritten characters and the language they are written in. the pre-processing apparatus scales and re-parameterizes the handwritten strokes, encodes the scaled and re-parameterizd strokes into fuzzy membership vectors and binary pointwise data, and combines the vectors and data into a sparse data structure of features. the (nonconvolutional) neural network performs a matrix-vector multiply on the sparse data structure, using only the data for nonzero features collected in that structure, and, for a first layer of that neural network, using only successive chunks of the neural weights. the post-processing apparatus adjusts the confidence values for character symbols using a set of expert rules embodying common-sense knowledge, from which it generates a set of character probabilities for each character position; these character probabilities are combined with a markov model of character sequence transitions and a dictionary of known words, to produce a final work output for a sequence of handwritten characters. dated 1998-09-22"
5812700,data compression neural network with winner-take-all function,"the invention is embodied in an image data system including a lossy image compressor having an image compression ratio in excess of 10 for producing first compressed image data from an original image, the first compressed image data specifying a corresponding one of a set of predetermined images, apparatus for computing an difference between the original image and the predetermined image specified by the first compressed image data and a lossless image compressor for compressing at least the difference to produce second compressed image data.",1998-09-22,"The title of the patent is data compression neural network with winner-take-all function and its abstract is the invention is embodied in an image data system including a lossy image compressor having an image compression ratio in excess of 10 for producing first compressed image data from an original image, the first compressed image data specifying a corresponding one of a set of predetermined images, apparatus for computing an difference between the original image and the predetermined image specified by the first compressed image data and a lossless image compressor for compressing at least the difference to produce second compressed image data. dated 1998-09-22"
5812992,method and system for training a neural network with adaptive weight updating and adaptive pruning in principal component space,"a signal processing system and method for accomplishing signal processing using a neural network that incorporates adaptive weight updating and adaptive pruning for tracking non-stationary signal is presented. the method updates the structural parameters of the neural network in principal component space (eigenspace) for every new available input sample. the non-stationary signal is recursively transformed into a matrix of eigenvectors with a corresponding matrix of eigenvalues. the method applies principal component pruning consisting of deleting the eigenmodes corresponding to the smallest saliencies, where a sum of the smallest saliencies is less than a predefined threshold level. removing eigenmodes with low saliencies reduces the effective number of parameters and generally improves generalization. the output is then computed by using the remaining eigenmodes and the weights of the neural network are updated using adaptive filtering techniques.",1998-09-22,"The title of the patent is method and system for training a neural network with adaptive weight updating and adaptive pruning in principal component space and its abstract is a signal processing system and method for accomplishing signal processing using a neural network that incorporates adaptive weight updating and adaptive pruning for tracking non-stationary signal is presented. the method updates the structural parameters of the neural network in principal component space (eigenspace) for every new available input sample. the non-stationary signal is recursively transformed into a matrix of eigenvectors with a corresponding matrix of eigenvalues. the method applies principal component pruning consisting of deleting the eigenmodes corresponding to the smallest saliencies, where a sum of the smallest saliencies is less than a predefined threshold level. removing eigenmodes with low saliencies reduces the effective number of parameters and generally improves generalization. the output is then computed by using the remaining eigenmodes and the weights of the neural network are updated using adaptive filtering techniques. dated 1998-09-22"
5812993,digital hardware architecture for realizing neural network,"a digital neural network architecture including a forward cascade of layers of neurons, having one input channel and one output channel, for forward processing of data examples that include many data packets. backward cascade of layers of neurons, having one input channel and one output channel, for backward propagation learning of errors of the processed data examples. each packet being of a given size. the forward cascade is adapted to be fed, through the input channel, with a succession of data examples and to deliver a succession of partially and fully processed data examples each consisting of a plurality of packets. the fully processed data examples are delivered through the one output channel. each one of the layers is adapted to receive as input in its input channel a first number of data packets per time unit and to deliver as output in its output channel a second number of data packets per time unit. the forward cascade of layers is inter-connected to the backward cascade of layers by means that include inter-layer structure, such that, during processing phase of the forward cascade of neurons, any given data example that is fed from a given layer in the forward cascade to a corresponding layer in the backward cascade, through the means, is synchronized with the error of the given processed data example that is fed to the corresponding layer from a preceding layer in the backward cascade. the first number of data packets and the second number of data packets being the same for all the layers.",1998-09-22,"The title of the patent is digital hardware architecture for realizing neural network and its abstract is a digital neural network architecture including a forward cascade of layers of neurons, having one input channel and one output channel, for forward processing of data examples that include many data packets. backward cascade of layers of neurons, having one input channel and one output channel, for backward propagation learning of errors of the processed data examples. each packet being of a given size. the forward cascade is adapted to be fed, through the input channel, with a succession of data examples and to deliver a succession of partially and fully processed data examples each consisting of a plurality of packets. the fully processed data examples are delivered through the one output channel. each one of the layers is adapted to receive as input in its input channel a first number of data packets per time unit and to deliver as output in its output channel a second number of data packets per time unit. the forward cascade of layers is inter-connected to the backward cascade of layers by means that include inter-layer structure, such that, during processing phase of the forward cascade of neurons, any given data example that is fed from a given layer in the forward cascade to a corresponding layer in the backward cascade, through the means, is synchronized with the error of the given processed data example that is fed to the corresponding layer from a preceding layer in the backward cascade. the first number of data packets and the second number of data packets being the same for all the layers. dated 1998-09-22"
5815608,apparatus and method for sensing and processing images,"an apparatus for sensing and processing images is provided with a photo detector array arranged in a matrix form, a control circuit for feeding a row of the array with voltage for sensitivity control, and a neural network for processing current flowing from a column of the array to the ground in order to obtain an apparatus for sensing and processing images having a simple configuration, a high frame speed, the capability of forming a focus of attention, and high throughput of data and possibility of connecting to the neural network.",1998-09-29,"The title of the patent is apparatus and method for sensing and processing images and its abstract is an apparatus for sensing and processing images is provided with a photo detector array arranged in a matrix form, a control circuit for feeding a row of the array with voltage for sensitivity control, and a neural network for processing current flowing from a column of the array to the ground in order to obtain an apparatus for sensing and processing images having a simple configuration, a high frame speed, the capability of forming a focus of attention, and high throughput of data and possibility of connecting to the neural network. dated 1998-09-29"
5815638,project estimator,"a system for estimating the effort necessary to complete a project comprises a rule-based expert system including groups of related estimation rules (rule groups) and a question table, a neural network, the neurodes of which represent groups of the detailed tasks (task groups), a rule indirection module for coupling the outputs of the rule-based expert system to selected neurodes, and an inference engine for implementing the rules of the expert system. the expert system, neural network, and rule indirection module are organized as an advisor module, which communicates information between an advisor interface and the question table. each rule group represents a quantifiable feature of the specific client/server implementation, such as its size or fault tolerance requirements. the inference engine applies the rule groups to user provided data, and each rule group generates an effort factor (ef) that provides a measure of the associated feature's potential impact on the task groups. the rule indirection module couples each effort factor to selected task group neurodes according to influence factors (if), which represent the correlation between the feature characterized by the rule group and the selected task groups. the influence factors are initially estimated as part of the expert system but may be further optimized by training the estimator using actual measured efforts and efforts estimated by the present invention.",1998-09-29,"The title of the patent is project estimator and its abstract is a system for estimating the effort necessary to complete a project comprises a rule-based expert system including groups of related estimation rules (rule groups) and a question table, a neural network, the neurodes of which represent groups of the detailed tasks (task groups), a rule indirection module for coupling the outputs of the rule-based expert system to selected neurodes, and an inference engine for implementing the rules of the expert system. the expert system, neural network, and rule indirection module are organized as an advisor module, which communicates information between an advisor interface and the question table. each rule group represents a quantifiable feature of the specific client/server implementation, such as its size or fault tolerance requirements. the inference engine applies the rule groups to user provided data, and each rule group generates an effort factor (ef) that provides a measure of the associated feature's potential impact on the task groups. the rule indirection module couples each effort factor to selected task group neurodes according to influence factors (if), which represent the correlation between the feature characterized by the rule group and the selected task groups. the influence factors are initially estimated as part of the expert system but may be further optimized by training the estimator using actual measured efforts and efforts estimated by the present invention. dated 1998-09-29"
5816247,monitoring an eeg,"an apparatus and method for eeg monitoring provides multi-dimensional classification of eeg samples, using a neural network having multiple outputs trained upon a training set of samples to define an n-dimensional space in which to classify the samples and provide to the user a display of that space.",1998-10-06,"The title of the patent is monitoring an eeg and its abstract is an apparatus and method for eeg monitoring provides multi-dimensional classification of eeg samples, using a neural network having multiple outputs trained upon a training set of samples to define an n-dimensional space in which to classify the samples and provide to the user a display of that space. dated 1998-10-06"
5818081,semiconductor device,"synapse can be formed from a smaller number of elements in a low-power semiconductor device, which realize a highly integrated neural network. precise modifications of synapse weighting become possible and a neuron computer chip of a practical level can be accomplished. the semiconductor device includes a first electrode for charge injection, connected to a floating gate through a first insulating film; a second electrode for applying programming pulses, connected to the floating gate through a second insulating film, and a mos transistor using the floating gate as its gate electrode, wherein the charge supplied from the source electrode of the mos transistor sets the potential at the first electrode to a predetermined value determined by the potential of the floating gate, and charges are transferred between the floating gate and the first electrode through the first insulating film by applying a predetermined pulsating voltage to the second electrode.",1998-10-06,"The title of the patent is semiconductor device and its abstract is synapse can be formed from a smaller number of elements in a low-power semiconductor device, which realize a highly integrated neural network. precise modifications of synapse weighting become possible and a neuron computer chip of a practical level can be accomplished. the semiconductor device includes a first electrode for charge injection, connected to a floating gate through a first insulating film; a second electrode for applying programming pulses, connected to the floating gate through a second insulating film, and a mos transistor using the floating gate as its gate electrode, wherein the charge supplied from the source electrode of the mos transistor sets the potential at the first electrode to a predetermined value determined by the potential of the floating gate, and charges are transferred between the floating gate and the first electrode through the first insulating film by applying a predetermined pulsating voltage to the second electrode. dated 1998-10-06"
5819006,method for operating a neural network with missing and/or incomplete data,"a neural network system is provided that models the system in a system model (12) with the output thereof providing a predicted output. this predicted output is modified or controlled by an output control (14). input data is processed in a data preprocess step (10) to reconcile the data for input to the system model (12). additionally, the error resulted from the reconciliation is input to an uncertainty model to predict the uncertainty in the predicted output. this is input to a decision processor (20) which is utilized to control the output control (14). the output control (14) is controlled to either vary the predicted output or to inhibit the predicted output whenever the output of the uncertainty model (18) exceeds a predetermined decision threshold, input by a decision threshold block (22). additionally, a validity model (16) is also provided which represents the reliability or validity of the output as a function of the number of data points in a given data region during training of the system model (12). this predicts the confidence in the predicted output which is also input to the decision processor (20). the decision processor (20) therefore bases its decision on the predicted confidence and the predicted uncertainty. additionally, the uncertainty output by the data preprocess block (10) can be utilized to train the system model (12).",1998-10-06,"The title of the patent is method for operating a neural network with missing and/or incomplete data and its abstract is a neural network system is provided that models the system in a system model (12) with the output thereof providing a predicted output. this predicted output is modified or controlled by an output control (14). input data is processed in a data preprocess step (10) to reconcile the data for input to the system model (12). additionally, the error resulted from the reconciliation is input to an uncertainty model to predict the uncertainty in the predicted output. this is input to a decision processor (20) which is utilized to control the output control (14). the output control (14) is controlled to either vary the predicted output or to inhibit the predicted output whenever the output of the uncertainty model (18) exceeds a predetermined decision threshold, input by a decision threshold block (22). additionally, a validity model (16) is also provided which represents the reliability or validity of the output as a function of the number of data points in a given data region during training of the system model (12). this predicts the confidence in the predicted output which is also input to the decision processor (20). the decision processor (20) therefore bases its decision on the predicted confidence and the predicted uncertainty. additionally, the uncertainty output by the data preprocess block (10) can be utilized to train the system model (12). dated 1998-10-06"
5819226,fraud detection using predictive modeling,an automated system and method detects fraudulent transactions using a predictive model such as a neural network to evaluate individual customer accounts and identify potentially fraudulent transactions based on learned relationships among known variables. the system may also output reason codes indicating relative contributions of various variables to a particular result. the system periodically monitors its performance and redevelops the model when performance drops below a predetermined level.,1998-10-06,The title of the patent is fraud detection using predictive modeling and its abstract is an automated system and method detects fraudulent transactions using a predictive model such as a neural network to evaluate individual customer accounts and identify potentially fraudulent transactions based on learned relationships among known variables. the system may also output reason codes indicating relative contributions of various variables to a particular result. the system periodically monitors its performance and redevelops the model when performance drops below a predetermined level. dated 1998-10-06
5819242,fuzzy-neural network system and a learning method therein,"a fuzzy-neural network system includes: an input layer outputting values of input parameters; a membership layer wherein a multiple number of regions for each of the input parameters are formed by dividing the probable range of the input parameter and a membership function is defined for each of the regions, the membership layer producing membership values as to the regions for each of the input parameters, in accordance with the output values from the input layer; a rule layer wherein specific rules are formed between regions belonging to different input parameters, the rule layer outputting a suitability for each of the rules; an outputting layer producing an output parameter or parameters in accordance with the output values from the rule layer; and a membership value setup means which, if some of the input parameters are unknown, sets up prescribed values as membership values corresponding to the unknown parameters.",1998-10-06,"The title of the patent is fuzzy-neural network system and a learning method therein and its abstract is a fuzzy-neural network system includes: an input layer outputting values of input parameters; a membership layer wherein a multiple number of regions for each of the input parameters are formed by dividing the probable range of the input parameter and a membership function is defined for each of the regions, the membership layer producing membership values as to the regions for each of the input parameters, in accordance with the output values from the input layer; a rule layer wherein specific rules are formed between regions belonging to different input parameters, the rule layer outputting a suitability for each of the rules; an outputting layer producing an output parameter or parameters in accordance with the output values from the rule layer; and a membership value setup means which, if some of the input parameters are unknown, sets up prescribed values as membership values corresponding to the unknown parameters. dated 1998-10-06"
5819245,method of organizing data into a graphically oriented format,"a neural network (10) organizes the data items into a graphically oriented format by retrieving data items from a database (68) where each data item has a plurality of attributes. the neural network is organized (102) such that data items having similar attributes are assigned to neurons located closer together. the neurons of the neural network are matched (104) with the data items from the database and stored in a cross reference table. the cross reference table is displayed (106) on a computer screen (108) in a graphical format so that user visually relates the food items and sees the similarities and differences in their attribute data by the proximity of the data items to one another. the graphic format allows easy visual interpretation of the data items. for large databases, multiple neural networks (110, 112) can be organized hierarchically.",1998-10-06,"The title of the patent is method of organizing data into a graphically oriented format and its abstract is a neural network (10) organizes the data items into a graphically oriented format by retrieving data items from a database (68) where each data item has a plurality of attributes. the neural network is organized (102) such that data items having similar attributes are assigned to neurons located closer together. the neurons of the neural network are matched (104) with the data items from the database and stored in a cross reference table. the cross reference table is displayed (106) on a computer screen (108) in a graphical format so that user visually relates the food items and sees the similarities and differences in their attribute data by the proximity of the data items to one another. the graphic format allows easy visual interpretation of the data items. for large databases, multiple neural networks (110, 112) can be organized hierarchically. dated 1998-10-06"
5819246,non-linear model automatic generating method,"the present invention is intended to automatically select least required input items for a non-linear model and improve an efficiency of building up the non-linear model. for building up, for example, a neural network as the non-linear model, a group-by rule 105 and dividing point information 106 of the data are automatically generated by a group-by rule induction device 104 by selecting a dividing method option 101 and a group selection information 103 if data for learning 101 is given. an initial neural network model generating device 107 automatically generates an initial neural network model 108 from the group-by rule 105. the initial neural network model is learned in a neural network model learning device 111 and outputted as a post-learning neural network model 112. data for learning with group information 110 which is the data for learning is generated in a data classification device 109 by using data for learning 102, group-by rule 105 and dividing point information 106. input-output variables can be automatically selected from the data for learning according to selection of the group and a neural network model of respective groups can be built up.",1998-10-06,"The title of the patent is non-linear model automatic generating method and its abstract is the present invention is intended to automatically select least required input items for a non-linear model and improve an efficiency of building up the non-linear model. for building up, for example, a neural network as the non-linear model, a group-by rule 105 and dividing point information 106 of the data are automatically generated by a group-by rule induction device 104 by selecting a dividing method option 101 and a group selection information 103 if data for learning 101 is given. an initial neural network model generating device 107 automatically generates an initial neural network model 108 from the group-by rule 105. the initial neural network model is learned in a neural network model learning device 111 and outputted as a post-learning neural network model 112. data for learning with group information 110 which is the data for learning is generated in a data classification device 109 by using data for learning 102, group-by rule 105 and dividing point information 106. input-output variables can be automatically selected from the data for learning according to selection of the group and a neural network model of respective groups can be built up. dated 1998-10-06"
5821860,driving condition-monitoring apparatus for automotive vehicles,"a driving condition-monitoring apparatus for an automotive vehicle monitors a driving condition of a driver of the automotive vehicle. at least one of behavior of the vehicle, a driving operation of the driver, and at least one condition of the driver is detected to thereby generate driving condition-indicative data indicative of the driving condition of the driver. it is determined whether the driving condition of the driver is abnormal, based on the driving condition-indicative data generated. when it is not determined that the driving condition of the driver is abnormal, a degree of normality of the driving condition of the driver is determined by inputting a plurality of pieces of the driving condition-indicative data to a neural network. at least one of warning and control of the vehicle is carried out depending on a result of the determination as to whether the driving condition of the driver is abnormal and the degree of normality of the driving condition of the driver.",1998-10-13,"The title of the patent is driving condition-monitoring apparatus for automotive vehicles and its abstract is a driving condition-monitoring apparatus for an automotive vehicle monitors a driving condition of a driver of the automotive vehicle. at least one of behavior of the vehicle, a driving operation of the driver, and at least one condition of the driver is detected to thereby generate driving condition-indicative data indicative of the driving condition of the driver. it is determined whether the driving condition of the driver is abnormal, based on the driving condition-indicative data generated. when it is not determined that the driving condition of the driver is abnormal, a degree of normality of the driving condition of the driver is determined by inputting a plurality of pieces of the driving condition-indicative data to a neural network. at least one of warning and control of the vehicle is carried out depending on a result of the determination as to whether the driving condition of the driver is abnormal and the degree of normality of the driving condition of the driver. dated 1998-10-13"
5822220,process for controlling the efficiency of the causticizing process,"continuous measurements are made of a characteristic of the individual components of green liquor fed to a slaker and white liquor exiting from the slaker. the liquor component measurements provide a precise characterization of the liquors allowing for a more efficient control of the causticizing reaction in the kraft process. the individual component measurements are provided as inputs, along with certain ambient measurements, to a non-linear controller. the controller produces a causticizing control signal which is used to control the amount of lime introduced to the slaker. the controller is adapted to a particular process installation through the application of data collected from that installation. the controller, for example a neural network or fuzzy logic controller, produces a causticizing control signal according to unique parameters developed for the specific installation.",1998-10-13,"The title of the patent is process for controlling the efficiency of the causticizing process and its abstract is continuous measurements are made of a characteristic of the individual components of green liquor fed to a slaker and white liquor exiting from the slaker. the liquor component measurements provide a precise characterization of the liquors allowing for a more efficient control of the causticizing reaction in the kraft process. the individual component measurements are provided as inputs, along with certain ambient measurements, to a non-linear controller. the controller produces a causticizing control signal which is used to control the amount of lime introduced to the slaker. the controller is adapted to a particular process installation through the application of data collected from that installation. the controller, for example a neural network or fuzzy logic controller, produces a causticizing control signal according to unique parameters developed for the specific installation. dated 1998-10-13"
5822741,neural network/conceptual clustering fraud detection architecture,the invention relates to an apparatus for detecting fraud using a neural network. the architecture of the system involves first employing a conceptual clustering technique to generate a collection of classes from historical data. neural networks are provided for each class created by the clustering step and the networks are trained using the same historical data. this apparatus is particularly useful for detecting the incidence of fraudulent activity from very large amounts of data such as tax returns or insurance claims.,1998-10-13,The title of the patent is neural network/conceptual clustering fraud detection architecture and its abstract is the invention relates to an apparatus for detecting fraud using a neural network. the architecture of the system involves first employing a conceptual clustering technique to generate a collection of classes from historical data. neural networks are provided for each class created by the clustering step and the networks are trained using the same historical data. this apparatus is particularly useful for detecting the incidence of fraudulent activity from very large amounts of data such as tax returns or insurance claims. dated 1998-10-13
5822742,dynamically stable associative learning neural network system,"a dynamically stable associative learning neural system includes a plurality of neural network architectural units. a neural network architectural unit has as input both condition stimuli and unconditioned stimulus, an output neuron for accepting the input, and patch elements interposed between each input and the output neuron. the patches in the architectural unit can be modified and added. a neural network can be formed from a single unit, a layer of units, or multiple layers of units.",1998-10-13,"The title of the patent is dynamically stable associative learning neural network system and its abstract is a dynamically stable associative learning neural system includes a plurality of neural network architectural units. a neural network architectural unit has as input both condition stimuli and unconditioned stimulus, an output neuron for accepting the input, and patch elements interposed between each input and the output neuron. the patches in the architectural unit can be modified and added. a neural network can be formed from a single unit, a layer of units, or multiple layers of units. dated 1998-10-13"
5824937,signal analysis device having at least one stretched string and one pickup,a signal analysis device for determining the pitch of a plucked string in which propagation times of plucking transients are evaluated to determine pitch. a neural network may be employed to perform evaluation based upon groups of pulses.,1998-10-20,The title of the patent is signal analysis device having at least one stretched string and one pickup and its abstract is a signal analysis device for determining the pitch of a plucked string in which propagation times of plucking transients are evaluated to determine pitch. a neural network may be employed to perform evaluation based upon groups of pulses. dated 1998-10-20
5825645,two-level system identifier apparatus with optimization,"an apparatus for identifying the structure and estimating the parameters in the structure of a controlled system is described. the apparatus uses a structure identifier, in a preferred embodiment, employing a neural network, to match the relationship of the system inputs and outputs to a mathematical relation. this identified structure is communicated to a parameter estimator. the parameter estimator may employ a neural network to give an initial parameter vector which can be used to generate, iteratively, better parameter estimates. this increases the efficacy of the parameter estimator. the apparatus can be inserted into controllers for various systems to improve performance. using a neural network for parameter estimation as shown can be generalized to any nonlinear optimization problem.",1998-10-20,"The title of the patent is two-level system identifier apparatus with optimization and its abstract is an apparatus for identifying the structure and estimating the parameters in the structure of a controlled system is described. the apparatus uses a structure identifier, in a preferred embodiment, employing a neural network, to match the relationship of the system inputs and outputs to a mathematical relation. this identified structure is communicated to a parameter estimator. the parameter estimator may employ a neural network to give an initial parameter vector which can be used to generate, iteratively, better parameter estimates. this increases the efficacy of the parameter estimator. the apparatus can be inserted into controllers for various systems to improve performance. using a neural network for parameter estimation as shown can be generalized to any nonlinear optimization problem. dated 1998-10-20"
5825646,method and apparatus for determining the sensitivity of inputs to a neural network on output parameters,"a distributed control system (14) receives on the input thereof the control inputs and then outputs control signals to a plant (10) for the operation thereof. the measured variables of the plant and the control inputs are input to a predictive model (34) that operates in conjunction with an inverse model (36) to generate predicted control inputs. the predicted control inputs are processed through a filter (46) to apply hard constraints and sensitivity modifiers, the values of which are received from a control parameter block (22). during operation, the sensitivity of output variables on various input variables is determined. this information can be displayed and then the user allowed to select which of the input variables constitute the most sensitive input variables. these can then be utilized with a control network (470) to modify the predicted values of the input variables. additionally, a neural network (406) can be trained on only the selected input variables that are determined to be the most sensitive. in this operation, the network is first configured and trained with all input nodes and with all training data. this provides a learned representation of the output wherein the combined effects of all other input variables are taken into account in the determination of the effect of each of the input variables thereon. the network (406) is then reconfigured with only the selected inputs and then the network (406) again trained on only the input/output pairs associated with the select input variables.",1998-10-20,"The title of the patent is method and apparatus for determining the sensitivity of inputs to a neural network on output parameters and its abstract is a distributed control system (14) receives on the input thereof the control inputs and then outputs control signals to a plant (10) for the operation thereof. the measured variables of the plant and the control inputs are input to a predictive model (34) that operates in conjunction with an inverse model (36) to generate predicted control inputs. the predicted control inputs are processed through a filter (46) to apply hard constraints and sensitivity modifiers, the values of which are received from a control parameter block (22). during operation, the sensitivity of output variables on various input variables is determined. this information can be displayed and then the user allowed to select which of the input variables constitute the most sensitive input variables. these can then be utilized with a control network (470) to modify the predicted values of the input variables. additionally, a neural network (406) can be trained on only the selected input variables that are determined to be the most sensitive. in this operation, the network is first configured and trained with all input nodes and with all training data. this provides a learned representation of the output wherein the combined effects of all other input variables are taken into account in the determination of the effect of each of the input variables thereon. the network (406) is then reconfigured with only the selected inputs and then the network (406) again trained on only the input/output pairs associated with the select input variables. dated 1998-10-20"
5825907,neural network system for classifying fingerprints,"a system for automatically classification of human fingerprints. an unidentified fingerprint is processed to produce a direction map. the direction map is processed to generate a course direction map. the coarse direction map is input to a locally connected, highly constrained feed-forward neural network. the neural network has a highly structured architecture well-suited to exploit the rotational symmetries and asymmetries of human fingerprints. the neural network classifies the unidentified fingerprint into one of five classifications: whorl, double loop, left loop, right arch and arch.",1998-10-20,"The title of the patent is neural network system for classifying fingerprints and its abstract is a system for automatically classification of human fingerprints. an unidentified fingerprint is processed to produce a direction map. the direction map is processed to generate a course direction map. the coarse direction map is input to a locally connected, highly constrained feed-forward neural network. the neural network has a highly structured architecture well-suited to exploit the rotational symmetries and asymmetries of human fingerprints. the neural network classifies the unidentified fingerprint into one of five classifications: whorl, double loop, left loop, right arch and arch. dated 1998-10-20"
5825936,image analyzing device using adaptive criteria,""" a hybrid filter architecture and an artificial neural network is proposed for image enhancement and detection of suspicious areas in digital x-ray images that is operator, image, and digital x-ray sensor independent. the hybrid filter architecture includes an adaptive multistage nonlinear filter (amnf) cascaded with an m-channel tree structured wavelet transform (tswt). the amnf shares the advantages of an array of linear and nonlinear filters and is adaptively supervised using either an order statistic or linear operator. the filter is used for noise suppression and image enhancement and adapts to the noise characteristics of the sensor. the tswt is used for multiresolution image decomposition and reconstruction of subimages for further image enhancement of diagnostic features of interest. a multistage artificial neural network (mann) is proposed, together with kalman filtering for network training, for both improved detection or classification of suspicious areas and computational efficiency to allow the mann to be applied to full digital images without operator input. the hybrid filter architecture and mann may be applied to any gray scale image in medical imaging. the specific application of the proposed method includes: (a) improved enhancement or detection of suspicious areas as a """"second opinion"""" strategy using a computer workstation or (b) mass screening of large image databases such as that used for medical screening programs. """,1998-10-20,"The title of the patent is image analyzing device using adaptive criteria and its abstract is "" a hybrid filter architecture and an artificial neural network is proposed for image enhancement and detection of suspicious areas in digital x-ray images that is operator, image, and digital x-ray sensor independent. the hybrid filter architecture includes an adaptive multistage nonlinear filter (amnf) cascaded with an m-channel tree structured wavelet transform (tswt). the amnf shares the advantages of an array of linear and nonlinear filters and is adaptively supervised using either an order statistic or linear operator. the filter is used for noise suppression and image enhancement and adapts to the noise characteristics of the sensor. the tswt is used for multiresolution image decomposition and reconstruction of subimages for further image enhancement of diagnostic features of interest. a multistage artificial neural network (mann) is proposed, together with kalman filtering for network training, for both improved detection or classification of suspicious areas and computational efficiency to allow the mann to be applied to full digital images without operator input. the hybrid filter architecture and mann may be applied to any gray scale image in medical imaging. the specific application of the proposed method includes: (a) improved enhancement or detection of suspicious areas as a """"second opinion"""" strategy using a computer workstation or (b) mass screening of large image databases such as that used for medical screening programs. "" dated 1998-10-20"
5826249,historical database training method for neural networks,"an on-line training neural network for process control system and method trains by retrieving training sets from the stream of process data. the neural network detects the availability of new training data, and constructs a training set by retrieving the corresponding input data. the neural network is trained using the training set. over time, many training sets are presented to the neural network. when multiple presentations are needed to effectively train, a buffer of training sets is filled-and updated as new training data becomes available. the size of the buffer is selected in accordance with the training needs of the neural network. once the buffer is full, a new training set bumps the oldest training set off the top of the buffer stack. the training sets in the buffer stack can be presented one or more times each time a new training set is constructed. an historical database of timestamped data can be used to construct training sets when training input data has a time delay from sample time to availability for the neural network. the timestamps of the training input data are used to select the appropriate timestamp at which input data is retrieved for use in the training set. using the historical database, the neural network can be trained retrospectively by searching the historical database and constructing training sets based on past data.",1998-10-20,"The title of the patent is historical database training method for neural networks and its abstract is an on-line training neural network for process control system and method trains by retrieving training sets from the stream of process data. the neural network detects the availability of new training data, and constructs a training set by retrieving the corresponding input data. the neural network is trained using the training set. over time, many training sets are presented to the neural network. when multiple presentations are needed to effectively train, a buffer of training sets is filled-and updated as new training data becomes available. the size of the buffer is selected in accordance with the training needs of the neural network. once the buffer is full, a new training set bumps the oldest training set off the top of the buffer stack. the training sets in the buffer stack can be presented one or more times each time a new training set is constructed. an historical database of timestamped data can be used to construct training sets when training input data has a time delay from sample time to availability for the neural network. the timestamps of the training input data are used to select the appropriate timestamp at which input data is retrieved for use in the training set. using the historical database, the neural network can be trained retrospectively by searching the historical database and constructing training sets based on past data. dated 1998-10-20"
5828775,"method and apparatus for adjusting read-out conditions and/or image processing conditions for radiation images , radiation image read-out apparatus, and radiation image analyzing method and apparatus","a first image signal representing a radiation image of an object is obtained by exposing a stimulable phosphor sheet, on which the radiation image has been stored, to stimulating rays, which cause the stimulable phosphor sheet to emit light in proportion to the amount of energy stored thereon during its exposure to radiation, the emitted light being detected. a second image signal representing the radiation image is thereafter obtained by again exposing the stimulable phosphor sheet to stimulating rays, the light emitted by the stimulable phosphor sheet being detected. read-out conditions, under which the second image signal is to be obtained, and/or image processing conditions, under which the second image signal having been obtained is to be image processed, are adjusted on the basis of the first image signal. a storage device stores information representing a standard pattern of radiation images. a signal transforming device transforms the first image signal representing the radiation image into a transformed image signal representing the radiation image, which has been transformed into the standard pattern. a condition adjuster is provided with a neural network, which receives the transformed image signal and feeds out information representing the read-out conditions and/or the image processing conditions.",1998-10-27,"The title of the patent is method and apparatus for adjusting read-out conditions and/or image processing conditions for radiation images , radiation image read-out apparatus, and radiation image analyzing method and apparatus and its abstract is a first image signal representing a radiation image of an object is obtained by exposing a stimulable phosphor sheet, on which the radiation image has been stored, to stimulating rays, which cause the stimulable phosphor sheet to emit light in proportion to the amount of energy stored thereon during its exposure to radiation, the emitted light being detected. a second image signal representing the radiation image is thereafter obtained by again exposing the stimulable phosphor sheet to stimulating rays, the light emitted by the stimulable phosphor sheet being detected. read-out conditions, under which the second image signal is to be obtained, and/or image processing conditions, under which the second image signal having been obtained is to be image processed, are adjusted on the basis of the first image signal. a storage device stores information representing a standard pattern of radiation images. a signal transforming device transforms the first image signal representing the radiation image into a transformed image signal representing the radiation image, which has been transformed into the standard pattern. a condition adjuster is provided with a neural network, which receives the transformed image signal and feeds out information representing the read-out conditions and/or the image processing conditions. dated 1998-10-27"
5828812,recurrent neural network-based fuzzy logic system and method,"a recurrent, neural network-based fuzzy logic system includes in a rule base layer and a membership function layer neurons which each have a recurrent architecture with an output-to-input feedback path including a time delay element and a neural weight. further included is a recurrent, neural network-based fuzzy logic rule generator wherein a neural network receives and fuzzifies input data and provides data corresponding to fuzzy logic membership functions and recurrent fuzzy logic rules.",1998-10-27,"The title of the patent is recurrent neural network-based fuzzy logic system and method and its abstract is a recurrent, neural network-based fuzzy logic system includes in a rule base layer and a membership function layer neurons which each have a recurrent architecture with an output-to-input feedback path including a time delay element and a neural weight. further included is a recurrent, neural network-based fuzzy logic rule generator wherein a neural network receives and fuzzifies input data and provides data corresponding to fuzzy logic membership functions and recurrent fuzzy logic rules. dated 1998-10-27"
5828817,neural network recognizer for pdls,"apparatus and method for recognizing the language type of the page description language (pdl) of a print document that is not dependent on the presence of a `dead-ringer`, standard identifying sequence of characters for language type identification.",1998-10-27,"The title of the patent is neural network recognizer for pdls and its abstract is apparatus and method for recognizing the language type of the page description language (pdl) of a print document that is not dependent on the presence of a `dead-ringer`, standard identifying sequence of characters for language type identification. dated 1998-10-27"
5828981,generating pore types and synthetic capillary pressure curves from wireline logs using neural networks,"methods of directly analyzing wireline well logging data to derive pore types, pore volumes and capillary pressure curves from the wireline logs are disclosed. a trained and validated neural network is applied to wireline log data on porosity, bulk density and shallow, medium and deep conductivity to derive synthetic pore type proportions as a function of depth. these synthetic data are then applied through a derived and validated capillary pressure curve data model to derive pore volume and pressure data as a function of borehole depth.",1998-10-27,"The title of the patent is generating pore types and synthetic capillary pressure curves from wireline logs using neural networks and its abstract is methods of directly analyzing wireline well logging data to derive pore types, pore volumes and capillary pressure curves from the wireline logs are disclosed. a trained and validated neural network is applied to wireline log data on porosity, bulk density and shallow, medium and deep conductivity to derive synthetic pore type proportions as a function of depth. these synthetic data are then applied through a derived and validated capillary pressure curve data model to derive pore volume and pressure data as a function of borehole depth. dated 1998-10-27"
5832106,method for camera calibration of range imaging system by use of neural network,"a method and an apparatus for acquisition of calibrated three dimensional data from camera image. the apparatus for acquisition of calibrated three dimensional data from camera image includes a cameral, alight source and an image processing computer. the camera acquires a light strip image of a target object. the light illuminates light strip to the target object and provides information about the illumination angle form base line (or reference line). the image processing computer obtains image and information about the angle .theta. of light plane from base line; computes connection strength of neural network and acquires calibrated three dimensional data in neutral network based on the obtained information. the mapping relationship between a control point in the three-dimensional space and a control point projected onto the two-dimensional image plane and illumination angle of light source to control point are derived by the neural network circuit.",1998-11-03,"The title of the patent is method for camera calibration of range imaging system by use of neural network and its abstract is a method and an apparatus for acquisition of calibrated three dimensional data from camera image. the apparatus for acquisition of calibrated three dimensional data from camera image includes a cameral, alight source and an image processing computer. the camera acquires a light strip image of a target object. the light illuminates light strip to the target object and provides information about the illumination angle form base line (or reference line). the image processing computer obtains image and information about the angle .theta. of light plane from base line; computes connection strength of neural network and acquires calibrated three dimensional data in neutral network based on the obtained information. the mapping relationship between a control point in the three-dimensional space and a control point projected onto the two-dimensional image plane and illumination angle of light source to control point are derived by the neural network circuit. dated 1998-11-03"
5832108,pattern recognition method using a network and system therefor,"an interval 0, 1! of the output of neural network is equally divided into m (m being an integer of two or more), and the numbers or frequencies of data for the correct/incorrect patterns contained in the i-th interval (i-1)/m, i/m! are .mu.1i and .mu.0i, respectively (where, i=1 . . . m). in this case, if this network provides an output contained in the i-th interval to unknown pattern data, this pattern is stored as a likelihood conversion table so that the pattern outputs likelihood p1i, which is a category, in an equation p1i=(.mu.1i+1)/(.mu.1i+.mu.0i+2). then, when a value contained in the i-th interval (i-1)/m, i/m! is output from a neural network, the likelihood convertor receives it as an input and outputs p1i which is so to speak normalized likelihood.",1998-11-03,"The title of the patent is pattern recognition method using a network and system therefor and its abstract is an interval 0, 1! of the output of neural network is equally divided into m (m being an integer of two or more), and the numbers or frequencies of data for the correct/incorrect patterns contained in the i-th interval (i-1)/m, i/m! are .mu.1i and .mu.0i, respectively (where, i=1 . . . m). in this case, if this network provides an output contained in the i-th interval to unknown pattern data, this pattern is stored as a likelihood conversion table so that the pattern outputs likelihood p1i, which is a category, in an equation p1i=(.mu.1i+1)/(.mu.1i+.mu.0i+2). then, when a value contained in the i-th interval (i-1)/m, i/m! is output from a neural network, the likelihood convertor receives it as an input and outputs p1i which is so to speak normalized likelihood. dated 1998-11-03"
5832132,image processing using neural network,"use is made of a neural network in order to restore a binary image to an original multi-level image, by way of example. using the neural network makes it possible to raise the accuracy of restoration and the speed of processing.",1998-11-03,"The title of the patent is image processing using neural network and its abstract is use is made of a neural network in order to restore a binary image to an original multi-level image, by way of example. using the neural network makes it possible to raise the accuracy of restoration and the speed of processing. dated 1998-11-03"
5832183,information recognition system and control system using same,"an information recognition circuit comprises a plurality of recognition processing units each composed of a neural network. teacher signals and information signals to be processed are supplied to a plurality of the units, individually so as to obtain output signals by executing individual learning. thereafter, the plural units are connected to each other so as to construct a large scale information recognition system. further, in the man-machine interface system, a plurality of operating instruction data are prepared. an operator's face is sensed by a tv camera to extract the factors related to the operator's facial expression. the neural network analogizes operator's feeling on the basis of the extracted factors. in accordance with the guessed results, a specific sort of the operating instruction is selected from a plurality of sorts of the operating instructions, and the selected instruction is displayed as an appropriate instruction for the operator. further, the one- loop controller for automatizing operation comprises an input interface section for acquiring image information, an image recognition section for recognizing the image using the acquired image information, a control section for calculating control commands according to the image recognition results, and an output interface for outputting control commands to process actuators or subordinate controllers, respectively.",1998-11-03,"The title of the patent is information recognition system and control system using same and its abstract is an information recognition circuit comprises a plurality of recognition processing units each composed of a neural network. teacher signals and information signals to be processed are supplied to a plurality of the units, individually so as to obtain output signals by executing individual learning. thereafter, the plural units are connected to each other so as to construct a large scale information recognition system. further, in the man-machine interface system, a plurality of operating instruction data are prepared. an operator's face is sensed by a tv camera to extract the factors related to the operator's facial expression. the neural network analogizes operator's feeling on the basis of the extracted factors. in accordance with the guessed results, a specific sort of the operating instruction is selected from a plurality of sorts of the operating instructions, and the selected instruction is displayed as an appropriate instruction for the operator. further, the one- loop controller for automatizing operation comprises an input interface section for acquiring image information, an image recognition section for recognizing the image using the acquired image information, a control section for calculating control commands according to the image recognition results, and an output interface for outputting control commands to process actuators or subordinate controllers, respectively. dated 1998-11-03"
5832421,method for blade temperature estimation in a steam turbine,"a method for blade temp estimation in a steam turbine utilizes measurement values including pressure and temperature at locations other than directly at the blades, principally at the input and output stages. initially, blade temperature is simulated by using a water/steam cycle analysis program as well as by directed experiments. an artificial neural network (ann) is trained by presenting the measurement values and the blade temp values. in a hybrid approach, 5 measured values are utilized. a subset of 4 parameter values is used for training the ann and another subset of 3 values is used for performing a calculation for another intermediate parameter. using the intermediate parameter and one of the 5 measured values, a blade temperature is calculated.",1998-11-03,"The title of the patent is method for blade temperature estimation in a steam turbine and its abstract is a method for blade temp estimation in a steam turbine utilizes measurement values including pressure and temperature at locations other than directly at the blades, principally at the input and output stages. initially, blade temperature is simulated by using a water/steam cycle analysis program as well as by directed experiments. an artificial neural network (ann) is trained by presenting the measurement values and the blade temp values. in a hybrid approach, 5 measured values are utilized. a subset of 4 parameter values is used for training the ann and another subset of 3 values is used for performing a calculation for another intermediate parameter. using the intermediate parameter and one of the 5 measured values, a blade temperature is calculated. dated 1998-11-03"
5832466,system and method for dynamic learning control in genetically enhanced back-propagation neural networks,"in the design and implementation of neural networks, training is determined by a series of architectural and parametric decisions. a method is disclosed that, using genetic algorithms, improves the training characteristics of a neural network. the method begins with a population and iteratively modifies one or more parameters in each generation based on the network with the best training response in the previous generation.",1998-11-03,"The title of the patent is system and method for dynamic learning control in genetically enhanced back-propagation neural networks and its abstract is in the design and implementation of neural networks, training is determined by a series of architectural and parametric decisions. a method is disclosed that, using genetic algorithms, improves the training characteristics of a neural network. the method begins with a population and iteratively modifies one or more parameters in each generation based on the network with the best training response in the previous generation. dated 1998-11-03"
5832468,method for improving process control by reducing lag time of sensors using artificial neural networks,"an artificial neural network is used to predict the current state of a process based upon sensor measurements of the process variables at previous times. the output of the neural network provides the process control system with the predicted process state, thereby reducing the time lag of the sensors and providing improved control of the process.",1998-11-03,"The title of the patent is method for improving process control by reducing lag time of sensors using artificial neural networks and its abstract is an artificial neural network is used to predict the current state of a process based upon sensor measurements of the process variables at previous times. the output of the neural network provides the process control system with the predicted process state, thereby reducing the time lag of the sensors and providing improved control of the process. dated 1998-11-03"
5835901,perceptive system including a neural network,"a real-time learning (rtl) neural network is capable of indicating when an input feature vector is novel with respect to feature vectors contained within its training data set, and is capable of learning to generate a correct response to a new data vector while maintaining correct responses to previously learned data vectors without requiring that the neural network be retrained on the previously learned data. the neural network has a sensor for inputting a feature vector, a first layer and a second layer. the feature vector is supplied to the first layer which may have one or more declared and unused nodes. during training, the input feature vector is clustered to a declared node only if it lies within a hypervolume defined by the declared node's automatically selectable reject radius, else the input feature vector is clustered to an unused node. clustering in overlapping hypervolumes is determined by a decision surface. during testing of the rtl network, the same strategy is applied to cluster an input feature vector to declared (existing) nodes. if clustering occurs, then a classification signal corresponding to the node is generated. however, if the input feature vector is not clustered to a declared node, then the second layer outputs a signal indicating novelty. the rtl neural network is used in a perceptive system which alternatively selects the rtl network novelty output or the output of a classifier trained on historical target data if the input vector is a subset of the historical target data.",1998-11-10,"The title of the patent is perceptive system including a neural network and its abstract is a real-time learning (rtl) neural network is capable of indicating when an input feature vector is novel with respect to feature vectors contained within its training data set, and is capable of learning to generate a correct response to a new data vector while maintaining correct responses to previously learned data vectors without requiring that the neural network be retrained on the previously learned data. the neural network has a sensor for inputting a feature vector, a first layer and a second layer. the feature vector is supplied to the first layer which may have one or more declared and unused nodes. during training, the input feature vector is clustered to a declared node only if it lies within a hypervolume defined by the declared node's automatically selectable reject radius, else the input feature vector is clustered to an unused node. clustering in overlapping hypervolumes is determined by a decision surface. during testing of the rtl network, the same strategy is applied to cluster an input feature vector to declared (existing) nodes. if clustering occurs, then a classification signal corresponding to the node is generated. however, if the input feature vector is not clustered to a declared node, then the second layer outputs a signal indicating novelty. the rtl neural network is used in a perceptive system which alternatively selects the rtl network novelty output or the output of a classifier trained on historical target data if the input vector is a subset of the historical target data. dated 1998-11-10"
5838881,system and method for mitigation of streaming electrification in power transformers by intelligent cooling system control,"an intelligent controller for on-line monitoring of circulation system in order to minimize a problem of streaming electrification and to avoid degradation of the electrical insulation of the system. the controller is designed to combine advantages of fuzzy logic and neural networks, and includes a fuzzy logic (employing a pseudo-neural network) for acquiring analog input data from a plurality of sensors and for interpreting this input data into fuzzyfier outputs. a main processor consisting of a fully connected feed-forward neural network serves for processing the fuzzifier outputs in order to obtain controlling instructions for the system.",1998-11-17,"The title of the patent is system and method for mitigation of streaming electrification in power transformers by intelligent cooling system control and its abstract is an intelligent controller for on-line monitoring of circulation system in order to minimize a problem of streaming electrification and to avoid degradation of the electrical insulation of the system. the controller is designed to combine advantages of fuzzy logic and neural networks, and includes a fuzzy logic (employing a pseudo-neural network) for acquiring analog input data from a plurality of sensors and for interpreting this input data into fuzzyfier outputs. a main processor consisting of a fully connected feed-forward neural network serves for processing the fuzzifier outputs in order to obtain controlling instructions for the system. dated 1998-11-17"
5839438,computer-based neural network system and method for medical diagnosis and interpretation,""" a neural network system and method for diagnosing patients' medical conditions provide an efficient aid in identifying and interpreting factors which are significant in the medical diagnosis. the neural network is trained to recognize medical conditions by being provided with input data that is available for a number of patients, and diagnosis made by physicians in each case. upon completion of the training period the neural network system uses input measurement and interview data to produce a score, or a graded classification, of a patient's medical condition that is accompanied with a diagnosis interpretation. the interpretation is a sorted catalogue of individual factors and interactions that influenced the score. the interpretive facility of the present invention is based on comparison with a set of nominal values for each input factor or interaction. it can assist the physician in making a diagnosis of the patient's condition and can further provide a """"second opinion"""" that may confirm the physician's findings or point to ambiguities that call for a more detailed analysis. """,1998-11-24,"The title of the patent is computer-based neural network system and method for medical diagnosis and interpretation and its abstract is "" a neural network system and method for diagnosing patients' medical conditions provide an efficient aid in identifying and interpreting factors which are significant in the medical diagnosis. the neural network is trained to recognize medical conditions by being provided with input data that is available for a number of patients, and diagnosis made by physicians in each case. upon completion of the training period the neural network system uses input measurement and interview data to produce a score, or a graded classification, of a patient's medical condition that is accompanied with a diagnosis interpretation. the interpretation is a sorted catalogue of individual factors and interactions that influenced the score. the interpretive facility of the present invention is based on comparison with a set of nominal values for each input factor or interaction. it can assist the physician in making a diagnosis of the patient's condition and can further provide a """"second opinion"""" that may confirm the physician's findings or point to ambiguities that call for a more detailed analysis. "" dated 1998-11-24"
5841651,closed loop adaptive control of spectrum-producing step using neural networks,"characteristics of the plasma in a plasma-based manufacturing process step are monitored directly and in real time by observing the spectrum which it produces. an artificial neural network analyzes the plasma spectrum and generates control signals to control one or more of the process input parameters in response to any deviation of the spectrum beyond a narrow range. in an embodiment, a plasma reaction chamber forms a plasma in response to input parameters such as gas flow, pressure and power. the chamber includes a window through which the electromagnetic spectrum produced by a plasma in the chamber, just above the subject surface, may be viewed. the spectrum is conducted to an optical spectrometer which measures the intensity of the incoming optical spectrum at different wavelengths. the output of optical spectrometer is provided to an analyzer which produces a plurality of error signals, each indicating whether a respective one of the input parameters to the chamber is to be increased or decreased. the microcontroller provides signals to control respective controls, but these lines are intercepted and first added to the error signals, before being provided to the controls for the chamber. the analyzer can include a neural network and an optional spectrum preprocessor to reduce background noise, as well as a comparator which compares the parameter values predicted by the neural network with a set of desired values provided by the microcontroller.",1998-11-24,"The title of the patent is closed loop adaptive control of spectrum-producing step using neural networks and its abstract is characteristics of the plasma in a plasma-based manufacturing process step are monitored directly and in real time by observing the spectrum which it produces. an artificial neural network analyzes the plasma spectrum and generates control signals to control one or more of the process input parameters in response to any deviation of the spectrum beyond a narrow range. in an embodiment, a plasma reaction chamber forms a plasma in response to input parameters such as gas flow, pressure and power. the chamber includes a window through which the electromagnetic spectrum produced by a plasma in the chamber, just above the subject surface, may be viewed. the spectrum is conducted to an optical spectrometer which measures the intensity of the incoming optical spectrum at different wavelengths. the output of optical spectrometer is provided to an analyzer which produces a plurality of error signals, each indicating whether a respective one of the input parameters to the chamber is to be increased or decreased. the microcontroller provides signals to control respective controls, but these lines are intercepted and first added to the error signals, before being provided to the controls for the chamber. the analyzer can include a neural network and an optional spectrum preprocessor to reduce background noise, as well as a comparator which compares the parameter values predicted by the neural network with a set of desired values provided by the microcontroller. dated 1998-11-24"
5841671,apparatus for the operation of a plant for producing deinked pulp with state analysers constructed in the form of neural networks for the waste paper suspension,"apparatus for the operation of a plant for producing deinked pulp with state analysers constructed in the form of neural networks for the waste paper suspension. at least one measuring device (me) records spectral and/or physical characteristic values (imf, mp) of a waste paper suspension (ps). furthermore, there are closed-loop or open-loop control devices (rs1 . . . rs8) for operating means of a waste paper preparation (aaa) in the plant. according to the invention, there is at least one state analyser (za), configured in the form of a neural network (nng) or a plurality of parallel neural networks (nn1 . . . nn4), for the waste paper suspension (ps). this analyser forms from the characteristic values (imf, mp) controlled variables (st: ab, az, fl, aa, as, at) for process control of the closed-loop or open-loop control devices (rs1 . . . rs8) of operating means at least of the waste paper preparation (aaa). as controlled variables, the ratio of white to coloured papers (ab), the ratio of illustrated-magazine paper to newsprint paper (az), the average fibre length (fl), the ash content (aa), the content of dirt (as) and/or the content of adhesive contaminants (at) in the waste paper suspension (ps) are suitable with preference.",1998-11-24,"The title of the patent is apparatus for the operation of a plant for producing deinked pulp with state analysers constructed in the form of neural networks for the waste paper suspension and its abstract is apparatus for the operation of a plant for producing deinked pulp with state analysers constructed in the form of neural networks for the waste paper suspension. at least one measuring device (me) records spectral and/or physical characteristic values (imf, mp) of a waste paper suspension (ps). furthermore, there are closed-loop or open-loop control devices (rs1 . . . rs8) for operating means of a waste paper preparation (aaa) in the plant. according to the invention, there is at least one state analyser (za), configured in the form of a neural network (nng) or a plurality of parallel neural networks (nn1 . . . nn4), for the waste paper suspension (ps). this analyser forms from the characteristic values (imf, mp) controlled variables (st: ab, az, fl, aa, as, at) for process control of the closed-loop or open-loop control devices (rs1 . . . rs8) of operating means at least of the waste paper preparation (aaa). as controlled variables, the ratio of white to coloured papers (ab), the ratio of illustrated-magazine paper to newsprint paper (az), the average fibre length (fl), the ash content (aa), the content of dirt (as) and/or the content of adhesive contaminants (at) in the waste paper suspension (ps) are suitable with preference. dated 1998-11-24"
5841904,image processing method and apparatus,"in coding color still image data into multi-valued image data, the image quality deteriorates when a large compression ratio is used. in the present invention, a transmitted image is classified using a parameter transmitted from the transmitting station, and the deterioration of image quality caused by compression at the transmitting station is estimated from the image classification. in an image reproduction, a quantization table is used in the case of an adaptive discrete cosine transform (adct) method to select an optimum neural network suitable for image reproduction and improve the image quality.",1998-11-24,"The title of the patent is image processing method and apparatus and its abstract is in coding color still image data into multi-valued image data, the image quality deteriorates when a large compression ratio is used. in the present invention, a transmitted image is classified using a parameter transmitted from the transmitting station, and the deterioration of image quality caused by compression at the transmitting station is estimated from the image classification. in an image reproduction, a quantization table is used in the case of an adaptive discrete cosine transform (adct) method to select an optimum neural network suitable for image reproduction and improve the image quality. dated 1998-11-24"
5842189,method for operating a neural network with missing and/or incomplete data,"a neural network system is provided that models the system in a system model (12) with the output thereof providing a predicted output. this predicted output is modified or controlled by an output control (14). input data is processed in a data preprocess step (10) to reconcile the data for input to the system model (12). additionally, the error resulted from the reconciliation is input to an uncertainty model to predict the uncertainty in the predicted output. this is input to a decision processor (20) which is utilized to control the output control (14). the output control (14) is controlled to either vary the predicted output or to inhibit the predicted output whenever the output of the uncertainty model (18) exceeds a predetermined decision threshold, input by a decision threshold block (22). additionally, a validity model (16) is also provided which represents the reliability or validity of the output as a function of the number of data points in a given data region during training of the system model (12). this predicts the confidence in the predicted output which is also input to the decision processor (20). the decision processor (20) therefore bases its decision on the predicted confidence and the predicted uncertainty. additionally, the uncertainty output by the data preprocess block (10) can be utilized to train the system model (12).",1998-11-24,"The title of the patent is method for operating a neural network with missing and/or incomplete data and its abstract is a neural network system is provided that models the system in a system model (12) with the output thereof providing a predicted output. this predicted output is modified or controlled by an output control (14). input data is processed in a data preprocess step (10) to reconcile the data for input to the system model (12). additionally, the error resulted from the reconciliation is input to an uncertainty model to predict the uncertainty in the predicted output. this is input to a decision processor (20) which is utilized to control the output control (14). the output control (14) is controlled to either vary the predicted output or to inhibit the predicted output whenever the output of the uncertainty model (18) exceeds a predetermined decision threshold, input by a decision threshold block (22). additionally, a validity model (16) is also provided which represents the reliability or validity of the output as a function of the number of data points in a given data region during training of the system model (12). this predicts the confidence in the predicted output which is also input to the decision processor (20). the decision processor (20) therefore bases its decision on the predicted confidence and the predicted uncertainty. additionally, the uncertainty output by the data preprocess block (10) can be utilized to train the system model (12). dated 1998-11-24"
5842191,neural network that incorporates direct optical imaging,"a compact neural network architecture is trainable to sense and classify an optical image directly projected onto it. the system is based upon the combination of a two-dimensional amorphous silicon photoconductor array and a liquid-crystal spatial light modulator. appropriate filtering of the incident optical image upon capture is incorporated into the net work training rules, through a modification of the standard backpropagation training algorithm. training of the network on two image classification problems is described: the recognition of handprinted digits, and facial recognition. the network, once trained is capable of standalone operation, sensing an incident image and outputting a final classification signal in real time.",1998-11-24,"The title of the patent is neural network that incorporates direct optical imaging and its abstract is a compact neural network architecture is trainable to sense and classify an optical image directly projected onto it. the system is based upon the combination of a two-dimensional amorphous silicon photoconductor array and a liquid-crystal spatial light modulator. appropriate filtering of the incident optical image upon capture is incorporated into the net work training rules, through a modification of the standard backpropagation training algorithm. training of the network on two image classification problems is described: the recognition of handprinted digits, and facial recognition. the network, once trained is capable of standalone operation, sensing an incident image and outputting a final classification signal in real time. dated 1998-11-24"
5842194,method of recognizing images of faces or general images using fuzzy combination of multiple resolutions,"a system comprising a neural network, or computer, implementing a feature detection and a statistical procedure, together with fuzzy logic for solving the problem of recognition of faces or other objects) at multiple resolutions is described. a plurality of previously described systems for recognizing faces (or other objects) which use local autocorrelation coefficients and linear discriminant analysis are trained on a data set to recognize facial images each at a particular resolution. in a second training stage, each of the previously described systems is tested on a second training set in which the images presented to the previously described recognition systems have a matching resolution to those of the first training set, the statistical performance of this second training stage being used to train a fuzzy combination technique, that of fuzzy integrals. finally, in a test stage, the results from the classifiers at the multiple resolutions are combined using fuzzy combination to produce an aggregated system whose performance is higher than that of any of the individual systems and shows very good performance relative to all known face recognitior systems which operate on similar types of training and testing data, this aggregated system, however, not being limited to the recognition of faces and being able to be applied to the recognition of other objects.",1998-11-24,"The title of the patent is method of recognizing images of faces or general images using fuzzy combination of multiple resolutions and its abstract is a system comprising a neural network, or computer, implementing a feature detection and a statistical procedure, together with fuzzy logic for solving the problem of recognition of faces or other objects) at multiple resolutions is described. a plurality of previously described systems for recognizing faces (or other objects) which use local autocorrelation coefficients and linear discriminant analysis are trained on a data set to recognize facial images each at a particular resolution. in a second training stage, each of the previously described systems is tested on a second training set in which the images presented to the previously described recognition systems have a matching resolution to those of the first training set, the statistical performance of this second training stage being used to train a fuzzy combination technique, that of fuzzy integrals. finally, in a test stage, the results from the classifiers at the multiple resolutions are combined using fuzzy combination to produce an aggregated system whose performance is higher than that of any of the individual systems and shows very good performance relative to all known face recognitior systems which operate on similar types of training and testing data, this aggregated system, however, not being limited to the recognition of faces and being able to be applied to the recognition of other objects. dated 1998-11-24"
5845049,neural network system with n-gram term weighting method for molecular sequence classification and motif identification,"a method for rapid and sensitive protein family identification is disclosed. the new designs include an n-gram term weighting algorithm for extracting local motif patterns, an enhanced n-gram method for extracting residues of long-range correlation, and integrated neural networks for combining global and motif sequence information.",1998-12-01,"The title of the patent is neural network system with n-gram term weighting method for molecular sequence classification and motif identification and its abstract is a method for rapid and sensitive protein family identification is disclosed. the new designs include an n-gram term weighting algorithm for extracting local motif patterns, an enhanced n-gram method for extracting residues of long-range correlation, and integrated neural networks for combining global and motif sequence information. dated 1998-12-01"
5845051,learning method for multilayer perceptron neural network with n-bit data representation,"a multilayer perceptron neural network with n-bit (8-bit) data representation generates weighted sums in forward and backward calculations having 2n-bit data precision. during n-bit digital learning of a multilayer perceptron, the maximum value represented with n-bits is set to a value corresponding to the sigmoidal saturated region when the result of the weighted sum having 2n-bit data precision in the forward calculation of the multilayer perceptron is represented with n-bit data for a sigmoidal nonlinear transformation. the maximum value of n-bit presentation is set to a value comparatively smaller than that represented in 2n bits when the result of the weighted sum having the 2n-bit data precision in the backward calculation of the multilayer perceptron is represented with n-bit data. with the representation range of the weights being small, if a predetermined ratio of the weights approaches a maximum value according to the learning progress, the weight range is expanded. the efficiency of 8-bit digital learning can be enhanced to approach that of 16-bit digital learning.",1998-12-01,"The title of the patent is learning method for multilayer perceptron neural network with n-bit data representation and its abstract is a multilayer perceptron neural network with n-bit (8-bit) data representation generates weighted sums in forward and backward calculations having 2n-bit data precision. during n-bit digital learning of a multilayer perceptron, the maximum value represented with n-bits is set to a value corresponding to the sigmoidal saturated region when the result of the weighted sum having 2n-bit data precision in the forward calculation of the multilayer perceptron is represented with n-bit data for a sigmoidal nonlinear transformation. the maximum value of n-bit presentation is set to a value comparatively smaller than that represented in 2n bits when the result of the weighted sum having the 2n-bit data precision in the backward calculation of the multilayer perceptron is represented with n-bit data. with the representation range of the weights being small, if a predetermined ratio of the weights approaches a maximum value according to the learning progress, the weight range is expanded. the efficiency of 8-bit digital learning can be enhanced to approach that of 16-bit digital learning. dated 1998-12-01"
5845271,non-algorithmically implemented artificial neural networks and components thereof,"constructing and simulating artificial neural networks and components thereof within a spreadsheet environment results in user friendly neural networks which do not require algorithmic based software in order to train or operate. such neural networks can be easily cascaded to form complex neural networks and neural network systems, including neural networks capable of self-organizing so as to self-train within a spreadsheet, neural networks which train simultaneously within a spreadsheet, and neural networks capable of autonomously moving, monitoring, analyzing, and altering data within a spreadsheet. neural networks can also be cascaded together in self training neural network form to achieve a device prototyping system.",1998-12-01,"The title of the patent is non-algorithmically implemented artificial neural networks and components thereof and its abstract is constructing and simulating artificial neural networks and components thereof within a spreadsheet environment results in user friendly neural networks which do not require algorithmic based software in order to train or operate. such neural networks can be easily cascaded to form complex neural networks and neural network systems, including neural networks capable of self-organizing so as to self-train within a spreadsheet, neural networks which train simultaneously within a spreadsheet, and neural networks capable of autonomously moving, monitoring, analyzing, and altering data within a spreadsheet. neural networks can also be cascaded together in self training neural network form to achieve a device prototyping system. dated 1998-12-01"
5845285,computer system and method of data analysis,"a neural network based data comparison system compares data stored within a database against each other to determine duplicative, fraudulent, defective and/or irregular data. the system includes a database storing data therein, and a pattern database storing pattern data therein. the system further includes a data pattern build system, responsively connected to the database and to the pattern database. the data pattern build system retrieves the data from the database and generates the pattern data formatted in accordance a predetermined patten. the predetermined pattern includes an array having array locations corresponding to each character in a defined character set. the data pattern build system increments a value in each of the array locations responsive to the number of occurrences of each character in the data and stores the array as the pattern data in the pattern database. the comparison system also includes a neural network, responsively connected to the pattern database, which retrieves the pattern data stored therein and compares the pattern data to each other and determines responsive to the comparing when different pattern data match in accordance with predetermined criteria indicating that the different pattern data are duplicative, fraudulent, defective and/or irregular.",1998-12-01,"The title of the patent is computer system and method of data analysis and its abstract is a neural network based data comparison system compares data stored within a database against each other to determine duplicative, fraudulent, defective and/or irregular data. the system includes a database storing data therein, and a pattern database storing pattern data therein. the system further includes a data pattern build system, responsively connected to the database and to the pattern database. the data pattern build system retrieves the data from the database and generates the pattern data formatted in accordance a predetermined patten. the predetermined pattern includes an array having array locations corresponding to each character in a defined character set. the data pattern build system increments a value in each of the array locations responsive to the number of occurrences of each character in the data and stores the array as the pattern data in the pattern database. the comparison system also includes a neural network, responsively connected to the pattern database, which retrieves the pattern data stored therein and compares the pattern data to each other and determines responsive to the comparing when different pattern data match in accordance with predetermined criteria indicating that the different pattern data are duplicative, fraudulent, defective and/or irregular. dated 1998-12-01"
5846208,method and apparatus for the evaluation of eeg data,"a method for the evaluation of eeg data for medical purposes includes the steps of acquiring eeg data, recognizing artifacts and determining an output data value with a neural network. in a training method for a neural network, training vectors are determined to which a respective data value is allocated, the neural network is trained with these training vectors, and an output data value is determined for each neuron that is based on the allocated data values of those training vectors that are contained in the data cluster represented by the neuron. a processing apparatus and an eeg monitor are configured for implementing the evaluation method for eeg data.",1998-12-08,"The title of the patent is method and apparatus for the evaluation of eeg data and its abstract is a method for the evaluation of eeg data for medical purposes includes the steps of acquiring eeg data, recognizing artifacts and determining an output data value with a neural network. in a training method for a neural network, training vectors are determined to which a respective data value is allocated, the neural network is trained with these training vectors, and an output data value is determined for each neuron that is based on the allocated data values of those training vectors that are contained in the data cluster represented by the neuron. a processing apparatus and an eeg monitor are configured for implementing the evaluation method for eeg data. dated 1998-12-08"
5848073,method and apparatus for predicting transmission system errors and failures,"errors or failures in a transmission system (10) can be accurately predicted by a system (18) which processes error or failure data collected by the transmission system during uniform intervals to yield data indicative of the errors or failures during each of a plurality of non-uniform, overlapping intervals. the errors or failures measured during the non-uniform intervals are selectively weighted and summed within a neural network (22), trained with historical error or failure data, to yield a set of predictions, each representing the predicted number of errors or failures during each non-uniform interval. the predictions from the neural network (22) are then thresholded by a control unit (24) and used to raise an alarm which signals the possibility of error or failure in the transmission system.",1998-12-08,"The title of the patent is method and apparatus for predicting transmission system errors and failures and its abstract is errors or failures in a transmission system (10) can be accurately predicted by a system (18) which processes error or failure data collected by the transmission system during uniform intervals to yield data indicative of the errors or failures during each of a plurality of non-uniform, overlapping intervals. the errors or failures measured during the non-uniform intervals are selectively weighted and summed within a neural network (22), trained with historical error or failure data, to yield a set of predictions, each representing the predicted number of errors or failures during each non-uniform interval. the predictions from the neural network (22) are then thresholded by a control unit (24) and used to raise an alarm which signals the possibility of error or failure in the transmission system. dated 1998-12-08"
5850470,neural network for locating and recognizing a deformable object,"a system for automatically detecting and recognizing the identity of a deformable object such as a human face, within an arbitrary image scene. the system comprises an object detector implemented as a probabilistic dbnn, for determining whether the object is within the arbitrary image scene and a feature localizer also implemented as a probabilistic dbnn, for determining the position of an identifying feature on the object such as the eyes. a feature extractor is coupled to the feature localizer and receives coordinates sent from the feature localizer which are indicative of the position of the identifying feature and also extracts from the coordinates information relating to other features of the object such as the eyebrows and nose, which are used to create a low resolution image of the object. a probabilistic dbnn based object recognizer for determining the identity of the object receives the low resolution image of the object inputted from the feature extractor to identify the object.",1998-12-15,"The title of the patent is neural network for locating and recognizing a deformable object and its abstract is a system for automatically detecting and recognizing the identity of a deformable object such as a human face, within an arbitrary image scene. the system comprises an object detector implemented as a probabilistic dbnn, for determining whether the object is within the arbitrary image scene and a feature localizer also implemented as a probabilistic dbnn, for determining the position of an identifying feature on the object such as the eyes. a feature extractor is coupled to the feature localizer and receives coordinates sent from the feature localizer which are indicative of the position of the identifying feature and also extracts from the coordinates information relating to other features of the object such as the eyebrows and nose, which are used to create a low resolution image of the object. a probabilistic dbnn based object recognizer for determining the identity of the object receives the low resolution image of the object inputted from the feature extractor to identify the object. dated 1998-12-15"
5852815,neural network based prototyping system and method,"constructing and simulating artificial neural networks and components thereof within a spreadsheet environment results in user friendly neural networks which do not require algorithmic based software in order to train or operate. such neural networks can be easily cascaded to form complex neural networks and neural network systems, including neural networks capable of self-organizing so as to self-train within a spreadsheet, neural networks which train simultaneously within a spreadsheet, and neural networks capable of autonomously moving, monitoring, analyzing, and altering data within a spreadsheet. neural networks can also be cascaded together in self training neural network form to achieve a device prototyping system.",1998-12-22,"The title of the patent is neural network based prototyping system and method and its abstract is constructing and simulating artificial neural networks and components thereof within a spreadsheet environment results in user friendly neural networks which do not require algorithmic based software in order to train or operate. such neural networks can be easily cascaded to form complex neural networks and neural network systems, including neural networks capable of self-organizing so as to self-train within a spreadsheet, neural networks which train simultaneously within a spreadsheet, and neural networks capable of autonomously moving, monitoring, analyzing, and altering data within a spreadsheet. neural networks can also be cascaded together in self training neural network form to achieve a device prototyping system. dated 1998-12-22"
5852816,neural network based database scanning system,"constructing and simulating artificial neural networks and components thereof within a spreadsheet environment results in user friendly neural networks which do not require algorithmic based software in order to train or operate. such neural networks can be easily cascaded to form complex neural networks and neural network systems, including neural networks capable of self-organizing so as to self-train within a spreadsheet, neural networks which train simultaneously within a spreadsheet, and neural networks capable of autonomously moving, monitoring, analyzing, and altering data within a spreadsheet. neural networks can also be cascaded together in self training neural network form to achieve a device prototyping system.",1998-12-22,"The title of the patent is neural network based database scanning system and its abstract is constructing and simulating artificial neural networks and components thereof within a spreadsheet environment results in user friendly neural networks which do not require algorithmic based software in order to train or operate. such neural networks can be easily cascaded to form complex neural networks and neural network systems, including neural networks capable of self-organizing so as to self-train within a spreadsheet, neural networks which train simultaneously within a spreadsheet, and neural networks capable of autonomously moving, monitoring, analyzing, and altering data within a spreadsheet. neural networks can also be cascaded together in self training neural network form to achieve a device prototyping system. dated 1998-12-22"
5853351,method of determining an optimum workload corresponding to user's target heart rate and exercise device therefor,"an optimum workload corresponding to a target heart rate of an user is determined by the following method, which is preferably utilized for providing a safe training or an accurate examination of physical strength to the user. after the target heart rate is set, a first steady heart rate of the user is measured during an initial exercise cycle in which a first workload is applied to the user. the first workload is derived in accordance with the target heart rate and a statistically obtained workload versus heart rate correlation corresponding to at least one factor selected from the group consisting of the user's age, gender, body weight, and body height, etc. in addition, a second steady heart rate of the user is measured during at least one subsequent exercise cycle in which a second workload is applied to the user. the second workload is derived by entering as input parameters the applied workload and the measured heart rate at the immediately previous exercise cycle into a multiple variate model equation. consequently, the optimum workload is determined by entering the applied workload and the measure heart rate at the last exercise cycle into the multiple variate model equation. the model equation is prepared by utilizing a neural network analysis or a multiple variate analysis.",1998-12-29,"The title of the patent is method of determining an optimum workload corresponding to user's target heart rate and exercise device therefor and its abstract is an optimum workload corresponding to a target heart rate of an user is determined by the following method, which is preferably utilized for providing a safe training or an accurate examination of physical strength to the user. after the target heart rate is set, a first steady heart rate of the user is measured during an initial exercise cycle in which a first workload is applied to the user. the first workload is derived in accordance with the target heart rate and a statistically obtained workload versus heart rate correlation corresponding to at least one factor selected from the group consisting of the user's age, gender, body weight, and body height, etc. in addition, a second steady heart rate of the user is measured during at least one subsequent exercise cycle in which a second workload is applied to the user. the second workload is derived by entering as input parameters the applied workload and the measured heart rate at the immediately previous exercise cycle into a multiple variate model equation. consequently, the optimum workload is determined by entering the applied workload and the measure heart rate at the last exercise cycle into the multiple variate model equation. the model equation is prepared by utilizing a neural network analysis or a multiple variate analysis. dated 1998-12-29"
5854590,method for generating a fault indication signal,"the invention concerns a method for generating a fault indication signal in the event of a fault in an electric power supply network being monitored through a neural network arrangement. in order to rapidly generate fault indication signals capable of distinguishing between permanent metal contact short-circuits and short circuits due to arcing, a neural network (26) is used for each phase of the power supply network being monitored. each neural network (26) is taught, while being coached through simulated voltages (u.sub.r (t)) of each phase during short-circuits caused by metal contact and arcing, so that the output signal (s) assumes a pre-defined value (0.8) for arcing and another pre-defined value (0.1) for a metal contact short-circuit. at the same time, the sequentially sampled normalized values (u.sub.rn1 (t) through u.sub.rn20 (t)) of the respective phase voltages (r.sub.r (t)) of the power supply network are applied to the different neurons (31) of the input layer (27) of each neural network (26) at the same time.",1998-12-29,"The title of the patent is method for generating a fault indication signal and its abstract is the invention concerns a method for generating a fault indication signal in the event of a fault in an electric power supply network being monitored through a neural network arrangement. in order to rapidly generate fault indication signals capable of distinguishing between permanent metal contact short-circuits and short circuits due to arcing, a neural network (26) is used for each phase of the power supply network being monitored. each neural network (26) is taught, while being coached through simulated voltages (u.sub.r (t)) of each phase during short-circuits caused by metal contact and arcing, so that the output signal (s) assumes a pre-defined value (0.8) for arcing and another pre-defined value (0.1) for a metal contact short-circuit. at the same time, the sequentially sampled normalized values (u.sub.rn1 (t) through u.sub.rn20 (t)) of the respective phase voltages (r.sub.r (t)) of the power supply network are applied to the different neurons (31) of the input layer (27) of each neural network (26) at the same time. dated 1998-12-29"
5854993,component machine testing using neural network processed vibration data analysis,"a component machine testing technique is provided that performs diagnostic analysis on a vibration signal of the component machine that has been separated from power and load machine background noise in a first neural network. the diagnostic analysis, with operator direction through an interactive interface, uses a second neural network in performing a series of diagnostic operations followed by archival of any experience acquired in the testing operation being performed. in the diagnostic analysis, both time based and frequency based vibration signal information from the component machine under test are used together through a simultaneous multiple display interactive interface under operator direction.",1998-12-29,"The title of the patent is component machine testing using neural network processed vibration data analysis and its abstract is a component machine testing technique is provided that performs diagnostic analysis on a vibration signal of the component machine that has been separated from power and load machine background noise in a first neural network. the diagnostic analysis, with operator direction through an interactive interface, uses a second neural network in performing a series of diagnostic operations followed by archival of any experience acquired in the testing operation being performed. in the diagnostic analysis, both time based and frequency based vibration signal information from the component machine under test are used together through a simultaneous multiple display interactive interface under operator direction. dated 1998-12-29"
5857030,automated method and system for digital image processing of radiologic images utilizing artificial neural networks,"an automated method and system for digital imaging processing of radiologic images, wherein digital image data is acquired and subjected to multiple phases of digital imaging processing. during the pre-processing stage, simultaneous box-rim filtering and k-nearest neighbor processing and subsequent global thresholding are performed on the image data to enhance object-to-background contrast, merge subclusters and determine gray scale thresholds for further processing. next, during the preliminary selection phase, body part segmentation, morphological erosion processing, connected component analysis and image block segmentation occurs to subtract unwanted image data preliminarily identify potentials areas including abnormalities. during the pattern classification phase, feature patterns are developed for each area of interest, a supervised, back propagation neural network is trained, a feed forward neural network is developed and employed to detect true and several false positive categories, and two types of pruned neural networks are utilized in connection with a heuristic decision tree to finally determine whether the regions of interest are abnormalities or false positives.",1999-01-05,"The title of the patent is automated method and system for digital image processing of radiologic images utilizing artificial neural networks and its abstract is an automated method and system for digital imaging processing of radiologic images, wherein digital image data is acquired and subjected to multiple phases of digital imaging processing. during the pre-processing stage, simultaneous box-rim filtering and k-nearest neighbor processing and subsequent global thresholding are performed on the image data to enhance object-to-background contrast, merge subclusters and determine gray scale thresholds for further processing. next, during the preliminary selection phase, body part segmentation, morphological erosion processing, connected component analysis and image block segmentation occurs to subtract unwanted image data preliminarily identify potentials areas including abnormalities. during the pattern classification phase, feature patterns are developed for each area of interest, a supervised, back propagation neural network is trained, a feed forward neural network is developed and employed to detect true and several false positive categories, and two types of pruned neural networks are utilized in connection with a heuristic decision tree to finally determine whether the regions of interest are abnormalities or false positives. dated 1999-01-05"
5857177,neural network,"a neural network has a plurality of network neurons and a plurality of network connections which connect each network neuron to one or more other network neurons. each network neuron has control parameters in the form of an associated threshold value and/or signal distribution when a signal is supplied to other network neurons, and supplies a signal to the output in response to a comparison between said threshold value and a signal received on the input. one or more network neurons serve as network inputs, which supply an output representation in dependence on the sensed parameters applied to the network input. the network has associated with it sensor neurons which register changes in the conditions under which the network works, and the control parameters of the network neurons are regulated in dependence on this.",1999-01-05,"The title of the patent is neural network and its abstract is a neural network has a plurality of network neurons and a plurality of network connections which connect each network neuron to one or more other network neurons. each network neuron has control parameters in the form of an associated threshold value and/or signal distribution when a signal is supplied to other network neurons, and supplies a signal to the output in response to a comparison between said threshold value and a signal received on the input. one or more network neurons serve as network inputs, which supply an output representation in dependence on the sensed parameters applied to the network input. the network has associated with it sensor neurons which register changes in the conditions under which the network works, and the control parameters of the network neurons are regulated in dependence on this. dated 1999-01-05"
5857178,neural network apparatus and learning method thereof,"a neural network apparatus includes a neural network including at least two neuron layers each having a plurality of neurons and at least one synapse layer having a plurality of synapses each arranged between the neuron layers, each synapse storing a weight value between the neurons and multiplying the weight value with an output value from each of the neurons in the previous-stage neuron layer to output a product to the next-stage neuron layer, a section for causing an error signal between an output from the neural network and a desired output to back-propagate from an output-side neuron layer to an input-side neuron layer of the neural network, a learning control section for updating the weight value in the synapse on the basis of the error signal and the output value from the previous-stage neuron, and a selecting section for selecting a synapse whose weight value is to be updated by the learning control section when the learning control section is to update the weight values of a predetermined number of synapses in a predetermined order.",1999-01-05,"The title of the patent is neural network apparatus and learning method thereof and its abstract is a neural network apparatus includes a neural network including at least two neuron layers each having a plurality of neurons and at least one synapse layer having a plurality of synapses each arranged between the neuron layers, each synapse storing a weight value between the neurons and multiplying the weight value with an output value from each of the neurons in the previous-stage neuron layer to output a product to the next-stage neuron layer, a section for causing an error signal between an output from the neural network and a desired output to back-propagate from an output-side neuron layer to an input-side neuron layer of the neural network, a learning control section for updating the weight value in the synapse on the basis of the error signal and the output value from the previous-stage neuron, and a selecting section for selecting a synapse whose weight value is to be updated by the learning control section when the learning control section is to update the weight values of a predetermined number of synapses in a predetermined order. dated 1999-01-05"
5857321,controller with neural network for estimating gas turbine internal cycle parameters,"a gas turbine control system includes a controller that is coupled to actuator systems that govern operation of the gas turbine. the controller includes a processor for generating respective actuator control signals in correspondence with a plurality of turbine operating condition signals; the controller includes at least one neural network estimator that is trained to generate an estimated turbine operating condition signal. the neural network estimator typically has one or more hidden neuron layers that are coupled together in a feedforward structure, a recurrent neural network architecture. the estimated turbine operating condition signal generated by the neural network estimator typically, but not necessarily, represents a turbine internal cycle operating parameter for which the turbine has no corresponding operating parameter sensor.",1999-01-12,"The title of the patent is controller with neural network for estimating gas turbine internal cycle parameters and its abstract is a gas turbine control system includes a controller that is coupled to actuator systems that govern operation of the gas turbine. the controller includes a processor for generating respective actuator control signals in correspondence with a plurality of turbine operating condition signals; the controller includes at least one neural network estimator that is trained to generate an estimated turbine operating condition signal. the neural network estimator typically has one or more hidden neuron layers that are coupled together in a feedforward structure, a recurrent neural network architecture. the estimated turbine operating condition signal generated by the neural network estimator typically, but not necessarily, represents a turbine internal cycle operating parameter for which the turbine has no corresponding operating parameter sensor. dated 1999-01-12"
5859773,residual activation neural network,"a plant (72) is operable to receive control inputs c(t) and provide an output y(t). the plant (72) has associated therewith state variables s(t) that are not variable. a control network (74) is provided that accurately models the plant (72). the output of the control network (74) provides a predicted output which is combined with a desired output to generate an error. this error is back propagated through an inverse control network (76), which is the inverse of the control network (74) to generate a control error signal that is input to a distributed control system (73) to vary the control inputs to the plant (72) in order to change the output y(t) to meet the desired output. the control network (74) is comprised of a first network net 1 that is operable to store a representation of the dependency of the control variables on the state variables. the predicted result is subtracted from the actual state variable input and stored as a residual in a residual layer (102). the output of the residual layer (102) is input to a hidden layer (108) which also receives the control inputs to generate a predicted output in an output layer (106). during back propagation of error, the residual values in the residual layer (102) are latched and only the control inputs allowed to vary.",1999-01-12,"The title of the patent is residual activation neural network and its abstract is a plant (72) is operable to receive control inputs c(t) and provide an output y(t). the plant (72) has associated therewith state variables s(t) that are not variable. a control network (74) is provided that accurately models the plant (72). the output of the control network (74) provides a predicted output which is combined with a desired output to generate an error. this error is back propagated through an inverse control network (76), which is the inverse of the control network (74) to generate a control error signal that is input to a distributed control system (73) to vary the control inputs to the plant (72) in order to change the output y(t) to meet the desired output. the control network (74) is comprised of a first network net 1 that is operable to store a representation of the dependency of the control variables on the state variables. the predicted result is subtracted from the actual state variable input and stored as a residual in a residual layer (102). the output of the residual layer (102) is input to a hidden layer (108) which also receives the control inputs to generate a predicted output in an output layer (106). during back propagation of error, the residual values in the residual layer (102) are latched and only the control inputs allowed to vary. dated 1999-01-12"
5859925,classifying system having a single neural network architecture for multiple input representations,"a classification system is provided for combining multiple input representations by a single neural network architecture. in such a classification system having a single neural network architecture, classification channels corresponding to various input representations may be integrated through their own and shared hidden layers of the network to produce highly accurate classification. the classification system is particularly applicable to character classifying applications which use stroke and character image features as the main classification criteria, along with scalar features such as stroke count and aspect ratio features as secondary classification. the classification channels corresponding to the scalar features may be cross wired to the classification channels corresponding to the main input representations for further improving the accuracy of the classification output. because a single neural network architecture is used, only one, standard training technique is needed for this classification system, special data handling is minimized, and the training time can be reduced, while highly accurate classification is achieved.",1999-01-12,"The title of the patent is classifying system having a single neural network architecture for multiple input representations and its abstract is a classification system is provided for combining multiple input representations by a single neural network architecture. in such a classification system having a single neural network architecture, classification channels corresponding to various input representations may be integrated through their own and shared hidden layers of the network to produce highly accurate classification. the classification system is particularly applicable to character classifying applications which use stroke and character image features as the main classification criteria, along with scalar features such as stroke count and aspect ratio features as secondary classification. the classification channels corresponding to the scalar features may be cross wired to the classification channels corresponding to the main input representations for further improving the accuracy of the classification output. because a single neural network architecture is used, only one, standard training technique is needed for this classification system, special data handling is minimized, and the training time can be reduced, while highly accurate classification is achieved. dated 1999-01-12"
5859964,"system and method for performing real time data acquisition, process modeling and fault detection of wafer fabrication processes","a system and method for detecting faults in wafer fabrication process tools by acquiring real-time process parameter signal data samples used to model the process performed by the process tool. the system includes a computer system including a daq device, which acquires the data samples, and a fault detector program which employs a process model program to analyze the data samples for the purpose of detecting faults. the model uses data samples in a reference database acquired from previous known good runs of the process tool. the fault detector notifies a process tool operator of any faults which occur thus potentially avoiding wafer scrap and potentially improving mean time between failures. the fault detector also receives notification of the occurrence of process events from the process tool, such as the start or end of processing a wafer, which the fault detector uses to start and stop the data acquisition, respectively. the fault detector also receives notification of the occurrence of a new process recipe and uses the recipe information to select the appropriate model for modeling the data samples. the fault detector employs a standard data exchange interface, such as dde, between the fault detector and the model, thus facilitating modular selection of models best suited to the particular fabrication process being modeled. embodiments are contemplated which use a upm model, a pca model, or a neural network model.",1999-01-12,"The title of the patent is system and method for performing real time data acquisition, process modeling and fault detection of wafer fabrication processes and its abstract is a system and method for detecting faults in wafer fabrication process tools by acquiring real-time process parameter signal data samples used to model the process performed by the process tool. the system includes a computer system including a daq device, which acquires the data samples, and a fault detector program which employs a process model program to analyze the data samples for the purpose of detecting faults. the model uses data samples in a reference database acquired from previous known good runs of the process tool. the fault detector notifies a process tool operator of any faults which occur thus potentially avoiding wafer scrap and potentially improving mean time between failures. the fault detector also receives notification of the occurrence of process events from the process tool, such as the start or end of processing a wafer, which the fault detector uses to start and stop the data acquisition, respectively. the fault detector also receives notification of the occurrence of a new process recipe and uses the recipe information to select the appropriate model for modeling the data samples. the fault detector employs a standard data exchange interface, such as dde, between the fault detector and the model, thus facilitating modular selection of models best suited to the particular fabrication process being modeled. embodiments are contemplated which use a upm model, a pca model, or a neural network model. dated 1999-01-12"
5860285,system for monitoring outdoor heat exchanger coil,a system for monitoring an outdoor heat exchange coil of a heating or cooling system includes a neural network for computing the status of the coil. the neural network is trained during a development mode to learn certain characteristics of the heating or cooling system that will allow it to accurately compute the status of the coil. the thus trained neural network timely computes the status of the outdoor heat exchange coil during a run time mode of operation. information as to the status of the coil is made available for assessment during the run time mode of operation.,1999-01-19,The title of the patent is system for monitoring outdoor heat exchanger coil and its abstract is a system for monitoring an outdoor heat exchange coil of a heating or cooling system includes a neural network for computing the status of the coil. the neural network is trained during a development mode to learn certain characteristics of the heating or cooling system that will allow it to accurately compute the status of the coil. the thus trained neural network timely computes the status of the outdoor heat exchange coil during a run time mode of operation. information as to the status of the coil is made available for assessment during the run time mode of operation. dated 1999-01-19
5860286,system monitoring refrigeration charge,a refrigerant monitoring system for a heating or cooling system includes a neural network that is used to compute the refrigerant charge in at least one refrigeration circuit of the system. the neural network is trained to learn certain characteristics of the heating or cooling system during a development mode of operation. the thus trained neural network timely computes refrigerant charge during a run time mode of operation. information as to the computed refrigerant charge being at variance with the nominal amount of refrigerant charge for at least one refrigeration circuit is made available for assessment during the run time mode of operation.,1999-01-19,The title of the patent is system monitoring refrigeration charge and its abstract is a refrigerant monitoring system for a heating or cooling system includes a neural network that is used to compute the refrigerant charge in at least one refrigeration circuit of the system. the neural network is trained to learn certain characteristics of the heating or cooling system during a development mode of operation. the thus trained neural network timely computes refrigerant charge during a run time mode of operation. information as to the computed refrigerant charge being at variance with the nominal amount of refrigerant charge for at least one refrigeration circuit is made available for assessment during the run time mode of operation. dated 1999-01-19
5861936,regulating focus in accordance with relationship of features of a person's eyes,"the focus of variable-focus lenses in eyeglasses worn by a person is regulated in accordance with an indication of a spatial relationship of features of the person's eyes to provide an at least partially in-focus image of a feature at which the person is gazing.. the lenses are nematic liquid crystal display devices. light reflected off of each of the person's eyes is detected to provide data indicating the respective dispositions of the person's eyes; and the combined data for both eyes is processed by a neural network in a computer to provide focus-control signals for setting the focus of each lens. at least partially in-focus images are also provided of features at which a person in gazing (a) through goggles of a three-dimensional virtual reality system, (b) on a monitor that perceptually simultaneously displays different features with different degrees of focus as though they were at different distances from a person viewing the monitor, (c) on a monitor displaying remotely generated images and (d) through a view finder of a camera.",1999-01-19,"The title of the patent is regulating focus in accordance with relationship of features of a person's eyes and its abstract is the focus of variable-focus lenses in eyeglasses worn by a person is regulated in accordance with an indication of a spatial relationship of features of the person's eyes to provide an at least partially in-focus image of a feature at which the person is gazing.. the lenses are nematic liquid crystal display devices. light reflected off of each of the person's eyes is detected to provide data indicating the respective dispositions of the person's eyes; and the combined data for both eyes is processed by a neural network in a computer to provide focus-control signals for setting the focus of each lens. at least partially in-focus images are also provided of features at which a person in gazing (a) through goggles of a three-dimensional virtual reality system, (b) on a monitor that perceptually simultaneously displays different features with different degrees of focus as though they were at different distances from a person viewing the monitor, (c) on a monitor displaying remotely generated images and (d) through a view finder of a camera. dated 1999-01-19"
5862304,method for predicting the future occurrence of clinically occult or non-existent medical conditions,"a method is presented for evaluating data to predict the future occurrence of a medical condition that is presently clinically occult or which has not yet occurred. specifically, the method uses a neural network to analyze and interpret dna flow cytometric histograms. a first set of dna histograms taken from tumors from patients having known relapse rates are used to train the neural network, an then the trained network is applied to predict the relapse rates of patients using dna histograms of tumors from those patients. prognosis according to this method can be performed using only diploid histograms, using only aneuploid histograms, or using a combination of diploid and aneuploid histograms.",1999-01-19,"The title of the patent is method for predicting the future occurrence of clinically occult or non-existent medical conditions and its abstract is a method is presented for evaluating data to predict the future occurrence of a medical condition that is presently clinically occult or which has not yet occurred. specifically, the method uses a neural network to analyze and interpret dna flow cytometric histograms. a first set of dna histograms taken from tumors from patients having known relapse rates are used to train the neural network, an then the trained network is applied to predict the relapse rates of patients using dna histograms of tumors from those patients. prognosis according to this method can be performed using only diploid histograms, using only aneuploid histograms, or using a combination of diploid and aneuploid histograms. dated 1999-01-19"
5862513,systems and methods for forward modeling of well logging tool responses,"a method for producing synthetic tool responses for a well logging tool for an earth formation, the method including, in one aspect, generating wellbore logging data for a particular part of an earth formation with a wellbore logging system with a wellbore logging tool, the earth formation having at least one layer, producing an input earth model of the particular part of the earth formation based on the wellbore logging data, inputting the input earth model to a trained artificial neural network, e.g. resident in a computer, the computer with the trained artificial neural network processing the input earth model and producing synthetic tool responses for the wellbore logging tool for one point or for a plurality of points in the earth formation.",1999-01-19,"The title of the patent is systems and methods for forward modeling of well logging tool responses and its abstract is a method for producing synthetic tool responses for a well logging tool for an earth formation, the method including, in one aspect, generating wellbore logging data for a particular part of an earth formation with a wellbore logging system with a wellbore logging tool, the earth formation having at least one layer, producing an input earth model of the particular part of the earth formation based on the wellbore logging data, inputting the input earth model to a trained artificial neural network, e.g. resident in a computer, the computer with the trained artificial neural network processing the input earth model and producing synthetic tool responses for the wellbore logging tool for one point or for a plurality of points in the earth formation. dated 1999-01-19"
5864255,four quadrant square law analog multiplier using floating gate mos transitions,"a four quadrant multiplier using multiple input floating-gate mos transistors is provided. it is based on the square law characteristics of the mos transistor and can be realised with only four floating gate mos transistors, two resistors and a current source. the four floating gate transistors are configured with their sources connected in common and biased by a single current source. output is taken between two common drain connections. each transistor has three control gates with two being provided for selected ones of the two input signals and one for a biasing signal (optional). input signals can be connected to the control gates in either a differential or single ended configuration. in one application, a programmable synaptic cell for neural networks employs the multi-input floating-gate mos four-quadrant analog multiplier. varying of the neural weight connection strength of each synaptic cell is achieved by two possible methods. one method involves programming charges into or out of the primary floating-gate of the mfmos devices associated with the multiplier. the other method is to configure the third input gate of each mfmos device of the multiplier as another (secondary) floating-gate structure whereby charge can be programmed into or out of this secondary floating-gate structure and its coupling area to the primary floating-gate would determine the neural weight. the differential output current is proportional to the product of the input signal and the programmed charge difference. in a natural extension an array of individually programmable synaptic cells form a neural network.",1999-01-26,"The title of the patent is four quadrant square law analog multiplier using floating gate mos transitions and its abstract is a four quadrant multiplier using multiple input floating-gate mos transistors is provided. it is based on the square law characteristics of the mos transistor and can be realised with only four floating gate mos transistors, two resistors and a current source. the four floating gate transistors are configured with their sources connected in common and biased by a single current source. output is taken between two common drain connections. each transistor has three control gates with two being provided for selected ones of the two input signals and one for a biasing signal (optional). input signals can be connected to the control gates in either a differential or single ended configuration. in one application, a programmable synaptic cell for neural networks employs the multi-input floating-gate mos four-quadrant analog multiplier. varying of the neural weight connection strength of each synaptic cell is achieved by two possible methods. one method involves programming charges into or out of the primary floating-gate of the mfmos devices associated with the multiplier. the other method is to configure the third input gate of each mfmos device of the multiplier as another (secondary) floating-gate structure whereby charge can be programmed into or out of this secondary floating-gate structure and its coupling area to the primary floating-gate would determine the neural weight. the differential output current is proportional to the product of the input signal and the programmed charge difference. in a natural extension an array of individually programmable synaptic cells form a neural network. dated 1999-01-26"
5864693,"movement control method and chaotic information processing unit using chaotic neural network, and group movement control method","this invention discloses a method of realizing a variety of movements which are controlled while maintaining their complexity by generating complex movements, which appear irregular at a glance, by the deterministic method. a neural network which connects a plurality of neurons to each other is set. the neurons perform calculations that generate chaotic oscillation, and are set in correspondence with the moving direction of an object. the movement vector of the object is calculated from time-series data obtained by the calculations.",1999-01-26,"The title of the patent is movement control method and chaotic information processing unit using chaotic neural network, and group movement control method and its abstract is this invention discloses a method of realizing a variety of movements which are controlled while maintaining their complexity by generating complex movements, which appear irregular at a glance, by the deterministic method. a neural network which connects a plurality of neurons to each other is set. the neurons perform calculations that generate chaotic oscillation, and are set in correspondence with the moving direction of an object. the movement vector of the object is calculated from time-series data obtained by the calculations. dated 1999-01-26"
5864803,signal processing and training by a neural network for phoneme recognition,"an automated speech recognition system converts a speech signal into a compact, coded representation that correlates to a speech phoneme set. a number of different neural network pattern matching schemes may be used to perform the necessary speech coding. an integrated user interface guides a user unfamiliar with the details of speech recognition or neural networks to quickly develop and test a neural network for phoneme recognition. to train the neural network, digitized voice data containing known phonemes that the user wants the neural network to ultimately recognize are processed by the integrated user interface. the digitized speech is segmented into phonemes with each segment being labelled with a corresponding phoneme code. based on a user selected transformation method and transformation parameters, each segment is transformed into a series of multiple dimension vectors representative of the speech characteristics of that segment. these vectors are iteratively presented to a neural network to train/adapt that neural network to consistently distinguish and recognize these vectors and assign an appropriate phoneme code to each vector. simultaneous display of the digitized speech, segments, vector sets, and a representation of the trained neural network assist the user in visually confirming the acceptability of the phoneme training set. a user may also selectively audibly confirm the acceptability of the digitization scheme, the segments, and the transform vectors so that satisfactory training data are presented to the neural network. if the user finds a particular step or parameter produces an unacceptable result, the user may modify one or more of the parameters and verify whether the modification effected an improvement in performance. the trained neural network is also automatically tested by presenting a test speech signal to the integrated user interface and observing both audibly and visually automatic segmentation of the speech, transformation into multidimensional vectors, and the resulting neural network assigned phoneme codes. a method of decoding such phoneme codes using the neural network is also disclosed.",1999-01-26,"The title of the patent is signal processing and training by a neural network for phoneme recognition and its abstract is an automated speech recognition system converts a speech signal into a compact, coded representation that correlates to a speech phoneme set. a number of different neural network pattern matching schemes may be used to perform the necessary speech coding. an integrated user interface guides a user unfamiliar with the details of speech recognition or neural networks to quickly develop and test a neural network for phoneme recognition. to train the neural network, digitized voice data containing known phonemes that the user wants the neural network to ultimately recognize are processed by the integrated user interface. the digitized speech is segmented into phonemes with each segment being labelled with a corresponding phoneme code. based on a user selected transformation method and transformation parameters, each segment is transformed into a series of multiple dimension vectors representative of the speech characteristics of that segment. these vectors are iteratively presented to a neural network to train/adapt that neural network to consistently distinguish and recognize these vectors and assign an appropriate phoneme code to each vector. simultaneous display of the digitized speech, segments, vector sets, and a representation of the trained neural network assist the user in visually confirming the acceptability of the phoneme training set. a user may also selectively audibly confirm the acceptability of the digitization scheme, the segments, and the transform vectors so that satisfactory training data are presented to the neural network. if the user finds a particular step or parameter produces an unacceptable result, the user may modify one or more of the parameters and verify whether the modification effected an improvement in performance. the trained neural network is also automatically tested by presenting a test speech signal to the integrated user interface and observing both audibly and visually automatic segmentation of the speech, transformation into multidimensional vectors, and the resulting neural network assigned phoneme codes. a method of decoding such phoneme codes using the neural network is also disclosed. dated 1999-01-26"
5864807,method and apparatus for training a speaker recognition system,a method and apparatus for training a system to assess the identity of a person through the audio characteristics of their voice. the system inserts an audio input (10) into an a/d converter (20) for processing in a digital signal processor (30). the system then applies neural network type processing by using a polynomial pattern classifier (60) for training the speaker recognition system.,1999-01-26,The title of the patent is method and apparatus for training a speaker recognition system and its abstract is a method and apparatus for training a system to assess the identity of a person through the audio characteristics of their voice. the system inserts an audio input (10) into an a/d converter (20) for processing in a digital signal processor (30). the system then applies neural network type processing by using a polynomial pattern classifier (60) for training the speaker recognition system. dated 1999-01-26
5864832,method and apparatus for selecting higher order terms for a holographic neural network,"a method and apparatus for selecting terms, preferably higher order terms, for use in a stimulus vector of a holographic neural network. the method includes an evolutionary type algorithm in which stimulus vector arrangements are preferably modelled as chromosomes containing genes that each uniquely identify one of a plurality of terms in a corresponding stimulus vector. techniques for introducing variation and for selecting fit chromosome/stimulus vector arrangements are also disclosed, as is an apparatus for executing the term selecting method.",1999-01-26,"The title of the patent is method and apparatus for selecting higher order terms for a holographic neural network and its abstract is a method and apparatus for selecting terms, preferably higher order terms, for use in a stimulus vector of a holographic neural network. the method includes an evolutionary type algorithm in which stimulus vector arrangements are preferably modelled as chromosomes containing genes that each uniquely identify one of a plurality of terms in a corresponding stimulus vector. techniques for introducing variation and for selecting fit chromosome/stimulus vector arrangements are also disclosed, as is an apparatus for executing the term selecting method. dated 1999-01-26"
5864834,method and apparatus for estimating a special reflectance distribution or a spectral transmittance distribution using illuminant-independent characteristic parameters,"a neural network is made to undergo learning such that at least three characteristic parameters, which correspond to an inputted set of color information values when the set of color information values, including at least three color information values, is inputted and which are obtained by multivariate analysis of the spectral reflectance distribution or the spectral transmittance distribution, are outputted. a subject set of color information values is transformed into the characteristic parameters by using the neural network which has completed learning, and a spectral reflectance distribution or a spectral transmittance distribution is estimated by linear polynomial approximation using the transformed characteristic parameters, eigenvectors obtained by the multivariate analysis, and a mean vector of the spectral reflectance distribution or the spectral transmittance distribution.",1999-01-26,"The title of the patent is method and apparatus for estimating a special reflectance distribution or a spectral transmittance distribution using illuminant-independent characteristic parameters and its abstract is a neural network is made to undergo learning such that at least three characteristic parameters, which correspond to an inputted set of color information values when the set of color information values, including at least three color information values, is inputted and which are obtained by multivariate analysis of the spectral reflectance distribution or the spectral transmittance distribution, are outputted. a subject set of color information values is transformed into the characteristic parameters by using the neural network which has completed learning, and a spectral reflectance distribution or a spectral transmittance distribution is estimated by linear polynomial approximation using the transformed characteristic parameters, eigenvectors obtained by the multivariate analysis, and a mean vector of the spectral reflectance distribution or the spectral transmittance distribution. dated 1999-01-26"
5864836,optically programmable optoelectronic cellular neural network,"an optoelectronic cellular neural network, which can be programmed optically, in amorphous or polycrystalline silicon, which makes up a monolithic image processing system with optical input and output which can be programmed by optical signals. its layered structure for both the upper and lower surfaces are occupied, without discontinuity, by photosensitive and photoemissive devices. the network basically consists of an input photosensitive layer, a processing layer, a photosensitive control layer and also an optical control mask.",1999-01-26,"The title of the patent is optically programmable optoelectronic cellular neural network and its abstract is an optoelectronic cellular neural network, which can be programmed optically, in amorphous or polycrystalline silicon, which makes up a monolithic image processing system with optical input and output which can be programmed by optical signals. its layered structure for both the upper and lower surfaces are occupied, without discontinuity, by photosensitive and photoemissive devices. the network basically consists of an input photosensitive layer, a processing layer, a photosensitive control layer and also an optical control mask. dated 1999-01-26"
5867813,method and apparatus for automatically and reproducibly rating the transmission quality of a speech transmission system,"a recorded voice test signal is transmitted from a transmitting end of a mobile communication system, and in a test unit at the receiving end a frame generator is synchronized with the received signal. each frame of the signal is evaluated in computing circuits on the outputs of the frame generator, which calculate characteristic values of each frame which are then subtracted from calculated characteristic reference values of the frames stored in a memory. the differences between these characteristic values are fed to a neural network which classifies the quality of the difference signals as good, medium and bad, and a defuzzyfication logic circuit further refines the quality classification output, whereby the transmission quality throughout the mobile communication system can be reprodicbly rated for a system quality rating.",1999-02-02,"The title of the patent is method and apparatus for automatically and reproducibly rating the transmission quality of a speech transmission system and its abstract is a recorded voice test signal is transmitted from a transmitting end of a mobile communication system, and in a test unit at the receiving end a frame generator is synchronized with the received signal. each frame of the signal is evaluated in computing circuits on the outputs of the frame generator, which calculate characteristic values of each frame which are then subtracted from calculated characteristic reference values of the frames stored in a memory. the differences between these characteristic values are fed to a neural network which classifies the quality of the difference signals as good, medium and bad, and a defuzzyfication logic circuit further refines the quality classification output, whereby the transmission quality throughout the mobile communication system can be reprodicbly rated for a system quality rating. dated 1999-02-02"
5867816,operator interactions for developing phoneme recognition by neural networks,"an automated speech recognition system converts a speech signal into a compact, coded representation that correlates to a speech phoneme set. a number of different neural network pattern matching schemes may be used to perform the necessary speech coding. an integrated user interface guides a user unfamiliar with the details of speech recognition or neural networks to quickly develop and test a neural network for phoneme recognition. to train the neural network, digitized voice data containing known phonemes that the user wants the neural network to ultimately recognize are processed by the integrated user interface. the digitized speech is segmented into phonemes with each segment being labelled with a corresponding phoneme code. based on a user selected transformation method and transformation parameters, each segment is transformed into a series of multiple dimension vectors representative of the speech characteristics of that segment. these vectors are iteratively presented to a neural network to train/adapt that neural network to consistently distinguish and recognize these vectors and assign an appropriate phoneme code to each vector. simultaneous display of the digitized speech, segments, vector sets, and a representation of the trained neural network assist the user in visually confirming the acceptability of the phoneme training set. a user may also selectively audibly confirm the acceptability of the digitization scheme, the segments, and the transform vectors so that satisfactory training data are presented to the neural network. if the user finds a particular step or parameter produces an unacceptable result, the user may modify one or more of the parameters and verify whether the modification effected an improvement in performance. the trained neural network is also automatically tested by presenting a test speech signal to the integrated user interface and observing both audibly and visually automatic segmentation of the speech, transformation into multidimensional vectors, and the resulting neural network assigned phoneme codes. a method of decoding such phoneme codes using the neural network is also disclosed.",1999-02-02,"The title of the patent is operator interactions for developing phoneme recognition by neural networks and its abstract is an automated speech recognition system converts a speech signal into a compact, coded representation that correlates to a speech phoneme set. a number of different neural network pattern matching schemes may be used to perform the necessary speech coding. an integrated user interface guides a user unfamiliar with the details of speech recognition or neural networks to quickly develop and test a neural network for phoneme recognition. to train the neural network, digitized voice data containing known phonemes that the user wants the neural network to ultimately recognize are processed by the integrated user interface. the digitized speech is segmented into phonemes with each segment being labelled with a corresponding phoneme code. based on a user selected transformation method and transformation parameters, each segment is transformed into a series of multiple dimension vectors representative of the speech characteristics of that segment. these vectors are iteratively presented to a neural network to train/adapt that neural network to consistently distinguish and recognize these vectors and assign an appropriate phoneme code to each vector. simultaneous display of the digitized speech, segments, vector sets, and a representation of the trained neural network assist the user in visually confirming the acceptability of the phoneme training set. a user may also selectively audibly confirm the acceptability of the digitization scheme, the segments, and the transform vectors so that satisfactory training data are presented to the neural network. if the user finds a particular step or parameter produces an unacceptable result, the user may modify one or more of the parameters and verify whether the modification effected an improvement in performance. the trained neural network is also automatically tested by presenting a test speech signal to the integrated user interface and observing both audibly and visually automatic segmentation of the speech, transformation into multidimensional vectors, and the resulting neural network assigned phoneme codes. a method of decoding such phoneme codes using the neural network is also disclosed. dated 1999-02-02"
5870314,method for randomly accessing stored video and a field inspection system employing the same,"the present invention relates to a field inspection system for compressing video signals received from a field inspection video camera into compressed video data and for burning the compressed video data on a compact disc along with an electronic logsheet. the electronic logsheet includes a listing of suspected defects or anomalies and associated pointers to reference frames in the compressed video data. the electronic logsheet may be displayed and an operator may access a portion of the field inspection video showing a listed defect by clicking a mouse button when a pointer icon is posited on the listed defect. to perform this task, the present invention utilizes a technique for randomly accessing the compressed video data in which reference frames included therein are used as access points to the video footage. the field inspection system of the present invention may also utilize a neural network and an artificial intelligence system to detect, identify, and log defects without human intervention. the artificial intelligence system may also automatically generate a report making recommendations for repairing each detected defect.",1999-02-09,"The title of the patent is method for randomly accessing stored video and a field inspection system employing the same and its abstract is the present invention relates to a field inspection system for compressing video signals received from a field inspection video camera into compressed video data and for burning the compressed video data on a compact disc along with an electronic logsheet. the electronic logsheet includes a listing of suspected defects or anomalies and associated pointers to reference frames in the compressed video data. the electronic logsheet may be displayed and an operator may access a portion of the field inspection video showing a listed defect by clicking a mouse button when a pointer icon is posited on the listed defect. to perform this task, the present invention utilizes a technique for randomly accessing the compressed video data in which reference frames included therein are used as access points to the video footage. the field inspection system of the present invention may also utilize a neural network and an artificial intelligence system to detect, identify, and log defects without human intervention. the artificial intelligence system may also automatically generate a report making recommendations for repairing each detected defect. dated 1999-02-09"
5870399,telecommunications system comprising a communications protocol analyzer which comprises a neural network that obviates the need for a database,"telecommunications systems embody means for establishing communications links between entities within that telecommunications system. these links are usually established following an exchange of digital messages in accordance with a communications protocol. communications protocols are often sophisticated, involving a sequence of complex message exchanges. known protocol analysers for testing communications protocols involve the use of data bases, which are pre-programmed with all possible combinations of messages sequences which correspond with the protocol to be tested. this is time consuming, expensive and inflexible. by using a neural network within the protocol analyser, a sequence of messages exchanged in accordance with a correctly operating communications protocol may be learned, obviating the need for a data base.",1999-02-09,"The title of the patent is telecommunications system comprising a communications protocol analyzer which comprises a neural network that obviates the need for a database and its abstract is telecommunications systems embody means for establishing communications links between entities within that telecommunications system. these links are usually established following an exchange of digital messages in accordance with a communications protocol. communications protocols are often sophisticated, involving a sequence of complex message exchanges. known protocol analysers for testing communications protocols involve the use of data bases, which are pre-programmed with all possible combinations of messages sequences which correspond with the protocol to be tested. this is time consuming, expensive and inflexible. by using a neural network within the protocol analyser, a sequence of messages exchanged in accordance with a correctly operating communications protocol may be learned, obviating the need for a data base. dated 1999-02-09"
5870493,top down preprocessor for a machine vision system,""" an image recognition and classification system includes a preprocessor in which a """"top-down"""" method is used to extract features from an image; an associative learning neural network system, which groups the features into patterns and classifies the patterns: and a feedback mechanism which improves system performance by tuning preprocessor scale, feature detection, and feature selection. """,1999-02-09,"The title of the patent is top down preprocessor for a machine vision system and its abstract is "" an image recognition and classification system includes a preprocessor in which a """"top-down"""" method is used to extract features from an image; an associative learning neural network system, which groups the features into patterns and classifies the patterns: and a feedback mechanism which improves system performance by tuning preprocessor scale, feature detection, and feature selection. "" dated 1999-02-09"
5870721,system and method for real time loan approval,"a method and apparatus for closed loop, automatic processing a loan, including completion of the application, underwriting, and transferring funds, includes use of a programmed computer to interface with an applicant, obtain the information needed to process the loan, determine whether to approve the loan, and effect electronic fund transfers to the applicant's deposit account and arrange for automatic withdrawals to repay the loan. information is received from the applicant preferably by using voice recognition technology but alternatively by entering the alpha-numeric information using a personal computer keyboard or using the buttons on a telephone. the loan approval determination is made using a neural network with input obtained in part from the applicant and in part from databases accessed by the computer, such as a credit bureau, to obtain a credit report. the loan agreement is transmitted by facsimile to and from the applicant when the applicant has access to a facsimile machine or datafile to be printed or to an agent who delivers the agreement to the applicant when the applicant does not have access. in a preferred embodiment, the applicant accesses the computer from a kiosk where the complete transaction can take place as the applicant waits.",1999-02-09,"The title of the patent is system and method for real time loan approval and its abstract is a method and apparatus for closed loop, automatic processing a loan, including completion of the application, underwriting, and transferring funds, includes use of a programmed computer to interface with an applicant, obtain the information needed to process the loan, determine whether to approve the loan, and effect electronic fund transfers to the applicant's deposit account and arrange for automatic withdrawals to repay the loan. information is received from the applicant preferably by using voice recognition technology but alternatively by entering the alpha-numeric information using a personal computer keyboard or using the buttons on a telephone. the loan approval determination is made using a neural network with input obtained in part from the applicant and in part from databases accessed by the computer, such as a credit bureau, to obtain a credit report. the loan agreement is transmitted by facsimile to and from the applicant when the applicant has access to a facsimile machine or datafile to be printed or to an agent who delivers the agreement to the applicant when the applicant does not have access. in a preferred embodiment, the applicant accesses the computer from a kiosk where the complete transaction can take place as the applicant waits. dated 1999-02-09"
5870728,learning procedure for multi-level neural network,"a reiterative learning procedure with training and test processes for a binary supervised neural network includes at least an error signal generator for weighting factor updating in the training process, which generates an error signal that is perturbed in polarity and amplitude in the difference derived by subtracting an output unit signal from corresponding binary teacher signal and then generates the difference as an error signal after a maximum absolute value of differences among erroneous binary output signals has become smaller than a threshold once. a training signal memory stores a set of training signals and adds test signals providing erroneous binary output signals that are transferred from a test signal memory in the test process to the set of training input signals as incremental training input signals. an affordable signal memory stores input signals with sufficiently large margins providing correct binary output signals that are transferred from the training signal memory in the training process and the test signal memory in the test process. the reiterative learning procedure, with minimum necessary training and test input signals and control of the error perturbation in the training process, can provide a binary space to obtain a desired binary output, and also realizes an extremely high generalization ability.",1999-02-09,"The title of the patent is learning procedure for multi-level neural network and its abstract is a reiterative learning procedure with training and test processes for a binary supervised neural network includes at least an error signal generator for weighting factor updating in the training process, which generates an error signal that is perturbed in polarity and amplitude in the difference derived by subtracting an output unit signal from corresponding binary teacher signal and then generates the difference as an error signal after a maximum absolute value of differences among erroneous binary output signals has become smaller than a threshold once. a training signal memory stores a set of training signals and adds test signals providing erroneous binary output signals that are transferred from a test signal memory in the test process to the set of training input signals as incremental training input signals. an affordable signal memory stores input signals with sufficiently large margins providing correct binary output signals that are transferred from the training signal memory in the training process and the test signal memory in the test process. the reiterative learning procedure, with minimum necessary training and test input signals and control of the error perturbation in the training process, can provide a binary space to obtain a desired binary output, and also realizes an extremely high generalization ability. dated 1999-02-09"
5870729,self-organizing neural network for pattern classification,"a neural network includes a plurality of input nodes for receiving the respective elements of the input vector. a copy of all of the elements of the input vector is sent to the next level of nodes in the neural network denoted as intermediate nodes. the intermediate nodes each encode a separate template pattern. they compare the actual input pattern with the template and generate a signal indicative of the difference between the input pattern and the template pattern. each of the templates encoded in the intermediate nodes has a class associated with it. the difference calculated by the intermediate nodes is passed to an output node for each of the intermediate nodes at a given class. the output node then selects the minimum difference amongst the values sent from the intermediate nodes. this lowest difference for the class represented by the output node is then forwarded to a selector. the selector receives such values from each of the output nodes of all of the classes and then selects that to output value which is a minimum difference. the selector in turn, generates a signal indicative of the class of the intermediate node that sent the smallest difference value.",1999-02-09,"The title of the patent is self-organizing neural network for pattern classification and its abstract is a neural network includes a plurality of input nodes for receiving the respective elements of the input vector. a copy of all of the elements of the input vector is sent to the next level of nodes in the neural network denoted as intermediate nodes. the intermediate nodes each encode a separate template pattern. they compare the actual input pattern with the template and generate a signal indicative of the difference between the input pattern and the template pattern. each of the templates encoded in the intermediate nodes has a class associated with it. the difference calculated by the intermediate nodes is passed to an output node for each of the intermediate nodes at a given class. the output node then selects the minimum difference amongst the values sent from the intermediate nodes. this lowest difference for the class represented by the output node is then forwarded to a selector. the selector receives such values from each of the output nodes of all of the classes and then selects that to output value which is a minimum difference. the selector in turn, generates a signal indicative of the class of the intermediate node that sent the smallest difference value. dated 1999-02-09"
5872864,image processing apparatus for performing adaptive data processing in accordance with kind of image,"an image processing apparatus includes an image input section, a color image/monochrome image converting section for performing image area division, a binarization circuit for binarizing a converted monochrome image, a reducing section for reducing a binary image, a boundary extracting section for extracting the boundaries between the areas of constituent elements constituting an input image, e.g., a binary image and a continuous gradation image, and a kind-of-image determining section for determining the kinds of images in partial areas defined by the extracted boundaries. for example, the image processing apparatus further includes a data compressing section. in the image processing apparatus, pre-processing is performed to efficiently performing pattern determination in consideration of localization of frequencies of occurrence of edges and black pixel patterns of an image, and kind-of-image determination is performed by a neural network on the basis of the data obtained by pre-processing, thus performing suitable processing such as data compression in accordance with the determination result.",1999-02-16,"The title of the patent is image processing apparatus for performing adaptive data processing in accordance with kind of image and its abstract is an image processing apparatus includes an image input section, a color image/monochrome image converting section for performing image area division, a binarization circuit for binarizing a converted monochrome image, a reducing section for reducing a binary image, a boundary extracting section for extracting the boundaries between the areas of constituent elements constituting an input image, e.g., a binary image and a continuous gradation image, and a kind-of-image determining section for determining the kinds of images in partial areas defined by the extracted boundaries. for example, the image processing apparatus further includes a data compressing section. in the image processing apparatus, pre-processing is performed to efficiently performing pattern determination in consideration of localization of frequencies of occurrence of edges and black pixel patterns of an image, and kind-of-image determination is performed by a neural network on the basis of the data obtained by pre-processing, thus performing suitable processing such as data compression in accordance with the determination result. dated 1999-02-16"
5875284,neuro-fuzzy-integrated data processing system,"the present invention relates to a data processing system in a hierarchical network configuration for executing applicable data processes in a comprehensible and executable form. the present invention allows data processing capabilities to be established with high precision in a short time based on a fuzzy-neuro-integrated concept. a fuzzy model is generated by a data processing system in the form of membership functions and fuzzy rules as technical information relating to a control target. according to this fuzzy model, a weight value of the connection between neurons is set and a pre-wired neural network is established. then, the data of the control target are learned by the neural network. the connection state and a weight value of the neural network after the learning enable tuning of the fuzzy model.",1999-02-23,"The title of the patent is neuro-fuzzy-integrated data processing system and its abstract is the present invention relates to a data processing system in a hierarchical network configuration for executing applicable data processes in a comprehensible and executable form. the present invention allows data processing capabilities to be established with high precision in a short time based on a fuzzy-neuro-integrated concept. a fuzzy model is generated by a data processing system in the form of membership functions and fuzzy rules as technical information relating to a control target. according to this fuzzy model, a weight value of the connection between neurons is set and a pre-wired neural network is established. then, the data of the control target are learned by the neural network. the connection state and a weight value of the neural network after the learning enable tuning of the fuzzy model. dated 1999-02-23"
5875347,neural network processing system using semiconductor memories,"herein disclosed is a data processing system having a memory packaged therein for realizing a largescale and high-speed parallel distributed processing and, especially, a data processing system for the neural network processing. the neural network processing system according to the present invention comprises: a memory circuit for storing neuron output values, connection weights, the desired values of outputs, and data necessary for learning; an input/output circuit for writing or reading data in or out of said memory circuit; a processing circuit for performing a processing for determining the neuron outputs such as the product, sum and nonlinear conversion of the data stored in said memory circuit, a comparison of the output value and its desired value, and a processing necessary for learning; and a control circuit for controlling the operations of said memory circuit, said input/output circuit and said processing circuit. the processing circuit is constructed to include at least one of an adder, a multiplier, a nonlinear transfer function circuit and a comparator so that at least a portion of the processing necessary for determining the neutron output values such as the product or sum may be accomplished in parallel. moreover, these circuits are shared among a plurality of neutrons and are operated in a time sharing manner to determine the plural neuron output values. still moreover, the aforementioned comparator compares the neuron output value determined and the desired value of the output in parallel.",1999-02-23,"The title of the patent is neural network processing system using semiconductor memories and its abstract is herein disclosed is a data processing system having a memory packaged therein for realizing a largescale and high-speed parallel distributed processing and, especially, a data processing system for the neural network processing. the neural network processing system according to the present invention comprises: a memory circuit for storing neuron output values, connection weights, the desired values of outputs, and data necessary for learning; an input/output circuit for writing or reading data in or out of said memory circuit; a processing circuit for performing a processing for determining the neuron outputs such as the product, sum and nonlinear conversion of the data stored in said memory circuit, a comparison of the output value and its desired value, and a processing necessary for learning; and a control circuit for controlling the operations of said memory circuit, said input/output circuit and said processing circuit. the processing circuit is constructed to include at least one of an adder, a multiplier, a nonlinear transfer function circuit and a comparator so that at least a portion of the processing necessary for determining the neutron output values such as the product or sum may be accomplished in parallel. moreover, these circuits are shared among a plurality of neutrons and are operated in a time sharing manner to determine the plural neuron output values. still moreover, the aforementioned comparator compares the neuron output value determined and the desired value of the output in parallel. dated 1999-02-23"
5875439,nonrecurrent binary code recognizer,"a nonrecurrent version of the neural network binary code recognizer is disclosed. this nonrecurrent binary code recognizer, which decodes an input vector of n analog components into a decoded binary word of n bits, comprises an analog-to-digital converter, an inverter circuit, a digital summing circuit and a comparator circuit.",1999-02-23,"The title of the patent is nonrecurrent binary code recognizer and its abstract is a nonrecurrent version of the neural network binary code recognizer is disclosed. this nonrecurrent binary code recognizer, which decodes an input vector of n analog components into a decoded binary word of n bits, comprises an analog-to-digital converter, an inverter circuit, a digital summing circuit and a comparator circuit. dated 1999-02-23"
5877954,hybrid linear-neural network process control,"a hybrid analyzer having a data derived primary analyzer and an error correction analyzer connected in parallel is disclosed. the primary analyzer, preferably a data derived linear model such as a partial least squares model, is trained using training data to generate major predictions of defined output variables. the error correction analyzer, preferably a neural network model is trained to capture the residuals between the primary analyzer outputs and the target process variables. the residuals generated by the error correction analyzer is summed with the output of the primary analyzer to compensate for the error residuals of the primary analyzer to arrive at a more accurate overall model of the target process. additionally, an adaptive filter can be applied to the output of the primary analyzer to further capture the process dynamics. the data derived hybrid analyzer provides a readily adaptable framework to build the process model without requiring up-front knowledge. additionally, the primary analyzer, which incorporates the pls model, is well accepted by process control engineers. further, the hybrid analyzer also addresses the reliability of the process model output over the operating range since the primary analyzer can extrapolate data in a predictable way beyond the data used to train the model. together, the primary and the error correction analyzers provide a more accurate hybrid process analyzer which mitigates the disadvantages, and enhances the advantages, of each modeling methodology when used alone.",1999-03-02,"The title of the patent is hybrid linear-neural network process control and its abstract is a hybrid analyzer having a data derived primary analyzer and an error correction analyzer connected in parallel is disclosed. the primary analyzer, preferably a data derived linear model such as a partial least squares model, is trained using training data to generate major predictions of defined output variables. the error correction analyzer, preferably a neural network model is trained to capture the residuals between the primary analyzer outputs and the target process variables. the residuals generated by the error correction analyzer is summed with the output of the primary analyzer to compensate for the error residuals of the primary analyzer to arrive at a more accurate overall model of the target process. additionally, an adaptive filter can be applied to the output of the primary analyzer to further capture the process dynamics. the data derived hybrid analyzer provides a readily adaptable framework to build the process model without requiring up-front knowledge. additionally, the primary analyzer, which incorporates the pls model, is well accepted by process control engineers. further, the hybrid analyzer also addresses the reliability of the process model output over the operating range since the primary analyzer can extrapolate data in a predictable way beyond the data used to train the model. together, the primary and the error correction analyzers provide a more accurate hybrid process analyzer which mitigates the disadvantages, and enhances the advantages, of each modeling methodology when used alone. dated 1999-03-02"
5878112,medical system having movable components and a control device for preventing component collisions,"a medical system, such as an x-ray diagnostic device, having adjustable components, including a patient table, and having a monitoring device for determining the relative positions of the components and preventing collisions between them. signals indicating the positions, the directions of motion and the speeds of movement of the components are fed directly from the drives and sensors of the components to the monitoring device. calculations for controlling the components are carried out by a neural network, which has been appropriately trained in a learning phase.",1999-03-02,"The title of the patent is medical system having movable components and a control device for preventing component collisions and its abstract is a medical system, such as an x-ray diagnostic device, having adjustable components, including a patient table, and having a monitoring device for determining the relative positions of the components and preventing collisions between them. signals indicating the positions, the directions of motion and the speeds of movement of the components are fed directly from the drives and sensors of the components to the monitoring device. calculations for controlling the components are carried out by a neural network, which has been appropriately trained in a learning phase. dated 1999-03-02"
5878165,method for extracting object images and method for detecting movements thereof,"in a method for extracting an object image, an extraction area for extraction of a candidate for a predetermined object image from an image is determined. the center point of a view window, which has a predetermined size, is caused to travel to the position of the candidate for the predetermined object image. the extraction area is determined in accordance with the size and/or the shape of the candidate for the predetermined object image, the center point of the view window being taken as a reference during the determination of the extraction area. the extraction of the candidate for the predetermined object image is carried out by using a neural network. even if a plurality of object images, which are to be extracted, are embedded in a given image, the object images are extracted efficiently such that an object image, which has already been extracted, may not be extracted again.",1999-03-02,"The title of the patent is method for extracting object images and method for detecting movements thereof and its abstract is in a method for extracting an object image, an extraction area for extraction of a candidate for a predetermined object image from an image is determined. the center point of a view window, which has a predetermined size, is caused to travel to the position of the candidate for the predetermined object image. the extraction area is determined in accordance with the size and/or the shape of the candidate for the predetermined object image, the center point of the view window being taken as a reference during the determination of the extraction area. the extraction of the candidate for the predetermined object image is carried out by using a neural network. even if a plurality of object images, which are to be extracted, are embedded in a given image, the object images are extracted efficiently such that an object image, which has already been extracted, may not be extracted again. dated 1999-03-02"
5884294,system and method for functional recognition of emitters,"the present invention relates to a radar emitter recognition system and method for classifying incoming emitters into their functional roles as disclosed by their electronic signatures and the context within which the radar transmissions are received. an electronic support system provides incoming data, including unknown attribute information, to a neural network which has been synthesized or trained to calculate a network solution indicative of an emitter's classification within a range of attributes. similarly the electronic support system provides incoming data, including unknown context data, to a fuzzy logic system that has been provided with possibility distributions to classify the emitter as originating from one of several strategic contexts. the resultant categorizations from the neural network and the fuzzy logic system are combined in a classifier to yield an improved recognition of the emitter under observation. this system and method of using a neural network and fuzzy logic is also applicable to other recognition problems in a number of fields.",1999-03-16,"The title of the patent is system and method for functional recognition of emitters and its abstract is the present invention relates to a radar emitter recognition system and method for classifying incoming emitters into their functional roles as disclosed by their electronic signatures and the context within which the radar transmissions are received. an electronic support system provides incoming data, including unknown attribute information, to a neural network which has been synthesized or trained to calculate a network solution indicative of an emitter's classification within a range of attributes. similarly the electronic support system provides incoming data, including unknown context data, to a fuzzy logic system that has been provided with possibility distributions to classify the emitter as originating from one of several strategic contexts. the resultant categorizations from the neural network and the fuzzy logic system are combined in a classifier to yield an improved recognition of the emitter under observation. this system and method of using a neural network and fuzzy logic is also applicable to other recognition problems in a number of fields. dated 1999-03-16"
5884295,system for neural network interpretation of aeromagnetic data,"a system for processing aeromagnetic survey data to determine depth to basement rock is disclosed. the system uses neural networks having an input layer of elements, a hidden layer of elements and an output layer of elements which are interconnected by a weighted system of interconnections. a training session using known input and output data is used to train the neural network by adjusting the weighting functions repetitively to minimize any error in the output of the neural network.",1999-03-16,"The title of the patent is system for neural network interpretation of aeromagnetic data and its abstract is a system for processing aeromagnetic survey data to determine depth to basement rock is disclosed. the system uses neural networks having an input layer of elements, a hidden layer of elements and an output layer of elements which are interconnected by a weighted system of interconnections. a training session using known input and output data is used to train the neural network by adjusting the weighting functions repetitively to minimize any error in the output of the neural network. dated 1999-03-16"
5884296,network and image area attribute discriminating device and method for use with said neural network,a device for discriminating an attribute of an image in a block area contained in a document image includes a device for performing a fourier transformation based on image data in the block area and for determining a spatial frequency spectrum relating to the image in the block area; and a neural network for outputting a discrimination result as to whether or not the attribute of the image in the block area is a halftone dot image based on the spatial frequency spectrum output from the fourier transform device.,1999-03-16,The title of the patent is network and image area attribute discriminating device and method for use with said neural network and its abstract is a device for discriminating an attribute of an image in a block area contained in a document image includes a device for performing a fourier transformation based on image data in the block area and for determining a spatial frequency spectrum relating to the image in the block area; and a neural network for outputting a discrimination result as to whether or not the attribute of the image in the block area is a halftone dot image based on the spatial frequency spectrum output from the fourier transform device. dated 1999-03-16
5884626,apparatus and method for analyzing information relating to physical and mental condition,"an apparatus and method are provided for analyzing information relating to the physiological and psychological conditions of a driver. psychological conditions such as comfortableness or degree of alertness are estimated on the basis of physical data such as fluctuation in brain waves. this apparatus comprises a first neural network having a pre-processed 1/f fluctuation signal for brain waves as an input and for estimating a degree of alertness of the driver, and a second neural network receiving the estimated degree of alertness and the pre-processed 1/f fluctuation signal, for estimating and outputting driving comfortableness. by employing a neural network, which has a mapping ability as well as flexible adaptability even for non-linear data, based on the learning function, more accurate estimation of mental conditions can be achieved in comparison with conventional statistical analysis.",1999-03-23,"The title of the patent is apparatus and method for analyzing information relating to physical and mental condition and its abstract is an apparatus and method are provided for analyzing information relating to the physiological and psychological conditions of a driver. psychological conditions such as comfortableness or degree of alertness are estimated on the basis of physical data such as fluctuation in brain waves. this apparatus comprises a first neural network having a pre-processed 1/f fluctuation signal for brain waves as an input and for estimating a degree of alertness of the driver, and a second neural network receiving the estimated degree of alertness and the pre-processed 1/f fluctuation signal, for estimating and outputting driving comfortableness. by employing a neural network, which has a mapping ability as well as flexible adaptability even for non-linear data, based on the learning function, more accurate estimation of mental conditions can be achieved in comparison with conventional statistical analysis. dated 1999-03-23"
5887078,apparatus and method for classifying and recognizing image patterns using neural network,"the present invention provides an apparatus and a method for classifying and recognizing image patterns using a second-order neural network, thereby achieving high-rate parallel processing while lowering the complexity. the second-order neural network, which is made of adders and multipliers, corrects positional translations generated in a complex-log mapping unit to output the same result for the same object irrespective of the scale and/or rotation of the object. the present invention enables high-rate image pattern classification and recognition based on parallel processing, which is the advantage obtained in neural network models, because consistent neural networks and consistent network structure computation models are applied to all steps from the image input step to the pattern classifying and recognizing step.",1999-03-23,"The title of the patent is apparatus and method for classifying and recognizing image patterns using neural network and its abstract is the present invention provides an apparatus and a method for classifying and recognizing image patterns using a second-order neural network, thereby achieving high-rate parallel processing while lowering the complexity. the second-order neural network, which is made of adders and multipliers, corrects positional translations generated in a complex-log mapping unit to output the same result for the same object irrespective of the scale and/or rotation of the object. the present invention enables high-rate image pattern classification and recognition based on parallel processing, which is the advantage obtained in neural network models, because consistent neural networks and consistent network structure computation models are applied to all steps from the image input step to the pattern classifying and recognizing step. dated 1999-03-23"
5890101,neural network based method for estimating helicopter low airspeed,"the invention is directed to a system, utilizing a neural network, for estimating helicopter airspeed in the low airspeed flight range of below about 50 knots using only fixed system parameters as inputs to the neural network. the method includes the steps of: (a) defining input parameters derivable from variable state parameters generated during flight of the helicopter and measurable in a nonrotating reference frame associated with the helicopter; (b) determining the input parameters and a corresponding helicopter airspeed at a plurality of flight conditions representing a predetermined low airspeed flight domain of the helicopter; (c) establishing a learned relationship between the determined input parameters and the corresponding helicopter airspeed wherein the relationship is represented by at least one nonlinear equation; (d) storing the at least one nonlinear equation in a memory onboard the helicopter; (e) measuring real time values of the variable state parameters during low airspeed flight of the helicopter; (f) calculating real time values of the input parameters; (g) storing the real time values of the input parameters in the memory; (h) processing the real time values of the input parameters in accordance with the at least one nonlinear equation to determine real time airspeed; and (i) displaying the real time airspeed.",1999-03-30,"The title of the patent is neural network based method for estimating helicopter low airspeed and its abstract is the invention is directed to a system, utilizing a neural network, for estimating helicopter airspeed in the low airspeed flight range of below about 50 knots using only fixed system parameters as inputs to the neural network. the method includes the steps of: (a) defining input parameters derivable from variable state parameters generated during flight of the helicopter and measurable in a nonrotating reference frame associated with the helicopter; (b) determining the input parameters and a corresponding helicopter airspeed at a plurality of flight conditions representing a predetermined low airspeed flight domain of the helicopter; (c) establishing a learned relationship between the determined input parameters and the corresponding helicopter airspeed wherein the relationship is represented by at least one nonlinear equation; (d) storing the at least one nonlinear equation in a memory onboard the helicopter; (e) measuring real time values of the variable state parameters during low airspeed flight of the helicopter; (f) calculating real time values of the input parameters; (g) storing the real time values of the input parameters in the memory; (h) processing the real time values of the input parameters in accordance with the at least one nonlinear equation to determine real time airspeed; and (i) displaying the real time airspeed. dated 1999-03-30"
5892592,image processing apparatus,"an image processing apparatus inputs an image signal obtained by scanning a document in which a character region, a photographic region and a dot region are mixed and stores image data in a local block composed of a target picture element and plural picture elements surrounding the target picture element. the image processing apparatus computes first and second feature parameters p.sub.0 and p.sub.1 representing the features of each region based on the image data in the local block, and inputs the resulting first and second feature parameters p.sub.0 and p.sub.1 to an identification circuit adopting a neural network. the identification circuit outputs the region identification information o.sub.0 and o.sub.1 of the target picture element, and the filter processing circuit performs a spatial filtering process according to the region identification information o.sub.0 and o.sub.1. as described, in the image processing apparatus, since a multi-dimensional identification process is performed using the neural circuit which receives inputs of plural feature parameters, an image identification with very high precision is permitted, and an optimum spatial-filtering process can be applied according to the feature of each picture element,",1999-04-06,"The title of the patent is image processing apparatus and its abstract is an image processing apparatus inputs an image signal obtained by scanning a document in which a character region, a photographic region and a dot region are mixed and stores image data in a local block composed of a target picture element and plural picture elements surrounding the target picture element. the image processing apparatus computes first and second feature parameters p.sub.0 and p.sub.1 representing the features of each region based on the image data in the local block, and inputs the resulting first and second feature parameters p.sub.0 and p.sub.1 to an identification circuit adopting a neural network. the identification circuit outputs the region identification information o.sub.0 and o.sub.1 of the target picture element, and the filter processing circuit performs a spatial filtering process according to the region identification information o.sub.0 and o.sub.1. as described, in the image processing apparatus, since a multi-dimensional identification process is performed using the neural circuit which receives inputs of plural feature parameters, an image identification with very high precision is permitted, and an optimum spatial-filtering process can be applied according to the feature of each picture element, dated 1999-04-06"
5892838,biometric recognition using a classification neural network,""" a biometric recognition system involves two phases: creation of a master pattern set of authorized users biometric indicia and authentication using a classification neural network. to create the master pattern set, an image of an authorized biometric user's indicia is divided into a plurality of regions of interest or """"features"""". the system determines which features are the most useful for identification purposes. master patterns are then created from these master features, thus creating a master pattern set. during authentication, a sample pattern set of a user to be authenticated is similarly created. a neural network is used to compare the sample pattern set with the master pattern set to determine whether the user should be authenticated. """,1999-04-06,"The title of the patent is biometric recognition using a classification neural network and its abstract is "" a biometric recognition system involves two phases: creation of a master pattern set of authorized users biometric indicia and authentication using a classification neural network. to create the master pattern set, an image of an authorized biometric user's indicia is divided into a plurality of regions of interest or """"features"""". the system determines which features are the most useful for identification purposes. master patterns are then created from these master features, thus creating a master pattern set. during authentication, a sample pattern set of a user to be authenticated is similarly created. a neural network is used to compare the sample pattern set with the master pattern set to determine whether the user should be authenticated. "" dated 1999-04-06"
5892962,fpga-based processor,"a multiprocessor having an input/output controller, a process controller, and a multidimensional arrays of field programmable gate arrays (fpgas), each fpga having its own local memory. the multiprocessor may be programmed to function as a single-instruction, multiple-data (simd) parallel processor having a matrix of processing elements (pes), where each fpga may be programmed to operate as a submatrix array of pes. the multiprocessor is especially useful for image processing, pattern recognition, and neural network applications.",1999-04-06,"The title of the patent is fpga-based processor and its abstract is a multiprocessor having an input/output controller, a process controller, and a multidimensional arrays of field programmable gate arrays (fpgas), each fpga having its own local memory. the multiprocessor may be programmed to function as a single-instruction, multiple-data (simd) parallel processor having a matrix of processing elements (pes), where each fpga may be programmed to operate as a submatrix array of pes. the multiprocessor is especially useful for image processing, pattern recognition, and neural network applications. dated 1999-04-06"
5895435,"vehicle drive mode estimating device, and vehicle control apparatus, transmission shift control apparatus and vehicle drive force control apparatus including drive mode estimating device","apparatus for estimating a drive mode of a motor vehicle desired by an operator of the motor vehicle, including a variable calculating device and a drive mode estimating device, wherein the variable calculating device calculates at least one of drive mode indicating variables such as a drive force of the vehicle desired by the operator upon starting of the vehicle, a maximum rate of increase of the drive force, a maximum deceleration of the vehicle upon operation of a manually operated member for brake application to the vehicle, a coasting run time of the vehicle and a steady run time of the vehicle, and wherein the drive mode estimating device includes a neural network which receives the drive mode indicating variable or variables calculated by the variable calculating device, so that the drive mode estimating device estimates the drive mode of the motor vehicle desired by the operator on the basis of an output of the neural network. the output of the neural network may be used to control a desired controllable system of the vehicle such as an automatic transmission and a vehicle drive force control device.",1999-04-20,"The title of the patent is vehicle drive mode estimating device, and vehicle control apparatus, transmission shift control apparatus and vehicle drive force control apparatus including drive mode estimating device and its abstract is apparatus for estimating a drive mode of a motor vehicle desired by an operator of the motor vehicle, including a variable calculating device and a drive mode estimating device, wherein the variable calculating device calculates at least one of drive mode indicating variables such as a drive force of the vehicle desired by the operator upon starting of the vehicle, a maximum rate of increase of the drive force, a maximum deceleration of the vehicle upon operation of a manually operated member for brake application to the vehicle, a coasting run time of the vehicle and a steady run time of the vehicle, and wherein the drive mode estimating device includes a neural network which receives the drive mode indicating variable or variables calculated by the variable calculating device, so that the drive mode estimating device estimates the drive mode of the motor vehicle desired by the operator on the basis of an output of the neural network. the output of the neural network may be used to control a desired controllable system of the vehicle such as an automatic transmission and a vehicle drive force control device. dated 1999-04-20"
5896294,method and apparatus for inspecting manufactured products for defects in response to in-situ monitoring,"an apparatus and method for selecting products to inspect for defects performs in-situ monitoring of a processing tool during a manufacturing processing step. the data from the in-situ monitoring for a test run of products is correlated by a neural network with data collected during inspection of the test products for defects. during a production run of products, the in-situ monitor data is provided to the neural network which, based on the input data and the correlation, predicts the values of the data that would be collected upon inspection of the products. specific products from the production run are selected for inspection based upon the predicted values.",1999-04-20,"The title of the patent is method and apparatus for inspecting manufactured products for defects in response to in-situ monitoring and its abstract is an apparatus and method for selecting products to inspect for defects performs in-situ monitoring of a processing tool during a manufacturing processing step. the data from the in-situ monitoring for a test run of products is correlated by a neural network with data collected during inspection of the test products for defects. during a production run of products, the in-situ monitor data is provided to the neural network which, based on the input data and the correlation, predicts the values of the data that would be collected upon inspection of the products. specific products from the production run are selected for inspection based upon the predicted values. dated 1999-04-20"
5898304,sensor arrangement including a neural network and detection method using same,"a sensor arrangement (1) comprising at least one measuring coil (2), at least one voltage source (3) for the measuring coil (2), and an evaluation unit (4) with means for detecting, processing, and evaluating measured signals. this sensor arrangement (1) is used to measure distances and thicknesses substantially independently of the material involved, without the user having to know the physicomathematical relations between the influencing quantities and the measured values. in order to evaluate the measured signals, the evaluation unit (4) of the sensor arrangement comprises a neural network (5) with an input layer, at least one hidden layer, an output layer, and connection weights for the individual layers. the connection weights are determined and stored in a learning phase by measurements taken on a plurality of different suitable learning objects with known actual values.",1999-04-27,"The title of the patent is sensor arrangement including a neural network and detection method using same and its abstract is a sensor arrangement (1) comprising at least one measuring coil (2), at least one voltage source (3) for the measuring coil (2), and an evaluation unit (4) with means for detecting, processing, and evaluating measured signals. this sensor arrangement (1) is used to measure distances and thicknesses substantially independently of the material involved, without the user having to know the physicomathematical relations between the influencing quantities and the measured values. in order to evaluate the measured signals, the evaluation unit (4) of the sensor arrangement comprises a neural network (5) with an input layer, at least one hidden layer, an output layer, and connection weights for the individual layers. the connection weights are determined and stored in a learning phase by measurements taken on a plurality of different suitable learning objects with known actual values. dated 1999-04-27"
5898603,method and apparatus for approximating a sigmoidal response using digital circuitry,"a method and apparatus for approximating a sigmoidal response using digital circuitry for neural network computations. the digital circuitry in processing element (16) produces the neuron output signal (110) by performing a squashing operation which determines an approximation of a sigmoid function. in one form, the present invention uses digital circuitry (16) in data processor (10) to approximate a sigmoid function of a neuron (100) using a plurality of parabolas. in an alternate embodiment, the sigmoid function of neuron (100) is approximated using a quasi-log.sub.2 function.",1999-04-27,"The title of the patent is method and apparatus for approximating a sigmoidal response using digital circuitry and its abstract is a method and apparatus for approximating a sigmoidal response using digital circuitry for neural network computations. the digital circuitry in processing element (16) produces the neuron output signal (110) by performing a squashing operation which determines an approximation of a sigmoid function. in one form, the present invention uses digital circuitry (16) in data processor (10) to approximate a sigmoid function of a neuron (100) using a plurality of parabolas. in an alternate embodiment, the sigmoid function of neuron (100) is approximated using a quasi-log.sub.2 function. dated 1999-04-27"
5898792,methods and devices for automatic assessment of corn,"the flour yield, protein content and bulk density of cereal kernels can be determined by producing images of the cereal kernels, at least one color parameter and/or at least one geometric parameter are determined for the cereal kernels, and input signals to a neural network are produced by means of the color parameter and/or the geometric parameter. if the neural network is trained in some suitable manner, it can determine the flour yield, protein content and bulk density on the basis of the input signals.",1999-04-27,"The title of the patent is methods and devices for automatic assessment of corn and its abstract is the flour yield, protein content and bulk density of cereal kernels can be determined by producing images of the cereal kernels, at least one color parameter and/or at least one geometric parameter are determined for the cereal kernels, and input signals to a neural network are produced by means of the color parameter and/or the geometric parameter. if the neural network is trained in some suitable manner, it can determine the flour yield, protein content and bulk density on the basis of the input signals. dated 1999-04-27"
5899005,system and method for predicting the dryness of clothing articles,"the present invention discloses a system and method for predicting the dryness of clothing articles in a clothes dryer 10. in one embodiment of this invention, the clothes dryer 10 uses a temperature sensor 52, a phase angle sensor 54, and a humidity sensor 56 to generate signal representations of the temperature of the clothing articles, the motor phase angle, and the humidity of the heated air in the duct, respectively. a controller 58 receives the signal representations and determines a feature vector. a neural network 168 uses the feature vector to predict a percentage of moisture content and a degree of dryness of the clothing articles in the clothes dryer 10. in another embodiment of this invention, the clothes dryer uses a combination of sensors to predicts a percentage of moisture content and a degree of dryness of the clothing articles.",1999-05-04,"The title of the patent is system and method for predicting the dryness of clothing articles and its abstract is the present invention discloses a system and method for predicting the dryness of clothing articles in a clothes dryer 10. in one embodiment of this invention, the clothes dryer 10 uses a temperature sensor 52, a phase angle sensor 54, and a humidity sensor 56 to generate signal representations of the temperature of the clothing articles, the motor phase angle, and the humidity of the heated air in the duct, respectively. a controller 58 receives the signal representations and determines a feature vector. a neural network 168 uses the feature vector to predict a percentage of moisture content and a degree of dryness of the clothing articles in the clothes dryer 10. in another embodiment of this invention, the clothes dryer uses a combination of sensors to predicts a percentage of moisture content and a degree of dryness of the clothing articles. dated 1999-05-04"
5899984,apparatus and method for detection of molecular vapors in an atmospheric region,apparatus for detecting molecular vapors in an atmospheric region includes an interferometer which monitors light parameter data signals received and provides an interferometer light parameter signal corresponding to the light parameter data signals at a plurality of frequencies. the apparatus further includes an interferogram detector/converter which records and digitizes the interferometer light parameter signal to generate a plurality of discrete data points wherein each discrete data point corresponds to the interferometer light parameter signal at a specific frequency. the apparatus also includes a fourier transform circuit for receiving the discrete interferometer light parameter signal and providing a fourier transformed molecular parameter data signal. the apparatus further includes a probabilistic neural network for receiving and sorting the fourier transformed molecular parameter data signals without the use of a priori training data.,1999-05-04,The title of the patent is apparatus and method for detection of molecular vapors in an atmospheric region and its abstract is apparatus for detecting molecular vapors in an atmospheric region includes an interferometer which monitors light parameter data signals received and provides an interferometer light parameter signal corresponding to the light parameter data signals at a plurality of frequencies. the apparatus further includes an interferogram detector/converter which records and digitizes the interferometer light parameter signal to generate a plurality of discrete data points wherein each discrete data point corresponds to the interferometer light parameter signal at a specific frequency. the apparatus also includes a fourier transform circuit for receiving the discrete interferometer light parameter signal and providing a fourier transformed molecular parameter data signal. the apparatus further includes a probabilistic neural network for receiving and sorting the fourier transformed molecular parameter data signals without the use of a priori training data. dated 1999-05-04
5900634,"real-time on-line analysis of organic and non-organic compounds for food, fertilizers, and pharmaceutical products",the device is an apparatus for infrared spectroscopy. a succession of collimated light beams throughout the middle and near infrared spectrum are impinged against a sample or samples and the diffuse component of the reflected light is measured throughout the spectrum. this diffuse component is analyzed by a neural network to determine such characteristics as content of the sample.,1999-05-04,"The title of the patent is real-time on-line analysis of organic and non-organic compounds for food, fertilizers, and pharmaceutical products and its abstract is the device is an apparatus for infrared spectroscopy. a succession of collimated light beams throughout the middle and near infrared spectrum are impinged against a sample or samples and the diffuse component of the reflected light is measured throughout the spectrum. this diffuse component is analyzed by a neural network to determine such characteristics as content of the sample. dated 1999-05-04"
5901272,neural network based helicopter low airspeed indicator,"the invention is directed to means, utilizing a neural network, for estimating helicopter airspeed at speeds below about 50 knots using only fixed system parameters as inputs to the neural network. the system includes: means for entering at least one initial parameter; means for measuring, in a nonrotating reference frame associated with the helicopter, a plurality of variable state parameters generated during flight of the helicopter; means for determining a plurality of input parameters based on the at least one initial parameter and the plurality of variable state parameters and for generating successive signals representing the input parameters; at least one equation representing a nonlinear input-output relationship between the input parameters and airspeed; memory means for storing the at least one equation and for successively receiving and storing signals from the determining means; and processing means responsive to signals received from the memory means for generating airspeed information based on the input parameters and the at least one equation.",1999-05-04,"The title of the patent is neural network based helicopter low airspeed indicator and its abstract is the invention is directed to means, utilizing a neural network, for estimating helicopter airspeed at speeds below about 50 knots using only fixed system parameters as inputs to the neural network. the system includes: means for entering at least one initial parameter; means for measuring, in a nonrotating reference frame associated with the helicopter, a plurality of variable state parameters generated during flight of the helicopter; means for determining a plurality of input parameters based on the at least one initial parameter and the plurality of variable state parameters and for generating successive signals representing the input parameters; at least one equation representing a nonlinear input-output relationship between the input parameters and airspeed; memory means for storing the at least one equation and for successively receiving and storing signals from the determining means; and processing means responsive to signals received from the memory means for generating airspeed information based on the input parameters and the at least one equation. dated 1999-05-04"
5903884,method for training a statistical classifier with reduced tendency for overfitting,"to prevent overfitting a neural network to a finite set of training samples, random distortions are dynamically applied to the samples each time they are applied to the network during a training session. a plurality of different types of distortions can be applied, which are randomly selected each time a sample is applied to the network. alternatively, a combination of two or more types of distortion can be applied each time, with the amount of distortion being randomly varied for each type.",1999-05-11,"The title of the patent is method for training a statistical classifier with reduced tendency for overfitting and its abstract is to prevent overfitting a neural network to a finite set of training samples, random distortions are dynamically applied to the samples each time they are applied to the network during a training session. a plurality of different types of distortions can be applied, which are randomly selected each time a sample is applied to the network. alternatively, a combination of two or more types of distortion can be applied each time, with the amount of distortion being randomly varied for each type. dated 1999-05-11"
5904215,automatic brake control system and the method thereof,"an automatic brake control system brakes a driving vehicle according to a target brake control amount calculated by a neural network. the system detects vehicle speed of the driving vehicle, and detects the actual distance between the driving vehicle and a vehicle ahead of the driving vehicle. the system calculates a reference distance defined as a physically preferred distance between the driving vehicle and the vehicle ahead of the driving vehicle according to the vehicle speed, and normalizes the actual distance with the reference distance in order to obtain a normalized distance in a dimensionless quantity. the neural network calculates target brake control amount according to the vehicle speed and the normalized distance, and the automatic brake control system brakes the driving vehicle according to the target brake control amount.",1999-05-18,"The title of the patent is automatic brake control system and the method thereof and its abstract is an automatic brake control system brakes a driving vehicle according to a target brake control amount calculated by a neural network. the system detects vehicle speed of the driving vehicle, and detects the actual distance between the driving vehicle and a vehicle ahead of the driving vehicle. the system calculates a reference distance defined as a physically preferred distance between the driving vehicle and the vehicle ahead of the driving vehicle according to the vehicle speed, and normalizes the actual distance with the reference distance in order to obtain a normalized distance in a dimensionless quantity. the neural network calculates target brake control amount according to the vehicle speed and the normalized distance, and the automatic brake control system brakes the driving vehicle according to the target brake control amount. dated 1999-05-18"
5904227,method for continuously adjusting the architecture of a neural network used in elevator dispatching,"a method for adapting to observed special use patterns a neural network used to estimate quantities needed by an elevator dispatching system responsible for assigning the elevator or another elevator to a hall call. rather than simply refining values of existing connection weights to train the neural network to provide acceptable outputs for predetermined inputs, the method analyzes use information to determine whether additional inputs to the neural network might be advantageous and what those inputs might be. if so, the method alters the neural network architecture by providing new input nodes and corresponding connection weights, the connection weights having initially relatively small values. all connection weights can then be adjusted during actual operation of the elevator to accommodate the new input nodes.",1999-05-18,"The title of the patent is method for continuously adjusting the architecture of a neural network used in elevator dispatching and its abstract is a method for adapting to observed special use patterns a neural network used to estimate quantities needed by an elevator dispatching system responsible for assigning the elevator or another elevator to a hall call. rather than simply refining values of existing connection weights to train the neural network to provide acceptable outputs for predetermined inputs, the method analyzes use information to determine whether additional inputs to the neural network might be advantageous and what those inputs might be. if so, the method alters the neural network architecture by providing new input nodes and corresponding connection weights, the connection weights having initially relatively small values. all connection weights can then be adjusted during actual operation of the elevator to accommodate the new input nodes. dated 1999-05-18"
5907629,method of estimating chromaticity of illumination using neural networks,"a method of estimating the chromaticity of illumination of a colored image consisting of a plurality of color-encoded pixels. the image colors are first mapped into an intensity-independent chromaticity space which is then divided into a plurality of separate regions. for each region, a first binary value is assigned to the region if the region contains no chromaticity value; or, a second binary value is assigned to the region if it does contain a chromaticity value. the assigned values are then applied as inputs to a pre-trained neural network having two output ports and at least one intermediate layer containing a plurality rality of ports connectible between selected input ports and the output ports. the chromaticity space values which characterize the input image's chromaticity of illumination are then derived at the output ports. the network is pretrained trained by initially connecting an arbitrary number of the intermediate layer ports to selected input layer ports. a weight value is associated with each connection. the weight values, which have the effect of altering signals transmitted along each connection by a selected amount, are initialized with random values. each one of a plurality of pre-stored data sets, each containing values characterizing presence or absence of color in selected regions of one of a corresponding plurality of known colored images, are sequentially presented as inputs to the network and the chromaticity space values derived at the output ports are compared with known chromaticity space values characterizing illumination of the known colored image to derive an error value representative of difference therebetween. the weight values are adjusted in response to the inputs in accordance with the well known back propagation algorithm. after the weights are adjusted the intermediate layer ports are adaptively reconnected to the input layer ports to eliminate connections to input layer ports which repeatedly receive zero value inputs. the training process continues until the error value is less than a selected threshold.",1999-05-25,"The title of the patent is method of estimating chromaticity of illumination using neural networks and its abstract is a method of estimating the chromaticity of illumination of a colored image consisting of a plurality of color-encoded pixels. the image colors are first mapped into an intensity-independent chromaticity space which is then divided into a plurality of separate regions. for each region, a first binary value is assigned to the region if the region contains no chromaticity value; or, a second binary value is assigned to the region if it does contain a chromaticity value. the assigned values are then applied as inputs to a pre-trained neural network having two output ports and at least one intermediate layer containing a plurality rality of ports connectible between selected input ports and the output ports. the chromaticity space values which characterize the input image's chromaticity of illumination are then derived at the output ports. the network is pretrained trained by initially connecting an arbitrary number of the intermediate layer ports to selected input layer ports. a weight value is associated with each connection. the weight values, which have the effect of altering signals transmitted along each connection by a selected amount, are initialized with random values. each one of a plurality of pre-stored data sets, each containing values characterizing presence or absence of color in selected regions of one of a corresponding plurality of known colored images, are sequentially presented as inputs to the network and the chromaticity space values derived at the output ports are compared with known chromaticity space values characterizing illumination of the known colored image to derive an error value representative of difference therebetween. the weight values are adjusted in response to the inputs in accordance with the well known back propagation algorithm. after the weights are adjusted the intermediate layer ports are adaptively reconnected to the input layer ports to eliminate connections to input layer ports which repeatedly receive zero value inputs. the training process continues until the error value is less than a selected threshold. dated 1999-05-25"
5907822,loss tolerant speech decoder for telecommunications,"a method and device for extrapolating past signal-history data for insertion into missing data segments in order to conceal digital speech frame errors. the extrapolation method uses past-signal history that is stored in a buffer. the method is implemented with a device that utilizes a finite-impulse response (fir) multi-layer feed-forward artificial neural network that is trained by back-propagation for one-step extrapolation of speech compression algorithm (sca) parameters. once a speech connection has been established, the speech compression algorithm device begins sending encoded speech frames. as the speech frames are received, they are decoded and converted back into speech signal voltages. during the normal decoding process, pre-processing of the required sca parameters will occur and the results stored in the past-history buffer. if a speech frame is detected to be lost or in error, then extrapolation modules are executed and replacement sca parameters are generated and sent as the parameters required by the sca. in this way, the information transfer to the sca is transparent, and the sca processing continues as usual. the listener will not normally notice that a speech frame has been lost because of the smooth transition between the last-received, lost, and next-received speech frames.",1999-05-25,"The title of the patent is loss tolerant speech decoder for telecommunications and its abstract is a method and device for extrapolating past signal-history data for insertion into missing data segments in order to conceal digital speech frame errors. the extrapolation method uses past-signal history that is stored in a buffer. the method is implemented with a device that utilizes a finite-impulse response (fir) multi-layer feed-forward artificial neural network that is trained by back-propagation for one-step extrapolation of speech compression algorithm (sca) parameters. once a speech connection has been established, the speech compression algorithm device begins sending encoded speech frames. as the speech frames are received, they are decoded and converted back into speech signal voltages. during the normal decoding process, pre-processing of the required sca parameters will occur and the results stored in the past-history buffer. if a speech frame is detected to be lost or in error, then extrapolation modules are executed and replacement sca parameters are generated and sent as the parameters required by the sca. in this way, the information transfer to the sca is transparent, and the sca processing continues as usual. the listener will not normally notice that a speech frame has been lost because of the smooth transition between the last-received, lost, and next-received speech frames. dated 1999-05-25"
5909675,device for recognizing information conveyed by a received signal,"for recognizing information conveyed by a received signal, represented by convention by possible elementary forms of the signal to be transmitted, a device includes a correlator for establishing a correlation between the received signal and various possible forms of signal, in accordance with the convention. a neural network using correlation coefficients obtained from the correlator is trained by application to its input of correlation coefficients corresponding to a received signal conveying given information whilst imposing the given information at its output. the network supplies recognized information.",1999-06-01,"The title of the patent is device for recognizing information conveyed by a received signal and its abstract is for recognizing information conveyed by a received signal, represented by convention by possible elementary forms of the signal to be transmitted, a device includes a correlator for establishing a correlation between the received signal and various possible forms of signal, in accordance with the convention. a neural network using correlation coefficients obtained from the correlator is trained by application to its input of correlation coefficients corresponding to a received signal conveying given information whilst imposing the given information at its output. the network supplies recognized information. dated 1999-06-01"
5909676,system for controlling an object and medium using neural networks,"a function for compensating an error between a teacher signal and an output signal with a weight value is defined. a multilayered neural network is changed so that the function becomes minimum. thus, the multilayered neural network can be adaptively controlled.",1999-06-01,"The title of the patent is system for controlling an object and medium using neural networks and its abstract is a function for compensating an error between a teacher signal and an output signal with a weight value is defined. a multilayered neural network is changed so that the function becomes minimum. thus, the multilayered neural network can be adaptively controlled. dated 1999-06-01"
5909681,computer system and computerized method for partitioning data for parallel processing,"a computer system splits a data space to partition data between processors or processes. the data space may be split into sub-regions which need not be orthogonal to the axes defined by the data space's parameters, using a decision tree. the decision tree can have neural networks in each of its non-terminal nodes that are trained on, and are used to partition, training data. each terminal, or leaf, node can have a hidden layer neural network trained on the training data that reaches the terminal node. the training of the non-terminal nodes' neural networks can be performed on one processor and the training of the leaf nodes' neural networks can be run on separate processors. different target values can be used for the training of the networks of different non-terminal nodes. the non-terminal node networks may be hidden layer neural networks. each non-terminal node automatically may send a desired ratio of the training records it receives to each of its child nodes, so the leaf node networks each receives approximately the same number of training records. the system may automatically configures the tree to have a number of leaf nodes equal to the number of separate processors available to train leaf node networks. after the non-terminal and leaf node networks have been trained, the records of a large data base can be passed through the tree for classification or for estimation of certain parameter values.",1999-06-01,"The title of the patent is computer system and computerized method for partitioning data for parallel processing and its abstract is a computer system splits a data space to partition data between processors or processes. the data space may be split into sub-regions which need not be orthogonal to the axes defined by the data space's parameters, using a decision tree. the decision tree can have neural networks in each of its non-terminal nodes that are trained on, and are used to partition, training data. each terminal, or leaf, node can have a hidden layer neural network trained on the training data that reaches the terminal node. the training of the non-terminal nodes' neural networks can be performed on one processor and the training of the leaf nodes' neural networks can be run on separate processors. different target values can be used for the training of the networks of different non-terminal nodes. the non-terminal node networks may be hidden layer neural networks. each non-terminal node automatically may send a desired ratio of the training records it receives to each of its child nodes, so the leaf node networks each receives approximately the same number of training records. the system may automatically configures the tree to have a number of leaf nodes equal to the number of separate processors available to train leaf node networks. after the non-terminal and leaf node networks have been trained, the records of a large data base can be passed through the tree for classification or for estimation of certain parameter values. dated 1999-06-01"
5910109,non-invasive glucose measuring device and method for measuring blood glucose,"a glucose measuring device for determining the concentration of glucose in intravascular blood within a body part of a subject. the device includes light sources having a wavelength of 650, 880, 940 or 1300 nm to illuminate the fluid. receptors associated with the light sources for receiving light and generating a transmission signal representing the light transmitted are also provided. a support piece is including for supporting the light sources associated with their respective light sources. the support piece is adapted to engage a body part of a subject. finally, a signal analyzer, which includes a trained neural network, determines the glucose concentration in the blood of the subject. a method for determining the glucose concentration is also provided which calibrates a measuring device and sets the operating current for illuminating the light sources during operation of the device. once a transmission signal is generated by receptors receiving light via the light sources and illuminated blood, and the high and low values from each of the signals are stored in the device, and averaged to obtain a single transmission value for each of the light sources. the averaged values are then analyzed to determine the glucose concentration, which value is displayed on the device.",1999-06-08,"The title of the patent is non-invasive glucose measuring device and method for measuring blood glucose and its abstract is a glucose measuring device for determining the concentration of glucose in intravascular blood within a body part of a subject. the device includes light sources having a wavelength of 650, 880, 940 or 1300 nm to illuminate the fluid. receptors associated with the light sources for receiving light and generating a transmission signal representing the light transmitted are also provided. a support piece is including for supporting the light sources associated with their respective light sources. the support piece is adapted to engage a body part of a subject. finally, a signal analyzer, which includes a trained neural network, determines the glucose concentration in the blood of the subject. a method for determining the glucose concentration is also provided which calibrates a measuring device and sets the operating current for illuminating the light sources during operation of the device. once a transmission signal is generated by receptors receiving light via the light sources and illuminated blood, and the high and low values from each of the signals are stored in the device, and averaged to obtain a single transmission value for each of the light sources. the averaged values are then analyzed to determine the glucose concentration, which value is displayed on the device. dated 1999-06-08"
5910765,sensor module,"a device, and a method for using the device, to generate discrete information about an environmental event includes an array of sensors. all sensors in the array have a determinable detection capability and, preferably, some of these detection capabilities are redundant while others are overlapping. the individual sensors are positioned in the particular environment to detect characteristics of the event from different perspectives. the outputs which are generated by the various sensors in the array are selectively segmented and joined to create a convolved pattern of data which explicitly and implicitly includes information about the characteristics of the event. the convolved pattern is then presented to a pattern recognition unit, such as a neural network, where the characteristics are interpreted from the convolved pattern to generate the desired discrete information about the environmental event.",1999-06-08,"The title of the patent is sensor module and its abstract is a device, and a method for using the device, to generate discrete information about an environmental event includes an array of sensors. all sensors in the array have a determinable detection capability and, preferably, some of these detection capabilities are redundant while others are overlapping. the individual sensors are positioned in the particular environment to detect characteristics of the event from different perspectives. the outputs which are generated by the various sensors in the array are selectively segmented and joined to create a convolved pattern of data which explicitly and implicitly includes information about the characteristics of the event. the convolved pattern is then presented to a pattern recognition unit, such as a neural network, where the characteristics are interpreted from the convolved pattern to generate the desired discrete information about the environmental event. dated 1999-06-08"
5912986,evidential confidence measure and rejection technique for use in a neural network based optical character recognition system,"apparatus, and an accompanying method, for use in, e.g., a neural network-based optical character recognition (ocr) system (5) for accurately classifying each individual character extracted from a string of characters, and specifically for generating a highly reliable confidence measure that would be used in deciding whether to accept or reject each classified character. specifically, a confidence measure, associated with each output of, e.g., a neural classifier (165), is generated through use of all the neural activation output values. each individual neural activation output provides information for a corresponding atomic hypothesis of an evidence function. this hypothesis is that a pattern belongs to a particular class. each neural output is transformed (1650) through a pre-defined monotonic function into a degree of support in its associated evidence function. these degrees of support are then combined (1680, 1690) through an orthogonal sum to yield a single confidence measure associated with the specific classification then being produced by the neural classifier.",1999-06-15,"The title of the patent is evidential confidence measure and rejection technique for use in a neural network based optical character recognition system and its abstract is apparatus, and an accompanying method, for use in, e.g., a neural network-based optical character recognition (ocr) system (5) for accurately classifying each individual character extracted from a string of characters, and specifically for generating a highly reliable confidence measure that would be used in deciding whether to accept or reject each classified character. specifically, a confidence measure, associated with each output of, e.g., a neural classifier (165), is generated through use of all the neural activation output values. each individual neural activation output provides information for a corresponding atomic hypothesis of an evidence function. this hypothesis is that a pattern belongs to a particular class. each neural output is transformed (1650) through a pre-defined monotonic function into a degree of support in its associated evidence function. these degrees of support are then combined (1680, 1690) through an orthogonal sum to yield a single confidence measure associated with the specific classification then being produced by the neural classifier. dated 1999-06-15"
5913194,"method, device and system for using statistical information to reduce computation and memory requirements of a neural network based speech synthesis system","a method (400), device and system (300) provide, in response to linguistic information, efficient generation of a parametric representation of speech using a neural network. the method provides, in response to linguistic information efficient generation of a refined parametric representation of speech, comprising the steps of: a) using a data selection module to retrieve representative parameter vectors for each segment description according to the phonetic segment type and the phonetic segment types included in adjacent segment descriptions; b) interpolating between the representative parameter vectors according to the segment descriptions and duration to provide interpolated statistical parameters; c) converting the interpolated statistical parameters and linguistic information to neural network input parameters; d) utilizing a statistically enhanced neural network/neural network with post-processor to provide neural network output parameters that correspond to a parametric representation of speech; and converting the neural network output parameters to a refined parametric representation of speech.",1999-06-15,"The title of the patent is method, device and system for using statistical information to reduce computation and memory requirements of a neural network based speech synthesis system and its abstract is a method (400), device and system (300) provide, in response to linguistic information, efficient generation of a parametric representation of speech using a neural network. the method provides, in response to linguistic information efficient generation of a refined parametric representation of speech, comprising the steps of: a) using a data selection module to retrieve representative parameter vectors for each segment description according to the phonetic segment type and the phonetic segment types included in adjacent segment descriptions; b) interpolating between the representative parameter vectors according to the segment descriptions and duration to provide interpolated statistical parameters; c) converting the interpolated statistical parameters and linguistic information to neural network input parameters; d) utilizing a statistically enhanced neural network/neural network with post-processor to provide neural network output parameters that correspond to a parametric representation of speech; and converting the neural network output parameters to a refined parametric representation of speech. dated 1999-06-15"
5914868,multiplier and neural network synapse using current mirror having low-power mosfets,"a multiplier and a neural network synapse capable of removing nonlinear current using current mirror circuits. the multiplier produces a linear current by using mos transistors operating in a nonsaturation region. the multiplier includes a first current mirror including a plurality of mos transistors to form a first current and a second current mirror including a plurality of mos transistors to form a second current, wherein the second current mirror is coupled in parallel to the first current mirror. as a result, the multiplier outputs an output current by subtracting a second current from said first current.",1999-06-22,"The title of the patent is multiplier and neural network synapse using current mirror having low-power mosfets and its abstract is a multiplier and a neural network synapse capable of removing nonlinear current using current mirror circuits. the multiplier produces a linear current by using mos transistors operating in a nonsaturation region. the multiplier includes a first current mirror including a plurality of mos transistors to form a first current and a second current mirror including a plurality of mos transistors to form a second current, wherein the second current mirror is coupled in parallel to the first current mirror. as a result, the multiplier outputs an output current by subtracting a second current from said first current. dated 1999-06-22"
5915368,air/fuel ratio control apparatus that uses a neural network,"an air/fuel ratio control apparatus for executing auxiliary control of an air/fuel ratio by compensating an injected fuel amount set by a control system for maintaining the air/fuel ratio at a preset value. the air/fuel ratio control apparatus includes a state detecting unit for detecting a plurality of physical values which can be measured at low temperature and which show a state of an engine, an air/fuel ratio estimating unit for receiving a plurality of physical values detected by the state detecting means as input parameters and for estimating the air/fuel ratio using a neural network, and a compensatory fuel amount calculating unit for calculating a compensatory fuel amount for the injected fuel amount from the estimated air/fuel ratio. here, low temperature refers to a temperature at which an air/fuel sensor cannot operate.",1999-06-29,"The title of the patent is air/fuel ratio control apparatus that uses a neural network and its abstract is an air/fuel ratio control apparatus for executing auxiliary control of an air/fuel ratio by compensating an injected fuel amount set by a control system for maintaining the air/fuel ratio at a preset value. the air/fuel ratio control apparatus includes a state detecting unit for detecting a plurality of physical values which can be measured at low temperature and which show a state of an engine, an air/fuel ratio estimating unit for receiving a plurality of physical values detected by the state detecting means as input parameters and for estimating the air/fuel ratio using a neural network, and a compensatory fuel amount calculating unit for calculating a compensatory fuel amount for the injected fuel amount from the estimated air/fuel ratio. here, low temperature refers to a temperature at which an air/fuel sensor cannot operate. dated 1999-06-29"
5917891,voice-dialing system using adaptive model of calling behavior,"a method and apparatus for assisting voice-dialing using a model of an individual's calling behavior to improve recognition of an input name corresponding a desired telephone number. when the individual picks up a telephone, activity is initiated in a neural network model of the individual's calling behavior that predicts the likelihood that different numbers will be called, given such predictors as the day of the week and the time of day. the model is constructed by training the neural network with data from the user's history of making and receiving telephone calls. the auditory output from an automatic speech recognition system and the output from the user model are integrated together so as to select the number that is most likely to be the number desired by the speaker. the system can also provide automatic directory assistance, by speaking the number aloud rather than dialing it. in one version, the system is a personal directory for an individual maintained on that individual's personal computer. in another version, the system serves as a directory for a given physical or virtual site, with information about the institutional organization at the site in addition to individual calling histories used to track calling patterns and make predictions about the likelihood of calls within the site.",1999-06-29,"The title of the patent is voice-dialing system using adaptive model of calling behavior and its abstract is a method and apparatus for assisting voice-dialing using a model of an individual's calling behavior to improve recognition of an input name corresponding a desired telephone number. when the individual picks up a telephone, activity is initiated in a neural network model of the individual's calling behavior that predicts the likelihood that different numbers will be called, given such predictors as the day of the week and the time of day. the model is constructed by training the neural network with data from the user's history of making and receiving telephone calls. the auditory output from an automatic speech recognition system and the output from the user model are integrated together so as to select the number that is most likely to be the number desired by the speaker. the system can also provide automatic directory assistance, by speaking the number aloud rather than dialing it. in one version, the system is a personal directory for an individual maintained on that individual's personal computer. in another version, the system serves as a directory for a given physical or virtual site, with information about the institutional organization at the site in addition to individual calling histories used to track calling patterns and make predictions about the likelihood of calls within the site. dated 1999-06-29"
5919267,neural network fault diagnostics systems and related method,"a fault diagnostics system for monitoring the operating condition of a host system, e.g., an aircraft, which includes a plurality of subsystems. the fault diagnostics system is preferably implemented in software running on a high-speed neural network processor. the fault diagnostics system constructs a neural network model of the performance of each subsystem in a normal operating mode and each of a plurality of different possible failure modes. the system preferably dynamically predicts the performance of each subsystem based upon the response of each of the neural network models to dynamically changing operating conditions, compares the actual performance of each subsystem with the dynamically predicted performance thereof in each of the normal and possible failure modes, and determines the operating condition of the host system on the basis of these comparisons. in a preferred embodiment, the determining step is carried out by performing a statistical analysis of the comparisons made in the comparing step, e.g., by emloying a comparison voting technique. a related method is also disclosed.",1999-07-06,"The title of the patent is neural network fault diagnostics systems and related method and its abstract is a fault diagnostics system for monitoring the operating condition of a host system, e.g., an aircraft, which includes a plurality of subsystems. the fault diagnostics system is preferably implemented in software running on a high-speed neural network processor. the fault diagnostics system constructs a neural network model of the performance of each subsystem in a normal operating mode and each of a plurality of different possible failure modes. the system preferably dynamically predicts the performance of each subsystem based upon the response of each of the neural network models to dynamically changing operating conditions, compares the actual performance of each subsystem with the dynamically predicted performance thereof in each of the normal and possible failure modes, and determines the operating condition of the host system on the basis of these comparisons. in a preferred embodiment, the determining step is carried out by performing a statistical analysis of the comparisons made in the comparing step, e.g., by emloying a comparison voting technique. a related method is also disclosed. dated 1999-07-06"
5920212,control-type continuous ramp converting apparatus and method therefore,"control-type continuous ramp converting apparatus and method therefore. the present invention provides real-time processing of neurons in the neural network, easy implementation and reduction of manufacture cost of high density neurons in the neural network. the present invention comprises a first voltage controlling part for receiving a first voltage from an outside, and for non-linearly increasing a charged voltage in accordance with a differential continuous function of an exponential function; a second voltage controlling part for receiving a second voltage from an outside, and for non-linearly reducing a charged voltage in accordance with a differential continuous function of an exponential function; a charging part for charging an input current, and for providing the charged voltage of the charging part with the second voltage controlling part and an outside; and a plurality of switches for coupling outside and the first and the second voltage controlling part to the charging part, for selectively providing a third voltage from outside, an increased voltage and a decreased voltage based on the voltage of the charging part.",1999-07-06,"The title of the patent is control-type continuous ramp converting apparatus and method therefore and its abstract is control-type continuous ramp converting apparatus and method therefore. the present invention provides real-time processing of neurons in the neural network, easy implementation and reduction of manufacture cost of high density neurons in the neural network. the present invention comprises a first voltage controlling part for receiving a first voltage from an outside, and for non-linearly increasing a charged voltage in accordance with a differential continuous function of an exponential function; a second voltage controlling part for receiving a second voltage from an outside, and for non-linearly reducing a charged voltage in accordance with a differential continuous function of an exponential function; a charging part for charging an input current, and for providing the charged voltage of the charging part with the second voltage controlling part and an outside; and a plurality of switches for coupling outside and the first and the second voltage controlling part to the charging part, for selectively providing a third voltage from outside, an increased voltage and a decreased voltage based on the voltage of the charging part. dated 1999-07-06"
5923004,method for continuous learning by a neural network used in an elevator dispatching system,"a method for training a neural network used to estimate for an elevator the remaining response time for the elevator to service a hall call. the training, which results in adjusting connection weights between nodes of the neural network, is performed while the elevator is in actual operation. the method is not restricted to any particular architecture of neural network. the method uses a cutoff to limit changes to the connection weights, and provides for scaling the different inputs to the neural network so that all inputs lie in a predetermined range. the method also provides for training in case the elevator is diverted from servicing the hall call by an intervening hall call.",1999-07-13,"The title of the patent is method for continuous learning by a neural network used in an elevator dispatching system and its abstract is a method for training a neural network used to estimate for an elevator the remaining response time for the elevator to service a hall call. the training, which results in adjusting connection weights between nodes of the neural network, is performed while the elevator is in actual operation. the method is not restricted to any particular architecture of neural network. the method uses a cutoff to limit changes to the connection weights, and provides for scaling the different inputs to the neural network so that all inputs lie in a predetermined range. the method also provides for training in case the elevator is diverted from servicing the hall call by an intervening hall call. dated 1999-07-13"
5924066,system and method for classifying a speech signal,"a system and method for classifying a speech signal within a likely speech signal class of a plurality of speech signal classes are provided. stochastic models include a plurality of states having state transitions and output probabilities to generate state sequences which model evolutionary characteristics and durational variability of a speech signal. the method includes extracting a frame sequence, and determining a state sequence for each stochastic model with each state sequence having full state segmentation. representative frames are determined to provide speech signal time normalization. a likely speech signal class is determined from a neural network having a plurality of inputs receiving the representative frames and a plurality of outputs corresponding to the plurality of speech signal classes. an output signal is generated based on the likely stochastic model.",1999-07-13,"The title of the patent is system and method for classifying a speech signal and its abstract is a system and method for classifying a speech signal within a likely speech signal class of a plurality of speech signal classes are provided. stochastic models include a plurality of states having state transitions and output probabilities to generate state sequences which model evolutionary characteristics and durational variability of a speech signal. the method includes extracting a frame sequence, and determining a state sequence for each stochastic model with each state sequence having full state segmentation. representative frames are determined to provide speech signal time normalization. a likely speech signal class is determined from a neural network having a plurality of inputs receiving the representative frames and a plurality of outputs corresponding to the plurality of speech signal classes. an output signal is generated based on the likely stochastic model. dated 1999-07-13"
5924085,stochastic encoder/decoder/predictor,""" an artificial intelligence system is provided which makes use of a dual subroutine to adapt weights. elastic fuzzy logic (""""elf"""") system is provided in which classical neural network learning techniques are combined with fuzzy logic techniques in order to accomplish artificial intelligence tasks such as pattern recognition, expert cloning and trajectory control. the system may be implemented in a computer provided with multiplier means and storage means for storing a vector of weights to be used as multiplier factors in an apparatus for fuzzy control. """,1999-07-13,"The title of the patent is stochastic encoder/decoder/predictor and its abstract is "" an artificial intelligence system is provided which makes use of a dual subroutine to adapt weights. elastic fuzzy logic (""""elf"""") system is provided in which classical neural network learning techniques are combined with fuzzy logic techniques in order to accomplish artificial intelligence tasks such as pattern recognition, expert cloning and trajectory control. the system may be implemented in a computer provided with multiplier means and storage means for storing a vector of weights to be used as multiplier factors in an apparatus for fuzzy control. "" dated 1999-07-13"
5924086,method for developing a neural network tool for process identification,"a tool, and the method of making the tool, for process system identification that is based on the general purpose learning capabilities of neural networks. the tool and method can be used for a wide variety of system identification problems with little or no analytic effort. a neural network is trained using a process model to approximate a function which relates process input and output data to process parameter values. once trained, the network can be used as a system identification tool. in principle, this approach can be used for linear or nonlinear processes, for open or closed loop identification, and for identifying any or all process parameters.",1999-07-13,"The title of the patent is method for developing a neural network tool for process identification and its abstract is a tool, and the method of making the tool, for process system identification that is based on the general purpose learning capabilities of neural networks. the tool and method can be used for a wide variety of system identification problems with little or no analytic effort. a neural network is trained using a process model to approximate a function which relates process input and output data to process parameter values. once trained, the network can be used as a system identification tool. in principle, this approach can be used for linear or nonlinear processes, for open or closed loop identification, and for identifying any or all process parameters. dated 1999-07-13"
5926773,system for identifying known materials within a mixture of unknowns,one or both of two methods and systems are used to determine concentration of a known material in an unknown mixture on the basis of the measured interaction of electromagnetic waves upon the mixture. one technique is to utilize a multivariate analysis patch technique to develop a library of optimized patches of spectral signatures of known materials containing only those pixels most descriptive of the known materials by an evolutionary algorithm. identity and concentration of the known materials within the unknown mixture is then determined by minimizing the residuals between the measurements from the library of optimized patches and the measurements from the same pixels from the unknown mixture. another technique is to train a neural network by the genetic algorithm to determine the identity and concentration of known materials in the unknown mixture. the two techniques may be combined into an expert system providing cross checks for accuracy.,1999-07-20,The title of the patent is system for identifying known materials within a mixture of unknowns and its abstract is one or both of two methods and systems are used to determine concentration of a known material in an unknown mixture on the basis of the measured interaction of electromagnetic waves upon the mixture. one technique is to utilize a multivariate analysis patch technique to develop a library of optimized patches of spectral signatures of known materials containing only those pixels most descriptive of the known materials by an evolutionary algorithm. identity and concentration of the known materials within the unknown mixture is then determined by minimizing the residuals between the measurements from the library of optimized patches and the measurements from the same pixels from the unknown mixture. another technique is to train a neural network by the genetic algorithm to determine the identity and concentration of known materials in the unknown mixture. the two techniques may be combined into an expert system providing cross checks for accuracy. dated 1999-07-20
5926804,discriminant neural networks,"a discriminant neural network and a method of training the network are disclosed. the network includes a set of hidden nodes having associated weights, and the number of hidden nodes is minimized by the training method of the invention. the training method includes the steps of 1) loading a training data set and assigning it to a residual data set, 2) computing a vector associated with a first hidden node using the residual data set, 3) projecting training data onto a hyperplane associated with said first hidden node, 4) determining the number and locations of hard-limiter thresholds associated with the first node, and 5) repeating the above for successive hidden nodes after removing satisfied subsets from the training data until all partitioned regions of the input data space are satisfied.",1999-07-20,"The title of the patent is discriminant neural networks and its abstract is a discriminant neural network and a method of training the network are disclosed. the network includes a set of hidden nodes having associated weights, and the number of hidden nodes is minimized by the training method of the invention. the training method includes the steps of 1) loading a training data set and assigning it to a residual data set, 2) computing a vector associated with a first hidden node using the residual data set, 3) projecting training data onto a hyperplane associated with said first hidden node, 4) determining the number and locations of hard-limiter thresholds associated with the first node, and 5) repeating the above for successive hidden nodes after removing satisfied subsets from the training data until all partitioned regions of the input data space are satisfied. dated 1999-07-20"
5929906,color correcting method and apparatus,"a color correcting unit receives color separation values such as cmy values from an image input unit. under the control of a control portion, inputs to a first conversion portion constituted by a neural network which has been trained in advance on the basis of the spectral distribution of an arbitrary illuminant are corrected so that outputs from the first conversion portion satisfy the color separation values and a predetermined requirement. the input values to the first conversion portion, which satisfy the predetermined requirement, are sent to an image output unit. the image output unit outputs an image in accordance with these input values.",1999-07-27,"The title of the patent is color correcting method and apparatus and its abstract is a color correcting unit receives color separation values such as cmy values from an image input unit. under the control of a control portion, inputs to a first conversion portion constituted by a neural network which has been trained in advance on the basis of the spectral distribution of an arbitrary illuminant are corrected so that outputs from the first conversion portion satisfy the color separation values and a predetermined requirement. the input values to the first conversion portion, which satisfy the predetermined requirement, are sent to an image output unit. the image output unit outputs an image in accordance with these input values. dated 1999-07-27"
5930754,"method, device and article of manufacture for neural-network based orthography-phonetics transformation","a method (2000), device (2200) and article of manufacture (2300) provide, in response to orthographic information, efficient generation of a phonetic representation. the method provides for, in response to orthographic information, efficient generation of a phonetic representation, using the steps of: inputting an orthography of a word and a predetermined set of input letter features; utilizing a neural network that has been trained using automatic letter phone alignment and predetermined letter features to provide a neural network hypothesis of a word pronunciation.",1999-07-27,"The title of the patent is method, device and article of manufacture for neural-network based orthography-phonetics transformation and its abstract is a method (2000), device (2200) and article of manufacture (2300) provide, in response to orthographic information, efficient generation of a phonetic representation. the method provides for, in response to orthographic information, efficient generation of a phonetic representation, using the steps of: inputting an orthography of a word and a predetermined set of input letter features; utilizing a neural network that has been trained using automatic letter phone alignment and predetermined letter features to provide a neural network hypothesis of a word pronunciation. dated 1999-07-27"
5930781,neural network training by integration of adjoint systems of equations forward in time,"a method and apparatus for supervised neural learning of time dependent trajectories exploits the concepts of adjoint operators to enable computation of the gradient of an objective functional with respect to the various parameters of the network architecture in a highly efficient manner. specifically, it combines the advantage of dramatic reductions in computational complexity inherent in adjoint methods with the ability to solve two adjoint systems of equations together forward in time. not only is a large amount of computation and storage saved, but the handling of real-time applications becomes also possible. the invention has been applied it to two examples of representative complexity which have recently been analyzed in the open literature and demonstrated that a circular trajectory can be learned in approximately 200 iterations compared to the 12000 reported in the literature. a figure eight trajectory was achieved in under 500 iterations compared to 20000 previously required. the trajectories computed using our new method are much closer to the target trajectories than was reported in previous studies.",1999-07-27,"The title of the patent is neural network training by integration of adjoint systems of equations forward in time and its abstract is a method and apparatus for supervised neural learning of time dependent trajectories exploits the concepts of adjoint operators to enable computation of the gradient of an objective functional with respect to the various parameters of the network architecture in a highly efficient manner. specifically, it combines the advantage of dramatic reductions in computational complexity inherent in adjoint methods with the ability to solve two adjoint systems of equations together forward in time. not only is a large amount of computation and storage saved, but the handling of real-time applications becomes also possible. the invention has been applied it to two examples of representative complexity which have recently been analyzed in the open literature and demonstrated that a circular trajectory can be learned in approximately 200 iterations compared to the 12000 reported in the literature. a figure eight trajectory was achieved in under 500 iterations compared to 20000 previously required. the trajectories computed using our new method are much closer to the target trajectories than was reported in previous studies. dated 1999-07-27"
5933818,autonomous knowledge discovery system and method,"a knowlege discovery system (10) is provided for autonomously discovering knowlege from a database. the system includes a data reduction module (50) which reduces data into one or more clusters. this is accomplished by the use of one or more functions including a genetic clustering function, a hierarchical valley formation function, a symbolic exspansion reduction function, a fuzzy case clustering function, a relational clustering function, a k-means clustering function, a kohonen neural network clustering function, and a minimum distance classifier clustering function. a data analysis modual (60) autonomously determines one or more correlations among the clusters. the corrolations represent knowlege.",1999-08-03,"The title of the patent is autonomous knowledge discovery system and method and its abstract is a knowlege discovery system (10) is provided for autonomously discovering knowlege from a database. the system includes a data reduction module (50) which reduces data into one or more clusters. this is accomplished by the use of one or more functions including a genetic clustering function, a hierarchical valley formation function, a symbolic exspansion reduction function, a fuzzy case clustering function, a relational clustering function, a k-means clustering function, a kohonen neural network clustering function, and a minimum distance classifier clustering function. a data analysis modual (60) autonomously determines one or more correlations among the clusters. the corrolations represent knowlege. dated 1999-08-03"
5933819,prediction of relative binding motifs of biologically active peptides and peptide mimetics,"a general neural network based method and system for identifying peptide binding motifs from limited experimental data. in particular, an artificial neural network (ann) is trained with peptides with known sequence and function (i.e., binding strength) identified from a phage display library. the ann is then challenged with unknown peptides, and predicts relative binding motifs. analysis of the unknown peptides validate the predictive capability of the ann.",1999-08-03,"The title of the patent is prediction of relative binding motifs of biologically active peptides and peptide mimetics and its abstract is a general neural network based method and system for identifying peptide binding motifs from limited experimental data. in particular, an artificial neural network (ann) is trained with peptides with known sequence and function (i.e., binding strength) identified from a phage display library. the ann is then challenged with unknown peptides, and predicts relative binding motifs. analysis of the unknown peptides validate the predictive capability of the ann. dated 1999-08-03"
5935177,pilot-induced oscillation detection and compensation apparatus and method,"the apparatus of this invention includes a pilot-induced oscillation (pio) detector, a pio compensator and a pilot input modifier. the pio detector is coupled to receive aircraft state signal including the aircraft's pitch, roll and yaw attitudes. the pio detector is also coupled to receive pilot control signal generated by the aircraft's pilot by manipulation of flight control instruments. preferably, the pio detector includes a feature calculator and a discriminator. based on the aircraft state signal and the pilot control signal, the feature calculator generates at least one feature signal indicative of whether a pio or non-pio condition exists in the aircraft. the feature calculator supplies the feature signal to the discriminator, that uses the feature signal to determine whether or not a pio condition exists. the discriminator preferably uses a discrimination function determined by a neural network trained to discriminate pio and non-pio conditions, to generate the pio indicator signal based on the feature signal(s). the discriminator generates the pio indicator signal that indicates the likelihood of the existence of a pio condition in the aircraft. the pio indicator signal can be coupled to an audio and/or visual alarm to warn the pilot that the pilot's control actions are driving the pio condition. the pio compensator is coupled to receive the pio indicator signal, and generates the compensation signal using the pio indicator signal. the pio compensator is coupled to output the pio compensation signal to a pilot input modifier. in addition to the compensation signal, the pilot input modifier receives and modifies the pilot control signal, based on the compensation signal. the pilot input modifier can accomplish appropriate modification through gain attenuation or phase shift of the pilot control signal. the modified pilot control signal is used to control the aircraft's flight control actuators to eliminate the pio condition. accordingly, the invented apparatus prevents aircraft damage or pilot injury or death, that could otherwise result from a pio condition. the invention also includes a related method.",1999-08-10,"The title of the patent is pilot-induced oscillation detection and compensation apparatus and method and its abstract is the apparatus of this invention includes a pilot-induced oscillation (pio) detector, a pio compensator and a pilot input modifier. the pio detector is coupled to receive aircraft state signal including the aircraft's pitch, roll and yaw attitudes. the pio detector is also coupled to receive pilot control signal generated by the aircraft's pilot by manipulation of flight control instruments. preferably, the pio detector includes a feature calculator and a discriminator. based on the aircraft state signal and the pilot control signal, the feature calculator generates at least one feature signal indicative of whether a pio or non-pio condition exists in the aircraft. the feature calculator supplies the feature signal to the discriminator, that uses the feature signal to determine whether or not a pio condition exists. the discriminator preferably uses a discrimination function determined by a neural network trained to discriminate pio and non-pio conditions, to generate the pio indicator signal based on the feature signal(s). the discriminator generates the pio indicator signal that indicates the likelihood of the existence of a pio condition in the aircraft. the pio indicator signal can be coupled to an audio and/or visual alarm to warn the pilot that the pilot's control actions are driving the pio condition. the pio compensator is coupled to receive the pio indicator signal, and generates the compensation signal using the pio indicator signal. the pio compensator is coupled to output the pio compensation signal to a pilot input modifier. in addition to the compensation signal, the pilot input modifier receives and modifies the pilot control signal, based on the compensation signal. the pilot input modifier can accomplish appropriate modification through gain attenuation or phase shift of the pilot control signal. the modified pilot control signal is used to control the aircraft's flight control actuators to eliminate the pio condition. accordingly, the invented apparatus prevents aircraft damage or pilot injury or death, that could otherwise result from a pio condition. the invention also includes a related method. dated 1999-08-10"
5936212,"adjustment of elevator response time for horizon effect, including the use of a simple neural network","for elevator car dispatching scenarios, the standard remaining response time (rrt) estimation is augmented by a selected amount of time where a horizon floor is to be served before reaching the hall call to be assigned. the augmentation may be made by adding a fixed rrt penalty, by assuming a car call stop after the horizon floor at a selected floor, or by using a neural network.",1999-08-10,"The title of the patent is adjustment of elevator response time for horizon effect, including the use of a simple neural network and its abstract is for elevator car dispatching scenarios, the standard remaining response time (rrt) estimation is augmented by a selected amount of time where a horizon floor is to be served before reaching the hall call to be assigned. the augmentation may be made by adding a fixed rrt penalty, by assuming a car call stop after the horizon floor at a selected floor, or by using a neural network. dated 1999-08-10"
5937318,monocrystalline three-dimensional integrated circuit,""" a monocrystalline monolith contains a 3-d array of interconnected lattice-matched devices (which may be of one kind exclusively, or that kind in combination with one or more other kinds) performing digital, analog, image-processing, or neural-network functions, singly or in combination. localized inclusions of lattice-matched metal and (or) insulator can exist in the monolith, but monolith-wide layers of insulator are avoided. the devices may be self-isolated, junction-isolated, or insulator-isolated, and may include but not be limited to mosfets, bjts, jfets, mfets, ccds, resistors, and capacitors. the monolith is fabricated in a single apparatus using a process such as mbe or sputter epitaxy executed in a continuous or quasicontinuous manner under automatic control, and supplanting hundreds of discrete steps with handling and storage steps interpolated. """"writing"""" on the growing crystal is done during crystal growth by methods that may include but not be limited to ion beams, laser beams, patterned light exposures, and physical masks. the interior volume of the fabrication apparatus is far cleaner and more highly controlled than that of a clean room. the apparatus is highly, replicated and is amenable to mass production. the product has unprecedented volumetric function density, and high performance stems from short signal paths, low parasitic loading, and 3-d architecture. high reliability stems from contamination-free fabrication, small signal-arrival skew, and generous noise margins. economy stems from mass-produced factory apparatus, automatic ic manufacture, and high ic yield. among the ic products are fast and efficient memories with equally fast and efficient error-correction abilities, crosstalk-free operational amplifiers, and highly paralleled and copiously interconnected neural networks. """,1999-08-10,"The title of the patent is monocrystalline three-dimensional integrated circuit and its abstract is "" a monocrystalline monolith contains a 3-d array of interconnected lattice-matched devices (which may be of one kind exclusively, or that kind in combination with one or more other kinds) performing digital, analog, image-processing, or neural-network functions, singly or in combination. localized inclusions of lattice-matched metal and (or) insulator can exist in the monolith, but monolith-wide layers of insulator are avoided. the devices may be self-isolated, junction-isolated, or insulator-isolated, and may include but not be limited to mosfets, bjts, jfets, mfets, ccds, resistors, and capacitors. the monolith is fabricated in a single apparatus using a process such as mbe or sputter epitaxy executed in a continuous or quasicontinuous manner under automatic control, and supplanting hundreds of discrete steps with handling and storage steps interpolated. """"writing"""" on the growing crystal is done during crystal growth by methods that may include but not be limited to ion beams, laser beams, patterned light exposures, and physical masks. the interior volume of the fabrication apparatus is far cleaner and more highly controlled than that of a clean room. the apparatus is highly, replicated and is amenable to mass production. the product has unprecedented volumetric function density, and high performance stems from short signal paths, low parasitic loading, and 3-d architecture. high reliability stems from contamination-free fabrication, small signal-arrival skew, and generous noise margins. economy stems from mass-produced factory apparatus, automatic ic manufacture, and high ic yield. among the ic products are fast and efficient memories with equally fast and efficient error-correction abilities, crosstalk-free operational amplifiers, and highly paralleled and copiously interconnected neural networks. "" dated 1999-08-10"
5940777,automatic seismic pattern recognition method,"an automatic seismic pattern recognition method includes the steps of: determining a given number of seismic patterns to be recognized; providing a set of seismic trace portions for the region; defining a pattern recognition parameter common to all the trace portions, and determining the value of the parameters for each of the traces portions of the set. the method also includes the steps of: selecting trace portions of the set; selecting a one-dimensional neural network containing as many cells as there are patterns to be recognized where each cell is assigned a value of the recognition parameter; and submitting the neural network to a learning process with the selected trace portions so that at the end of the process each cell matches a pattern to be recognized and so that the patterns are progressively ordered. the method also includes the steps of: presenting each trace portion of the set to be processed to the classified and ordered neural network and attributing to each trace portion presented to the network the number of the cell closest to it.",1999-08-17,"The title of the patent is automatic seismic pattern recognition method and its abstract is an automatic seismic pattern recognition method includes the steps of: determining a given number of seismic patterns to be recognized; providing a set of seismic trace portions for the region; defining a pattern recognition parameter common to all the trace portions, and determining the value of the parameters for each of the traces portions of the set. the method also includes the steps of: selecting trace portions of the set; selecting a one-dimensional neural network containing as many cells as there are patterns to be recognized where each cell is assigned a value of the recognition parameter; and submitting the neural network to a learning process with the selected trace portions so that at the end of the process each cell matches a pattern to be recognized and so that the patterns are progressively ordered. the method also includes the steps of: presenting each trace portion of the set to be processed to the classified and ordered neural network and attributing to each trace portion presented to the network the number of the cell closest to it. dated 1999-08-17"
5942689,system and method for predicting a web break in a paper machine,"in this invention there is disclosed a system and method for predicting web breaks in a paper machine. in particular, this invention uses a neural network to predict web break tendencies from sensor measurements taken from the paper machine. also, an induction tree model is used to isolate the root cause of the predicted web break tendencies.",1999-08-24,"The title of the patent is system and method for predicting a web break in a paper machine and its abstract is in this invention there is disclosed a system and method for predicting web breaks in a paper machine. in particular, this invention uses a neural network to predict web break tendencies from sensor measurements taken from the paper machine. also, an induction tree model is used to isolate the root cause of the predicted web break tendencies. dated 1999-08-24"
5943659,deterministic encoding of fuzzy finite state automata in continuous recurrent neural networks,"based on the encoding of deterministic finite-state automata (dfa) in discrete-time, second-order recurrent neural networks, an algorithm constructs an augmented recurrent neural network that encodes a ffa and recognizes a given fuzzy regular language with arbitrary accuracy.",1999-08-24,"The title of the patent is deterministic encoding of fuzzy finite state automata in continuous recurrent neural networks and its abstract is based on the encoding of deterministic finite-state automata (dfa) in discrete-time, second-order recurrent neural networks, an algorithm constructs an augmented recurrent neural network that encodes a ffa and recognizes a given fuzzy regular language with arbitrary accuracy. dated 1999-08-24"
5943660,method for feedback linearization of neural networks and neural network incorporating same,"a method for linearization of feedback in neural networks, and a neural network incorporating the feedback linearization method are presented. control action is used to achieve tracking performance for a state-feedback linearizable, but unknown nonlinear control system. the control signal comprises a feedback linearization portion provided by neural networks, plus a robustifying portion that keep the control magnitude bounded. proofs are provided to show that all of the signals in the closed-loop system are semi-globally uniformly ultimately bounded. this eliminates an off-line learning phase, and simplifies the initialization of neural network weights.",1999-08-24,"The title of the patent is method for feedback linearization of neural networks and neural network incorporating same and its abstract is a method for linearization of feedback in neural networks, and a neural network incorporating the feedback linearization method are presented. control action is used to achieve tracking performance for a state-feedback linearizable, but unknown nonlinear control system. the control signal comprises a feedback linearization portion provided by neural networks, plus a robustifying portion that keep the control magnitude bounded. proofs are provided to show that all of the signals in the closed-loop system are semi-globally uniformly ultimately bounded. this eliminates an off-line learning phase, and simplifies the initialization of neural network weights. dated 1999-08-24"
5943661,"hybrid neural network classifier, systems and methods","a method of and system for parallelizing an program, comprising the steps of inputting an algorithm, operating said algorithm on selected data inputs to generate representative outputs, inputting representative outputs into parallelizing algorithms, and outputting a parallel implementation of said algorithm. in particular, this provides a parallel framework for target classification and pattern recognition procedures.",1999-08-24,"The title of the patent is hybrid neural network classifier, systems and methods and its abstract is a method of and system for parallelizing an program, comprising the steps of inputting an algorithm, operating said algorithm on selected data inputs to generate representative outputs, inputting representative outputs into parallelizing algorithms, and outputting a parallel implementation of said algorithm. in particular, this provides a parallel framework for target classification and pattern recognition procedures. dated 1999-08-24"
5946640,composition analysis,"a method and apparatus for analysing a sample, in which a neural network is trained to correct for measurement drift of a given analytical instrument (e.g., a mass spectrometer). the training is carried out using first and second sets of data obtained by the instrument from samples of known compositions at initial and subsequent instants of time, respectively. the trained neural network is used to transform data, obtained by the instrument from a sample of unknown composition at said subsequent instant of time, to an estimate of the data which would have been obtained by the instrument from that sample at the initial instant of time. the transformed dasta is then analysed to analyse the sample of unknown composition.",1999-08-31,"The title of the patent is composition analysis and its abstract is a method and apparatus for analysing a sample, in which a neural network is trained to correct for measurement drift of a given analytical instrument (e.g., a mass spectrometer). the training is carried out using first and second sets of data obtained by the instrument from samples of known compositions at initial and subsequent instants of time, respectively. the trained neural network is used to transform data, obtained by the instrument from a sample of unknown composition at said subsequent instant of time, to an estimate of the data which would have been obtained by the instrument from that sample at the initial instant of time. the transformed dasta is then analysed to analyse the sample of unknown composition. dated 1999-08-31"
5947909,neural network polymorphic qrs detector,a large universe of signals can be rapidly examined for a specific phenomenon by searching for a morphology representative of a subgroup of the universe. a detector for each such representative morphology is arranged in parallel to examine a signal input and provide an indication of positive identification a specific morphology.,1999-09-07,The title of the patent is neural network polymorphic qrs detector and its abstract is a large universe of signals can be rapidly examined for a specific phenomenon by searching for a morphology representative of a subgroup of the universe. a detector for each such representative morphology is arranged in parallel to examine a signal input and provide an indication of positive identification a specific morphology. dated 1999-09-07
5949367,device and method for classifying objects in an environmentally adaptive manner,"neural networks are used to classify objects automatically by means of doppler-broadened radar echo signals. the classification device kk contains a neural network (net, net2) which has an input layer (il) of input nodes (in1, . . . , in57) for features (m) of the doppler-broadened radar echo signals, and an output layer (ol) of output nodes (on1, on2, on3) for predetermined classes to which the objects can be allocated. the neural network (net, net2) is adapted to the external conditions prevailing at the time of the classification operation. the adaptation takes place either via accessible input nodes (zn1, zn2) into which control information (si) can be entered, and which cause the neural network (net) to adapt to one or to several external influence factors, or via a selection device (sel) which, from the parameters (p1, . . . , p4) of several neural networks stored in a memory (mem), which are trained with training data under respectively different conditions of external influence factors, selects the one most similar to the prevailing conditions.",1999-09-07,"The title of the patent is device and method for classifying objects in an environmentally adaptive manner and its abstract is neural networks are used to classify objects automatically by means of doppler-broadened radar echo signals. the classification device kk contains a neural network (net, net2) which has an input layer (il) of input nodes (in1, . . . , in57) for features (m) of the doppler-broadened radar echo signals, and an output layer (ol) of output nodes (on1, on2, on3) for predetermined classes to which the objects can be allocated. the neural network (net, net2) is adapted to the external conditions prevailing at the time of the classification operation. the adaptation takes place either via accessible input nodes (zn1, zn2) into which control information (si) can be entered, and which cause the neural network (net) to adapt to one or to several external influence factors, or via a selection device (sel) which, from the parameters (p1, . . . , p4) of several neural networks stored in a memory (mem), which are trained with training data under respectively different conditions of external influence factors, selects the one most similar to the prevailing conditions. dated 1999-09-07"
5949890,active noise control apparatus and waveform transforming apparatus through neural network,"noise data from a noise source is provided for a neural network. an output signal from the neural network is provided for a node of a hidden layer h of the neural network. the weight of the neural network is updated by an update unit according to an error signal e.sub.j0 detected by a microphone, thereby outputting a deadening sound from a speaker.",1999-09-07,"The title of the patent is active noise control apparatus and waveform transforming apparatus through neural network and its abstract is noise data from a noise source is provided for a neural network. an output signal from the neural network is provided for a node of a hidden layer h of the neural network. the weight of the neural network is updated by an update unit according to an error signal e.sub.j0 detected by a microphone, thereby outputting a deadening sound from a speaker. dated 1999-09-07"
5950181,apparatus and method for detecting and assessing a spatially discrete dot pattern,"in an apparatus and a method for detecting and assessing a spatially discrete dot pattern disposed in a multidimensional coordinate system, each dot in the pattern assumes at least two differentiatable status values. a measuring device records the coordinate values and status values of each dot of the multidimensional spatial dot pattern. a memory stores data that correspond to the recorded coordinate values and status values of each dot of the multidimensional spatial dot pattern. a computer into which the stored data are entered and in which a coordinate counter for each coordinate value of a coordinate axis is determined from the stored data, is associated with the memory. the value of the coordinate counter is formed from the number of detected dots of the coordinates that have a predetermined status value. a neural network is associated with the computer. an n-dimensional input vector with components formed from the calculated coordinate counters of each dot of the spatially discrete dot pattern is entered in the neural network. the neural network calculates an output vector by comparing the calculated input vector of the measured dot pattern with stored set-point vectors obtained on the basis of exemplary dot patterns. the neural network assigns a classification value of the measured dot pattern from the output vector ascertained and outputs it.",1999-09-07,"The title of the patent is apparatus and method for detecting and assessing a spatially discrete dot pattern and its abstract is in an apparatus and a method for detecting and assessing a spatially discrete dot pattern disposed in a multidimensional coordinate system, each dot in the pattern assumes at least two differentiatable status values. a measuring device records the coordinate values and status values of each dot of the multidimensional spatial dot pattern. a memory stores data that correspond to the recorded coordinate values and status values of each dot of the multidimensional spatial dot pattern. a computer into which the stored data are entered and in which a coordinate counter for each coordinate value of a coordinate axis is determined from the stored data, is associated with the memory. the value of the coordinate counter is formed from the number of detected dots of the coordinates that have a predetermined status value. a neural network is associated with the computer. an n-dimensional input vector with components formed from the calculated coordinate counters of each dot of the spatially discrete dot pattern is entered in the neural network. the neural network calculates an output vector by comparing the calculated input vector of the measured dot pattern with stored set-point vectors obtained on the basis of exemplary dot patterns. the neural network assigns a classification value of the measured dot pattern from the output vector ascertained and outputs it. dated 1999-09-07"
5953452,optical-digital method and processor for pattern recognition,the invention is an optical-digital method and processor which uses micro-optical lenslet arrays and fixed masks to implement an angular correlation algorithm and the hough transform for extracting amplitude and geometric features from objects embedded in video imagery. the optical-digital processor can be interfaced to a variety of sensors and can be used to classify objects when used in conjunction with a neural network.,1999-09-14,The title of the patent is optical-digital method and processor for pattern recognition and its abstract is the invention is an optical-digital method and processor which uses micro-optical lenslet arrays and fixed masks to implement an angular correlation algorithm and the hough transform for extracting amplitude and geometric features from objects embedded in video imagery. the optical-digital processor can be interfaced to a variety of sensors and can be used to classify objects when used in conjunction with a neural network. dated 1999-09-14
5953713,method and apparatus for treatment of sleep disorder breathing employing artificial neural network,"a method and apparatus for treating sleep disorder breathing is disclosed having improved ability to accurately detect pharyngeal wall vibration or other apneic events. an interface or mask is placed over a patient's airway. the interface is coupled to a source of pressurized gas. a respiration-related variable, namely the total pressure in the interface, is measured or sampled. the respiration-related variable is input into an artificial neural network trained to recognize patterns characterizing sleep disorder breathing. responsive to recognition by the artificial neural network of sleep disorder breathing, pressurized gas is supplied to the patient's airways through the interface.",1999-09-14,"The title of the patent is method and apparatus for treatment of sleep disorder breathing employing artificial neural network and its abstract is a method and apparatus for treating sleep disorder breathing is disclosed having improved ability to accurately detect pharyngeal wall vibration or other apneic events. an interface or mask is placed over a patient's airway. the interface is coupled to a source of pressurized gas. a respiration-related variable, namely the total pressure in the interface, is measured or sampled. the respiration-related variable is input into an artificial neural network trained to recognize patterns characterizing sleep disorder breathing. responsive to recognition by the artificial neural network of sleep disorder breathing, pressurized gas is supplied to the patient's airways through the interface. dated 1999-09-14"
5956413,method and device for automatic evaluation of cereal grains and other granular products,"in automatic evaluation of cereal kernels or like granular products handled in bulk, the kernels are conveyed on a vibrating conveyor belt (15). owing to the vibrations, the kernels are spread and settled in grooves (14) in the belt so as to be oriented in essentially the same direction. a video camera (40) produces digital images of all the kernels on the belt. the kernels are identified in the images, and for each kernel input signals are produced and then sent to a neural network based on picture element values for the picture elements representing each kernel. a neural network then determines which of a plurality of predetermined classes that each kernel belongs.",1999-09-21,"The title of the patent is method and device for automatic evaluation of cereal grains and other granular products and its abstract is in automatic evaluation of cereal kernels or like granular products handled in bulk, the kernels are conveyed on a vibrating conveyor belt (15). owing to the vibrations, the kernels are spread and settled in grooves (14) in the belt so as to be oriented in essentially the same direction. a video camera (40) produces digital images of all the kernels on the belt. the kernels are identified in the images, and for each kernel input signals are produced and then sent to a neural network based on picture element values for the picture elements representing each kernel. a neural network then determines which of a plurality of predetermined classes that each kernel belongs. dated 1999-09-21"
5956463,audio monitoring system for assessing wildlife biodiversity,"the invention relates to an automated system for monitoring wildlife auditory data and recording same for subsequent analysis and identification. the system comprises one or more microphones coupled to a recording apparatus for recording wildlife vocalizations in digital format. the resultant recorded data is preprocessed, segmented, and analyzed by means of a neural network to identify the respective species. the system minimizes the need for human intervention and subjective interpretation of the recorded sounds.",1999-09-21,"The title of the patent is audio monitoring system for assessing wildlife biodiversity and its abstract is the invention relates to an automated system for monitoring wildlife auditory data and recording same for subsequent analysis and identification. the system comprises one or more microphones coupled to a recording apparatus for recording wildlife vocalizations in digital format. the resultant recorded data is preprocessed, segmented, and analyzed by means of a neural network to identify the respective species. the system minimizes the need for human intervention and subjective interpretation of the recorded sounds. dated 1999-09-21"
5956702,time-series trend estimating system and method using column-structured recurrent neural network,"each neural element of a column-structured recurrent neural network generates an output from input data and recurrent data provided from a context layer of a corresponding column. one or more candidates for an estimated value is obtained, and an occurrence probability is computed using an internal state by solving an estimation equation determined by the internal state output from the neural network. a candidate having the highest occurrence probability is an estimated value for unknown data. thus, the internal state of the recurrent neural network is explicitly associated with the estimated value for data, and a data change can be efficiently estimated.",1999-09-21,"The title of the patent is time-series trend estimating system and method using column-structured recurrent neural network and its abstract is each neural element of a column-structured recurrent neural network generates an output from input data and recurrent data provided from a context layer of a corresponding column. one or more candidates for an estimated value is obtained, and an occurrence probability is computed using an internal state by solving an estimation equation determined by the internal state output from the neural network. a candidate having the highest occurrence probability is an estimated value for unknown data. thus, the internal state of the recurrent neural network is explicitly associated with the estimated value for data, and a data change can be efficiently estimated. dated 1999-09-21"
5956703,configurable neural network integrated circuit,"a neural network ic 31 includes n dedicated processing elements (pes) 62, an output register 66 for storing the pes' outputs so that they are immediately accessible to all of the pes, a number of output circuits 78 that are connected to selected pes to provide binary outputs, and a timing circuit 74. each of the pes includes a weight memory 90 for storing input, output and bias weight arrays, a first in first out (fifo) memory 88 for storing input data, a dot product circuit 92 and an activation circuit 94. the dot product circuit computes a dot product of the input weight array and the contents of the fifo memory, a dot product of the output weight array and the contents of the output register, a dot product of the bias value and a constant, and sums the three results. the activation circuit maps the output of the dot product circuit through an activation function to produce the pe's output. the inclusion of a memory 90 that stores both input and output weight arrays in conjunction with the output register 66 allows the pes to be configured to implement arbitrary feed-forward and recurrent neural network architectures.",1999-09-21,"The title of the patent is configurable neural network integrated circuit and its abstract is a neural network ic 31 includes n dedicated processing elements (pes) 62, an output register 66 for storing the pes' outputs so that they are immediately accessible to all of the pes, a number of output circuits 78 that are connected to selected pes to provide binary outputs, and a timing circuit 74. each of the pes includes a weight memory 90 for storing input, output and bias weight arrays, a first in first out (fifo) memory 88 for storing input data, a dot product circuit 92 and an activation circuit 94. the dot product circuit computes a dot product of the input weight array and the contents of the fifo memory, a dot product of the output weight array and the contents of the output register, a dot product of the bias value and a constant, and sums the three results. the activation circuit maps the output of the dot product circuit through an activation function to produce the pe's output. the inclusion of a memory 90 that stores both input and output weight arrays in conjunction with the output register 66 allows the pes to be configured to implement arbitrary feed-forward and recurrent neural network architectures. dated 1999-09-21"
5958001,output-processing circuit for a neural network and method of using same,"an output-processing circuit for a neural network, which may be implemented on an integrated circuit, comprises at least one latch and at least one adder. outputs from a plurality of neurons are sequentially received by the output-processing circuit. the output-processing circuit uses gating functions to determine which neuron outputs are summed together to produce neural network outputs.",1999-09-28,"The title of the patent is output-processing circuit for a neural network and method of using same and its abstract is an output-processing circuit for a neural network, which may be implemented on an integrated circuit, comprises at least one latch and at least one adder. outputs from a plurality of neurons are sequentially received by the output-processing circuit. the output-processing circuit uses gating functions to determine which neuron outputs are summed together to produce neural network outputs. dated 1999-09-28"
5960111,method and apparatus for segmenting images prior to coding,"to segment moving foreground from background, where the moving foreground is of most interest to the viewer, this method uses three detection algorithms as the input to a neural network. the multiple cues used are focus, intensity, and motion. the neural network consists of a two-layered neural network. focus and motion measurements are taken from high frequency data, edges; whereas, intensity measurements are taken from low frequency data, object interiors. combined, these measurements are used to segment a complete object. results indicate that moving foreground can be segmented from stationary foreground and moving or stationary background. the neural network segments the entire object, both interior and exterior, in this integrated approach. results also demonstrate that combining cues allows flexibility in both type and complexity of scenes. integration of cues improves accuracy in segmenting complex scenes containing both moving foreground and background. good segmentation yields bit rate savings when coding the object of interest, also called the video object in mpeg4. this method combines simple measurements to increase segmentation robustness.",1999-09-28,"The title of the patent is method and apparatus for segmenting images prior to coding and its abstract is to segment moving foreground from background, where the moving foreground is of most interest to the viewer, this method uses three detection algorithms as the input to a neural network. the multiple cues used are focus, intensity, and motion. the neural network consists of a two-layered neural network. focus and motion measurements are taken from high frequency data, edges; whereas, intensity measurements are taken from low frequency data, object interiors. combined, these measurements are used to segment a complete object. results indicate that moving foreground can be segmented from stationary foreground and moving or stationary background. the neural network segments the entire object, both interior and exterior, in this integrated approach. results also demonstrate that combining cues allows flexibility in both type and complexity of scenes. integration of cues improves accuracy in segmenting complex scenes containing both moving foreground and background. good segmentation yields bit rate savings when coding the object of interest, also called the video object in mpeg4. this method combines simple measurements to increase segmentation robustness. dated 1999-09-28"
5960391,"signal extraction system, system and method for speech restoration, learning method for neural network model, constructing method of neural network model, and signal processing system","a signal extraction system for extracting one or more signal components from an input signal including a plurality of signal components. this system is equipped with a neural network arithmetic section designed to process information through the use of a recurrent neural network. the neural network arithmetic section extracts one or more signal components, for example, a speech signal component and a noise signal component from an input signal including a plurality of signal components such as a speech and noises and outputs the extracted signal components. owing to the presence of this neural network arithmetic section, the signal extraction becomes possible with a high accuracy.",1999-09-28,"The title of the patent is signal extraction system, system and method for speech restoration, learning method for neural network model, constructing method of neural network model, and signal processing system and its abstract is a signal extraction system for extracting one or more signal components from an input signal including a plurality of signal components. this system is equipped with a neural network arithmetic section designed to process information through the use of a recurrent neural network. the neural network arithmetic section extracts one or more signal components, for example, a speech signal component and a noise signal component from an input signal including a plurality of signal components such as a speech and noises and outputs the extracted signal components. owing to the presence of this neural network arithmetic section, the signal extraction becomes possible with a high accuracy. dated 1999-09-28"
5963904,phoneme dividing method using multilevel neural network,"a phoneme dividing method using a multilevel neural network applied to a phoneme dividing apparatus having a voice input portion, a preprocessor, a multi-layer perceptron (mlp) phoneme dividing portion, and a phoneme border outputting portion includes the steps of: (a) sequentially segmenting and framing voice with digitalized voice samples, extracting characteristic vectors by vocal frames, and extracting an inter-frame characteristic vector of the difference between nearby frames of the characteristic vectors by frames, to thereby normalize the maximum and minimum of the characteristics; (b) storing information on the weight obtained through learning and the standard of the mlp; and (c) reading the weight obtained in the step (b), receiving the characteristic vectors, performing an operation of phoneme border discrimination to generate an output value, discriminating the phoneme border according to the output value, and if the current analyzed frame arrives two frames preceding the final frame of incoming voice, outputting a frame number indicative of the border of phoneme as a final result.",1999-10-05,"The title of the patent is phoneme dividing method using multilevel neural network and its abstract is a phoneme dividing method using a multilevel neural network applied to a phoneme dividing apparatus having a voice input portion, a preprocessor, a multi-layer perceptron (mlp) phoneme dividing portion, and a phoneme border outputting portion includes the steps of: (a) sequentially segmenting and framing voice with digitalized voice samples, extracting characteristic vectors by vocal frames, and extracting an inter-frame characteristic vector of the difference between nearby frames of the characteristic vectors by frames, to thereby normalize the maximum and minimum of the characteristics; (b) storing information on the weight obtained through learning and the standard of the mlp; and (c) reading the weight obtained in the step (b), receiving the characteristic vectors, performing an operation of phoneme border discrimination to generate an output value, discriminating the phoneme border according to the output value, and if the current analyzed frame arrives two frames preceding the final frame of incoming voice, outputting a frame number indicative of the border of phoneme as a final result. dated 1999-10-05"
5963929,recursive neural filters,"a recursive neurofilter comprising a recursive neural network (nn) is disclosed for processing an information process to estimate a signal process with respect to an estimation error criterion. the information process either consists of a measurement process, or if the signal and measurement processes are time-variant, consists of the measurement process as well as a time variance process, that describes the time-variant properties of the signal and measurement processes. the recursive neurofilter is synthesized from exemplary realizations of the signal and information processes. no assumptions such as gaussian distribution, linear dynamics, additive noise, and markov property are required. the synthesis is performed essentially through training recursive nns. the training criterion is constructed to reflect the mentioned estimation error criterion with the exemplary realizations. if an estimation error statistics process of a primary recursive neurofilter is needed, an ancillary recursive neurofilter is used to produce an approximate of this estimation error statistics process. an ancillary recursive neurofilter inputs either said primary recursive neurofilter's input process or said primary recursive neurofilter's input and output processes.",1999-10-05,"The title of the patent is recursive neural filters and its abstract is a recursive neurofilter comprising a recursive neural network (nn) is disclosed for processing an information process to estimate a signal process with respect to an estimation error criterion. the information process either consists of a measurement process, or if the signal and measurement processes are time-variant, consists of the measurement process as well as a time variance process, that describes the time-variant properties of the signal and measurement processes. the recursive neurofilter is synthesized from exemplary realizations of the signal and information processes. no assumptions such as gaussian distribution, linear dynamics, additive noise, and markov property are required. the synthesis is performed essentially through training recursive nns. the training criterion is constructed to reflect the mentioned estimation error criterion with the exemplary realizations. if an estimation error statistics process of a primary recursive neurofilter is needed, an ancillary recursive neurofilter is used to produce an approximate of this estimation error statistics process. an ancillary recursive neurofilter inputs either said primary recursive neurofilter's input process or said primary recursive neurofilter's input and output processes. dated 1999-10-05"
5966302,sheet processing system with neural network control,"a sheet processing apparatus, which may be a mailing machine, inserter or similar system for producing mail pieces or may be a copier or printer or the like. the system includes a control mechanism for reducing the likelihood of jams as sheets are fed through a sheet handling apparatus included in the system. preferably the control mechanism will include a neural network trained to response to a characteristic signal generated by a sheet feeder as a sheet is input to the apparatus. after training the network will produce a control signal output for controlling the processing rate of the apparatus to reduce the rate if there is a likelihood that the input sheet will jam. the network may also produce an outstacking signal for diverting sheets, in extreme cases, for corrective action. the drive current of a motor used to output the sheet from a sheet feeder is monitored to provide the characteristic signal which is sampled as an input to the network. an additional signal representative of an external condition, such as temperature or humidity or operating history, may also be input to the network.",1999-10-12,"The title of the patent is sheet processing system with neural network control and its abstract is a sheet processing apparatus, which may be a mailing machine, inserter or similar system for producing mail pieces or may be a copier or printer or the like. the system includes a control mechanism for reducing the likelihood of jams as sheets are fed through a sheet handling apparatus included in the system. preferably the control mechanism will include a neural network trained to response to a characteristic signal generated by a sheet feeder as a sheet is input to the apparatus. after training the network will produce a control signal output for controlling the processing rate of the apparatus to reduce the rate if there is a likelihood that the input sheet will jam. the network may also produce an outstacking signal for diverting sheets, in extreme cases, for corrective action. the drive current of a motor used to output the sheet from a sheet feeder is monitored to provide the characteristic signal which is sampled as an input to the network. an additional signal representative of an external condition, such as temperature or humidity or operating history, may also be input to the network. dated 1999-10-12"
5966460,on-line learning for neural net-based character recognition systems,"a neural network based improving the performance of an omni-font classifier by using recognized characters for additional training is presented. the invention applies the outputs of the hidden layer nodes of the neural net as the feature vector. characters that are recognized with high confidence are used to dynamically train a secondary classifier. after the secondary classifier is trained, it is combined with the original main classifier. the invention can re-adjust the partition or boundary of feature space, based on on-line learning, by utilizing the secondary classifier data to form an alternative partition location. the new partition can be referred to when a character conflict exists during character recognition.",1999-10-12,"The title of the patent is on-line learning for neural net-based character recognition systems and its abstract is a neural network based improving the performance of an omni-font classifier by using recognized characters for additional training is presented. the invention applies the outputs of the hidden layer nodes of the neural net as the feature vector. characters that are recognized with high confidence are used to dynamically train a secondary classifier. after the secondary classifier is trained, it is combined with the original main classifier. the invention can re-adjust the partition or boundary of feature space, based on on-line learning, by utilizing the secondary classifier data to form an alternative partition location. the new partition can be referred to when a character conflict exists during character recognition. dated 1999-10-12"
5966650,detecting mobile telephone misuse,"an arrangement for the detection of fraudulent use of a telephone subscriber's instrument in a mobile telephone system includes an input preprocessor (110), a neural network engine (111) coupled to the preprocessor, and an output postprocessor (112) coupled to the neural network engine. the preprocessor determines for each subscriber a first long term calling profile, a second short term calling profile, and a subscriber profile pattern comprising the difference between the first and second profiles. each calling profile and subscriber profile pattern comprises a set of values for a respective set of call attributes. the neural network engine comprises a self organizing map trained to effect pattern recognition of the subscriber profile patterns and a multilayer perceptron adapted to determine for each recognized pattern a value indicative of the probability of a fraud being associated with that pattern.",1999-10-12,"The title of the patent is detecting mobile telephone misuse and its abstract is an arrangement for the detection of fraudulent use of a telephone subscriber's instrument in a mobile telephone system includes an input preprocessor (110), a neural network engine (111) coupled to the preprocessor, and an output postprocessor (112) coupled to the neural network engine. the preprocessor determines for each subscriber a first long term calling profile, a second short term calling profile, and a subscriber profile pattern comprising the difference between the first and second profiles. each calling profile and subscriber profile pattern comprises a set of values for a respective set of call attributes. the neural network engine comprises a self organizing map trained to effect pattern recognition of the subscriber profile patterns and a multilayer perceptron adapted to determine for each recognized pattern a value indicative of the probability of a fraud being associated with that pattern. dated 1999-10-12"
5966682,system for calculating an output of a multi-stage forming process,a system for calculating an output of a multi-stage forming process using a model of the forming process with which the output of the forming process is determined as a function of properties of the forming process. selected properties of the forming process are determined using a neural network-based information processing arrangement.,1999-10-12,The title of the patent is system for calculating an output of a multi-stage forming process and its abstract is a system for calculating an output of a multi-stage forming process using a model of the forming process with which the output of the forming process is determined as a function of properties of the forming process. selected properties of the forming process are determined using a neural network-based information processing arrangement. dated 1999-10-12
5967981,time series prediction for event triggering,"delays in event detection in time-varying data can be reduced by predicting the time-varying data and then detecting the event in the predicted data. this finds application in the triggering of medical imaging devices, where physiological events can be detected in the time-varying data. an artificial neural network can be trained to predict data such as ecg signals from which a detection algorithm can accurately predict the occurrence of an event that will serve as a reference point for triggering.",1999-10-19,"The title of the patent is time series prediction for event triggering and its abstract is delays in event detection in time-varying data can be reduced by predicting the time-varying data and then detecting the event in the predicted data. this finds application in the triggering of medical imaging devices, where physiological events can be detected in the time-varying data. an artificial neural network can be trained to predict data such as ecg signals from which a detection algorithm can accurately predict the occurrence of an event that will serve as a reference point for triggering. dated 1999-10-19"
5970435,automatic load measuring device,"an automatic load measuring device includes an actual load calculating device. the actual load calculating device includes a multi-layer feed-forward type neural network having an input layer, an intermediate layer and an output layer arranged in a hierarchial manner. also, the actual load calculating device, by use of the multi-layer feed-forward type neural network, can previously execute learning relating to the correction of a carrying load of a vehicle to be measured by automatic load measuring sensors respectively used to measure the carrying loads of the vehicle using measured load information measured by the automatic load measuring sensors, and, based on the result of the learning, can correct the carrying loads measured by the automatic load measuring sensors so as to find the actual load of the vehicle.",1999-10-19,"The title of the patent is automatic load measuring device and its abstract is an automatic load measuring device includes an actual load calculating device. the actual load calculating device includes a multi-layer feed-forward type neural network having an input layer, an intermediate layer and an output layer arranged in a hierarchial manner. also, the actual load calculating device, by use of the multi-layer feed-forward type neural network, can previously execute learning relating to the correction of a carrying load of a vehicle to be measured by automatic load measuring sensors respectively used to measure the carrying loads of the vehicle using measured load information measured by the automatic load measuring sensors, and, based on the result of the learning, can correct the carrying loads measured by the automatic load measuring sensors so as to find the actual load of the vehicle. dated 1999-10-19"
5970482,system for data mining using neuroagents,"a neuroagent approach is used in an automated and unified data mining system to provide an explicitly predictive knowledge model. the neuroagent is a neural multi-agent approach based on macro-connectionism and comprises a double integration at the association and symbolic level as well as the knowledge model level. this data mining system permits discovery, evaluation and prediction of the correlative factors of data, i.e., the conjunctions, as corresponding to neuroexpressions (a semantic connection of neuroagents) connected to an output neuroagent which corresponds to the data output, the connection weights yielding the relative significance of these factors to the given output. the system takes data sets called domains, establishes candidate dimensions or parameters, categorizes parameters into discrete bins, and trains a neuroagent network composed of neuroagents allocated for each bin and each output based on a discovery data set, called a discovery domain, and by building up the various minimal and contextual neuroexpressions, and setting the appropriate connection weights, the results may therefore be compared with an optional evaluation data set, called an evaluation domain to establish the accuracy of the knowledge model, and thereafter applied with some degree of confidence to a prediction set or prediction domain. the ranking in importance of the composite parameters may be calculated as well as the discrimination between the various outputs, which permits the relevant factors of interest to a decision maker to come into focus.",1999-10-19,"The title of the patent is system for data mining using neuroagents and its abstract is a neuroagent approach is used in an automated and unified data mining system to provide an explicitly predictive knowledge model. the neuroagent is a neural multi-agent approach based on macro-connectionism and comprises a double integration at the association and symbolic level as well as the knowledge model level. this data mining system permits discovery, evaluation and prediction of the correlative factors of data, i.e., the conjunctions, as corresponding to neuroexpressions (a semantic connection of neuroagents) connected to an output neuroagent which corresponds to the data output, the connection weights yielding the relative significance of these factors to the given output. the system takes data sets called domains, establishes candidate dimensions or parameters, categorizes parameters into discrete bins, and trains a neuroagent network composed of neuroagents allocated for each bin and each output based on a discovery data set, called a discovery domain, and by building up the various minimal and contextual neuroexpressions, and setting the appropriate connection weights, the results may therefore be compared with an optional evaluation data set, called an evaluation domain to establish the accuracy of the knowledge model, and thereafter applied with some degree of confidence to a prediction set or prediction domain. the ranking in importance of the composite parameters may be calculated as well as the discrimination between the various outputs, which permits the relevant factors of interest to a decision maker to come into focus. dated 1999-10-19"
5971747,automatically optimized combustion control,"systems and methods are disclosed that optimize the combustion process in various reactors, furnaces, and internal combustion engines. video cameras are used to evaluate the combustion flame grade. depending on the desired form, standard or special video devices, or beam scanning devices, are used to image the combustion flame and by-products. the video device generates and outputs image signals during various phases of, and at various locations in, the combustion process. other forms of sensors monitor and generate data signals defining selected parameters of the combustion process, such as air flow, fuel flow, turbulence, exhaust and inlet valve openings, etc. in a preferred form, a neural networks initially processes the image data and characterizes the combustion flame. a fuzzy logic controller and associated fuzzy logic rule base analyzes the image data from the neural network, along with other sensor information. the fuzzy logic controller determines and generates control signals defining adjustments necessary to optimize the combustion process.",1999-10-26,"The title of the patent is automatically optimized combustion control and its abstract is systems and methods are disclosed that optimize the combustion process in various reactors, furnaces, and internal combustion engines. video cameras are used to evaluate the combustion flame grade. depending on the desired form, standard or special video devices, or beam scanning devices, are used to image the combustion flame and by-products. the video device generates and outputs image signals during various phases of, and at various locations in, the combustion process. other forms of sensors monitor and generate data signals defining selected parameters of the combustion process, such as air flow, fuel flow, turbulence, exhaust and inlet valve openings, etc. in a preferred form, a neural networks initially processes the image data and characterizes the combustion flame. a fuzzy logic controller and associated fuzzy logic rule base analyzes the image data from the neural network, along with other sensor information. the fuzzy logic controller determines and generates control signals defining adjustments necessary to optimize the combustion process. dated 1999-10-26"
5974404,method and apparatus for input classification using a neural network,the present invention is a classification method and apparatus for classifying an input into one of a plurality of possible outputs. the invention is also a method and apparatus for adjusting a neuron encompassing a plurality of feature vectors. the invention characterizes the spatial distribution of the feature vectors. the invention then spatially adjusts the neuron in accordance with that characterization.,1999-10-26,The title of the patent is method and apparatus for input classification using a neural network and its abstract is the present invention is a classification method and apparatus for classifying an input into one of a plurality of possible outputs. the invention is also a method and apparatus for adjusting a neuron encompassing a plurality of feature vectors. the invention characterizes the spatial distribution of the feature vectors. the invention then spatially adjusts the neuron in accordance with that characterization. dated 1999-10-26
5978505,system and method for image regularization in inhomogeneous environments using clustering in neural networks,"a regularization system and method for image restoration in homogeneous or inhomogeneous environments. the system and method includes features similar to a neural network with intermediate levels of structure including a pixel having processing capabilities; clusters consisting of a plurality of interconnected pixels and also having processing capabilities; and an image space comprised of a plurality of interconnected pixels and clusters and also having processing capabilities. the system and method also include means for assigning a regularization parameter to each pixel depending on the local variance of intensity of pixels; decomposing the image space into clusters of pixels, each cluster having the same regularization parameter; imposing a blurring function on each pixel; rapidly forming a regularized image by simultaneous local and global encoding of a regularization matrix onto each pixel directed through a process of gradient energy decent; and a means of assessing the output image.",1999-11-02,"The title of the patent is system and method for image regularization in inhomogeneous environments using clustering in neural networks and its abstract is a regularization system and method for image restoration in homogeneous or inhomogeneous environments. the system and method includes features similar to a neural network with intermediate levels of structure including a pixel having processing capabilities; clusters consisting of a plurality of interconnected pixels and also having processing capabilities; and an image space comprised of a plurality of interconnected pixels and clusters and also having processing capabilities. the system and method also include means for assigning a regularization parameter to each pixel depending on the local variance of intensity of pixels; decomposing the image space into clusters of pixels, each cluster having the same regularization parameter; imposing a blurring function on each pixel; rapidly forming a regularized image by simultaneous local and global encoding of a regularization matrix onto each pixel directed through a process of gradient energy decent; and a means of assessing the output image. dated 1999-11-02"
5978782,neural network signal processor for magnetic storage channels,"a neural network based signal processor for a magnetic storage channel, such as a magnetic disk drive for a computer, uses a multiple layer perceptron neural network to perform the symbol sequencing detection, equalization and decoding of information signals retrieved from the magnetic storage medium.",1999-11-02,"The title of the patent is neural network signal processor for magnetic storage channels and its abstract is a neural network based signal processor for a magnetic storage channel, such as a magnetic disk drive for a computer, uses a multiple layer perceptron neural network to perform the symbol sequencing detection, equalization and decoding of information signals retrieved from the magnetic storage medium. dated 1999-11-02"
5982403,potential estimating apparatus using a plurality of neural networks for carrying out an electrographic process,"a potential estimation apparatus estimates a potential of a photosensitive body of an image forming apparatus that carries out an electro-photography process using the photosensitive body. the potential estimation apparatus includes a sensor group for sensing and outputting data related to information which affects the electro-photography process, a storage unit for at least storing the data output from the sensor group and information related to charge of the photosensitive body, and an estimation circuit including a neural network for estimating a charged portion potential of the photosensitive body based on a charge retentivity of the photosensitive body learned by the neural network. the neural network in a learning mode receives at least one of the data output from the sensor group and time-sequentially sampled, and parameters which affect the charge retentivity of the photosensitive body as an input, and receives as a teaching value a charged portion potential which is obtained in advance with respect to at least an amount of charge and the charge retentivity of the photosensitive body.",1999-11-09,"The title of the patent is potential estimating apparatus using a plurality of neural networks for carrying out an electrographic process and its abstract is a potential estimation apparatus estimates a potential of a photosensitive body of an image forming apparatus that carries out an electro-photography process using the photosensitive body. the potential estimation apparatus includes a sensor group for sensing and outputting data related to information which affects the electro-photography process, a storage unit for at least storing the data output from the sensor group and information related to charge of the photosensitive body, and an estimation circuit including a neural network for estimating a charged portion potential of the photosensitive body based on a charge retentivity of the photosensitive body learned by the neural network. the neural network in a learning mode receives at least one of the data output from the sensor group and time-sequentially sampled, and parameters which affect the charge retentivity of the photosensitive body as an input, and receives as a teaching value a charged portion potential which is obtained in advance with respect to at least an amount of charge and the charge retentivity of the photosensitive body. dated 1999-11-09"
5983211,method and apparatus for the diagnosis of colorectal cancer,"a process is set forth in which cancer of the colon is assessed in a patient. the probabilities of developing cancer involves the initial step of extracting a set of sample body fluids from the patient. fluids can be evaluated to determine certain marker constituents in the body fluids. fluids which are extracted have some relationship to me development of cancer, precancer or tendency toward cancerous conditions. the body fluid markers are measured and other quantified. the marker data then is evaluated using a nonlinear technique exemplified through the use of a multiple input and multiple output neural network having a variable learning rate and training rate. the neural network is provided with data from other patients for the same or similar markers. data from other patients who did and did not have cancer is used in the learning of the neural network which thereby processes the data and provides a determination that the patient has a cancerous condition, precancer cells or a tendency towards cancer.",1999-11-09,"The title of the patent is method and apparatus for the diagnosis of colorectal cancer and its abstract is a process is set forth in which cancer of the colon is assessed in a patient. the probabilities of developing cancer involves the initial step of extracting a set of sample body fluids from the patient. fluids can be evaluated to determine certain marker constituents in the body fluids. fluids which are extracted have some relationship to me development of cancer, precancer or tendency toward cancerous conditions. the body fluid markers are measured and other quantified. the marker data then is evaluated using a nonlinear technique exemplified through the use of a multiple input and multiple output neural network having a variable learning rate and training rate. the neural network is provided with data from other patients for the same or similar markers. data from other patients who did and did not have cancer is used in the learning of the neural network which thereby processes the data and provides a determination that the patient has a cancerous condition, precancer cells or a tendency towards cancer. dated 1999-11-09"
5984870,method and system for the automated analysis of lesions in ultrasound images,"a method and apparatus for the computerized automatic analysis of lesions in ultrasound images, including the computerized analysis of lesions in the breast, using gradient, gray-level, and texture based measures. echogenicity features are developed to assess the characteristics of the lesions and in some cases give an estimate of the likelihood of malignancy or of prognosis. the output from the computerized analysis is used in making a diagnosis and/or prognosis. for example, with the analysis of the ultrasound images of the breast, the features can be used to either distinguish between malignant and benign lesions, or distinguish between (i.e., diagnosis) the types of benign lesions such as benign solid lesions (e.g., fibroadenoma), simple cysts, complex cysts, and benign cysts. the ultrasound image features can be merged with those from mammographic and/or magnetic resonance images of the same lesion for classification by means of a common artificial neural network.",1999-11-16,"The title of the patent is method and system for the automated analysis of lesions in ultrasound images and its abstract is a method and apparatus for the computerized automatic analysis of lesions in ultrasound images, including the computerized analysis of lesions in the breast, using gradient, gray-level, and texture based measures. echogenicity features are developed to assess the characteristics of the lesions and in some cases give an estimate of the likelihood of malignancy or of prognosis. the output from the computerized analysis is used in making a diagnosis and/or prognosis. for example, with the analysis of the ultrasound images of the breast, the features can be used to either distinguish between malignant and benign lesions, or distinguish between (i.e., diagnosis) the types of benign lesions such as benign solid lesions (e.g., fibroadenoma), simple cysts, complex cysts, and benign cysts. the ultrasound image features can be merged with those from mammographic and/or magnetic resonance images of the same lesion for classification by means of a common artificial neural network. dated 1999-11-16"
5987328,method and device for placement of transmitters in wireless networks,"from layout and attenuation data of an area and an initial guess, the placement and power levels of base stations and the resulting attenuation and base station ranges are calculated based on any chosen propagation model. a cost function is calculated which indicates the merit of the initial guess placement. the cost function is a function of the transmitter locations and power levels and can be calculated as a linear combination of the uncovered and interference areas. other cost functions can also be considered. the cost function is optimized by one of several optimization methods to give the optimal base station placement. the optimization can be continuous or discrete. continuous optimization methods include the modified steepest descent method and the downhill simplex method, while discrete optimization methods include the hopfield neural network method. the optimization can be performed several times for different initial guess placements to achieve a global, rather than simply local, optimization. the entire optimization process is packaged in the form of an interactive software tool that permits the designer to steer and adjust the solution according to any criteria that the designer may choose.",1999-11-16,"The title of the patent is method and device for placement of transmitters in wireless networks and its abstract is from layout and attenuation data of an area and an initial guess, the placement and power levels of base stations and the resulting attenuation and base station ranges are calculated based on any chosen propagation model. a cost function is calculated which indicates the merit of the initial guess placement. the cost function is a function of the transmitter locations and power levels and can be calculated as a linear combination of the uncovered and interference areas. other cost functions can also be considered. the cost function is optimized by one of several optimization methods to give the optimal base station placement. the optimization can be continuous or discrete. continuous optimization methods include the modified steepest descent method and the downhill simplex method, while discrete optimization methods include the hopfield neural network method. the optimization can be performed several times for different initial guess placements to achieve a global, rather than simply local, optimization. the entire optimization process is packaged in the form of an interactive software tool that permits the designer to steer and adjust the solution according to any criteria that the designer may choose. dated 1999-11-16"
5987397,neural network system for estimation of helicopter gross weight and center of gravity location,the invention is directed to a helicopter health and usage monitoring system utilizing a neural network for estimating gross weight and center of gravity location from measured flight condition parameter inputs; and includes means for measuring a plurality of variable flight condition parameters during flight of the helicopter; memory means for successively receiving and storing parameter input signals as well as estimates of gross weight and center of gravity location; and processing means responsive to the signals received from the measurement means for generating the gross weight and center of gravity location estimates.,1999-11-16,The title of the patent is neural network system for estimation of helicopter gross weight and center of gravity location and its abstract is the invention is directed to a helicopter health and usage monitoring system utilizing a neural network for estimating gross weight and center of gravity location from measured flight condition parameter inputs; and includes means for measuring a plurality of variable flight condition parameters during flight of the helicopter; memory means for successively receiving and storing parameter input signals as well as estimates of gross weight and center of gravity location; and processing means responsive to the signals received from the measurement means for generating the gross weight and center of gravity location estimates. dated 1999-11-16
5987444,robust neutral systems,"a robust neural system for robust processing is disclosed for averting unacceptable or disastrous processing performances. this robust neural system either comprises a neural network or comprises a neural network and at least one range transformer. at least one adjustable weight of the robust neural system is a nonlinear weight of the neural work determined in a nonadaptive training of the robust neural system with respect to a nonadaptive risk-sensitive training criterion. if all the adjustable weights of the robust neural system are nonadaptively adjustable, all these weights are held fixed during the robust neural system's operation. if said neural network is recursive, and the nonadaptive training data used to construct said nonadaptive risk-sensitive training criterion contain data for each of a number of typical values of an environmental parameter, the robust neural system is not only robust but also adaptive to the environmental parameter. if the robust neural system comprises both nonadaptively and adaptively adjustable weights, these adaptively adjustable weights are adjusted by an adaptor in the robust neural system during its operation. such a robust neural system is called a robust adaptive neural system. two types of adaptor are described.",1999-11-16,"The title of the patent is robust neutral systems and its abstract is a robust neural system for robust processing is disclosed for averting unacceptable or disastrous processing performances. this robust neural system either comprises a neural network or comprises a neural network and at least one range transformer. at least one adjustable weight of the robust neural system is a nonlinear weight of the neural work determined in a nonadaptive training of the robust neural system with respect to a nonadaptive risk-sensitive training criterion. if all the adjustable weights of the robust neural system are nonadaptively adjustable, all these weights are held fixed during the robust neural system's operation. if said neural network is recursive, and the nonadaptive training data used to construct said nonadaptive risk-sensitive training criterion contain data for each of a number of typical values of an environmental parameter, the robust neural system is not only robust but also adaptive to the environmental parameter. if the robust neural system comprises both nonadaptively and adaptively adjustable weights, these adaptively adjustable weights are adjusted by an adaptor in the robust neural system during its operation. such a robust neural system is called a robust adaptive neural system. two types of adaptor are described. dated 1999-11-16"
5989811,sextant core biopsy predictive mechanism for non-organ confined disease status,"a method for screening individuals at risk for the loss of organ confinement in prostate cancer is disclosed. the method is useful for evaluating cells from patients at risk for recurrence of prostate cancer following surgery for prostate cancer. specifically, the method uses specific markovian nuclear texture features, alone or in combination with other biomarkers, to determine whether the cancer will progress or lose organ confinement. in addition, methods of predicting the development of fatal metastatic disease by statistical analysis of selected biomarkers is also disclosed. the invention also contemplates a method that uses a neural network to analyze and interpret cell morphology data. utilizing markovian factors and other biomarkers as parameters, the network is first trained with a sets of cell data from known progressors and known non-progressors. the trained network is then used to predict the loss of organ confinement by evaluating patient samples.",1999-11-23,"The title of the patent is sextant core biopsy predictive mechanism for non-organ confined disease status and its abstract is a method for screening individuals at risk for the loss of organ confinement in prostate cancer is disclosed. the method is useful for evaluating cells from patients at risk for recurrence of prostate cancer following surgery for prostate cancer. specifically, the method uses specific markovian nuclear texture features, alone or in combination with other biomarkers, to determine whether the cancer will progress or lose organ confinement. in addition, methods of predicting the development of fatal metastatic disease by statistical analysis of selected biomarkers is also disclosed. the invention also contemplates a method that uses a neural network to analyze and interpret cell morphology data. utilizing markovian factors and other biomarkers as parameters, the network is first trained with a sets of cell data from known progressors and known non-progressors. the trained network is then used to predict the loss of organ confinement by evaluating patient samples. dated 1999-11-23"
5993194,automatically optimized combustion control,"systems and methods are disclosed that optimize the combustion process in various reactors, furnaces, and internal combustion engines. video cameras are used to evaluate the combustion flame grade. depending on the desired form, standard or special video devices, or beam scanning devices, are used to image the combustion flame and by-products. the video device generates and outputs image signals during various phases of, and at various locations in, the combustion process. other forms of sensors monitor and generate data signals defining selected parameters of the combustion process, such as air flow, fuel flow, turbulence, exhaust and inlet valve openings, etc. in a preferred form, a neural networks initially processes the image data and characterizes the combustion flame. a fuzzy logic controller and associated fuzzy logic rule base analyzes the image data from the neural network, along with other sensor information. the fuzzy logic controller determines and generates control signals defining adjustments necessary to optimize the combustion process.",1999-11-30,"The title of the patent is automatically optimized combustion control and its abstract is systems and methods are disclosed that optimize the combustion process in various reactors, furnaces, and internal combustion engines. video cameras are used to evaluate the combustion flame grade. depending on the desired form, standard or special video devices, or beam scanning devices, are used to image the combustion flame and by-products. the video device generates and outputs image signals during various phases of, and at various locations in, the combustion process. other forms of sensors monitor and generate data signals defining selected parameters of the combustion process, such as air flow, fuel flow, turbulence, exhaust and inlet valve openings, etc. in a preferred form, a neural networks initially processes the image data and characterizes the combustion flame. a fuzzy logic controller and associated fuzzy logic rule base analyzes the image data from the neural network, along with other sensor information. the fuzzy logic controller determines and generates control signals defining adjustments necessary to optimize the combustion process. dated 1999-11-30"
5995644,robust and automatic adjustment of display window width and center for mr images,"a system for deriving final display parameters for a wide range of mr images consists of a feature generator utilizing both histogram and spatial information computed from an input mr image, a wavelet transform within the feature generator for compressing the size of the feature vector, a competitive layer based neural network for clustering mr images into different subclasses, a bi-modal linear estimation network and a radial bases function network based non-linear estimator for each subclass, as well as a data fusion system using estimates from both estimators to compute the final display parameters.",1999-11-30,"The title of the patent is robust and automatic adjustment of display window width and center for mr images and its abstract is a system for deriving final display parameters for a wide range of mr images consists of a feature generator utilizing both histogram and spatial information computed from an input mr image, a wavelet transform within the feature generator for compressing the size of the feature vector, a competitive layer based neural network for clustering mr images into different subclasses, a bi-modal linear estimation network and a radial bases function network based non-linear estimator for each subclass, as well as a data fusion system using estimates from both estimators to compute the final display parameters. dated 1999-11-30"
5995651,"image content classification methods, systems and computer programs using texture patterns","image content classification methods, systems and computer programs repeatedly scan an image having an array of image pixels, with at least one random neural network. each scan corresponds to one of multiple texture patterns. a corresponding texture pattern is compared to each of multiple image portions for each of the multiple scans. a value is assigned to each image portion, corresponding to the texture pattern having the highest coincidence. an array of pixels corresponding to the assigned values for the image portions may then be displayed. highly accurate results may be obtained, at high speed, without the need for lengthy expert analysis.",1999-11-30,"The title of the patent is image content classification methods, systems and computer programs using texture patterns and its abstract is image content classification methods, systems and computer programs repeatedly scan an image having an array of image pixels, with at least one random neural network. each scan corresponds to one of multiple texture patterns. a corresponding texture pattern is compared to each of multiple image portions for each of the multiple scans. a value is assigned to each image portion, corresponding to the texture pattern having the highest coincidence. an array of pixels corresponding to the assigned values for the image portions may then be displayed. highly accurate results may be obtained, at high speed, without the need for lengthy expert analysis. dated 1999-11-30"
5995652,pattern searching method using neural networks and correlation,"a pattern searching method using neural networks and correlation. this method combines the quickness and adaptiveness of neural networks with the accuracy of the mathematical correlation approach. images are divided into small sub-images which are presented to the trained neural network. sub-images that may contain the pattern or partial pattern are selected by the neural network. the neural network also provides the approximate location of the pattern, therefore the selected sub-images can be adjusted to contain the complete pattern. desired patterns can be located by measuring the new sub-images' correlation values against the reference models in a small area. experiments show that this superior method is able to find the desired patterns. moreover, this method is much faster than traditional pattern searching methods which use only correlation.",1999-11-30,"The title of the patent is pattern searching method using neural networks and correlation and its abstract is a pattern searching method using neural networks and correlation. this method combines the quickness and adaptiveness of neural networks with the accuracy of the mathematical correlation approach. images are divided into small sub-images which are presented to the trained neural network. sub-images that may contain the pattern or partial pattern are selected by the neural network. the neural network also provides the approximate location of the pattern, therefore the selected sub-images can be adjusted to contain the complete pattern. desired patterns can be located by measuring the new sub-images' correlation values against the reference models in a small area. experiments show that this superior method is able to find the desired patterns. moreover, this method is much faster than traditional pattern searching methods which use only correlation. dated 1999-11-30"
5995669,image processing method and apparatus,"an image processing apparatus includes an input unit for entering a plurality of color image signals, an image processing unit for subjecting the plurality of entered color image signals to processing based upon an algorithm of a cellular neural network, and a comparison decision unit for deciding output color data based upon results of processing by the image processing unit. since input multivalued color image data based upon the algorithm of a neural network are converted to output color image data, it is possible to obtain a high-quality output color image that is faithful to the input color image.",1999-11-30,"The title of the patent is image processing method and apparatus and its abstract is an image processing apparatus includes an input unit for entering a plurality of color image signals, an image processing unit for subjecting the plurality of entered color image signals to processing based upon an algorithm of a cellular neural network, and a comparison decision unit for deciding output color data based upon results of processing by the image processing unit. since input multivalued color image data based upon the algorithm of a neural network are converted to output color image data, it is possible to obtain a high-quality output color image that is faithful to the input color image. dated 1999-11-30"
5995910,method and system for synthesizing vibration data,"a system for training a neural network to synthesize vibration data relating to the operation of a machine. the system includes a first sensor operatively coupleable to the machine, the first sensor adapted to obtain at least one vibration signal relating to the operation of the machine. the system further includes a second sensor operatively coupleable to a power lead of the machine, the second sensor adapted to obtain at least one current signal relating to the operation of a machine. additionally, the system includes a neural network operatively coupleable to the second sensor, the neural network being trainable to generate at least one synthesized vibration signal from the current signal, wherein the synthesized vibration signal is substantially equivalent to the vibration signal obtained from the first sensor.",1999-11-30,"The title of the patent is method and system for synthesizing vibration data and its abstract is a system for training a neural network to synthesize vibration data relating to the operation of a machine. the system includes a first sensor operatively coupleable to the machine, the first sensor adapted to obtain at least one vibration signal relating to the operation of the machine. the system further includes a second sensor operatively coupleable to a power lead of the machine, the second sensor adapted to obtain at least one current signal relating to the operation of a machine. additionally, the system includes a neural network operatively coupleable to the second sensor, the neural network being trainable to generate at least one synthesized vibration signal from the current signal, wherein the synthesized vibration signal is substantially equivalent to the vibration signal obtained from the first sensor. dated 1999-11-30"
5995924,computer-based method and apparatus for classifying statement types based on intonation analysis,"a computer-based method and apparatus for classifying statement types using intonation analysis. the method and apparatus identify a user's potential query when the user responds to information during dialog with an automated dialog system. pitch information is extracted, via a cepstrum, from the speech signal. in one embodiment, the pitch intonation is processed to form a smoothed pitch or intonation contour. then the smoothed pitch contour is processed by a set of shape detectors and this output, together with statistical information, is sent to a rule-based algorithm which attempts to classify the statement type. in another embodiment, the smoothed pitch contour is processed by a pattern recognition system such as a neural network trained with a back-propagation learning algorithm.",1999-11-30,"The title of the patent is computer-based method and apparatus for classifying statement types based on intonation analysis and its abstract is a computer-based method and apparatus for classifying statement types using intonation analysis. the method and apparatus identify a user's potential query when the user responds to information during dialog with an automated dialog system. pitch information is extracted, via a cepstrum, from the speech signal. in one embodiment, the pitch intonation is processed to form a smoothed pitch or intonation contour. then the smoothed pitch contour is processed by a set of shape detectors and this output, together with statistical information, is sent to a rule-based algorithm which attempts to classify the statement type. in another embodiment, the smoothed pitch contour is processed by a pattern recognition system such as a neural network trained with a back-propagation learning algorithm. dated 1999-11-30"
5995952,neural network for providing hints to problem solving apparatus using tree search method,"a problem solving unit obtains a solution in a symbol process in response to a given problem. a neural network learning control unit makes a neural network unit perform a learning process on a solution output from the problem solving unit. after completing the learning process in response to the given problem, the neural network unit provides an output as a hint on solving the problem to the problem solving unit.",1999-11-30,"The title of the patent is neural network for providing hints to problem solving apparatus using tree search method and its abstract is a problem solving unit obtains a solution in a symbol process in response to a given problem. a neural network learning control unit makes a neural network unit perform a learning process on a solution output from the problem solving unit. after completing the learning process in response to the given problem, the neural network unit provides an output as a hint on solving the problem to the problem solving unit. dated 1999-11-30"
5995954,method and apparatus for associative memory,"a method and apparatus for an electronic artificial neural network, which serves as an associative memory that has a complete set of n-dimensional hadamard vectors as stored states, suitable for large n that are powers of 2. the neural net has nonlinear synapses, each of which processes signals from two neurons. these synapses can be implemented by simple passive circuits comprised of eight resistors and four diodes. the connections in the neural net are specified through a subset of a group that is defined over the integers from 1 to n. the subset is chosen such that the connections can be implemented in vlsi or wafer scale integration. an extension of the hadamard memory causes the memory to provide new hadamard vectors when these are needed for the purpose of hebb learning.",1999-11-30,"The title of the patent is method and apparatus for associative memory and its abstract is a method and apparatus for an electronic artificial neural network, which serves as an associative memory that has a complete set of n-dimensional hadamard vectors as stored states, suitable for large n that are powers of 2. the neural net has nonlinear synapses, each of which processes signals from two neurons. these synapses can be implemented by simple passive circuits comprised of eight resistors and four diodes. the connections in the neural net are specified through a subset of a group that is defined over the integers from 1 to n. the subset is chosen such that the connections can be implemented in vlsi or wafer scale integration. an extension of the hadamard memory causes the memory to provide new hadamard vectors when these are needed for the purpose of hebb learning. dated 1999-11-30"
5999638,"method and apparatus for adjusting read-out conditions and/or image processing conditions for radiation images, radiation image read-out apparatus, and radiation image analyzing method and apparatus","a first image signal representing a radiation image of an object is obtained by exposing a stimulable phosphor sheet, on which the radiation image has been stored, to stimulating rays, which cause the stimulable phosphor sheet to emit light in proportion to the amount of energy stored thereon during its exposure to radiation, the emitted light being detected. a second image signal representing the radiation image is thereafter obtained by again exposing the stimulable phosphor sheet to stimulating rays, the light emitted by the stimulable phosphor sheet being detected. read-out conditions, under which the second image signal is to be obtained, and/or image processing conditions, under which the second image signal having been obtained is to be image processed, are adjusted on the basis of the first image signal. a storage device stores information representing a standard pattern of radiation images. a signal transforming device transforms the first image signal representing the radiation image into a transformed image signal representing the radiation image, which has been transformed into the standard pattern. a condition adjuster is provided with a neural network, which receives the transformed image signal and feeds out information representing the read-out conditions and/or the image processing conditions.",1999-12-07,"The title of the patent is method and apparatus for adjusting read-out conditions and/or image processing conditions for radiation images, radiation image read-out apparatus, and radiation image analyzing method and apparatus and its abstract is a first image signal representing a radiation image of an object is obtained by exposing a stimulable phosphor sheet, on which the radiation image has been stored, to stimulating rays, which cause the stimulable phosphor sheet to emit light in proportion to the amount of energy stored thereon during its exposure to radiation, the emitted light being detected. a second image signal representing the radiation image is thereafter obtained by again exposing the stimulable phosphor sheet to stimulating rays, the light emitted by the stimulable phosphor sheet being detected. read-out conditions, under which the second image signal is to be obtained, and/or image processing conditions, under which the second image signal having been obtained is to be image processed, are adjusted on the basis of the first image signal. a storage device stores information representing a standard pattern of radiation images. a signal transforming device transforms the first image signal representing the radiation image into a transformed image signal representing the radiation image, which has been transformed into the standard pattern. a condition adjuster is provided with a neural network, which receives the transformed image signal and feeds out information representing the read-out conditions and/or the image processing conditions. dated 1999-12-07"
5999639,method and system for automated detection of clustered microcalcifications from digital mammograms,"a method and system for detecting and displaying clustered microcalcifications in a digital mammogram, wherein a single digital mammogram is first automatically cropped to a breast area sub-image which is then processed by means of an optimized difference of gaussians filter to enhance the appearance of potential microcalcifications in the sub-image. the potential microcalcifications are thresholded, clusters are detected, features are computed for the detected clusters, and the clusters are classified as either suspicious or not suspicious by means of a neural network. thresholding is preferably by sloping local thresholding, but may also be performed by global and dual-local thresholding. the locations in the original digital mammogram of the suspicious detected clustered microcalcifications are indicated. parameters for use in the detection and thresholding portions of the system are computer-optimized by means of a genetic algorithm. the results of the system are optimally combined with a radiologist's observation of the original mammogram by combining the observations with the results, after the radiologist has first accepted or rejected individual detections reported by the system.",1999-12-07,"The title of the patent is method and system for automated detection of clustered microcalcifications from digital mammograms and its abstract is a method and system for detecting and displaying clustered microcalcifications in a digital mammogram, wherein a single digital mammogram is first automatically cropped to a breast area sub-image which is then processed by means of an optimized difference of gaussians filter to enhance the appearance of potential microcalcifications in the sub-image. the potential microcalcifications are thresholded, clusters are detected, features are computed for the detected clusters, and the clusters are classified as either suspicious or not suspicious by means of a neural network. thresholding is preferably by sloping local thresholding, but may also be performed by global and dual-local thresholding. the locations in the original digital mammogram of the suspicious detected clustered microcalcifications are indicated. parameters for use in the detection and thresholding portions of the system are computer-optimized by means of a genetic algorithm. the results of the system are optimally combined with a radiologist's observation of the original mammogram by combining the observations with the results, after the radiologist has first accepted or rejected individual detections reported by the system. dated 1999-12-07"
5999643,switched-current type of hamming neural network system for pattern recognition,"disclosed is a two-layer switched-current type of a hamming neural network system. this hamming network system includes a matching rate computation circuit for modules on a first layer used to compute a matching rate between a to-be-identified pattern and each one of a plurality of standard patterns, a matching rate comparison circuit on a second layer for ranking an order of the matching rates including a switched-current type order-ranking circuit for receiving switched-current signals, finding a maximum value and outputting a time-division order-ranking output and an identification-rejection judgment circuit for performing an absolute and a relative judgment, and a pulse-generating circuit for generating sequential clock pulses, in which the circuit construction of the hamming network is simple and flexible due to a modular design with extendible circuit dimensions, and a high precision, improved performance and enhanced reliability of the network system is achieved. in addition, the hamming network can be easily integrated in a mixed analog/digital form due to a direct usage of a standard complementary metal-oxide-semiconductor device by using a switched-current technique.",1999-12-07,"The title of the patent is switched-current type of hamming neural network system for pattern recognition and its abstract is disclosed is a two-layer switched-current type of a hamming neural network system. this hamming network system includes a matching rate computation circuit for modules on a first layer used to compute a matching rate between a to-be-identified pattern and each one of a plurality of standard patterns, a matching rate comparison circuit on a second layer for ranking an order of the matching rates including a switched-current type order-ranking circuit for receiving switched-current signals, finding a maximum value and outputting a time-division order-ranking output and an identification-rejection judgment circuit for performing an absolute and a relative judgment, and a pulse-generating circuit for generating sequential clock pulses, in which the circuit construction of the hamming network is simple and flexible due to a modular design with extendible circuit dimensions, and a high precision, improved performance and enhanced reliability of the network system is achieved. in addition, the hamming network can be easily integrated in a mixed analog/digital form due to a direct usage of a standard complementary metal-oxide-semiconductor device by using a switched-current technique. dated 1999-12-07"
5999846,physiological monitoring,"an insomnia or vigilance monitor comprising one or more electrodes (1a,1b) for obtaining an electrical signal from a subject over a period of epochs, the electrical signal being related to the sleep or wakefulness stage type being experienced by the subject; and a processor (5) adapted to analyze the electrical signal and assign a sleep or wakefulness stage type to each epoch to generate a hypnogram. methods of monitoring sleep or vigilance using the mastoid site are also disclosed. further disclosures relate to a method of training and testing a first neural network for use in a physiological monitor, and a method of assigning a class to an epoch of a physiological signal obtained from a subject as a set of samples.",1999-12-07,"The title of the patent is physiological monitoring and its abstract is an insomnia or vigilance monitor comprising one or more electrodes (1a,1b) for obtaining an electrical signal from a subject over a period of epochs, the electrical signal being related to the sleep or wakefulness stage type being experienced by the subject; and a processor (5) adapted to analyze the electrical signal and assign a sleep or wakefulness stage type to each epoch to generate a hypnogram. methods of monitoring sleep or vigilance using the mastoid site are also disclosed. further disclosures relate to a method of training and testing a first neural network for use in a physiological monitor, and a method of assigning a class to an epoch of a physiological signal obtained from a subject as a set of samples. dated 1999-12-07"
5999922,neuroprocessing service,"a neuroprocessing center executes a neuroprocessing using a neurocomputer. the neuroprocessing center is a public facility available for a user having a user terminal and executes the neuroprocessing as requested by the user. the user is given a result of the neuroprocessing. it is unnecessary for a user to have a computer implementing the neural network and anyone can participate in the profits of the neuroprocessing service. examples of neuroprocessing service achieved by the neural network are graphic pattern generating services, character recognition services, sound synthesizing services, etc. a result of the neuroprocessing is effectively used by the user.",1999-12-07,"The title of the patent is neuroprocessing service and its abstract is a neuroprocessing center executes a neuroprocessing using a neurocomputer. the neuroprocessing center is a public facility available for a user having a user terminal and executes the neuroprocessing as requested by the user. the user is given a result of the neuroprocessing. it is unnecessary for a user to have a computer implementing the neural network and anyone can participate in the profits of the neuroprocessing service. examples of neuroprocessing service achieved by the neural network are graphic pattern generating services, character recognition services, sound synthesizing services, etc. a result of the neuroprocessing is effectively used by the user. dated 1999-12-07"
6000827,system identifying device and adaptive learning control device,"a system identifying device precisely represents as a mathematical model the features of a system such as a robot manipulator, various industrial plants, etc., and can be operated as hardware by identifying an object system over a neural network, thereby successfully identifying a non-linear system as well as a linear system. furthermore, an adaptive control device can perform an online learning by using the identifying device to obtain a teaching signal to be used in operating the identifying device in an adaptive control device.",1999-12-14,"The title of the patent is system identifying device and adaptive learning control device and its abstract is a system identifying device precisely represents as a mathematical model the features of a system such as a robot manipulator, various industrial plants, etc., and can be operated as hardware by identifying an object system over a neural network, thereby successfully identifying a non-linear system as well as a linear system. furthermore, an adaptive control device can perform an online learning by using the identifying device to obtain a teaching signal to be used in operating the identifying device in an adaptive control device. dated 1999-12-14"
6002985,method of controlling development of an oil or gas reservoir,"a method controlling development of an oil or gas reservoir uses a neural network and genetic algorithm program to define a neural network topology and the optimal inputs for that topology. the topology is defined from identified and selected (1) parameters associated with the formation or formations in which actual wells are drilled in the reservoir and (2) parameters associated with the drilling, completion and stimulation of those wells and (3) parameters associated with the oil or gas production from the wells. subsequent drilling, completion and stimulation of the reservoir is determined and applied based on hypothetical alternatives input to the topology and resulting outputs.",1999-12-14,"The title of the patent is method of controlling development of an oil or gas reservoir and its abstract is a method controlling development of an oil or gas reservoir uses a neural network and genetic algorithm program to define a neural network topology and the optimal inputs for that topology. the topology is defined from identified and selected (1) parameters associated with the formation or formations in which actual wells are drilled in the reservoir and (2) parameters associated with the drilling, completion and stimulation of those wells and (3) parameters associated with the oil or gas production from the wells. subsequent drilling, completion and stimulation of the reservoir is determined and applied based on hypothetical alternatives input to the topology and resulting outputs. dated 1999-12-14"
6003003,speech recognition system having a quantizer using a single robust codebook designed at multiple signal to noise ratios,"in one embodiment, a speech recognition system is organized with a fuzzy matrix quantizer with a single codebook representing u codewords. the single codebook is designed with entries from u codebooks which are designed with respective words at multiple signal to noise ratio levels. such entries are, in one embodiment, centroids of clustered training data. the training data is, in one embodiment, derived from line spectral frequency pairs representing respective speech input signals at various signal to noise ratios. the single codebook trained in this manner provides a codebook for a robust front end speech processor, such as the fuzzy matrix quantizer, for training a speech classifier such as a u hidden markov models and a speech post classifier such as a neural network. in one embodiment, a fuzzy viterbi algorithm is used with the hidden markov models to describe the speech input signal probabilistically.",1999-12-14,"The title of the patent is speech recognition system having a quantizer using a single robust codebook designed at multiple signal to noise ratios and its abstract is in one embodiment, a speech recognition system is organized with a fuzzy matrix quantizer with a single codebook representing u codewords. the single codebook is designed with entries from u codebooks which are designed with respective words at multiple signal to noise ratio levels. such entries are, in one embodiment, centroids of clustered training data. the training data is, in one embodiment, derived from line spectral frequency pairs representing respective speech input signals at various signal to noise ratios. the single codebook trained in this manner provides a codebook for a robust front end speech processor, such as the fuzzy matrix quantizer, for training a speech classifier such as a u hidden markov models and a speech post classifier such as a neural network. in one embodiment, a fuzzy viterbi algorithm is used with the hidden markov models to describe the speech input signal probabilistically. dated 1999-12-14"
6004267,method for diagnosing and staging prostate cancer,"the subject invention provides a method for diagnosing prostate cancer and determining preoperatively the pathological stage in patients with prostate cancer. the methods described herein can be used for prediction of margin positivity, seminal vesicle (s.v.) involvement, and lymph nodal (l.n.) involvement in patients with clinically localized prostate cancer. the method includes use of a neural network which provides prostate cancer stage information for a patient based upon input data which includes the patient's preoperative serum psa, biopsy gleason score, and systemic biopsy-based information. its positive predictive value (ppv), negative predictive value (npv), and accuracy are superior to that of current nomograms in use. use of this method can result in enormous cost savings by accurately diagnosing patients with prostate cancer and by avoiding multiple imaging tests and expensive surgery in unindicated patients.",1999-12-21,"The title of the patent is method for diagnosing and staging prostate cancer and its abstract is the subject invention provides a method for diagnosing prostate cancer and determining preoperatively the pathological stage in patients with prostate cancer. the methods described herein can be used for prediction of margin positivity, seminal vesicle (s.v.) involvement, and lymph nodal (l.n.) involvement in patients with clinically localized prostate cancer. the method includes use of a neural network which provides prostate cancer stage information for a patient based upon input data which includes the patient's preoperative serum psa, biopsy gleason score, and systemic biopsy-based information. its positive predictive value (ppv), negative predictive value (npv), and accuracy are superior to that of current nomograms in use. use of this method can result in enormous cost savings by accurately diagnosing patients with prostate cancer and by avoiding multiple imaging tests and expensive surgery in unindicated patients. dated 1999-12-21"
6009185,neural network based contact state estimator,"a method is described for providing an estimate of the state of a stationary or moving contact in a three dimensional ocean. the method comprises the steps of collecting information about a location of an observer during a sequence of time, information from at least one sensor about a position of the contact relative to the observer during the time sequence, and a knowledge vector. transforming the information into a series of three dimensional geographical grids. examining the grids to identify hypothesized contact paths and analyzing the hypothesized contact paths to produce an estimate of the state of the contact with respect to the location of the observer. a device for providing the estimate of the state of a stationary or moving contact includes a grid stimulation block for transforming the collected information into the three dimensional geographical grids. a fusion block where grids corresponding to similar time intervals are combined or fused. a correlation block for identifying possible contact paths and for producing path likelihood vectors and an estimation block for providing the estimate of the state of the moving contact. the device further includes a controller for providing knowledge to the aforementioned blocks.",1999-12-28,"The title of the patent is neural network based contact state estimator and its abstract is a method is described for providing an estimate of the state of a stationary or moving contact in a three dimensional ocean. the method comprises the steps of collecting information about a location of an observer during a sequence of time, information from at least one sensor about a position of the contact relative to the observer during the time sequence, and a knowledge vector. transforming the information into a series of three dimensional geographical grids. examining the grids to identify hypothesized contact paths and analyzing the hypothesized contact paths to produce an estimate of the state of the contact with respect to the location of the observer. a device for providing the estimate of the state of a stationary or moving contact includes a grid stimulation block for transforming the collected information into the three dimensional geographical grids. a fusion block where grids corresponding to similar time intervals are combined or fused. a correlation block for identifying possible contact paths and for producing path likelihood vectors and an estimation block for providing the estimate of the state of the moving contact. the device further includes a controller for providing knowledge to the aforementioned blocks. dated 1999-12-28"
6009418,method and apparatus for neural networking using semantic attractor architecture,"a semantic attractor memory uses an evolving neural network architecture and learning rules derived from the study of human language acquisition and change to store, process and retrieve information. the architecture is based on multiple layer channels, with random connections from one layer to the next. one or more layers are devoted to processing input information. at least one processing layer is provided. one or more layers are devoted to processing outputs and feedback is provided from the outputs back to the processing layer or layers. inputs from parallel channels are also provided to the one or more processing layers. with the exception of the feedback loop and central processing layers, the network is feedforward unless it is employed in a hybrid back-propagation configuration. the learning rules are based on non-stationary statistical processes, such as the polya process or the processes leading to bose-einstein statistics, again derived from considerations of human language acquisition. the invention provides rapid, unsupervised processing of complex data sets, such as imagery or continuous human speech, and a means to capture successful processing or pattern classification constellations for implementation in other networks.",1999-12-28,"The title of the patent is method and apparatus for neural networking using semantic attractor architecture and its abstract is a semantic attractor memory uses an evolving neural network architecture and learning rules derived from the study of human language acquisition and change to store, process and retrieve information. the architecture is based on multiple layer channels, with random connections from one layer to the next. one or more layers are devoted to processing input information. at least one processing layer is provided. one or more layers are devoted to processing outputs and feedback is provided from the outputs back to the processing layer or layers. inputs from parallel channels are also provided to the one or more processing layers. with the exception of the feedback loop and central processing layers, the network is feedforward unless it is employed in a hybrid back-propagation configuration. the learning rules are based on non-stationary statistical processes, such as the polya process or the processes leading to bose-einstein statistics, again derived from considerations of human language acquisition. the invention provides rapid, unsupervised processing of complex data sets, such as imagery or continuous human speech, and a means to capture successful processing or pattern classification constellations for implementation in other networks. dated 1999-12-28"
6009419,method for predicting cement properties,"a method of predicting a desired property, such as thickening time, of a cement slurry comprising measuring or determining other properties of a slurry composition and obtaining a predicted value of the desired property by applying values representative of the other properties to a model formed by determining the other properties for a series of cement compositions and correlating these with measured values of the desired property. in one embodiment the method comprises measuring and determining the other properties and inputting values corresponding to these properties to a neural network device configured to output a value representative of the desired property, the neural network device being a) configured to utilize each of said values as input values, b) provided with at least one hidden layer of at least one node, and c) trained with a dataset comprising series of values of said properties and a value corresponding to the desired property when measured for a slurry having the measured properties.",1999-12-28,"The title of the patent is method for predicting cement properties and its abstract is a method of predicting a desired property, such as thickening time, of a cement slurry comprising measuring or determining other properties of a slurry composition and obtaining a predicted value of the desired property by applying values representative of the other properties to a model formed by determining the other properties for a series of cement compositions and correlating these with measured values of the desired property. in one embodiment the method comprises measuring and determining the other properties and inputting values corresponding to these properties to a neural network device configured to output a value representative of the desired property, the neural network device being a) configured to utilize each of said values as input values, b) provided with at least one hidden layer of at least one node, and c) trained with a dataset comprising series of values of said properties and a value corresponding to the desired property when measured for a slurry having the measured properties. dated 1999-12-28"
6010604,neural network packing,"a column packing for use in the scrubbing of gases by aqueous liquid, the packing comprising material that is electrically conductive and material that is non-conductive with such materials being intimately mixed with each other, such that the packing as a whole provides the gas-liquid surface for absorption and the conductive material in particular serves as a bipolar electrode for electrolysis.",2000-01-04,"The title of the patent is neural network packing and its abstract is a column packing for use in the scrubbing of gases by aqueous liquid, the packing comprising material that is electrically conductive and material that is non-conductive with such materials being intimately mixed with each other, such that the packing as a whole provides the gas-liquid surface for absorption and the conductive material in particular serves as a bipolar electrode for electrolysis. dated 2000-01-04"
6011295,neural network active pixel cell,""" an active pixel image cell which includes a photosensor, active devices for control of the sensor and readout of a signal representing the intensity of light to which the sensor is exposed, and a neuron mosfet transistor which """"both amplifies the signal from the photosensor and"""" simulates the behavior of a human neuron. an integrated neural network and imaging array may be formed by interconnecting a group of such pixels. digital signal processing algorithms used for image processing may be implemented at the pixel level by appropriate interconnections between the output signals from the photosensor of surrounding pixels and the neuron mosfet. """,2000-01-04,"The title of the patent is neural network active pixel cell and its abstract is "" an active pixel image cell which includes a photosensor, active devices for control of the sensor and readout of a signal representing the intensity of light to which the sensor is exposed, and a neuron mosfet transistor which """"both amplifies the signal from the photosensor and"""" simulates the behavior of a human neuron. an integrated neural network and imaging array may be formed by interconnecting a group of such pixels. digital signal processing algorithms used for image processing may be implemented at the pixel level by appropriate interconnections between the output signals from the photosensor of surrounding pixels and the neuron mosfet. "" dated 2000-01-04"
6011557,method for obtaining a representation of a geological structure,"a method for obtaining a representation of the textures of a geological structure, characterized in that images characteristic of the sedimentology of the environment are formed, parameters corresponding to the nature of the images are estimated at every point of each image and in a spatial domain around the point so as to determine a texture vector for each of the points and to obtain a set of texture vectors. the method also includes the steps of selecting texture vectors representative of the characteristic textures of the geological environment in the set of texture vectors; and using a neural network formed of cells distributed in two dimensions which contains as many cells as characteristic textures. the selected texture vectors are used to submit the neural network to a learning process so that a final topology map of the textures characteristic of the geological environment is obtained.",2000-01-04,"The title of the patent is method for obtaining a representation of a geological structure and its abstract is a method for obtaining a representation of the textures of a geological structure, characterized in that images characteristic of the sedimentology of the environment are formed, parameters corresponding to the nature of the images are estimated at every point of each image and in a spatial domain around the point so as to determine a texture vector for each of the points and to obtain a set of texture vectors. the method also includes the steps of selecting texture vectors representative of the characteristic textures of the geological environment in the set of texture vectors; and using a neural network formed of cells distributed in two dimensions which contains as many cells as characteristic textures. the selected texture vectors are used to submit the neural network to a learning process so that a final topology map of the textures characteristic of the geological environment is obtained. dated 2000-01-04"
6011862,computer-aided method for automated image feature analysis and diagnosis of digitized medical images,"a computerized method for the detection and characterization of disease in an image derived from a chest radiograph, wherein an image in the chest radiograph is processed to determine the ribcage boundary, including lung top edges, right and left ribcage edges, and right and left hemidiaphragm edges. texture measures including rms variations of pixel values within regions of interest are converted to relative exposures and corrected for system noise existing in the system used to produce the image. texture and/or geometric pattern indices are produced. a histogram(s) of the produced index (indices) is produced and values of the histograms) are applied as inputs to a trained artificial neural network, which classifies the image as normal or abnormal. in one embodiment, obviously normal and obviously abnormal images are determined based on the ratio of abnormal regions of interest to the total number of regions of interest in a rule-based method, so that only difficult cases to diagnose are applied to the artificial neural network.",2000-01-04,"The title of the patent is computer-aided method for automated image feature analysis and diagnosis of digitized medical images and its abstract is a computerized method for the detection and characterization of disease in an image derived from a chest radiograph, wherein an image in the chest radiograph is processed to determine the ribcage boundary, including lung top edges, right and left ribcage edges, and right and left hemidiaphragm edges. texture measures including rms variations of pixel values within regions of interest are converted to relative exposures and corrected for system noise existing in the system used to produce the image. texture and/or geometric pattern indices are produced. a histogram(s) of the produced index (indices) is produced and values of the histograms) are applied as inputs to a trained artificial neural network, which classifies the image as normal or abnormal. in one embodiment, obviously normal and obviously abnormal images are determined based on the ratio of abnormal regions of interest to the total number of regions of interest in a rule-based method, so that only difficult cases to diagnose are applied to the artificial neural network. dated 2000-01-04"
6011865,hybrid on-line handwriting recognition and optical character recognition system,"the hybrid handwriting recognition method includes the steps of (a), in response to a handwriting input from a user, providing dynamic, time ordered stroke information; (b) determining a first list comprised of at least one probable character that the dynamic, time ordered stroke information is intended to represent; (c) converting the dynamic, time ordered stroke information to static stroke information; (d) determining a second list comprised of at least one probable character that the static stroke information represents; and (e) merging the first list and the second list to provide a third, unified list comprised of at least one element representing a most probable character that the dynamic, time ordered stroke information is intended to represent. the step of converting includes the steps of generating a static, bit-mapped representation of the dynamic stroke information, and generating one or more first stroke features based on contour directions of the bit-mapped stroke information. the first stroke features are applied as inputs to a plurality of neural network recognizers.",2000-01-04,"The title of the patent is hybrid on-line handwriting recognition and optical character recognition system and its abstract is the hybrid handwriting recognition method includes the steps of (a), in response to a handwriting input from a user, providing dynamic, time ordered stroke information; (b) determining a first list comprised of at least one probable character that the dynamic, time ordered stroke information is intended to represent; (c) converting the dynamic, time ordered stroke information to static stroke information; (d) determining a second list comprised of at least one probable character that the static stroke information represents; and (e) merging the first list and the second list to provide a third, unified list comprised of at least one element representing a most probable character that the dynamic, time ordered stroke information is intended to represent. the step of converting includes the steps of generating a static, bit-mapped representation of the dynamic stroke information, and generating one or more first stroke features based on contour directions of the bit-mapped stroke information. the first stroke features are applied as inputs to a plurality of neural network recognizers. dated 2000-01-04"
6014452,method and system for using local attention in the detection of abnormalities in digitized medical images,"a method an system for using a local attention threshold to aid in the detection of clustered abnormalities in digitized medical images is disclosed. the local attention threshold is applied to locate spots within a predetermined distance from previously identified spots. more specifically, seed pixels are identified by applying a first seed threshold function to the output of a shift-invariant neural network and adaptive threshold. the seed pixels are then segmented into spots by applying a segmentation threshold function to each seed pixel. false-positive spots are removed using various techniques. additional seed pixels are then identified by applying a local attention threshold to pixels within a predetermined distance to previously identified spots. the local attention threshold disclosed is less selective for pixels which are closer to the nearest spot than for pixels which are further from the nearest spot. the new seed pixels are then segmented into spots, and potential abnormalities are identified in the medical image based in part on the closeness of the identified spots.",2000-01-11,"The title of the patent is method and system for using local attention in the detection of abnormalities in digitized medical images and its abstract is a method an system for using a local attention threshold to aid in the detection of clustered abnormalities in digitized medical images is disclosed. the local attention threshold is applied to locate spots within a predetermined distance from previously identified spots. more specifically, seed pixels are identified by applying a first seed threshold function to the output of a shift-invariant neural network and adaptive threshold. the seed pixels are then segmented into spots by applying a segmentation threshold function to each seed pixel. false-positive spots are removed using various techniques. additional seed pixels are then identified by applying a local attention threshold to pixels within a predetermined distance to previously identified spots. the local attention threshold disclosed is less selective for pixels which are closer to the nearest spot than for pixels which are further from the nearest spot. the new seed pixels are then segmented into spots, and potential abnormalities are identified in the medical image based in part on the closeness of the identified spots. dated 2000-01-11"
6014652,object classification and identification system,an object classification and identification system including stored discriminator quantities indicative of the surface characteristics of known objects; an arm coupled to a detector which provides a voltage signal which varies relative to the surface characteristics of a detected object as the arm moves over the detected object; a computer programmed to calculate a plurality of discriminator quantities indicative of the surface characteristics of the detected object based on the voltage signal; and a neural network used for matching the calculated discriminator quantities of the detected object with the stored discriminator quantities indicative of the surface characteristics of known object.,2000-01-11,The title of the patent is object classification and identification system and its abstract is an object classification and identification system including stored discriminator quantities indicative of the surface characteristics of known objects; an arm coupled to a detector which provides a voltage signal which varies relative to the surface characteristics of a detected object as the arm moves over the detected object; a computer programmed to calculate a plurality of discriminator quantities indicative of the surface characteristics of the detected object based on the voltage signal; and a neural network used for matching the calculated discriminator quantities of the detected object with the stored discriminator quantities indicative of the surface characteristics of known object. dated 2000-01-11
6014653,non-algorithmically implemented artificial neural networks and components thereof,"constructing and simulating artificial neural networks and components thereof within a spreadsheet environment results in user friendly neural networks which do not require algorithmic based software in order to train or operate. such neural networks can be easily cascaded to form complex neural networks and neural network systems, including neural networks capable of self-organizing so as to self-train within a spreadsheet, neural networks which train simultaneously within a spreadsheet, and neural networks capable of autonomously moving, monitoring, analyzing, and altering data within a spreadsheet. neural networks can also be cascaded together in self training neural network form to achieve a device prototyping system.",2000-01-11,"The title of the patent is non-algorithmically implemented artificial neural networks and components thereof and its abstract is constructing and simulating artificial neural networks and components thereof within a spreadsheet environment results in user friendly neural networks which do not require algorithmic based software in order to train or operate. such neural networks can be easily cascaded to form complex neural networks and neural network systems, including neural networks capable of self-organizing so as to self-train within a spreadsheet, neural networks which train simultaneously within a spreadsheet, and neural networks capable of autonomously moving, monitoring, analyzing, and altering data within a spreadsheet. neural networks can also be cascaded together in self training neural network form to achieve a device prototyping system. dated 2000-01-11"
6016154,image forming apparatus,"an image forming apparatus, such as a laser printer, ink jet printer, or a thermal transfer, which comprises a neural network. the apparatus improves image quality by reducing a zigzag included, e.g., in input image data, reducing the circuit size by reducing the number of bits having small weights for an input combination, and concurrently correcting the size of center, left and right picture elements in a predetermined window. the neural network outputs correction data for the size and position of the center dot in a window in response to an input of dot image data in a window or subdot pattern exposure data for the center dot. additionally, the neural network uses any of the three values, +1, -1 and 0, for the coefficient of input combination for a hidden layer neuron, e.g., after a teacher pattern is learned. further, the neural network outputs correction data for 3.times.n subblocks of picture elements obtained by dividing blocks of picture elements in the center, left and right of a window.",2000-01-18,"The title of the patent is image forming apparatus and its abstract is an image forming apparatus, such as a laser printer, ink jet printer, or a thermal transfer, which comprises a neural network. the apparatus improves image quality by reducing a zigzag included, e.g., in input image data, reducing the circuit size by reducing the number of bits having small weights for an input combination, and concurrently correcting the size of center, left and right picture elements in a predetermined window. the neural network outputs correction data for the size and position of the center dot in a window in response to an input of dot image data in a window or subdot pattern exposure data for the center dot. additionally, the neural network uses any of the three values, +1, -1 and 0, for the coefficient of input combination for a hidden layer neuron, e.g., after a teacher pattern is learned. further, the neural network outputs correction data for 3.times.n subblocks of picture elements obtained by dividing blocks of picture elements in the center, left and right of a window. dated 2000-01-18"
6016384,method for speeding up the convergence of the back-propagation algorithm applied to realize the learning process in a neural network of the multilayer perceptron type,a method for speeding up the convergence of the back-propagation algorithm applied to realize the learning process in a neural network of the multilayer perceptron type intended for instance to recognize a set of samples. the method comprises a first stage based upon the elementary concept of progressively increasing the capability for learning of the network by progressively adding new samples as they are recognized by the network to a starting set of learning samples; a second stage based upon the concept of progressively increasing the learning capabilities of the network by progressively adding not previously recognized samples; and a third stage based upon the concept of progressively increasing the learning capabilities of the network by progressive corruption in the meaning of the assimilation between recognized samples and not recognized samples and their subsequent exposure to the network.,2000-01-18,The title of the patent is method for speeding up the convergence of the back-propagation algorithm applied to realize the learning process in a neural network of the multilayer perceptron type and its abstract is a method for speeding up the convergence of the back-propagation algorithm applied to realize the learning process in a neural network of the multilayer perceptron type intended for instance to recognize a set of samples. the method comprises a first stage based upon the elementary concept of progressively increasing the capability for learning of the network by progressively adding new samples as they are recognized by the network to a starting set of learning samples; a second stage based upon the concept of progressively increasing the learning capabilities of the network by progressively adding not previously recognized samples; and a third stage based upon the concept of progressively increasing the learning capabilities of the network by progressive corruption in the meaning of the assimilation between recognized samples and not recognized samples and their subsequent exposure to the network. dated 2000-01-18
6016763,submersible unit and diving position control method therefor,"a submersible unit having thrusters for changing a diving position based on a total work quantity, which is the sum of a first work quantity and a second work quantity. a proportional controller generates and outputs the first work quantity based on a difference between a position quantity indicating a desired target diving position and a diving position. a network controller uses a neural network data processing system to learn movement characteristics of the submersible unit based on the first work quantity and a diving position sampled over a plurality of times. the network controller generates a second work quantity using control coupling coefficients learned by minimizing an evaluation quantity determined from a difference between the learned movement characteristics and target movement characteristic values.",2000-01-25,"The title of the patent is submersible unit and diving position control method therefor and its abstract is a submersible unit having thrusters for changing a diving position based on a total work quantity, which is the sum of a first work quantity and a second work quantity. a proportional controller generates and outputs the first work quantity based on a difference between a position quantity indicating a desired target diving position and a diving position. a network controller uses a neural network data processing system to learn movement characteristics of the submersible unit based on the first work quantity and a diving position sampled over a plurality of times. the network controller generates a second work quantity using control coupling coefficients learned by minimizing an evaluation quantity determined from a difference between the learned movement characteristics and target movement characteristic values. dated 2000-01-25"
6018696,learning type position determining device,"the present invention relates to a neural network system for determining a position of a mobile object such as a robot, automobile, etc., which comprises a return information processing unit for generating return information corresponding to a relative position of the mobile object from the origin in a move space according to first sensor signal from the first sensor; a sense information processing unit for generating sense information corresponding to an environment state surrounding the moblie object at the relative position according to a second sensor signal from the second sensor; and an information integrating unit for inferring return information corresponding to sense information output from the sense information processing unit after learning a correlation between the return information and the sense information, and integrating the two items of information.",2000-01-25,"The title of the patent is learning type position determining device and its abstract is the present invention relates to a neural network system for determining a position of a mobile object such as a robot, automobile, etc., which comprises a return information processing unit for generating return information corresponding to a relative position of the mobile object from the origin in a move space according to first sensor signal from the first sensor; a sense information processing unit for generating sense information corresponding to an environment state surrounding the moblie object at the relative position according to a second sensor signal from the second sensor; and an information integrating unit for inferring return information corresponding to sense information output from the sense information processing unit after learning a correlation between the return information and the sense information, and integrating the two items of information. dated 2000-01-25"
6018727,device for the autonomous generation of useful information,"a device for generating useful information employing a first neural network trained to produce input-output maps within a predetermined initial knowledge domain, an apparatus for subjecting the neural network to perturbations which may produce changes in the predetermined knowledge domain, the neural network having an optional output for feeding the outputs of the first neural network to a second neural network that evaluates the outputs based on training within the second neural network. the device may also include a reciprocal feed back connection from the output of the second neural network to the first neural network to further influence and change what takes place in the aforesaid neural network.",2000-01-25,"The title of the patent is device for the autonomous generation of useful information and its abstract is a device for generating useful information employing a first neural network trained to produce input-output maps within a predetermined initial knowledge domain, an apparatus for subjecting the neural network to perturbations which may produce changes in the predetermined knowledge domain, the neural network having an optional output for feeding the outputs of the first neural network to a second neural network that evaluates the outputs based on training within the second neural network. the device may also include a reciprocal feed back connection from the output of the second neural network to the first neural network to further influence and change what takes place in the aforesaid neural network. dated 2000-01-25"
6018728,method and apparatus for training a neural network to learn hierarchical representations of objects and to detect and classify objects with uncertain training data,"a signal processing apparatus and concomitant method for learning and integrating features from multiple resolutions for detecting and/or classifying objects are presented. neural networks in a pattern tree structure with tree-structured descriptions of objects in terms of simple sub-patterns, are grown and trained to detect and integrate the sub-patterns. a plurality of objective functions and their approximations are presented to train the neural networks to detect sub-patterns of features of some class of objects. objective functions for training neural networks to detect objects whose positions in the training data are uncertain and for addressing supervised learning where there are potential errors in the training data are also presented.",2000-01-25,"The title of the patent is method and apparatus for training a neural network to learn hierarchical representations of objects and to detect and classify objects with uncertain training data and its abstract is a signal processing apparatus and concomitant method for learning and integrating features from multiple resolutions for detecting and/or classifying objects are presented. neural networks in a pattern tree structure with tree-structured descriptions of objects in terms of simple sub-patterns, are grown and trained to detect and integrate the sub-patterns. a plurality of objective functions and their approximations are presented to train the neural networks to detect sub-patterns of features of some class of objects. objective functions for training neural networks to detect objects whose positions in the training data are uncertain and for addressing supervised learning where there are potential errors in the training data are also presented. dated 2000-01-25"
6018729,neural network control of spot welding,"a spot welder comprises a neural network for processing, in real time, current and voltage energizing a weld in progress. the neural network generates a predicted time of optimal weld strength and/or nugget size for the weld in progress. a controller terminates the weld in progress at the predicted time. a method for controlling a spot welder comprises the steps of: sensing in real time current and voltage energizing a spot weld in progress; predicting a time of optimal weld strength and/or nugget size with a neural network responsive to the sensed current and voltage; and, terminating the weld in progress at the predicted time. a sensor for electromotive forces (emf) induced by the spot welder can generate a signal for canceling out a large fraction of emf components in at least one or both of the current and voltage signals. emf components are substantially precluded in the current signal if the current sensor uses a buried shunt. termination of the weld in progress at the predicted time is prevented when the predicted time precedes a predetermined minimum weld duration. the weld in progress is terminated at a predetermined maximum weld duration when the predicted time is after the predetermined maximum weld duration.",2000-01-25,"The title of the patent is neural network control of spot welding and its abstract is a spot welder comprises a neural network for processing, in real time, current and voltage energizing a weld in progress. the neural network generates a predicted time of optimal weld strength and/or nugget size for the weld in progress. a controller terminates the weld in progress at the predicted time. a method for controlling a spot welder comprises the steps of: sensing in real time current and voltage energizing a spot weld in progress; predicting a time of optimal weld strength and/or nugget size with a neural network responsive to the sensed current and voltage; and, terminating the weld in progress at the predicted time. a sensor for electromotive forces (emf) induced by the spot welder can generate a signal for canceling out a large fraction of emf components in at least one or both of the current and voltage signals. emf components are substantially precluded in the current signal if the current sensor uses a buried shunt. termination of the weld in progress at the predicted time is prevented when the predicted time precedes a predetermined minimum weld duration. the weld in progress is terminated at a predetermined maximum weld duration when the predicted time is after the predetermined maximum weld duration. dated 2000-01-25"
6021387,speech recognition apparatus for consumer electronic applications,"a spoken word or phrase recognition device. the device does not require a digital signal processor, large ram, or extensive analog circuitry. the input audio signal is digitized and passed recursively through a digital difference filter to produce a multiplicity of filtered output waveforms. these waveforms are processed in real time by a microprocessor to generate a pattern that is recognized by a neural network pattern classifier that operates in software in the microprocessor. by application of additional techniques, this device has been shown to recognize an unknown speaker saying a digit from zero through nine with an accuracy greater than 99%. because of the recognition accuracy and cost-effective design, the device may be used in cost sensitive applications such as toys, electronic learning aids, and consumer electronic products.",2000-02-01,"The title of the patent is speech recognition apparatus for consumer electronic applications and its abstract is a spoken word or phrase recognition device. the device does not require a digital signal processor, large ram, or extensive analog circuitry. the input audio signal is digitized and passed recursively through a digital difference filter to produce a multiplicity of filtered output waveforms. these waveforms are processed in real time by a microprocessor to generate a pattern that is recognized by a neural network pattern classifier that operates in software in the microprocessor. by application of additional techniques, this device has been shown to recognize an unknown speaker saying a digit from zero through nine with an accuracy greater than 99%. because of the recognition accuracy and cost-effective design, the device may be used in cost sensitive applications such as toys, electronic learning aids, and consumer electronic products. dated 2000-02-01"
6023663,method and apparatus for inspecting a solder joint using a correlation neural network,"a solder joint inspection process and system includes using a correlation values to classify the shape of the solder joint. the solder joint is illuminated by a multiple colors, and the image of the solder joint is captured. this multiple color image is then converted into a plurality of one-dimensional vectors. each one-dimensional vectors sequence is for one color of the multi-color illumination and for a one-dimensional spatial extent. correlation values among all combinations of the one-dimensional vector sequences are computed and used for training an automatic solder joint shape classifier.",2000-02-08,"The title of the patent is method and apparatus for inspecting a solder joint using a correlation neural network and its abstract is a solder joint inspection process and system includes using a correlation values to classify the shape of the solder joint. the solder joint is illuminated by a multiple colors, and the image of the solder joint is captured. this multiple color image is then converted into a plurality of one-dimensional vectors. each one-dimensional vectors sequence is for one color of the multi-color illumination and for a one-dimensional spatial extent. correlation values among all combinations of the one-dimensional vector sequences are computed and used for training an automatic solder joint shape classifier. dated 2000-02-08"
6025128,prediction of prostate cancer progression by analysis of selected predictive parameters,"a method for screening individuals at risk for prostate cancer progression is disclosed. the method is useful for evaluating cells from patients at risk for recurrence of prostate cancer following surgery for prostate cancer. specifically, the method uses specific markovian nuclear texture factors, alone or in combination with other biomarkers, to determine whether the cancer will progress or lose organ confinement. in addition, methods of predicting the development of fatal metastatic disease by statistical analysis of selected biomarkers is also disclosed. the invention also contemplates a method that uses a neural network to analyze and interpret cell morphology data. utilizing markovian factors and other biomarkers as parameters, the network is first trained with a sets of cell data from known progressors and known non-progressors. the trained network is then used to predict prostate cancer progression in patient samples.",2000-02-15,"The title of the patent is prediction of prostate cancer progression by analysis of selected predictive parameters and its abstract is a method for screening individuals at risk for prostate cancer progression is disclosed. the method is useful for evaluating cells from patients at risk for recurrence of prostate cancer following surgery for prostate cancer. specifically, the method uses specific markovian nuclear texture factors, alone or in combination with other biomarkers, to determine whether the cancer will progress or lose organ confinement. in addition, methods of predicting the development of fatal metastatic disease by statistical analysis of selected biomarkers is also disclosed. the invention also contemplates a method that uses a neural network to analyze and interpret cell morphology data. utilizing markovian factors and other biomarkers as parameters, the network is first trained with a sets of cell data from known progressors and known non-progressors. the trained network is then used to predict prostate cancer progression in patient samples. dated 2000-02-15"
6026177,method for identifying a sequence of alphanumeric characters,"a character recognition system is described, in particular a system suitable for use in monitoring cargo container codes or vehicle number plates. an image of the code is first analyzed to extract potential characters. as part of this process, long horizontal and vertical line segments are filtered out. the extracted potential characters are then input to a two-level character recognition means. the first level comprises a neural network classifier that classifies a character into a smaller set of possible characters; and then the second level comprises another neural network classifier which identifies which character among the smaller set of possible characters the extracted character is.",2000-02-15,"The title of the patent is method for identifying a sequence of alphanumeric characters and its abstract is a character recognition system is described, in particular a system suitable for use in monitoring cargo container codes or vehicle number plates. an image of the code is first analyzed to extract potential characters. as part of this process, long horizontal and vertical line segments are filtered out. the extracted potential characters are then input to a two-level character recognition means. the first level comprises a neural network classifier that classifies a character into a smaller set of possible characters; and then the second level comprises another neural network classifier which identifies which character among the smaller set of possible characters the extracted character is. dated 2000-02-15"
6026178,image processing apparatus using neural network,"in order to realize a neural network for image processing by an inexpensive hardware arrangement, a neural network arranged in an image processing apparatus is constituted by an input layer having neurons for receiving information from picture elements in a 7.times.7 area including an interesting picture element in an image, an intermediate layer having one neuron connected to all the 49 neurons in the input layer and five groups of nine neurons, the nine neurons in each group being connected to nine neurons in the input layer, which receive information from picture elements in at least one of five 3.times.3 areas (1a to 1e), and an output layer having one neuron, which is connected to all the neurons in the intermediate layer and outputs information corresponding to the interesting picture element.",2000-02-15,"The title of the patent is image processing apparatus using neural network and its abstract is in order to realize a neural network for image processing by an inexpensive hardware arrangement, a neural network arranged in an image processing apparatus is constituted by an input layer having neurons for receiving information from picture elements in a 7.times.7 area including an interesting picture element in an image, an intermediate layer having one neuron connected to all the 49 neurons in the input layer and five groups of nine neurons, the nine neurons in each group being connected to nine neurons in the input layer, which receive information from picture elements in at least one of five 3.times.3 areas (1a to 1e), and an output layer having one neuron, which is connected to all the neurons in the intermediate layer and outputs information corresponding to the interesting picture element. dated 2000-02-15"
6026358,"neural network, a method of learning of a neural network and phoneme recognition apparatus utilizing a neural network","a neuron device network is provided with a speech input layer, a context layer, a hidden layer, a speech output layer and a hypothesis layer. a phoneme to be learned is spectral-analyzed by an fft unit and a vector row at a time point t is input to a speech input layer. also, a vector state of the hidden layer at a time t-1 is input to the context layer, the vector row at a time t+1 is input to the speech output layer as an instructor signal, and a code row for hypothesizing the phoneme, or the code row, is input to the hypothesis layer. the time series relation of the vector rows and the phoneme are hypothetically learned. alternatively, a spectrum, a cepstrum or a speech vector row based on outputs from the hidden layer of an auto-associative neural network is input to the speech input layer, and the code row is output from the hypothesis layer, taking into account the time series relation. the speech is recognized when a cpu reads the stored output values of the hidden layer and the connection weights of the hidden layer and the hypothesis layer from a memory of the neuron device network and calculates output values of the respective neuron devices of the hypothesis layer based on the output values and the connection weights. the corresponding phoneme is determined by collating the output values of the respective neuron devices of the hypothesis layer with the code rows in an instructor signal table.",2000-02-15,"The title of the patent is neural network, a method of learning of a neural network and phoneme recognition apparatus utilizing a neural network and its abstract is a neuron device network is provided with a speech input layer, a context layer, a hidden layer, a speech output layer and a hypothesis layer. a phoneme to be learned is spectral-analyzed by an fft unit and a vector row at a time point t is input to a speech input layer. also, a vector state of the hidden layer at a time t-1 is input to the context layer, the vector row at a time t+1 is input to the speech output layer as an instructor signal, and a code row for hypothesizing the phoneme, or the code row, is input to the hypothesis layer. the time series relation of the vector rows and the phoneme are hypothetically learned. alternatively, a spectrum, a cepstrum or a speech vector row based on outputs from the hidden layer of an auto-associative neural network is input to the speech input layer, and the code row is output from the hypothesis layer, taking into account the time series relation. the speech is recognized when a cpu reads the stored output values of the hidden layer and the connection weights of the hidden layer and the hypothesis layer from a memory of the neuron device network and calculates output values of the respective neuron devices of the hypothesis layer based on the output values and the connection weights. the corresponding phoneme is determined by collating the output values of the respective neuron devices of the hypothesis layer with the code rows in an instructor signal table. dated 2000-02-15"
6027217,automated visual function testing via telemedicine,"a method and an apparatus for utilizing a central neural network and a central data bank to perform automatic interpretation of the visual function test parameters obtained in a plurality of visual field testing systems, for a plurality of patients, with control and response signals being transmitted via the internet. the data produced by the testing systems are automatically analyzed and compared with patterns on which the neural network was previously trained, and clinical diagnoses for pathological conditions are thereby suggested to the respective clinician for each patient.",2000-02-22,"The title of the patent is automated visual function testing via telemedicine and its abstract is a method and an apparatus for utilizing a central neural network and a central data bank to perform automatic interpretation of the visual function test parameters obtained in a plurality of visual field testing systems, for a plurality of patients, with control and response signals being transmitted via the internet. the data produced by the testing systems are automatically analyzed and compared with patterns on which the neural network was previously trained, and clinical diagnoses for pathological conditions are thereby suggested to the respective clinician for each patient. dated 2000-02-22"
6028994,method for predicting performance of microelectronic device based on electrical parameter test data using computer model,electrical parameter testing and performance testing are performed on a plurality of microelectronic devices to obtain parametric values and performance values respectively. the parametric values are applied as inputs to a computer program such as a back propagation neural network engine which generates a performance prediction model by self-learning that implements a function relating the performance values to the parametric values. the model is used to predict the performance of devices being fabricated by performing electrical parameter testing on these devices and applying the resulting parametric values to the model as inputs to produce predicted performance values as outputs. the model can be configured to produce predicted performance values as percentages of devices having speed or other parameters in predetermined respective ranges. the model can be further configured to produce predicted performance values as percentages of devices having different types of defects. the model can be improved by self-learning using additional test values. the model can also be used to identify parameters which result in low performance and improve devices being fabricated by adjusting the corresponding process parameters.,2000-02-22,The title of the patent is method for predicting performance of microelectronic device based on electrical parameter test data using computer model and its abstract is electrical parameter testing and performance testing are performed on a plurality of microelectronic devices to obtain parametric values and performance values respectively. the parametric values are applied as inputs to a computer program such as a back propagation neural network engine which generates a performance prediction model by self-learning that implements a function relating the performance values to the parametric values. the model is used to predict the performance of devices being fabricated by performing electrical parameter testing on these devices and applying the resulting parametric values to the model as inputs to produce predicted performance values as outputs. the model can be configured to produce predicted performance values as percentages of devices having speed or other parameters in predetermined respective ranges. the model can be further configured to produce predicted performance values as percentages of devices having different types of defects. the model can be improved by self-learning using additional test values. the model can also be used to identify parameters which result in low performance and improve devices being fabricated by adjusting the corresponding process parameters. dated 2000-02-22
6029139,method and apparatus for optimizing promotional sale of products based upon historical data,"a system for optimizing the promotional sale of a product, a product segment, or a category which may take into account related products or competing products comprising means for generating a three-dimensional data structure corresponding to the sales history for a product, the data structure dimensions corresponding to an event type domain, a time domain, and a unit of measurement domain, means for populating the three-dimensional data structure, a neural network, means for training the neural network and means for applying sales objectives and constraints to the neural network.",2000-02-22,"The title of the patent is method and apparatus for optimizing promotional sale of products based upon historical data and its abstract is a system for optimizing the promotional sale of a product, a product segment, or a category which may take into account related products or competing products comprising means for generating a three-dimensional data structure corresponding to the sales history for a product, the data structure dimensions corresponding to an event type domain, a time domain, and a unit of measurement domain, means for populating the three-dimensional data structure, a neural network, means for training the neural network and means for applying sales objectives and constraints to the neural network. dated 2000-02-22"
6029157,method for determining state parameters of a chemical reactor with artificial neural networks,a sequence of measured quantities is determined for a chemical reactor and a data generator generates a curve in normal coordinates from a respective predetermined normal form that describes a type of critical state. each normal form is imaged onto the sequence of measured quantities by neural networks whereby a respective neural network is allocated to a data generator. the imaging is optimized by applying parameter optimization methods. the neural network that converges best through use of parameter optimization method describes the critical state that lies closest to the actual state of the chemical reactor.,2000-02-22,The title of the patent is method for determining state parameters of a chemical reactor with artificial neural networks and its abstract is a sequence of measured quantities is determined for a chemical reactor and a data generator generates a curve in normal coordinates from a respective predetermined normal form that describes a type of critical state. each normal form is imaged onto the sequence of measured quantities by neural networks whereby a respective neural network is allocated to a data generator. the imaging is optimized by applying parameter optimization methods. the neural network that converges best through use of parameter optimization method describes the critical state that lies closest to the actual state of the chemical reactor. dated 2000-02-22
6031618,apparatus and method for attribute identification in color reproduction devices,"an apparatus and method identify at least one attribute of an article for use in accurately reproducing the article. the apparatus includes at least one detector having an array of cells. at least one cell, in the detector array, the sensor cell, is provided with an extra colored coating or is painted with an extra color. when the article is scanned by the detector, the at least one sensor cell will read a different color value from the other cells due to the extra coating applied to it. the color that the at least one sensor cell would have output without the extra coating is interpolated from the detection results of the neighboring cells. this color and the color actually detected by the at least one sensor cell are then input to a controller which determines the at least one attribute of the scanned article using a model, such as a neural network, an expert system, a fuzzy logic model, and the like. the controller performs appropriate processing based on the at least one determined attribute.",2000-02-29,"The title of the patent is apparatus and method for attribute identification in color reproduction devices and its abstract is an apparatus and method identify at least one attribute of an article for use in accurately reproducing the article. the apparatus includes at least one detector having an array of cells. at least one cell, in the detector array, the sensor cell, is provided with an extra colored coating or is painted with an extra color. when the article is scanned by the detector, the at least one sensor cell will read a different color value from the other cells due to the extra coating applied to it. the color that the at least one sensor cell would have output without the extra coating is interpolated from the detection results of the neighboring cells. this color and the color actually detected by the at least one sensor cell are then input to a controller which determines the at least one attribute of the scanned article using a model, such as a neural network, an expert system, a fuzzy logic model, and the like. the controller performs appropriate processing based on the at least one determined attribute. dated 2000-02-29"
6032116,distance measure in a speech recognition system for speech recognition using frequency shifting factors to compensate for input signal frequency shifts,"one embodiment of a speech recognition system is organized with speech input signal preprocessing and feature extraction followed by a fuzzy matrix quantizer (fmq). frames of the speech input signal are represented by a vector .function. of line spectral pair frequencies and are fuzzy matrix quantized to respective a vector .function. entries in a codebook of the fmq. a distance measure between .function. and .function., d(.function.,.function.), is defined as ##equ1## where the constants .alpha..sub.1, a.sub.2, .beta..sub.1 and .beta..sub.2 are set to substantially minimize quantization error, and e.sub.i is the error power spectrum of the speech input signal and a predicted speech input signal at the ith line spectral pair frequency of the speech input signal. the speech recognition system may also include hidden markov models and neural networks, such as a multilevel perceptron neural network, speech classifiers.",2000-02-29,"The title of the patent is distance measure in a speech recognition system for speech recognition using frequency shifting factors to compensate for input signal frequency shifts and its abstract is one embodiment of a speech recognition system is organized with speech input signal preprocessing and feature extraction followed by a fuzzy matrix quantizer (fmq). frames of the speech input signal are represented by a vector .function. of line spectral pair frequencies and are fuzzy matrix quantized to respective a vector .function. entries in a codebook of the fmq. a distance measure between .function. and .function., d(.function.,.function.), is defined as ##equ1## where the constants .alpha..sub.1, a.sub.2, .beta..sub.1 and .beta..sub.2 are set to substantially minimize quantization error, and e.sub.i is the error power spectrum of the speech input signal and a predicted speech input signal at the ith line spectral pair frequency of the speech input signal. the speech recognition system may also include hidden markov models and neural networks, such as a multilevel perceptron neural network, speech classifiers. dated 2000-02-29"
6032125,"demand forecasting method, demand forecasting system, and recording medium","a method and a system for forecasting the demand agreeing with the fluctuation trend of sales results at high and stable precision, without requiring user's maintenance, by using a model optimum for grasping the fluctuation trend of sales results, even if the products are diverse, by storing a plurality of models of neural network, for example, a model for forecasting the demand from data of the past several months, a model for forecasting the demand from data of the same period of the previous year, and a model for forecasting the demand from both the latest data and data of the same period of the previous year, and also by feeding sales results into a model of neural network to make it learn by the short period such as by the week, and a recording medium in which is recorded such program.",2000-02-29,"The title of the patent is demand forecasting method, demand forecasting system, and recording medium and its abstract is a method and a system for forecasting the demand agreeing with the fluctuation trend of sales results at high and stable precision, without requiring user's maintenance, by using a model optimum for grasping the fluctuation trend of sales results, even if the products are diverse, by storing a plurality of models of neural network, for example, a model for forecasting the demand from data of the past several months, a model for forecasting the demand from data of the same period of the previous year, and a model for forecasting the demand from both the latest data and data of the same period of the previous year, and also by feeding sales results into a model of neural network to make it learn by the short period such as by the week, and a recording medium in which is recorded such program. dated 2000-02-29"
6032140,"low-voltage, very-low-power conductance mode neuron","a neural network including a number of synaptic weighting elements, and a neuron stage; each of the synaptic weighting elements having a respective synaptic input connection supplied with a respective input signal; and the neuron stage having inputs connected to the synaptic weighting elements, and being connected to an output of the neural network supplying a digital output signal. the accumulated weighted inputs are represented as conductances, and a conductance-mode neuron is used to apply nonlinearity and produce an output. the synaptic weighting elements are formed by memory cells programmable to different threshold voltage levels, so that each presents a respective programmable conductance; and the neuron stage provides for measuring conductance on the basis of the current through the memory cells, and for generating a binary output signal on the basis of the total conductance of the synaptic elements.",2000-02-29,"The title of the patent is low-voltage, very-low-power conductance mode neuron and its abstract is a neural network including a number of synaptic weighting elements, and a neuron stage; each of the synaptic weighting elements having a respective synaptic input connection supplied with a respective input signal; and the neuron stage having inputs connected to the synaptic weighting elements, and being connected to an output of the neural network supplying a digital output signal. the accumulated weighted inputs are represented as conductances, and a conductance-mode neuron is used to apply nonlinearity and produce an output. the synaptic weighting elements are formed by memory cells programmable to different threshold voltage levels, so that each presents a respective programmable conductance; and the neuron stage provides for measuring conductance on the basis of the current through the memory cells, and for generating a binary output signal on the basis of the total conductance of the synaptic elements. dated 2000-02-29"
6035057,hierarchical data matrix pattern recognition and identification system,""" the present invention relates to a hierarchical artificial neural network (hann) for automating the recognition and identification of patterns in data matrices. it has particular, although not exclusive, application to the identification of severe storm events (sses) from spatial precipitation patterns, derived from conventional volumetric radar imagery. to identify characteristic features a data matrix, the data matrix is processed with a self organizing network to produce a self organizing feature space mapping. the self organizing feature space mapping is processed to produce a density characterization of the feature space mapping. the self organizing network is preferably completely unsupervised. it may, under some circumstances include a supervised layer, but it must include at least an unsupervised component for the purposes of the invention. the """"self organizing feature space"""" is intended to include any map with the self organizing characteristics of the kohonen self organizing feature map. the frequency vector of a cappi image that has been derived is a data abstraction that can be displayed directly for examination. in preferred embodiments, it is presented to a classification network, e.g. the standard cpn network, for classifying the density vector representation of the three dimensional data and displaying a representation of classified features in the three dimensional data. a novel methodology is preferably used for incorporating vigilance and conscience mechanisms in the forward counterpropagation network during training. """,2000-03-07,"The title of the patent is hierarchical data matrix pattern recognition and identification system and its abstract is "" the present invention relates to a hierarchical artificial neural network (hann) for automating the recognition and identification of patterns in data matrices. it has particular, although not exclusive, application to the identification of severe storm events (sses) from spatial precipitation patterns, derived from conventional volumetric radar imagery. to identify characteristic features a data matrix, the data matrix is processed with a self organizing network to produce a self organizing feature space mapping. the self organizing feature space mapping is processed to produce a density characterization of the feature space mapping. the self organizing network is preferably completely unsupervised. it may, under some circumstances include a supervised layer, but it must include at least an unsupervised component for the purposes of the invention. the """"self organizing feature space"""" is intended to include any map with the self organizing characteristics of the kohonen self organizing feature map. the frequency vector of a cappi image that has been derived is a data abstraction that can be displayed directly for examination. in preferred embodiments, it is presented to a classification network, e.g. the standard cpn network, for classifying the density vector representation of the three dimensional data and displaying a representation of classified features in the three dimensional data. a novel methodology is preferably used for incorporating vigilance and conscience mechanisms in the forward counterpropagation network during training. "" dated 2000-03-07"
6035246,method for identifying known materials within a mixture of unknowns,one or both of two methods and systems are used to determine concentration of a known material in an unknown mixture on the basis of the measured interaction of electromagnetic waves upon the mixture. one technique is to utilize a multivariate analysis patch technique to develop a library of optimized patches of spectral signatures of known materials containing only those pixels most descriptive of the known materials by an evolutionary algorithm. identity and concentration of the known materials within the unknown mixture is then determined by minimizing the residuals between the measurements from the library of optimized patches and the measurements from the same pixels from the unknown mixture. another technique is to train a neural network by the genetic algorithm to determine the identity and concentration of known materials in the unknown mixture. the two techniques may be combined into an expert system providing cross checks for accuracy.,2000-03-07,The title of the patent is method for identifying known materials within a mixture of unknowns and its abstract is one or both of two methods and systems are used to determine concentration of a known material in an unknown mixture on the basis of the measured interaction of electromagnetic waves upon the mixture. one technique is to utilize a multivariate analysis patch technique to develop a library of optimized patches of spectral signatures of known materials containing only those pixels most descriptive of the known materials by an evolutionary algorithm. identity and concentration of the known materials within the unknown mixture is then determined by minimizing the residuals between the measurements from the library of optimized patches and the measurements from the same pixels from the unknown mixture. another technique is to train a neural network by the genetic algorithm to determine the identity and concentration of known materials in the unknown mixture. the two techniques may be combined into an expert system providing cross checks for accuracy. dated 2000-03-07
6035270,trained artificial neural networks using an imperfect vocal tract model for assessment of speech signal quality,"a speech signal is subjected imperfect to vocal tract analysis model and the output therefrom is analyzed by a neural network. the output from the neural network is compared with the parameters stored in the network definition function, to derive measurement of the quality of the speech signal supplied to the source. the network definition function is determined by applying to the trainable processing apparatus a distortion perception measure indicative of the extent to which a distortion would be perceptible to a human listener.",2000-03-07,"The title of the patent is trained artificial neural networks using an imperfect vocal tract model for assessment of speech signal quality and its abstract is a speech signal is subjected imperfect to vocal tract analysis model and the output therefrom is analyzed by a neural network. the output from the neural network is compared with the parameters stored in the network definition function, to derive measurement of the quality of the speech signal supplied to the source. the network definition function is determined by applying to the trainable processing apparatus a distortion perception measure indicative of the extent to which a distortion would be perceptible to a human listener. dated 2000-03-07"
6035295,computer system and method of data analysis,"a neural network based data comparison system compares data stored within a database against each other to determine duplicative, fraudulent, defective and/or irregular data. the system includes a database storing data therein, and a pattern database storing pattern data therein. the system further includes a data pattern build system, responsively connected to the database and to the pattern database. the data pattern build system retrieves the data from the database and generates the pattern data formatted in accordance a predetermined patten. the predetermined pattern includes an array having array locations corresponding to each character in a defined character set. the data pattern build system increments a value in each of the array locations responsive to the number of occurrences of each character in the data and stores the array as the pattern data in the pattern database. the comparison system also includes a neural network, responsively connected to the pattern database, which retrieves the pattern data stored therein and compares the pattern data to each other and determines responsive to the comparing when different pattern data match in accordance with predetermined criteria indicating that the different pattern data are duplicative, fraudulent, defective and/or irregular.",2000-03-07,"The title of the patent is computer system and method of data analysis and its abstract is a neural network based data comparison system compares data stored within a database against each other to determine duplicative, fraudulent, defective and/or irregular data. the system includes a database storing data therein, and a pattern database storing pattern data therein. the system further includes a data pattern build system, responsively connected to the database and to the pattern database. the data pattern build system retrieves the data from the database and generates the pattern data formatted in accordance a predetermined patten. the predetermined pattern includes an array having array locations corresponding to each character in a defined character set. the data pattern build system increments a value in each of the array locations responsive to the number of occurrences of each character in the data and stores the array as the pattern data in the pattern database. the comparison system also includes a neural network, responsively connected to the pattern database, which retrieves the pattern data stored therein and compares the pattern data to each other and determines responsive to the comparing when different pattern data match in accordance with predetermined criteria indicating that the different pattern data are duplicative, fraudulent, defective and/or irregular. dated 2000-03-07"
6038337,method and apparatus for object recognition,"a hybrid neural network system for object recognition exhibiting local image sampling, a self-organizing map neural network, and a hybrid convolutional neural network. the self-organizing map provides a quantization of the image samples into a topological space where inputs that are nearby in the original space are also nearby in the output space, thereby providing dimensionality reduction and invariance to minor changes in the image sample, and the hybrid convolutional neural network provides for partial invariance to translation, rotation, scale, and deformation. the hybrid convolutional network extracts successively larger features in a hierarchical set of layers. alternative embodiments using the karhunen-loeve transform in place of the self-organizing map, and a multi-layer perceptron in place of the convolutional network are described.",2000-03-14,"The title of the patent is method and apparatus for object recognition and its abstract is a hybrid neural network system for object recognition exhibiting local image sampling, a self-organizing map neural network, and a hybrid convolutional neural network. the self-organizing map provides a quantization of the image samples into a topological space where inputs that are nearby in the original space are also nearby in the output space, thereby providing dimensionality reduction and invariance to minor changes in the image sample, and the hybrid convolutional neural network provides for partial invariance to translation, rotation, scale, and deformation. the hybrid convolutional network extracts successively larger features in a hierarchical set of layers. alternative embodiments using the karhunen-loeve transform in place of the self-organizing map, and a multi-layer perceptron in place of the convolutional network are described. dated 2000-03-14"
6038338,hybrid neural network for pattern recognition,"a system and a method for recognizing patterns comprises a first stage for xtracting features from inputted patterns and for providing topological representations of the characteristics of the inputted patterns and a second stage for classifying and recognizing the inputted patterns. the first stage comprises two one-layer neural networks and the second stage comprises a feedforward two-layer neural network. supplying signals representative of a set of inputted patterns to the input layers of the first and second neural networks, training the first and second neural networks using a competitive learning algorithm, and generating topological representations of the input patterns using the first and second neural networks the method further comprises providing a third neural network for classifying and recognizing the inputted patterns and training the third neural network with a back-propagation algorithm so that the third neural network recognizes at least one interested pattern.",2000-03-14,"The title of the patent is hybrid neural network for pattern recognition and its abstract is a system and a method for recognizing patterns comprises a first stage for xtracting features from inputted patterns and for providing topological representations of the characteristics of the inputted patterns and a second stage for classifying and recognizing the inputted patterns. the first stage comprises two one-layer neural networks and the second stage comprises a feedforward two-layer neural network. supplying signals representative of a set of inputted patterns to the input layers of the first and second neural networks, training the first and second neural networks using a competitive learning algorithm, and generating topological representations of the input patterns using the first and second neural networks the method further comprises providing a third neural network for classifying and recognizing the inputted patterns and training the third neural network with a back-propagation algorithm so that the third neural network recognizes at least one interested pattern. dated 2000-03-14"
6038555,generic processing capability,"a generic anomaly detection engine is described which provides neural network technology to process information such as call detail records and detect anomalies in the information, such as mobile phone fraud. information from call detail records is pre-processed to form signatures which represent the calling behavior of a subscriber. these signatures are formed in one of a plurality of predetermined formats. the generic anomaly detection engine is instantiated in order to create an anomaly detector which is suitable for a particular situation and during the instantiation process the topology of the neural network components is automatically adjusted to fit the required situation. the topology is adjusted to according to the format of the signatures.",2000-03-14,"The title of the patent is generic processing capability and its abstract is a generic anomaly detection engine is described which provides neural network technology to process information such as call detail records and detect anomalies in the information, such as mobile phone fraud. information from call detail records is pre-processed to form signatures which represent the calling behavior of a subscriber. these signatures are formed in one of a plurality of predetermined formats. the generic anomaly detection engine is instantiated in order to create an anomaly detector which is suitable for a particular situation and during the instantiation process the topology of the neural network components is automatically adjusted to fit the required situation. the topology is adjusted to according to the format of the signatures. dated 2000-03-14"
6039650,"card dispensing shoe with scanner apparatus, system and method therefor",""" the present invention is directed to a playing card dispensing shoe apparatus, system and method wherein the shoe has a card scanner which scans the indicia on a playing card as the card moves along and out of a chute of the shoe by operation of the dealer. the scanner comprises an optical-sensor used in combination with a neural network which is trained using error back-propagation to recognize the card suits and card values of the playing cards as they are moved past the scanner. the scanning process in combination with a central processing unit (cpu) determines the progress of the play of the game and, by identifying card counting systems or basic playing strategies in use by the players of the game, provides means to limit or prevent casino losses and calculate the theoretical win of the casino, thus also providing an accurate quality method of the amount of comps to be given a particular player. the shoe is also provided with additional devices which make it simple and easy to access, record and display other data relevant to the play of the game. these include means for accommodating a """"customer-tracking-card"""" which reads each player's account information from a magnetic stripe on the card, thus providing access to the player's customer data file stored on the casino's computer system and one or more alpha-numeric keyboards and lcd displays used to enter and retrieve player and game information. also included are keyboards on the game table so that each player can individually select various playing or wagering options using their own keyboard. """,2000-03-21,"The title of the patent is card dispensing shoe with scanner apparatus, system and method therefor and its abstract is "" the present invention is directed to a playing card dispensing shoe apparatus, system and method wherein the shoe has a card scanner which scans the indicia on a playing card as the card moves along and out of a chute of the shoe by operation of the dealer. the scanner comprises an optical-sensor used in combination with a neural network which is trained using error back-propagation to recognize the card suits and card values of the playing cards as they are moved past the scanner. the scanning process in combination with a central processing unit (cpu) determines the progress of the play of the game and, by identifying card counting systems or basic playing strategies in use by the players of the game, provides means to limit or prevent casino losses and calculate the theoretical win of the casino, thus also providing an accurate quality method of the amount of comps to be given a particular player. the shoe is also provided with additional devices which make it simple and easy to access, record and display other data relevant to the play of the game. these include means for accommodating a """"customer-tracking-card"""" which reads each player's account information from a magnetic stripe on the card, thus providing access to the player's customer data file stored on the casino's computer system and one or more alpha-numeric keyboards and lcd displays used to enter and retrieve player and game information. also included are keyboards on the game table so that each player can individually select various playing or wagering options using their own keyboard. "" dated 2000-03-21"
6041299,"apparatus for calculating a posterior probability of phoneme symbol, and speech recognition apparatus","there are disclosed an apparatus for calculating a posteriori probabilities of phoneme symbols and a speech recognition apparatus using the apparatus for calculating a posteriori probabilities of phoneme symbols. a feature extracting section extracts speech feature parameters from a speech signal of an uttered speech sentence composed of an inputted character series, and a calculating section calculates a a posteriori probability of a phoneme symbol of the speech signal, by using a bidirectional recurrent neural network. the bidirectional recurrent neural network includes (a) an input layer for receiving the speech feature parameters extracted by the feature extracting means and a plurality of hypothetical phoneme symbol series signals, (b) an intermediate layer of at least one layer having a plurality of units, and (c) an output layer for outputting a a posteriori probability of each phoneme symbol. the input layer includes (a) a first input neuron group having a plurality of units, for receiving a plurality of speech feature parameters and a plurality of phoneme symbol series signals, (b) a forward module, and (c) a backward module.",2000-03-21,"The title of the patent is apparatus for calculating a posterior probability of phoneme symbol, and speech recognition apparatus and its abstract is there are disclosed an apparatus for calculating a posteriori probabilities of phoneme symbols and a speech recognition apparatus using the apparatus for calculating a posteriori probabilities of phoneme symbols. a feature extracting section extracts speech feature parameters from a speech signal of an uttered speech sentence composed of an inputted character series, and a calculating section calculates a a posteriori probability of a phoneme symbol of the speech signal, by using a bidirectional recurrent neural network. the bidirectional recurrent neural network includes (a) an input layer for receiving the speech feature parameters extracted by the feature extracting means and a plurality of hypothetical phoneme symbol series signals, (b) an intermediate layer of at least one layer having a plurality of units, and (c) an output layer for outputting a a posteriori probability of each phoneme symbol. the input layer includes (a) a first input neuron group having a plurality of units, for receiving a plurality of speech feature parameters and a plurality of phoneme symbol series signals, (b) a forward module, and (c) a backward module. dated 2000-03-21"
6041322,method and apparatus for processing data in a neural network,"a digital artificial neural network (ann) reduces memory requirements by storing sample transfer function representing output values for multiple nodes. each nodes receives an input value representing the information to be processed by the network. additionally, the node determines threshold values indicative of boundaries for application of the sample transfer function for the node. from the input value received, the node generates an intermediate value. based on the threshold values and the intermediate value, the node determines an output value in accordance with the sample transfer function.",2000-03-21,"The title of the patent is method and apparatus for processing data in a neural network and its abstract is a digital artificial neural network (ann) reduces memory requirements by storing sample transfer function representing output values for multiple nodes. each nodes receives an input value representing the information to be processed by the network. additionally, the node determines threshold values indicative of boundaries for application of the sample transfer function for the node. from the input value received, the node generates an intermediate value. based on the threshold values and the intermediate value, the node determines an output value in accordance with the sample transfer function. dated 2000-03-21"
6044325,conductivity anisotropy estimation method for inversion processing of measurements made by a transverse electromagnetic induction logging instrument,"a method for generating an improved estimate of horizontal conductivity, dip angle, azimuth and anisotropy parameter of an earth formation penetrated by a wellbore from dual-frequency transverse electromagnetic induction measurements, comprising generating an initial estimate of the horizontal conductivity, dip angle, azimuth and anisotropy parameter from the dual-frequency transverse induction measurements made at each one of a plurality of base frequencies. the initial estimates from each of the plurality of base frequencies are input into a primary trained neural network, and the improved estimate is calculated by the trained neural network. the network is trained by generating models of earth formations each having a known value of horizontal conductivity, anisotropy parameter, dip angle and azimuth. voltages which would be measured by the transverse electromagnetic induction instrument in response to each model are synthesized. initial estimates from the synthesized voltages are calculated and the initial estimates and known values from each of the models are input to the neural network to cause it to learn a relationship between the initial estimates and the known values.",2000-03-28,"The title of the patent is conductivity anisotropy estimation method for inversion processing of measurements made by a transverse electromagnetic induction logging instrument and its abstract is a method for generating an improved estimate of horizontal conductivity, dip angle, azimuth and anisotropy parameter of an earth formation penetrated by a wellbore from dual-frequency transverse electromagnetic induction measurements, comprising generating an initial estimate of the horizontal conductivity, dip angle, azimuth and anisotropy parameter from the dual-frequency transverse induction measurements made at each one of a plurality of base frequencies. the initial estimates from each of the plurality of base frequencies are input into a primary trained neural network, and the improved estimate is calculated by the trained neural network. the network is trained by generating models of earth formations each having a known value of horizontal conductivity, anisotropy parameter, dip angle and azimuth. voltages which would be measured by the transverse electromagnetic induction instrument in response to each model are synthesized. initial estimates from the synthesized voltages are calculated and the initial estimates and known values from each of the models are input to the neural network to cause it to learn a relationship between the initial estimates and the known values. dated 2000-03-28"
6044375,automatic extraction of metadata using a neural network,"a method of automatically extracting metadata from a document. the method of the invention provides a computer readable document that includes blocks comprised of words, an authority list that includes common uses of a set of words, and a neural network trained to extract metadata from groupings of data called compounds. compounds are created with one compound describing each of the blocks. each compound includes the words making up the block, descriptive information about the blocks, and authority information associated with some of the words. the descriptive information may include such items as bounding box information, describing the size and position of the block, and font information, describing the size and type of font the words of the block use. the authority information is located by comparing each the words from the block to the authority list. the compounds are processed through the neural network to generate metadata guesses including word guesses, compound guesses and document guesses along with confidence factors associated with the guesses indicating the likelihood that each of the guesses is correct. the method may additionally include providing a document knowledge base of positioning information and size information for metadata in known documents. if the document knowledge base is provided, then the method includes deriving analysis data from the metadata guess and comparing the analysis data to the document knowledge base to determine metadata output.",2000-03-28,"The title of the patent is automatic extraction of metadata using a neural network and its abstract is a method of automatically extracting metadata from a document. the method of the invention provides a computer readable document that includes blocks comprised of words, an authority list that includes common uses of a set of words, and a neural network trained to extract metadata from groupings of data called compounds. compounds are created with one compound describing each of the blocks. each compound includes the words making up the block, descriptive information about the blocks, and authority information associated with some of the words. the descriptive information may include such items as bounding box information, describing the size and position of the block, and font information, describing the size and type of font the words of the block use. the authority information is located by comparing each the words from the block to the authority list. the compounds are processed through the neural network to generate metadata guesses including word guesses, compound guesses and document guesses along with confidence factors associated with the guesses indicating the likelihood that each of the guesses is correct. the method may additionally include providing a document knowledge base of positioning information and size information for metadata in known documents. if the document knowledge base is provided, then the method includes deriving analysis data from the metadata guess and comparing the analysis data to the document knowledge base to determine metadata output. dated 2000-03-28"
6047221,method for steady-state identification based upon identified dynamics,"a method for modeling a steady-state network in the absence of steady-state historical data. a steady-state neural network can be tied by impressing the dynamics of the system onto the input data during the training operation by first determining the dynamics in a local region of the input space, this providing a set of dynamic training data. this dynamic training data is then utilized to train a dynamic model, gain thereof then set equal to unity such that the dynamic model is now valid over the entire input space. this is a linear model, and the historical data over the entire input space is then processed through this model prior to input to the neural network during training thereof to remove the dynamic component from the data, leaving the steady-state component for the purpose of training. this provides a valid model in the presence of historical data that has a large content of dynamic behavior. a single dynamic model is required for each output variable in a multi-input multi-output steady-state model such that for each output there is a separate dynamic model required for pre-filtering. they are combined in a single network made up of multiple individual steady-state models for each output. the dynamic model can be identified utilizing a weighting factor for the gain to force the dynamic gain of the dynamic model to the steady-state gain by weighting the difference thereof during optimization of the dynamic model. the steady-state model is optimized utilizing gain constraints during the optimization procedure such that the gain of the network is prevented from exceeding the gain constraints.",2000-04-04,"The title of the patent is method for steady-state identification based upon identified dynamics and its abstract is a method for modeling a steady-state network in the absence of steady-state historical data. a steady-state neural network can be tied by impressing the dynamics of the system onto the input data during the training operation by first determining the dynamics in a local region of the input space, this providing a set of dynamic training data. this dynamic training data is then utilized to train a dynamic model, gain thereof then set equal to unity such that the dynamic model is now valid over the entire input space. this is a linear model, and the historical data over the entire input space is then processed through this model prior to input to the neural network during training thereof to remove the dynamic component from the data, leaving the steady-state component for the purpose of training. this provides a valid model in the presence of historical data that has a large content of dynamic behavior. a single dynamic model is required for each output variable in a multi-input multi-output steady-state model such that for each output there is a separate dynamic model required for pre-filtering. they are combined in a single network made up of multiple individual steady-state models for each output. the dynamic model can be identified utilizing a weighting factor for the gain to force the dynamic gain of the dynamic model to the steady-state gain by weighting the difference thereof during optimization of the dynamic model. the steady-state model is optimized utilizing gain constraints during the optimization procedure such that the gain of the network is prevented from exceeding the gain constraints. dated 2000-04-04"
6047276,cellular neural network to implement the unfolded chua's circuit,"a neural cellular network for implementing a so-called chua's circuit, and comprising at least first, second and third cells having respective first and second input terminals and respective state terminals, the first and second input terminals being to receive a first and a second reference signal, respectively, and the first cell, and the second and third cells being of mutually different types.",2000-04-04,"The title of the patent is cellular neural network to implement the unfolded chua's circuit and its abstract is a neural cellular network for implementing a so-called chua's circuit, and comprising at least first, second and third cells having respective first and second input terminals and respective state terminals, the first and second input terminals being to receive a first and a second reference signal, respectively, and the first cell, and the second and third cells being of mutually different types. dated 2000-04-04"
6047277,self-organizing neural network for plain text categorization,"a method and system for natural language processing including using a trained neural network having a plurality of baseline nodes. a connection weight between any selected pair of baseline nodes is described by text strings within the selected pair of nodes. a plurality of text messages are received from a preselected source. for each received message, a non-baseline node associated with a selected one of the baseline nodes is created. a connection weight between any non-baseline node and the associated baseline node is described by the text string within the baseline node and the received text message.",2000-04-04,"The title of the patent is self-organizing neural network for plain text categorization and its abstract is a method and system for natural language processing including using a trained neural network having a plurality of baseline nodes. a connection weight between any selected pair of baseline nodes is described by text strings within the selected pair of nodes. a plurality of text messages are received from a preselected source. for each received message, a non-baseline node associated with a selected one of the baseline nodes is created. a connection weight between any non-baseline node and the associated baseline node is described by the text string within the baseline node and the received text message. dated 2000-04-04"
6049773,automated method for identification of reinsurance claims,""" an automated method for identifying those claims of a raw claims insurance claims database for which reinsurance is applicable. the method develops a database of uniquely clustered catastrophic events such as storm reports. this """"catnodes"""" database may be developed manually or automatically through the use of a neural network approach. a fuzzy degree of belonging is employed to quantify the likelihood that a given insurance claim is properly associated with a given catastrophic event or storm. the assignment of a degree of belonging is derived by approaches which consider four factors: the date of the loss; the location of the loss; the type of the loss; and the presence of special keywords in the claim description. """,2000-04-11,"The title of the patent is automated method for identification of reinsurance claims and its abstract is "" an automated method for identifying those claims of a raw claims insurance claims database for which reinsurance is applicable. the method develops a database of uniquely clustered catastrophic events such as storm reports. this """"catnodes"""" database may be developed manually or automatically through the use of a neural network approach. a fuzzy degree of belonging is employed to quantify the likelihood that a given insurance claim is properly associated with a given catastrophic event or storm. the assignment of a degree of belonging is derived by approaches which consider four factors: the date of the loss; the location of the loss; the type of the loss; and the presence of special keywords in the claim description. "" dated 2000-04-11"
6049793,system for building an artificial neural network,"a system for building an artificial neural network is provided which precisely defines the network's structure of artificial neurons, and non-iteratively determines the synapse-weights and hard limiter threshold of each artificial neuron of the network. the system includes a computer for analyzing input data, which represents patterns of different classes of signals, to generate one or more data points in two or three dimensions representative of the signals in each of the different classes. a distribution of the data points is visualized on a map on an output device coupled to the computer. the data points are clustered on the map into clusters in accordance with the classes associated with the data points, and the map is then partitioned into regions by defining linear boundaries between clusters. the artificial neural network is configured in accordance with the data points, clusters, boundaries, and regions, such that each boundary represents a different artificial neuron of the artificial neural network, and the geometric relationship of the regions on the map to the classes defines the logic connectivity of the artificial neurons. the synaptic weights and threshold of each artificial neuron in the network are graphically determined based on the data points of the map.",2000-04-11,"The title of the patent is system for building an artificial neural network and its abstract is a system for building an artificial neural network is provided which precisely defines the network's structure of artificial neurons, and non-iteratively determines the synapse-weights and hard limiter threshold of each artificial neuron of the network. the system includes a computer for analyzing input data, which represents patterns of different classes of signals, to generate one or more data points in two or three dimensions representative of the signals in each of the different classes. a distribution of the data points is visualized on a map on an output device coupled to the computer. the data points are clustered on the map into clusters in accordance with the classes associated with the data points, and the map is then partitioned into regions by defining linear boundaries between clusters. the artificial neural network is configured in accordance with the data points, clusters, boundaries, and regions, such that each boundary represents a different artificial neuron of the artificial neural network, and the geometric relationship of the regions on the map to the classes defines the logic connectivity of the artificial neurons. the synaptic weights and threshold of each artificial neuron in the network are graphically determined based on the data points of the map. dated 2000-04-11"
6052349,waveform equalizer and memory device having a waveform equalizer,"a waveform equalizer comprises a preamplifier for amplifying the reproduced signal read out of a disk by means of an optical pick-up and the reproduced signal is converted into a digital signal by means of an a/d converter adapted to carrying out sampling operations with the data clock cycle. the reproduced signal waveform of the digital signal is subjected to a waveform equalizing process so as to minimize the mean square error relative to the amplitude of the target equalization waveform by means of linear operations. additionally, the sample values of the reproduced signal waveform that is linearly equalized are nonlinearly equalized by neural network type operations.",2000-04-18,"The title of the patent is waveform equalizer and memory device having a waveform equalizer and its abstract is a waveform equalizer comprises a preamplifier for amplifying the reproduced signal read out of a disk by means of an optical pick-up and the reproduced signal is converted into a digital signal by means of an a/d converter adapted to carrying out sampling operations with the data clock cycle. the reproduced signal waveform of the digital signal is subjected to a waveform equalizing process so as to minimize the mean square error relative to the amplitude of the target equalization waveform by means of linear operations. additionally, the sample values of the reproduced signal waveform that is linearly equalized are nonlinearly equalized by neural network type operations. dated 2000-04-18"
6052679,artificial neural networks including boolean-complete compartments,"artificial neural networks include a plurality of artificial neurons and a plurality of boolean-complete compartments, a respective one of which couples a respective pair of artificial neurons. by providing boolean-complete compartments, spurious complement memories can be avoided. a boolean-complete compartment includes a collection of at least four boolean functions that represent input vectors to the respective pair of artificial neurons. the collection of at least four boolean functions are selected from sixteen possible boolean functions that can represent input vectors to the respective pair of artificial neurons. a count for each of the at least four boolean functions is also provided. the count represents a number of occurrences of each of the at least four boolean functions in input vectors to the respective pair of artificial neurons. in order to read the artificial neural network, the network also includes a collection of transfer functions, a respective one of which is associated with a respective one the sixteen possible boolean functions.",2000-04-18,"The title of the patent is artificial neural networks including boolean-complete compartments and its abstract is artificial neural networks include a plurality of artificial neurons and a plurality of boolean-complete compartments, a respective one of which couples a respective pair of artificial neurons. by providing boolean-complete compartments, spurious complement memories can be avoided. a boolean-complete compartment includes a collection of at least four boolean functions that represent input vectors to the respective pair of artificial neurons. the collection of at least four boolean functions are selected from sixteen possible boolean functions that can represent input vectors to the respective pair of artificial neurons. a count for each of the at least four boolean functions is also provided. the count represents a number of occurrences of each of the at least four boolean functions in input vectors to the respective pair of artificial neurons. in order to read the artificial neural network, the network also includes a collection of transfer functions, a respective one of which is associated with a respective one the sixteen possible boolean functions. dated 2000-04-18"
6055524,model-free adaptive process control,"a model-free adaptive controller is disclosed, which uses a dynamic artificial neural network with a learning algorithm to control any single-variable or multivariable open-loop stable, controllable, and consistently direct-acting or reverse-acting industrial process without requiring any manual tuning, quantitative knowledge of the process, or process identifiers. the need for process knowledge is avoided by substituting 1 for the actual sensitivity function .differential.y(t)/.differential.u(t) of the process.",2000-04-25,"The title of the patent is model-free adaptive process control and its abstract is a model-free adaptive controller is disclosed, which uses a dynamic artificial neural network with a learning algorithm to control any single-variable or multivariable open-loop stable, controllable, and consistently direct-acting or reverse-acting industrial process without requiring any manual tuning, quantitative knowledge of the process, or process identifiers. the need for process knowledge is avoided by substituting 1 for the actual sensitivity function .differential.y(t)/.differential.u(t) of the process. dated 2000-04-25"
6058322,methods for improving the accuracy in differential diagnosis on radiologic examinations,"a computer-aided method for detecting, classifying, and displaying candidate abnormalities, such as microcalcifications and interstitial lung disease in digitized medical images, such as mammograms and chest radiographs, a computer programmed to implement the method, and a data structure for storing required parameters, wherein in the classifying method candidate abnormalities in a digitized medical image are located, regions are generated around one or more of the located candidate abnormalities, features are extracted from at least one of the located candidate abnormalities within the region and from the region itself, the extracted features are applied to a classification technique, such as an artificial neural network (ann) to produce a classification result (i.e., probability of malignancy in the form of a number and a bar graph), and the classification result is displayed along with the digitized medical image annotated with the region and the candidate abnormalities within the region. in the detecting method candidate abnormalities in each of a plurality of digitized medical images are located, regions around one or more of the located candidate abnormalities in each of a plurality of digitized medical images are generated, the plurality of digitized medical images annotated with respective regions and candidate abnormalities within the regions are displayed, and a first indicator (e.g., blue arrow) is superimposed over candidate abnormalities comprising of clusters and a second indicator (e.g., red arrow) is superimposed over candidate abnormalities comprising of masses. in a user modification mode, during classification, a user modifies the located candidate abnormalities, the determined regions, and/or the extracted features, so as to modify the extracted features applied to the classification technique and the displayed results, and, during detection, a user modifies the located candidate abnormalities, the determined regions, and the extracted features, so as to modify the displayed results.",2000-05-02,"The title of the patent is methods for improving the accuracy in differential diagnosis on radiologic examinations and its abstract is a computer-aided method for detecting, classifying, and displaying candidate abnormalities, such as microcalcifications and interstitial lung disease in digitized medical images, such as mammograms and chest radiographs, a computer programmed to implement the method, and a data structure for storing required parameters, wherein in the classifying method candidate abnormalities in a digitized medical image are located, regions are generated around one or more of the located candidate abnormalities, features are extracted from at least one of the located candidate abnormalities within the region and from the region itself, the extracted features are applied to a classification technique, such as an artificial neural network (ann) to produce a classification result (i.e., probability of malignancy in the form of a number and a bar graph), and the classification result is displayed along with the digitized medical image annotated with the region and the candidate abnormalities within the region. in the detecting method candidate abnormalities in each of a plurality of digitized medical images are located, regions around one or more of the located candidate abnormalities in each of a plurality of digitized medical images are generated, the plurality of digitized medical images annotated with respective regions and candidate abnormalities within the regions are displayed, and a first indicator (e.g., blue arrow) is superimposed over candidate abnormalities comprising of clusters and a second indicator (e.g., red arrow) is superimposed over candidate abnormalities comprising of masses. in a user modification mode, during classification, a user modifies the located candidate abnormalities, the determined regions, and/or the extracted features, so as to modify the extracted features applied to the classification technique and the displayed results, and, during detection, a user modifies the located candidate abnormalities, the determined regions, and the extracted features, so as to modify the displayed results. dated 2000-05-02"
6058351,management zones for precision farming,"a method and apparatus for determining management zones in a field for precision farming. the method and apparatus use a self-organizing network, such as a kohonen neural network, to perform the classification of site-specific farming data into management zones in a farm field. the self-organizing network is configured to learn the proper classifications through a learning or training process.",2000-05-02,"The title of the patent is management zones for precision farming and its abstract is a method and apparatus for determining management zones in a field for precision farming. the method and apparatus use a self-organizing network, such as a kohonen neural network, to perform the classification of site-specific farming data into management zones in a farm field. the self-organizing network is configured to learn the proper classifications through a learning or training process. dated 2000-05-02"
6058352,accurate tissue injury assessment using hybrid neural network analysis,systems and methods using a neural network based portable absorption spectrometer system for real-time automatic evaluation of tissue injury are described. an apparatus includes an electromagnetic signal generator; an optical fiber connected to the electromagnetic signal generator; a fiber optic probe connected to the optical fiber; a broad band spectrometer connected to the fiber optic probe; and a hybrid neural network connected to the broad band spectrometer. the hybrid neural network includes a principle component analyzer of broad band spectral data obtained from said broad band spectrometer.,2000-05-02,The title of the patent is accurate tissue injury assessment using hybrid neural network analysis and its abstract is systems and methods using a neural network based portable absorption spectrometer system for real-time automatic evaluation of tissue injury are described. an apparatus includes an electromagnetic signal generator; an optical fiber connected to the electromagnetic signal generator; a fiber optic probe connected to the optical fiber; a broad band spectrometer connected to the fiber optic probe; and a hybrid neural network connected to the broad band spectrometer. the hybrid neural network includes a principle component analyzer of broad band spectral data obtained from said broad band spectrometer. dated 2000-05-02
6058386,device for designing a neural network and neural network,"the invention relates to a device for designing a neural network, in which to determine the number of neurons (21 . . . 24) in the intermediate layer, the domain of the input signal (x1, x2) in question is subdivided into a predefinable number of subdomains, and in the case of a multiplicity n of input signals (x1, x2), the n-dimensional value space of the n input signals is subdivided in conformance with the subdomains in question into n-dimensional partial spaces, and the supporting values (xi, yi) of the training data are assigned to the subdomains or partial spaces, and the subdomains or partial spaces having the most supporting values are selected, and in which case, for each selected subdomain or partial space, provision is made for a neuron in the intermediate layer preceding the output layer. the device according to the invention can be advantageously used for designing neural networks where the training data are unevenly distributed. the invention is generally suited for applications in neural networks.",2000-05-02,"The title of the patent is device for designing a neural network and neural network and its abstract is the invention relates to a device for designing a neural network, in which to determine the number of neurons (21 . . . 24) in the intermediate layer, the domain of the input signal (x1, x2) in question is subdivided into a predefinable number of subdomains, and in the case of a multiplicity n of input signals (x1, x2), the n-dimensional value space of the n input signals is subdivided in conformance with the subdomains in question into n-dimensional partial spaces, and the supporting values (xi, yi) of the training data are assigned to the subdomains or partial spaces, and the subdomains or partial spaces having the most supporting values are selected, and in which case, for each selected subdomain or partial space, provision is made for a neuron in the intermediate layer preceding the output layer. the device according to the invention can be advantageously used for designing neural networks where the training data are unevenly distributed. the invention is generally suited for applications in neural networks. dated 2000-05-02"
6061672,fuzzy logic neural network modular architecture,"the invention relates to a modular architecture of a cellular network for improved large-scale integration, of the type which comprises a plurality of fuzzy cellular elements (c.sub.m,n) interconnected to form a matrix of elements having at least m rows and n columns, the row and column numbers describing the location of each element. each fuzzy processor is adapted for connection to other processors of the same type such that a parallel architecture of the modular type can be implemented. the management of the architecture is facilitated by each submatrix being controlled by an individually dedicated fuzzy processor device.",2000-05-09,"The title of the patent is fuzzy logic neural network modular architecture and its abstract is the invention relates to a modular architecture of a cellular network for improved large-scale integration, of the type which comprises a plurality of fuzzy cellular elements (c.sub.m,n) interconnected to form a matrix of elements having at least m rows and n columns, the row and column numbers describing the location of each element. each fuzzy processor is adapted for connection to other processors of the same type such that a parallel architecture of the modular type can be implemented. the management of the architecture is facilitated by each submatrix being controlled by an individually dedicated fuzzy processor device. dated 2000-05-09"
6064180,method and apparatus for determining battery state-of-charge using neural network architecture,a neural network characterized by a minimal architecture suitable for implementation in conventional microprocessor battery pack monitoring hardware includes linear and non-linear processing elements and battery parameter measurements representative of real time and temporal quantities whereby state of charge estimations actually converge with 100% and 0% states-of-charge.,2000-05-16,The title of the patent is method and apparatus for determining battery state-of-charge using neural network architecture and its abstract is a neural network characterized by a minimal architecture suitable for implementation in conventional microprocessor battery pack monitoring hardware includes linear and non-linear processing elements and battery parameter measurements representative of real time and temporal quantities whereby state of charge estimations actually converge with 100% and 0% states-of-charge. dated 2000-05-16
6064716,plain x-ray bone densitometry apparatus and method,"non-invasive quantitative plain radiographic evaluation of bone in a bony locale of a body is performed by subjecting the bony locale to a broadband collimated x-ray beam having energy in the range of about 20 kev to 150 kev. alongside the bony locale is a material phantom. an energy-selective multiple-film detector cassette containing at least two films is placed under the body and material phantom to receive the transmitted x-ray beam. the films in the cassette are developed and digitally scanned to produce sets of material phantom data and sets of bone data. the data sets are then processed with a feedforward neural network whereby to generate the indicated estimate of bone status, namely, bone-mineral density. in an alternative embodiment, an independent measurement is made of the total tissue thickness, and input to the neural network to achieve higher accuracy and precision.",2000-05-16,"The title of the patent is plain x-ray bone densitometry apparatus and method and its abstract is non-invasive quantitative plain radiographic evaluation of bone in a bony locale of a body is performed by subjecting the bony locale to a broadband collimated x-ray beam having energy in the range of about 20 kev to 150 kev. alongside the bony locale is a material phantom. an energy-selective multiple-film detector cassette containing at least two films is placed under the body and material phantom to receive the transmitted x-ray beam. the films in the cassette are developed and digitally scanned to produce sets of material phantom data and sets of bone data. the data sets are then processed with a feedforward neural network whereby to generate the indicated estimate of bone status, namely, bone-mineral density. in an alternative embodiment, an independent measurement is made of the total tissue thickness, and input to the neural network to achieve higher accuracy and precision. dated 2000-05-16"
6064997,discrete-time tuning of neural network controllers for nonlinear dynamical systems,"a family of novel multi-layer discrete-time neural net controllers is presented for the control of an multi-input multi-output (mimo) dynamical system. no learning phase is needed. the structure of the neural net (nn) controller is derived using a filtered error/passivity approach. for guaranteed stability, the upper bound on the constant learning rate parameter for the delta rule employed in standard back propagation is shown to decrease with the number of hidden-layer neurons so that learning must slow down. this major drawback is shown to be easily overcome by using a projection algorithm in each layer. the notion of persistency of excitation for multilayer nn is defined and explored. new on-line improved tuning algorithms for discrete-time systems are derived, which are similar to e-modification for the case of continuous-time systems, that include a modification to the learning rate parameter plus a correction term. these algorithms guarantee tracking as well as bounded nn weights. an extension of these novel weight tuning updates to nn with an arbitrary number of hidden layers is discussed. the notions of discrete-time passive nn, dissipative nn, and robust nn are introduced.",2000-05-16,"The title of the patent is discrete-time tuning of neural network controllers for nonlinear dynamical systems and its abstract is a family of novel multi-layer discrete-time neural net controllers is presented for the control of an multi-input multi-output (mimo) dynamical system. no learning phase is needed. the structure of the neural net (nn) controller is derived using a filtered error/passivity approach. for guaranteed stability, the upper bound on the constant learning rate parameter for the delta rule employed in standard back propagation is shown to decrease with the number of hidden-layer neurons so that learning must slow down. this major drawback is shown to be easily overcome by using a projection algorithm in each layer. the notion of persistency of excitation for multilayer nn is defined and explored. new on-line improved tuning algorithms for discrete-time systems are derived, which are similar to e-modification for the case of continuous-time systems, that include a modification to the learning rate parameter plus a correction term. these algorithms guarantee tracking as well as bounded nn weights. an extension of these novel weight tuning updates to nn with an arbitrary number of hidden layers is discussed. the notions of discrete-time passive nn, dissipative nn, and robust nn are introduced. dated 2000-05-16"
6067287,neural fuzzy connection admission controller and method in a node of an asynchronous transfer mode (atm) communication network,"a neural fuzzy connection admission controller and method is provided for controlling admission of a new connecting call in a node of a communication network. the neural fuzzy connection admission controller is an inference engine based on a multi-layered neural fuzzy network. the neural fuzzy connection admission controller provides a decision signal z in response to an available bandwidth capacity c.sub.a, a congestion control action y and a cell loss ratio p.sub.l. since the decision signal z is obtained via a learning process, so the decision is closer to a desired result.",2000-05-23,"The title of the patent is neural fuzzy connection admission controller and method in a node of an asynchronous transfer mode (atm) communication network and its abstract is a neural fuzzy connection admission controller and method is provided for controlling admission of a new connecting call in a node of a communication network. the neural fuzzy connection admission controller is an inference engine based on a multi-layered neural fuzzy network. the neural fuzzy connection admission controller provides a decision signal z in response to an available bandwidth capacity c.sub.a, a congestion control action y and a cell loss ratio p.sub.l. since the decision signal z is obtained via a learning process, so the decision is closer to a desired result. dated 2000-05-23"
6067371,method and system for non-invasive temperature mapping of tissue,"a method and system for mapping temperature from image data. the method includes the steps of: receiving an image of tissue comprised of multiple pixels, segregating the image into groups of pixels (104), each group of pixels having a set of descriptors (106), establishing a baseline set of descriptors corresponding to initial conditions of the imaged tissue (108), measuring a differential in the set of descriptors for a group of pixels (114), the differential corresponding to a change in pixel values for the group of pixels, correlating said measured differential to a temperature change for the tissue corresponding to the group of pixels (118-124), and overlaying an indication of temperature over the tissue image in response to said correlated temperature change indicating a change in temperature range for the tissue (130-132). the system includes a digital processor implementing a neural network for evaluating a set of descriptors against baseline values learned by the neural network form the initial conditions of the tissue.",2000-05-23,"The title of the patent is method and system for non-invasive temperature mapping of tissue and its abstract is a method and system for mapping temperature from image data. the method includes the steps of: receiving an image of tissue comprised of multiple pixels, segregating the image into groups of pixels (104), each group of pixels having a set of descriptors (106), establishing a baseline set of descriptors corresponding to initial conditions of the imaged tissue (108), measuring a differential in the set of descriptors for a group of pixels (114), the differential corresponding to a change in pixel values for the group of pixels, correlating said measured differential to a temperature change for the tissue corresponding to the group of pixels (118-124), and overlaying an indication of temperature over the tissue image in response to said correlated temperature change indicating a change in temperature range for the tissue (130-132). the system includes a digital processor implementing a neural network for evaluating a set of descriptors against baseline values learned by the neural network form the initial conditions of the tissue. dated 2000-05-23"
6067535,monitoring and retraining neural network,"a method of managing the processing of information using a first neural network, the information relating to the transmission of messages in a telecommunications network, uses the steps of: pa1 (i) monitoring the performance of the first neural network in processing the information; pa1 (ii) creating a second neural network of the same topology as the first when a predetermined performance threshold is reached, and pa1 (iii) retraining the second neural network while continuing to process the information using the first neural network. if the neural networks are implemented using objects, such retraining can be facilitated by using a persistance mechanism to enable the objects to be stored and moved. applications in fraud detection.",2000-05-23,"The title of the patent is monitoring and retraining neural network and its abstract is a method of managing the processing of information using a first neural network, the information relating to the transmission of messages in a telecommunications network, uses the steps of: pa1 (i) monitoring the performance of the first neural network in processing the information; pa1 (ii) creating a second neural network of the same topology as the first when a predetermined performance threshold is reached, and pa1 (iii) retraining the second neural network while continuing to process the information using the first neural network. if the neural networks are implemented using objects, such retraining can be facilitated by using a persistance mechanism to enable the objects to be stored and moved. applications in fraud detection. dated 2000-05-23"
6067536,neural network for voice and pattern recognition,"a neural network circuit for performing a processing of recognizing voices, images and the like comprises a weight memory for holding a lot of weight values (initial weight values) which correspond to a plurality of input terminals of each of a plurality of neurons forming a neural network and have been initially learned, and a difference value memory for storing difference values between the weight values of the weight memory and additionally learned weight values. the weight memory is formed by a rom. the difference value memory is formed by a sram, for example. during operation of recognizing input data, the initial weight values of the weight memory and the difference values of the difference value memory are added together. the added weight values are used to calculate an output value of each neuron of an output layer. accordingly, the initial weight values can be additionally learned at a high speed by existence of the difference value memory having a small capacity. thus, new numerals, characters and the like can be recognized well without error.",2000-05-23,"The title of the patent is neural network for voice and pattern recognition and its abstract is a neural network circuit for performing a processing of recognizing voices, images and the like comprises a weight memory for holding a lot of weight values (initial weight values) which correspond to a plurality of input terminals of each of a plurality of neurons forming a neural network and have been initially learned, and a difference value memory for storing difference values between the weight values of the weight memory and additionally learned weight values. the weight memory is formed by a rom. the difference value memory is formed by a sram, for example. during operation of recognizing input data, the initial weight values of the weight memory and the difference values of the difference value memory are added together. the added weight values are used to calculate an output value of each neuron of an output layer. accordingly, the initial weight values can be additionally learned at a high speed by existence of the difference value memory having a small capacity. thus, new numerals, characters and the like can be recognized well without error. dated 2000-05-23"
6070098,method of and apparatus for evaluation and mitigation of microsleep events,"a method and apparatus for determining, monitoring and predicting levels of alertness by detecting microsleep episodes includes a plurality of channel processing units and a channel combining unit. each of the channel processing units receives an information channel which conveys information associated with the mental and behavorial state of the subject, such as for example an eeg channel, and classifies the information into a distinct category. such categories may include microsleep, non-microsleep, one or more of a plurality of stages of sleep, one or more of a plurality of stages of wakefulness, or a transition state characterized by a transition from one of the aforementioned states to another. each of the channel processing units includes a neural network which has been trained with a set of example input/result vector pairs. the example input/result vector pairs are generated by correlating actual information channel outputs with observed fatigue related events such as nodding off, head snapping, multiple blinks, blank stares, wide eyes, yawning, partial and complete prolonged eyelid closures, and slow rolling eye movements.",2000-05-30,"The title of the patent is method of and apparatus for evaluation and mitigation of microsleep events and its abstract is a method and apparatus for determining, monitoring and predicting levels of alertness by detecting microsleep episodes includes a plurality of channel processing units and a channel combining unit. each of the channel processing units receives an information channel which conveys information associated with the mental and behavorial state of the subject, such as for example an eeg channel, and classifies the information into a distinct category. such categories may include microsleep, non-microsleep, one or more of a plurality of stages of sleep, one or more of a plurality of stages of wakefulness, or a transition state characterized by a transition from one of the aforementioned states to another. each of the channel processing units includes a neural network which has been trained with a set of example input/result vector pairs. the example input/result vector pairs are generated by correlating actual information channel outputs with observed fatigue related events such as nodding off, head snapping, multiple blinks, blank stares, wide eyes, yawning, partial and complete prolonged eyelid closures, and slow rolling eye movements. dated 2000-05-30"
6073046,heart monitor system,"a medical facility after discharge of a cardiovascular patient, can remain in contact with the patient. the patient is provided with a multiple lead ekg terminal spread placed on the body, and the signals therefrom are collected and transmitted. they are transmitted to a remote central location. at the central location, the transmitted ekg data is analyzed. it is compared with normal ekg signals and signals captured in time from the same patient as part of the patient history. the evaluation is done through a neural network which forms an output signal automatically or through intervention of a cardiologist sending an alarm condition signal to the patient instructing the patient to get immediate treatment at the patient's location or to otherwise go to a specific medical facility. signal preparation includes providing ekg signals through a multiplexer, conversation into a digital data, removal of bias signals, stabilization of this ekg base line, compression and data transmission through a modulator. the receive signal is reconstructed to provide an ekg signal of the patient which is then evaluated in the neural network. as appropriate, transmitter/receiver repeater stations and synchronous satellites are used to convey these signals.",2000-06-06,"The title of the patent is heart monitor system and its abstract is a medical facility after discharge of a cardiovascular patient, can remain in contact with the patient. the patient is provided with a multiple lead ekg terminal spread placed on the body, and the signals therefrom are collected and transmitted. they are transmitted to a remote central location. at the central location, the transmitted ekg data is analyzed. it is compared with normal ekg signals and signals captured in time from the same patient as part of the patient history. the evaluation is done through a neural network which forms an output signal automatically or through intervention of a cardiologist sending an alarm condition signal to the patient instructing the patient to get immediate treatment at the patient's location or to otherwise go to a specific medical facility. signal preparation includes providing ekg signals through a multiplexer, conversation into a digital data, removal of bias signals, stabilization of this ekg base line, compression and data transmission through a modulator. the receive signal is reconstructed to provide an ekg signal of the patient which is then evaluated in the neural network. as appropriate, transmitter/receiver repeater stations and synchronous satellites are used to convey these signals. dated 2000-06-06"
6075884,method and apparatus for training a neural network to learn and use fidelity metric as a control mechanism,"a signal processing apparatus and concomitant method for learning and using fidelity metric as a control mechanism and to process large quantities of fidelity metrics from a visual discrimination measure (vdm) to a manageable subjective image quality ratings. the signal processing apparatus incorporates a vdm and a neural network. the vdm receives input image sequences and generates fidelity metrics, which are received by a neural network. the neural network is trained to learn and use the fidelity metrics as a control mechanism, e.g., to control a video encoder.",2000-06-13,"The title of the patent is method and apparatus for training a neural network to learn and use fidelity metric as a control mechanism and its abstract is a signal processing apparatus and concomitant method for learning and using fidelity metric as a control mechanism and to process large quantities of fidelity metrics from a visual discrimination measure (vdm) to a manageable subjective image quality ratings. the signal processing apparatus incorporates a vdm and a neural network. the vdm receives input image sequences and generates fidelity metrics, which are received by a neural network. the neural network is trained to learn and use the fidelity metrics as a control mechanism, e.g., to control a video encoder. dated 2000-06-13"
6078410,image processing apparatus,"an image processing apparatus includes a feature data extracting circuit for detecting feature data indicative of density characteristics of a document based on image signals inputted from an input terminal, a density correction table selecting circuit composed of a neural circuit network which is learned beforehand so as to recognize image characteristics based on the feature data, and a density correcting circuit for selecting a density correction table in accordance with image characteristics based on a selection signal from the density correction table selecting circuit so that the density of image signals is corrected based on the density correction table. as a result, characteristics of the document are extracted, so that the density of the image signals can be corrected based thereon. as a result, the density characteristics of the document are extracted, and a density correction can be performed based on the extracted density characteristics, thereby obtaining a high quality recorded image with respect to documents of various kinds. furthermore, by adopting the neural circuit network, the density of the image signals can be corrected accurately at high speed.",2000-06-20,"The title of the patent is image processing apparatus and its abstract is an image processing apparatus includes a feature data extracting circuit for detecting feature data indicative of density characteristics of a document based on image signals inputted from an input terminal, a density correction table selecting circuit composed of a neural circuit network which is learned beforehand so as to recognize image characteristics based on the feature data, and a density correcting circuit for selecting a density correction table in accordance with image characteristics based on a selection signal from the density correction table selecting circuit so that the density of image signals is corrected based on the density correction table. as a result, characteristics of the document are extracted, so that the density of the image signals can be corrected based thereon. as a result, the density characteristics of the document are extracted, and a density correction can be performed based on the extracted density characteristics, thereby obtaining a high quality recorded image with respect to documents of various kinds. furthermore, by adopting the neural circuit network, the density of the image signals can be corrected accurately at high speed. dated 2000-06-20"
6078843,neural network including input normalization for use in a closed loop control system,"an apparatus and method for controlling a process using a neural network which operates as part of a closed loop control system. the state of the control system is defined by one or more process condition signals and monitored for a predetermined set of controller parameters. the output of the control system is one or more device control signals, used by a control device to alter a process being controlled. the neural network uses normalized values of process condition signals in combination with a predetermined set of controller parameters to produce correction control signals. the correction control signals are then used to the create device control signals. proper normalization of at least one of the process condition signals using the throttling range set by the controller parameters is necessary. the remaining input signals must be normalized as well, but the method of normalization is not as critical, except to create a common range for all process condition signals input to the neural network.",2000-06-20,"The title of the patent is neural network including input normalization for use in a closed loop control system and its abstract is an apparatus and method for controlling a process using a neural network which operates as part of a closed loop control system. the state of the control system is defined by one or more process condition signals and monitored for a predetermined set of controller parameters. the output of the control system is one or more device control signals, used by a control device to alter a process being controlled. the neural network uses normalized values of process condition signals in combination with a predetermined set of controller parameters to produce correction control signals. the correction control signals are then used to the create device control signals. proper normalization of at least one of the process condition signals using the throttling range set by the controller parameters is necessary. the remaining input signals must be normalized as well, but the method of normalization is not as critical, except to create a common range for all process condition signals input to the neural network. dated 2000-06-20"
6078857,apparatus for deciding a shift pattern suitable for a driver's driving habit using neural network operation and fuzzy inference and a control method thereof,"an apparatus for deciding a shift pattern suitable for a driver's habit using a neural network operation and fuzzy inference and a control method thereof which perform a neural network operation by inputting a driver's driving operation quantity as a deciding condition of a shift pattern, and decide an optimal shift pattern by performing fuzzy inference from the output from the a neural network operation.",2000-06-20,"The title of the patent is apparatus for deciding a shift pattern suitable for a driver's driving habit using neural network operation and fuzzy inference and a control method thereof and its abstract is an apparatus for deciding a shift pattern suitable for a driver's habit using a neural network operation and fuzzy inference and a control method thereof which perform a neural network operation by inputting a driver's driving operation quantity as a deciding condition of a shift pattern, and decide an optimal shift pattern by performing fuzzy inference from the output from the a neural network operation. dated 2000-06-20"
6078946,system and method for management of connection oriented networks,"the system and method for provisioning resources in a network provides real time, parallel evaluation of the best path within the network using neural network principles. elements of the network having a plurality of paths are assigned relative values according to a network user's requirements. attributes may include factors such as reliability, cost, speed, distance, expandability, etc., and may be static or dynamic. selection of a best path from the plurality of paths comprises application of fuzzy logic, using a threshold function to identify a best relative path value by providing an input to the function which is a combination of the attribute values of the elements within each path. the input to the function is the sum of weighted attribute values, where each attribute value is multiplied by a weight which is a relative value determined in accordance with the network user's priorities; the higher the priority, the greater the weight applied to that attribute. computation of the threshold function is performed for each path and the resulting values are compared to a pre-determined threshold value to determine if the threshold has been crossed. if a single optimal path has not been identified during this step, the weights of the various attributes are adjusted in order of their priority; upward if no paths have crossed the threshold and downward if multiple paths have crossed the threshold. the process is continued with increasingly smaller incremental changes in the weights until a single combination of elements provides a path value which crosses the threshold, indicating the best path for meeting the network user's criteria. the threshold function may be any algorithm which provides a threshold, including sigmoid, linear, exponential and quadratic functions.",2000-06-20,"The title of the patent is system and method for management of connection oriented networks and its abstract is the system and method for provisioning resources in a network provides real time, parallel evaluation of the best path within the network using neural network principles. elements of the network having a plurality of paths are assigned relative values according to a network user's requirements. attributes may include factors such as reliability, cost, speed, distance, expandability, etc., and may be static or dynamic. selection of a best path from the plurality of paths comprises application of fuzzy logic, using a threshold function to identify a best relative path value by providing an input to the function which is a combination of the attribute values of the elements within each path. the input to the function is the sum of weighted attribute values, where each attribute value is multiplied by a weight which is a relative value determined in accordance with the network user's priorities; the higher the priority, the greater the weight applied to that attribute. computation of the threshold function is performed for each path and the resulting values are compared to a pre-determined threshold value to determine if the threshold has been crossed. if a single optimal path has not been identified during this step, the weights of the various attributes are adjusted in order of their priority; upward if no paths have crossed the threshold and downward if multiple paths have crossed the threshold. the process is continued with increasingly smaller incremental changes in the weights until a single combination of elements provides a path value which crosses the threshold, indicating the best path for meeting the network user's criteria. the threshold function may be any algorithm which provides a threshold, including sigmoid, linear, exponential and quadratic functions. dated 2000-06-20"
6081766,machine-learning approach to modeling biological activity for molecular design and to modeling other characteristics,"explicit representation of molecular shape of molecules is combined with neural network learning methods to provide models with high predictive ability that generalize to different chemical classes where structurally diverse molecules exhibiting similar surface characteristics are treated as similar. a new machine-learning methodology is disclosed that can accept multiple representations of objects and construct models that predict characteristics of those objects. an extension of this methodology can be applied in cases where the representations of the objects are determined by a set of adjustable parameters. an iterative process applies intermediate models to generate new representations of the objects by adjusting said parameters and repeatedly. retrains the models to obtain better predictive models. this method can be applied to molecules because each molecule can have many orientations and conformations (representations) that are determined by a set of translation, rotation and torsion angle parameters.",2000-06-27,"The title of the patent is machine-learning approach to modeling biological activity for molecular design and to modeling other characteristics and its abstract is explicit representation of molecular shape of molecules is combined with neural network learning methods to provide models with high predictive ability that generalize to different chemical classes where structurally diverse molecules exhibiting similar surface characteristics are treated as similar. a new machine-learning methodology is disclosed that can accept multiple representations of objects and construct models that predict characteristics of those objects. an extension of this methodology can be applied in cases where the representations of the objects are determined by a set of adjustable parameters. an iterative process applies intermediate models to generate new representations of the objects by adjusting said parameters and repeatedly. retrains the models to obtain better predictive models. this method can be applied to molecules because each molecule can have many orientations and conformations (representations) that are determined by a set of translation, rotation and torsion angle parameters. dated 2000-06-27"
6083173,artificial neural network for predicting respiratory disturbances and method for developing the same,"a method for predicting respiratory disturbances, and a method for developing such an artificial neural network. the inputs to the method and to the artificial neural network of the present invention are the answers given by a person to a series of questions. the output of the artificial neural network is a predicted respiratory disturbance index.",2000-07-04,"The title of the patent is artificial neural network for predicting respiratory disturbances and method for developing the same and its abstract is a method for predicting respiratory disturbances, and a method for developing such an artificial neural network. the inputs to the method and to the artificial neural network of the present invention are the answers given by a person to a series of questions. the output of the artificial neural network is a predicted respiratory disturbance index. dated 2000-07-04"
6084510,danger warning and emergency response system and method,"surveillance platforms in airborne craft (8,10), land based vehicles (12), vessels at sea or fixed structures (14) detect dangers using conventional scanners and transmit information signals describing the dangers to a control center (2) which analyzes the data and determines the degree of danger and its geographic extent. the center generates a danger warning and emergency response including a danger index. the warning/response message identifies the degree of danger (danger index 144) and the gps coordinates (142) of the impacted geographic area for a wide region or regions of the earth (figs. 2-6). a vulnerability index (fig. 16) determined using neural networks (figs. 13-14) and fuzzy logic (figs. 15-20) enables a prioritized warning/response. the center broadcasts (18) the danger warning and emergency response (fig. 9) to a large population of remotely located warning devices (11), such as a network of pagers each of which has a gps receiver (6,28). the pagers compare the received danger coordinates with their own gps coordinates and each pager determines the extent to which it is in danger. the warning device automatically issues a warning signal or signals, which may be audible, visual or vibratory, appropriate to the degree of danger. emergency manned vehicles may also directly receive the broadcast warning/response and be immediately alerted to act appropriately relative to the degree of danger. one embodiment broadcasts (16) directly to home t.v.'s and radios (17) which have internal gps receivers and which display/annunciate an emergency message customized to that receiver resulting from the internal comparison of the danger coordinates versus the local receiver coordinates.",2000-07-04,"The title of the patent is danger warning and emergency response system and method and its abstract is surveillance platforms in airborne craft (8,10), land based vehicles (12), vessels at sea or fixed structures (14) detect dangers using conventional scanners and transmit information signals describing the dangers to a control center (2) which analyzes the data and determines the degree of danger and its geographic extent. the center generates a danger warning and emergency response including a danger index. the warning/response message identifies the degree of danger (danger index 144) and the gps coordinates (142) of the impacted geographic area for a wide region or regions of the earth (figs. 2-6). a vulnerability index (fig. 16) determined using neural networks (figs. 13-14) and fuzzy logic (figs. 15-20) enables a prioritized warning/response. the center broadcasts (18) the danger warning and emergency response (fig. 9) to a large population of remotely located warning devices (11), such as a network of pagers each of which has a gps receiver (6,28). the pagers compare the received danger coordinates with their own gps coordinates and each pager determines the extent to which it is in danger. the warning device automatically issues a warning signal or signals, which may be audible, visual or vibratory, appropriate to the degree of danger. emergency manned vehicles may also directly receive the broadcast warning/response and be immediately alerted to act appropriately relative to the degree of danger. one embodiment broadcasts (16) directly to home t.v.'s and radios (17) which have internal gps receivers and which display/annunciate an emergency message customized to that receiver resulting from the internal comparison of the danger coordinates versus the local receiver coordinates. dated 2000-07-04"
6084981,image processing apparatus for performing image converting process by neural network,"an image processing apparatus using a neural network having: an image supplying unit for supplying spatiotemporal data of a predetermined region including a target pixel of an image; and a neural network formed by coupling a plurality of artificial neuron models so as to have at least an input layer, a hidden layer, and an output layer, wherein in the output layer, an input/output converting process is executed by a linear function and data corresponding to a target pixel is outputted from the output layer.",2000-07-04,"The title of the patent is image processing apparatus for performing image converting process by neural network and its abstract is an image processing apparatus using a neural network having: an image supplying unit for supplying spatiotemporal data of a predetermined region including a target pixel of an image; and a neural network formed by coupling a plurality of artificial neuron models so as to have at least an input layer, a hidden layer, and an output layer, wherein in the output layer, an input/output converting process is executed by a linear function and data corresponding to a target pixel is outputted from the output layer. dated 2000-07-04"
6085178,apparatus and method for communicating between an intelligent agent and client computer process using disguised messages,"an intelligent agent and its client communicate using a selector known by both parties to generate and interpret messages and thereby effectively disguise confidential information transmitted in the messages from third parties. moreover, a neural network is used to implement the decision logic and/or the message disguising functions of an agent such that the logic employed in such functions is not readily reverse compiled or scanned by third parties.",2000-07-04,"The title of the patent is apparatus and method for communicating between an intelligent agent and client computer process using disguised messages and its abstract is an intelligent agent and its client communicate using a selector known by both parties to generate and interpret messages and thereby effectively disguise confidential information transmitted in the messages from third parties. moreover, a neural network is used to implement the decision logic and/or the message disguising functions of an agent such that the logic employed in such functions is not readily reverse compiled or scanned by third parties. dated 2000-07-04"
6088475,method and apparatus for forming and correcting color image,a color image forming and correcting method and apparatus employ a combination of neural network and fuzzy logic processing to reduce differences in reproducibility in the image system of device. an information value specifying a color gamut in coordinate system of the perceptual color space is entered and associated with an information value specifying a preset other color gamut by the combined techniques. the color gamut is converted into a preset color gamut in the coordinate system. the information value specifying the converted color gamut is outputted.,2000-07-11,The title of the patent is method and apparatus for forming and correcting color image and its abstract is a color image forming and correcting method and apparatus employ a combination of neural network and fuzzy logic processing to reduce differences in reproducibility in the image system of device. an information value specifying a color gamut in coordinate system of the perceptual color space is entered and associated with an information value specifying a preset other color gamut by the combined techniques. the color gamut is converted into a preset color gamut in the coordinate system. the information value specifying the converted color gamut is outputted. dated 2000-07-11
6090044,system for diagnosing medical conditions using a neural network,"a system for diagnosing medical conditions, such as low back pain (lbp), is provided, whereby a neural network is trained by presentation of large amounts of clinical data and diagnostic outcomes. following training, the system is able to produce the diagnosis from the clinical data. while the present invention may be useful in diagnosing lbp in one embodiment, other applications of the present invention, both in the medical field and in other fields, are also envisioned. this intelligent diagnostic system is less expensive and more accurate than conventional diagnostic methods, and has the unique capability to improve its accuracy over time as more data is analyzed.",2000-07-18,"The title of the patent is system for diagnosing medical conditions using a neural network and its abstract is a system for diagnosing medical conditions, such as low back pain (lbp), is provided, whereby a neural network is trained by presentation of large amounts of clinical data and diagnostic outcomes. following training, the system is able to produce the diagnosis from the clinical data. while the present invention may be useful in diagnosing lbp in one embodiment, other applications of the present invention, both in the medical field and in other fields, are also envisioned. this intelligent diagnostic system is less expensive and more accurate than conventional diagnostic methods, and has the unique capability to improve its accuracy over time as more data is analyzed. dated 2000-07-18"
6091841,method and system for segmenting desired regions in digital mammograms,"a method and system for detecting and displaying clustered microcalcifications in a digital mammogram, wherein a single digital mammogram is first automatically cropped to a breast area sub-image which is then processed by means of an optimized difference of gaussians filter to enhance the appearance of potential microcalcifications in the sub-image. the potential microcalcifications are thresholded, clusters are detected, features are computed for the detected clusters, and the clusters are classified as either suspicious or not suspicious by means of a neural network. thresholding is preferably by sloping local thresholding, but may also be performed by global and dual-local thresholding. the locations in the original digital mammogram of the suspicious detected clustered microcalcifications are indicated. parameters for use in the detection and thresholding portions of the system are computer-optimized by means of a genetic algorithm. the results of the system are optimally combined with a radiologist's observation of the original mammogram by combining the observations with the results, after the radiologist has first accepted or rejected individual detections reported by the system.",2000-07-18,"The title of the patent is method and system for segmenting desired regions in digital mammograms and its abstract is a method and system for detecting and displaying clustered microcalcifications in a digital mammogram, wherein a single digital mammogram is first automatically cropped to a breast area sub-image which is then processed by means of an optimized difference of gaussians filter to enhance the appearance of potential microcalcifications in the sub-image. the potential microcalcifications are thresholded, clusters are detected, features are computed for the detected clusters, and the clusters are classified as either suspicious or not suspicious by means of a neural network. thresholding is preferably by sloping local thresholding, but may also be performed by global and dual-local thresholding. the locations in the original digital mammogram of the suspicious detected clustered microcalcifications are indicated. parameters for use in the detection and thresholding portions of the system are computer-optimized by means of a genetic algorithm. the results of the system are optimally combined with a radiologist's observation of the original mammogram by combining the observations with the results, after the radiologist has first accepted or rejected individual detections reported by the system. dated 2000-07-18"
6091843,method of calibration and real-time analysis of particulates,"a method of analyzing particles for chemical or biological species. spectral images of the particles are acquired. targets are identified in the images and are classified according to morphology type and spectrum type. each target is assigned a value of an extensive property. a descriptor vector is formed, each element of the descriptor vector being the sum of the extensive property values for one target class. the descriptor vector is transformed to a vector of mass concentrations of chemical species of interest, or of number concentrations of biological species of interest, using a relationship determined in a calibration procedure. in the calibration procedure, spectral images of calibration samples of known composition are acquired, and empirical morphology types and spectrum types are inferred from the spectral images. targets are identified in the spectral images, classified according to morphology type and spectrum type, and assigned values of an extensive property. for each calibration sample, a calibration descriptor vector and a calibration concentration vector is formed. a collective relationship between the calibration descriptor vectors and the calibration concentration vectors is found, either by multivariate analysis or by training a neural network.",2000-07-18,"The title of the patent is method of calibration and real-time analysis of particulates and its abstract is a method of analyzing particles for chemical or biological species. spectral images of the particles are acquired. targets are identified in the images and are classified according to morphology type and spectrum type. each target is assigned a value of an extensive property. a descriptor vector is formed, each element of the descriptor vector being the sum of the extensive property values for one target class. the descriptor vector is transformed to a vector of mass concentrations of chemical species of interest, or of number concentrations of biological species of interest, using a relationship determined in a calibration procedure. in the calibration procedure, spectral images of calibration samples of known composition are acquired, and empirical morphology types and spectrum types are inferred from the spectral images. targets are identified in the spectral images, classified according to morphology type and spectrum type, and assigned values of an extensive property. for each calibration sample, a calibration descriptor vector and a calibration concentration vector is formed. a collective relationship between the calibration descriptor vectors and the calibration concentration vectors is found, either by multivariate analysis or by training a neural network. dated 2000-07-18"
6092017,parameter estimation apparatus,"an output parameter estimation apparatus for estimating an output parameter from an input data set that is composed of a plurality of input parameters and that is obtained whenever sampling time series input data. in this output parameter estimation apparatus, a fuzzy inference rule is used to calculate a fitness degree of the input data set in one of a plurality of fields included in a space that is formed using at least one input parameter. according to the calculated fitness degree, introduction routes through which the input data set is to be inputted into a neural network are selected. the neural network is set in a connection condition corresponding to the field to which the input data set belongs, the connection condition having been determined in advance as a result of learning process. with this connection condition, the neural network estimates the output parameter from the input data set inputted through the selected introduction routes.",2000-07-18,"The title of the patent is parameter estimation apparatus and its abstract is an output parameter estimation apparatus for estimating an output parameter from an input data set that is composed of a plurality of input parameters and that is obtained whenever sampling time series input data. in this output parameter estimation apparatus, a fuzzy inference rule is used to calculate a fitness degree of the input data set in one of a plurality of fields included in a space that is formed using at least one input parameter. according to the calculated fitness degree, introduction routes through which the input data set is to be inputted into a neural network are selected. the neural network is set in a connection condition corresponding to the field to which the input data set belongs, the connection condition having been determined in advance as a result of learning process. with this connection condition, the neural network estimates the output parameter from the input data set inputted through the selected introduction routes. dated 2000-07-18"
6092018,trained neural network engine idle speed control system,"a electronic engine control (eec) module executes a neural network processing program to control the idle speed of an internal combustion engine by controlling the bypass air (throttle duty cycle) and the engine's ignition timing. the neural network is defined by a unitary data structure which defmes the network architecture, including the number of node layers, the number of nodes per layer, and the interconnections between nodes. to achieve idle speed control, the neural network processes input signals indicating the current operating state of the engine, including engine speed, the intake mass air flow rate, a desired engine speed, engine temperature, and other variables which influence engine speed, including loads imposed by power steering and air conditioning systems. the network definition data structure holds weight values which determine the manner in which network signals, including the input signals, are combined. the network definition data structures are created by a network training system which utilizes an external training processor which employ dynamic gradient methods to derive network weight values in accordance with a cost function which quantitatively defines system objectives and an identification network which is pretined to provide gradient signals representative of the behavior of the physical plant. the training processor executes training cycles asynchronously with the operation of the eec module in a representative test vehicle.",2000-07-18,"The title of the patent is trained neural network engine idle speed control system and its abstract is a electronic engine control (eec) module executes a neural network processing program to control the idle speed of an internal combustion engine by controlling the bypass air (throttle duty cycle) and the engine's ignition timing. the neural network is defined by a unitary data structure which defmes the network architecture, including the number of node layers, the number of nodes per layer, and the interconnections between nodes. to achieve idle speed control, the neural network processes input signals indicating the current operating state of the engine, including engine speed, the intake mass air flow rate, a desired engine speed, engine temperature, and other variables which influence engine speed, including loads imposed by power steering and air conditioning systems. the network definition data structure holds weight values which determine the manner in which network signals, including the input signals, are combined. the network definition data structures are created by a network training system which utilizes an external training processor which employ dynamic gradient methods to derive network weight values in accordance with a cost function which quantitatively defines system objectives and an identification network which is pretined to provide gradient signals representative of the behavior of the physical plant. the training processor executes training cycles asynchronously with the operation of the eec module in a representative test vehicle. dated 2000-07-18"
6092919,system and method for adaptive control of uncertain nonlinear processes,"a process and neural network architecture for on-line adjustment of the weights of the neural network in a manner that corrects errors made by a nonlinear controller designed based on a model for the dynamics of a process under control. a computer system is provided for controlling the dynamic output response signal of a nonlinear physical process, where the physical process is represented by a fixed model of the process. the computer system includes a controlled device for responding to the output response signal of the system. the computer system also includes a linear controller for providing a pseudo control signal that is based on the fixed model for the process and provides a second controller, connected to the linear controller, for receiving the pseudo control signal and for providing a modified pseudo control signal to correct for the errors made in modeling the nonlinearities in the process. a response network is also included as part of the computer system. the response network receives the modified pseudo control signal and provides the output response signal to the controlled device. the second controller preferably is a neural network. the computer system may include a plurality of neural networks with each neural network designated to control a selected variable or degree of freedom within the system.",2000-07-25,"The title of the patent is system and method for adaptive control of uncertain nonlinear processes and its abstract is a process and neural network architecture for on-line adjustment of the weights of the neural network in a manner that corrects errors made by a nonlinear controller designed based on a model for the dynamics of a process under control. a computer system is provided for controlling the dynamic output response signal of a nonlinear physical process, where the physical process is represented by a fixed model of the process. the computer system includes a controlled device for responding to the output response signal of the system. the computer system also includes a linear controller for providing a pseudo control signal that is based on the fixed model for the process and provides a second controller, connected to the linear controller, for receiving the pseudo control signal and for providing a modified pseudo control signal to correct for the errors made in modeling the nonlinearities in the process. a response network is also included as part of the computer system. the response network receives the modified pseudo control signal and provides the output response signal to the controlled device. the second controller preferably is a neural network. the computer system may include a plurality of neural networks with each neural network designated to control a selected variable or degree of freedom within the system. dated 2000-07-25"
6098012,neural network based transient fuel control method,"a system and method for use in a motor vehicles is disclosed for calculating a fuel multiplier during transient engine operation. the fuel multiplier modifies the amount of fuel released from a fuel actuator into an engine. the fuel control system uses neural network logic to establish the fuel multiplier. the neural network logic involves taking inputs from engine sensors, processing the inputs through an input layer, a hidden layer and an output layer resulting in a fuel multiplier.",2000-08-01,"The title of the patent is neural network based transient fuel control method and its abstract is a system and method for use in a motor vehicles is disclosed for calculating a fuel multiplier during transient engine operation. the fuel multiplier modifies the amount of fuel released from a fuel actuator into an engine. the fuel control system uses neural network logic to establish the fuel multiplier. the neural network logic involves taking inputs from engine sensors, processing the inputs through an input layer, a hidden layer and an output layer resulting in a fuel multiplier. dated 2000-08-01"
6098060,process controlling method and device,"for controlling a process, manipulated variables for a plurality of actuators that act on the process are calculated in a control device from measured values for output quantities of the process. to optimize the control of the process by control device (7), the actuator efficiencies (w.sub.11, . . . w.sub.nm) that describe the dependence of the changes in the output quantities (dy.sub.1, . . . , dy.sub.m) on the changes in the manipulated variables (du.sub.1, . . . , du.sub.n) for each actuator are learned in a neural network (10) and sent to the control device (7) to improve the calculation of the manipulated variables (u.sub.1, . . . , u.sub.n)",2000-08-01,"The title of the patent is process controlling method and device and its abstract is for controlling a process, manipulated variables for a plurality of actuators that act on the process are calculated in a control device from measured values for output quantities of the process. to optimize the control of the process by control device (7), the actuator efficiencies (w.sub.11, . . . w.sub.nm) that describe the dependence of the changes in the output quantities (dy.sub.1, . . . , dy.sub.m) on the changes in the manipulated variables (du.sub.1, . . . , du.sub.n) for each actuator are learned in a neural network (10) and sent to the control device (7) to improve the calculation of the manipulated variables (u.sub.1, . . . , u.sub.n) dated 2000-08-01"
6098310,system and method for predicting the dryness of clothing articles,"a system and method for predicting the dryness of clothing articles in a clothes dryer. in one embodiment, the clothes dryer uses a temperature sensor, a phase angle sensor, and a humidity sensor to generate signal representations of the temperature of the clothing articles, the motor phase angle, and the humidity of the heated air in the duct, respectively. a controller receives the signal representations and determines a feature vector. a neural network uses the feature vector to predict a percentage of moisture content and a degree of dryness of the clothing articles in the clothes dryer. in another embodiment, the clothes dryer uses a combination of sensors to predict a percentage of moisture content and a degree of dryness of the clothing articles.",2000-08-08,"The title of the patent is system and method for predicting the dryness of clothing articles and its abstract is a system and method for predicting the dryness of clothing articles in a clothes dryer. in one embodiment, the clothes dryer uses a temperature sensor, a phase angle sensor, and a humidity sensor to generate signal representations of the temperature of the clothing articles, the motor phase angle, and the humidity of the heated air in the duct, respectively. a controller receives the signal representations and determines a feature vector. a neural network uses the feature vector to predict a percentage of moisture content and a degree of dryness of the clothing articles in the clothes dryer. in another embodiment, the clothes dryer uses a combination of sensors to predict a percentage of moisture content and a degree of dryness of the clothing articles. dated 2000-08-08"
6100989,method and device for detecting defects in textile webs,"the invention relates to a method and a device for detecting defects in textile webs. in order to rapidly adapt devices of this type to widely varying textile webs and to be able to operate such devices simply, brightness values are determined from the web and are supplied directly to a filter constructed as a neural network. the output results of the neural network can be displayed as grayscale values to indicate detected defects.",2000-08-08,"The title of the patent is method and device for detecting defects in textile webs and its abstract is the invention relates to a method and a device for detecting defects in textile webs. in order to rapidly adapt devices of this type to widely varying textile webs and to be able to operate such devices simply, brightness values are determined from the web and are supplied directly to a filter constructed as a neural network. the output results of the neural network can be displayed as grayscale values to indicate detected defects. dated 2000-08-08"
6101270,neural network architecture for recognition of upright and rotated characters,"a neural network architecture is provided for optical character recognition from an input image in which the target character may be rotated in the image plane. the architecture includes hidden units whose inputs receive image information from portions of the image which are rotationally distributed. that is, the local link between input units and the hidden units is adequate for rotation of the character in the image. therefore, regardless of the orientation of the image, it is right side up, or approximately so, with respect to one of the hidden units. the hidden units have corresponding inputs with corresponding weight factors, i.e., symmetric weight sharing. thus, regardless of the orientation of the image, one of the hidden units will produce a high output value indicative of an upright character. alternatively, a single hidden unit has groups of inputs, each group having a corresponding set of weight factors. the groups are coupled to input units for rotationally distributed portions of the image. therefore, for any orientation of the image corresponding with one of the groups of inputs of the hidden unit, the hidden unit produces the same output value. in an preferred embodiment, feature information such as local contour direction information is provided to the input units. the feature information is provided with respect to slices of the image taken in different directions.",2000-08-08,"The title of the patent is neural network architecture for recognition of upright and rotated characters and its abstract is a neural network architecture is provided for optical character recognition from an input image in which the target character may be rotated in the image plane. the architecture includes hidden units whose inputs receive image information from portions of the image which are rotationally distributed. that is, the local link between input units and the hidden units is adequate for rotation of the character in the image. therefore, regardless of the orientation of the image, it is right side up, or approximately so, with respect to one of the hidden units. the hidden units have corresponding inputs with corresponding weight factors, i.e., symmetric weight sharing. thus, regardless of the orientation of the image, one of the hidden units will produce a high output value indicative of an upright character. alternatively, a single hidden unit has groups of inputs, each group having a corresponding set of weight factors. the groups are coupled to input units for rotationally distributed portions of the image. therefore, for any orientation of the image corresponding with one of the groups of inputs of the hidden unit, the hidden unit produces the same output value. in an preferred embodiment, feature information such as local contour direction information is provided to the input units. the feature information is provided with respect to slices of the image taken in different directions. dated 2000-08-08"
6101450,stress analysis using a defect-free four-node finite element technique,"stress, strain and displacement in two-dimensional regions can be calculated by computerized techniques. in the x-y plane, a region is discretized into four-node finite elements such as quadrilaterals and triangles with side nodes. each such element has eight distinct deformation modes. these correspond to three rigid-body displacements, three uniform strain profiles for compressible materials or two deviatoric strain fields accompanied by an isotropic pressure for incompressible materials, and two flexures. pointwise equilibrium requires the bending shapes to be functions of poisson's ratio. nodal equilibrium and compatibility are satisfied for prescribed loads implementing exact differentiation and integration. element volume change parameters for compressible materials or isotropic pressure for incompressible materials are eliminated, and a linear system of equations is formed in terms of the seven remaining unknowns per element employing a neural-network solution strategy to pass simultaneously the patch and zero-locking tests. the strain and stress distributions including isotropic pressure for incompressible materials, and displacement profiles are solved as x-y polynomials. this technique, tessellica, can be used in computer-aided design, and can be integrated in subsequent manufacture or construction of buildings, bridges, dams, ships, aircraft and automobiles, for example, and in bioengineering applications.",2000-08-08,"The title of the patent is stress analysis using a defect-free four-node finite element technique and its abstract is stress, strain and displacement in two-dimensional regions can be calculated by computerized techniques. in the x-y plane, a region is discretized into four-node finite elements such as quadrilaterals and triangles with side nodes. each such element has eight distinct deformation modes. these correspond to three rigid-body displacements, three uniform strain profiles for compressible materials or two deviatoric strain fields accompanied by an isotropic pressure for incompressible materials, and two flexures. pointwise equilibrium requires the bending shapes to be functions of poisson's ratio. nodal equilibrium and compatibility are satisfied for prescribed loads implementing exact differentiation and integration. element volume change parameters for compressible materials or isotropic pressure for incompressible materials are eliminated, and a linear system of equations is formed in terms of the seven remaining unknowns per element employing a neural-network solution strategy to pass simultaneously the patch and zero-locking tests. the strain and stress distributions including isotropic pressure for incompressible materials, and displacement profiles are solved as x-y polynomials. this technique, tessellica, can be used in computer-aided design, and can be integrated in subsequent manufacture or construction of buildings, bridges, dams, ships, aircraft and automobiles, for example, and in bioengineering applications. dated 2000-08-08"
6105015,wavelet-based hybrid neurosystem for classifying a signal or an image represented by the signal in a data system,"the present invention relates to a system and a method for signal classification. the system comprises a sensor array for receiving a series of input signals such as acoustic signals, pixel-based image signal (such as from infrared images detectors), light signals, temperature signals, etc., a wavelet transform module for transforming the input signals so that characteristics of the signals are represented in the form of wavelet transform coefficients and an array of hybrid neural networks for classifying the signals into multiple distinct categories and generating a classification output signal. the hybrid neural networks each comprise a location neural network for processing data embedded in the frequency versus time location segment of the output of the transform module, a magnitude neural network for processing magnitude information embedded in the magnitude segment of the output of the transform module, and a classification neural network for processing the outputs from the location and magnitude neural networks. a method for processing the signal using the system of the present invention is also described.",2000-08-15,"The title of the patent is wavelet-based hybrid neurosystem for classifying a signal or an image represented by the signal in a data system and its abstract is the present invention relates to a system and a method for signal classification. the system comprises a sensor array for receiving a series of input signals such as acoustic signals, pixel-based image signal (such as from infrared images detectors), light signals, temperature signals, etc., a wavelet transform module for transforming the input signals so that characteristics of the signals are represented in the form of wavelet transform coefficients and an array of hybrid neural networks for classifying the signals into multiple distinct categories and generating a classification output signal. the hybrid neural networks each comprise a location neural network for processing data embedded in the frequency versus time location segment of the output of the transform module, a magnitude neural network for processing magnitude information embedded in the magnitude segment of the output of the transform module, and a classification neural network for processing the outputs from the location and magnitude neural networks. a method for processing the signal using the system of the present invention is also described. dated 2000-08-15"
6108616,process diagnosis system and method for the diagnosis of processes and states in an technical process,"the invention relates to a process diagnosis system and a method for the diagnosis of processes and states in a technical process, in particular for the diagnosis of a power station process. the invention is a structured multi-agent system which corresponds to the process and has a plurality of autonomous diagnostic agents. the diagnostic agents in each case contain a neural network, with whose aid a reference behavior of process components that are to be monitored can be learned. in addition, automatic adaptation to a new reference behavior is also possible by the diagnostic agents.",2000-08-22,"The title of the patent is process diagnosis system and method for the diagnosis of processes and states in an technical process and its abstract is the invention relates to a process diagnosis system and a method for the diagnosis of processes and states in a technical process, in particular for the diagnosis of a power station process. the invention is a structured multi-agent system which corresponds to the process and has a plurality of autonomous diagnostic agents. the diagnostic agents in each case contain a neural network, with whose aid a reference behavior of process components that are to be monitored can be learned. in addition, automatic adaptation to a new reference behavior is also possible by the diagnostic agents. dated 2000-08-22"
6108648,optimizer with neural network estimator,"a computer operated apparatus estimates values needed by an optimizer in a database management system (dbms). the dbms has one or more tables for storing data, each table having zero or more columns of user-definable data types and zero or more associated user-defined routines (udrs). the apparatus has a feature vector extractor connected to the database tables for converting the udr inputs into a base type representation. a neural network receives the feature vector and generates estimated values which are provided to the optimizer of the dbms. the neural network can be trained periodically using randomly generated queries, or it can be trained dynamically by capturing data generated during a query. during operation, the optimizer dynamically invokes the neural network to generate estimates such as selectivity and cost per call for determining optimum query search sequence.",2000-08-22,"The title of the patent is optimizer with neural network estimator and its abstract is a computer operated apparatus estimates values needed by an optimizer in a database management system (dbms). the dbms has one or more tables for storing data, each table having zero or more columns of user-definable data types and zero or more associated user-defined routines (udrs). the apparatus has a feature vector extractor connected to the database tables for converting the udr inputs into a base type representation. a neural network receives the feature vector and generates estimated values which are provided to the optimizer of the dbms. the neural network can be trained periodically using randomly generated queries, or it can be trained dynamically by capturing data generated during a query. during operation, the optimizer dynamically invokes the neural network to generate estimates such as selectivity and cost per call for determining optimum query search sequence. dated 2000-08-22"
6109270,multimodality instrument for tissue characterization,"a system with multimodality instrument for tissue identification includes a computer-controlled motor driven heuristic probe with a multisensory tip. for neurosurgical applications, the instrument is mounted on a stereotactic frame for the probe to penetrate the brain in a precisely controlled fashion. the resistance of the brain tissue being penetrated is continually monitored by a miniaturized strain gauge attached to the probe tip. other modality sensors may be mounted near the probe tip to provide real-time tissue characterizations and the ability to detect the proximity of blood vessels, thus eliminating errors normally associated with registration of pre-operative scans, tissue swelling, elastic tissue deformation, human judgement, etc., and rendering surgical procedures safer, more accurate, and efficient. a neural network program adaptively learns the information on resistance and other characteristic features of normal brain tissue during the surgery and provides near real-time modeling. a fuzzy logic interface to the neural network program incorporates expert medical knowledge in the learning process. identification of abnormal brain tissue is determined by the detection of change and comparison with previously learned models of abnormal brain tissues. the operation of the instrument is controlled through a user friendly graphical interface. patient data is presented in a 3d stereographics display. acoustic feedback of selected information may optionally be provided. upon detection of the close proximity to blood vessels or abnormal brain tissue, the computer-controlled motor immediately stops probe penetration.",2000-08-29,"The title of the patent is multimodality instrument for tissue characterization and its abstract is a system with multimodality instrument for tissue identification includes a computer-controlled motor driven heuristic probe with a multisensory tip. for neurosurgical applications, the instrument is mounted on a stereotactic frame for the probe to penetrate the brain in a precisely controlled fashion. the resistance of the brain tissue being penetrated is continually monitored by a miniaturized strain gauge attached to the probe tip. other modality sensors may be mounted near the probe tip to provide real-time tissue characterizations and the ability to detect the proximity of blood vessels, thus eliminating errors normally associated with registration of pre-operative scans, tissue swelling, elastic tissue deformation, human judgement, etc., and rendering surgical procedures safer, more accurate, and efficient. a neural network program adaptively learns the information on resistance and other characteristic features of normal brain tissue during the surgery and provides near real-time modeling. a fuzzy logic interface to the neural network program incorporates expert medical knowledge in the learning process. identification of abnormal brain tissue is determined by the detection of change and comparison with previously learned models of abnormal brain tissues. the operation of the instrument is controlled through a user friendly graphical interface. patient data is presented in a 3d stereographics display. acoustic feedback of selected information may optionally be provided. upon detection of the close proximity to blood vessels or abnormal brain tissue, the computer-controlled motor immediately stops probe penetration. dated 2000-08-29"
6111602,method and apparatus for inspecting solder joints,"a method and apparatus for inspecting solder joints are provided. a method for inspecting test solder joints includes the steps of illuminating the test solder joints on a printed circuit board to obtain sample images of the test solder joints by grabbing reflected light with a charge coupled device (ccd) camera, classifying the sample images by an inspector into a number of classes according to the soldering quality, inputting a specific sample image belonging to each class to a neural network to divide the class into a predetermined number of clusters; determining a synaptic weight of each class by learning each cluster, determining a confirmed synaptic weight by adjusting the synaptic weight according to a boundary condition between the clusters belonging to neighboring different classes, and selecting a similar cluster by comparing the similarities between the outputs of the neural network with respect to the test solder joints based on confirmed synaptic weights, and each output of the inspector's class. the method and apparatus for inspecting solder joints can guarantee satisfactory classification accuracy with the minimum number of prototype images.",2000-08-29,"The title of the patent is method and apparatus for inspecting solder joints and its abstract is a method and apparatus for inspecting solder joints are provided. a method for inspecting test solder joints includes the steps of illuminating the test solder joints on a printed circuit board to obtain sample images of the test solder joints by grabbing reflected light with a charge coupled device (ccd) camera, classifying the sample images by an inspector into a number of classes according to the soldering quality, inputting a specific sample image belonging to each class to a neural network to divide the class into a predetermined number of clusters; determining a synaptic weight of each class by learning each cluster, determining a confirmed synaptic weight by adjusting the synaptic weight according to a boundary condition between the clusters belonging to neighboring different classes, and selecting a similar cluster by comparing the similarities between the outputs of the neural network with respect to the test solder joints based on confirmed synaptic weights, and each output of the inspector's class. the method and apparatus for inspecting solder joints can guarantee satisfactory classification accuracy with the minimum number of prototype images. dated 2000-08-29"
6113548,method and apparatus for estimation of beat-to-beat pulmonary wedge pressure,"a medical device for estimation of pulmonary wedge pressure wherein a non-occluded pulmonary artery blood pressure measurement is utilized to directly estimate the pulmonary wedge pressure. a neural network is trained with occlusion-obtained data, whereafter the trained coefficients are utilized to implement the wedge pressure estimator. a flow-directed catheter is utilized to transduce the pressure waveform, which is then input to the processing computer through a analog-to-digital data acquisition board. the data is preprocessed in the computer in order to present the neural network with 11 samples of blood pressure data and 11 samples of time-correlated first derivatives of the blood pressure data as well as an indication of the length in time of the heartbeat. the trained neural network then directly outputs the estimated wedge pressure.",2000-09-05,"The title of the patent is method and apparatus for estimation of beat-to-beat pulmonary wedge pressure and its abstract is a medical device for estimation of pulmonary wedge pressure wherein a non-occluded pulmonary artery blood pressure measurement is utilized to directly estimate the pulmonary wedge pressure. a neural network is trained with occlusion-obtained data, whereafter the trained coefficients are utilized to implement the wedge pressure estimator. a flow-directed catheter is utilized to transduce the pressure waveform, which is then input to the processing computer through a analog-to-digital data acquisition board. the data is preprocessed in the computer in order to present the neural network with 11 samples of blood pressure data and 11 samples of time-correlated first derivatives of the blood pressure data as well as an indication of the length in time of the heartbeat. the trained neural network then directly outputs the estimated wedge pressure. dated 2000-09-05"
6114976,vehicle emergency warning and control system,"an automatic vehicle warning and control program is provided for determining if safety enhancing actions are appropriate. the on-line determination of an action that results in a preferred outcome (e.g., aircraft ejection) is made using a neural network controller. the neural network controller is trained off-line using appropriate preferred outcome data obtained via computer simulation or experimentation. appropriate actions are established for all conceivable sets of vehicle conditions. on-line, the neural network controller uses actual sensed vehicle conditions to determine the appropriate action. various actions can be performed based on the preferred outcome determination. appropriate actions can include commanding audible and visual warnings, guidance cues, automatic vehicle control, and aircraft automatic ejection.",2000-09-05,"The title of the patent is vehicle emergency warning and control system and its abstract is an automatic vehicle warning and control program is provided for determining if safety enhancing actions are appropriate. the on-line determination of an action that results in a preferred outcome (e.g., aircraft ejection) is made using a neural network controller. the neural network controller is trained off-line using appropriate preferred outcome data obtained via computer simulation or experimentation. appropriate actions are established for all conceivable sets of vehicle conditions. on-line, the neural network controller uses actual sensed vehicle conditions to determine the appropriate action. various actions can be performed based on the preferred outcome determination. appropriate actions can include commanding audible and visual warnings, guidance cues, automatic vehicle control, and aircraft automatic ejection. dated 2000-09-05"
6115488,method and system for combining automated detections from digital mammograms with observed detections of a human interpreter,"a method and system for detecting and displaying clustered microcalcifications in a digital mammogram, wherein a single digital mammogram is first automatically cropped to a breast area sub-image which is then processed by means of an optimized difference of gaussians filter to enhance the appearance of potential microcalcifications in the sub-image. the potential microcalcifications are thresholded, clusters are detected, features are computed for the detected clusters, and the clusters are classified as either suspicious or not suspicious by means of a neural network. thresholding is preferably by sloping local thresholding, but may also be performed by global and dual-local thresholding. the locations in the original digital mammogram of the suspicious detected clustered microcalcifications are indicated. parameters for use in the detection and thresholding portions of the system are computer-optimized by means of a genetic algorithm. the results of the system are optimally combined with a radiologist's observation of the original mammogram by combining the observations with the results, after the radiologist has first accepted or rejected individual detections reported by the system.",2000-09-05,"The title of the patent is method and system for combining automated detections from digital mammograms with observed detections of a human interpreter and its abstract is a method and system for detecting and displaying clustered microcalcifications in a digital mammogram, wherein a single digital mammogram is first automatically cropped to a breast area sub-image which is then processed by means of an optimized difference of gaussians filter to enhance the appearance of potential microcalcifications in the sub-image. the potential microcalcifications are thresholded, clusters are detected, features are computed for the detected clusters, and the clusters are classified as either suspicious or not suspicious by means of a neural network. thresholding is preferably by sloping local thresholding, but may also be performed by global and dual-local thresholding. the locations in the original digital mammogram of the suspicious detected clustered microcalcifications are indicated. parameters for use in the detection and thresholding portions of the system are computer-optimized by means of a genetic algorithm. the results of the system are optimally combined with a radiologist's observation of the original mammogram by combining the observations with the results, after the radiologist has first accepted or rejected individual detections reported by the system. dated 2000-09-05"
6115701,neural network-based target seeking system,"a system and process for readily determining, for a specified knowledge domain in a given field of endeavor, perturbations applicable to an artificial neural network embodying such a specified knowledge domain that will produce a desired output, comprising a first, previously trained, artificial neural network containing training in some problem domain, which neural network is responsive to the presentment of a set of data inputs at the input portion thereof to produce a set of data outputs at the output portion thereof, a monitoring portion which constantly monitors the outputs of the first neural network to identify the desired outputs, and a network perturbation portion for effecting the application of perturbations, either externally or internally, to the first neural network to thereby effect changes in the output thereof. the perturbations may be effected by any number of different means, including by, but not limited to, presentment of new, varied data inputs, alteration or fixed or previously applied data inputs, such as by the introduction of noise to the inputs, relaxation or degradation of the network, and so forth, either randomly or systematically, and may be accomplished autonomously or upon specific external authorization or control. identification of a desired output establishes an input/perturbation/output mapping relationship from which data inputs (external perturbations) and/or knowledge domain alterations (internal perturbations) that produce the desired output can be determined. the system and process can be employed in some instances and in some embodiments as a target seeking system for use with various design or problem solving applications, and can, in some embodiments, comprise or be comprised of a system and process for autonomously producing and identifying desirable design concepts through utilization of such a target seeking system.",2000-09-05,"The title of the patent is neural network-based target seeking system and its abstract is a system and process for readily determining, for a specified knowledge domain in a given field of endeavor, perturbations applicable to an artificial neural network embodying such a specified knowledge domain that will produce a desired output, comprising a first, previously trained, artificial neural network containing training in some problem domain, which neural network is responsive to the presentment of a set of data inputs at the input portion thereof to produce a set of data outputs at the output portion thereof, a monitoring portion which constantly monitors the outputs of the first neural network to identify the desired outputs, and a network perturbation portion for effecting the application of perturbations, either externally or internally, to the first neural network to thereby effect changes in the output thereof. the perturbations may be effected by any number of different means, including by, but not limited to, presentment of new, varied data inputs, alteration or fixed or previously applied data inputs, such as by the introduction of noise to the inputs, relaxation or degradation of the network, and so forth, either randomly or systematically, and may be accomplished autonomously or upon specific external authorization or control. identification of a desired output establishes an input/perturbation/output mapping relationship from which data inputs (external perturbations) and/or knowledge domain alterations (internal perturbations) that produce the desired output can be determined. the system and process can be employed in some instances and in some embodiments as a target seeking system for use with various design or problem solving applications, and can, in some embodiments, comprise or be comprised of a system and process for autonomously producing and identifying desirable design concepts through utilization of such a target seeking system. dated 2000-09-05"
6118066,autonomous undersea platform,"the invention provides an unmanned, autonomous, undersea platform which ends the sphere of influence of a host vessel from which it is launched. the platform is hydrodynamically and stealth shaped to minimize noise, wake and detectability of the platform. the platform includes advanced active and passive sensors for monitoring the undersea environment, high data rate rf and satellite communications capabilities for communication with the host vessel when necessary, forward deployed offensive and defensive weapons systems and sophisticated data processing to coordinate the sensing, communications and weapons systems with minimal direction from the host vessel. prior to launch, mission directives are input to the data processors. using artificial intelligence and neural network programming for decision making, the processors constantly update the platform's operating parameters to conform with changing environmental and threat conditions and to successfully complete the mission.",2000-09-12,"The title of the patent is autonomous undersea platform and its abstract is the invention provides an unmanned, autonomous, undersea platform which ends the sphere of influence of a host vessel from which it is launched. the platform is hydrodynamically and stealth shaped to minimize noise, wake and detectability of the platform. the platform includes advanced active and passive sensors for monitoring the undersea environment, high data rate rf and satellite communications capabilities for communication with the host vessel when necessary, forward deployed offensive and defensive weapons systems and sophisticated data processing to coordinate the sensing, communications and weapons systems with minimal direction from the host vessel. prior to launch, mission directives are input to the data processors. using artificial intelligence and neural network programming for decision making, the processors constantly update the platform's operating parameters to conform with changing environmental and threat conditions and to successfully complete the mission. dated 2000-09-12"
6118850,analysis methods for energy dispersive x-ray diffraction patterns,"energy dispersive x-ray diffraction spectra are obtained from numerous volume elements within an object. a feature set such as a set of cepstrum coefficients is extracted from each spectrum and classified by a trained classifier such as a neural network to provide an indication of whether or not contraband such as explosives is present in the volume element. indications for adjacent volume elements are evaluated in conjunction with one another, as by an erosion process, to suppress isolated indications and thereby suppress false alarms.",2000-09-12,"The title of the patent is analysis methods for energy dispersive x-ray diffraction patterns and its abstract is energy dispersive x-ray diffraction spectra are obtained from numerous volume elements within an object. a feature set such as a set of cepstrum coefficients is extracted from each spectrum and classified by a trained classifier such as a neural network to provide an indication of whether or not contraband such as explosives is present in the volume element. indications for adjacent volume elements are evaluated in conjunction with one another, as by an erosion process, to suppress isolated indications and thereby suppress false alarms. dated 2000-09-12"
6119111,neuro-parity pattern recognition system and method,"a method and system for monitoring a process and determining its condition. initial data is sensed, a first set of virtual data is produced by applying a system state analyzation to the initial data, a second set of virtual data is produced by applying a neural network analyzation to the initial data and a parity space analyzation is applied to the first and second set of virtual data and also to the initial data to provide a parity space decision about the condition of the process. a logic test can further be applied to produce a further system decision about the state of the process.",2000-09-12,"The title of the patent is neuro-parity pattern recognition system and method and its abstract is a method and system for monitoring a process and determining its condition. initial data is sensed, a first set of virtual data is produced by applying a system state analyzation to the initial data, a second set of virtual data is produced by applying a neural network analyzation to the initial data and a parity space analyzation is applied to the first and second set of virtual data and also to the initial data to provide a parity space decision about the condition of the process. a logic test can further be applied to produce a further system decision about the state of the process. dated 2000-09-12"
6119112,optimum cessation of training in neural networks,"a system and method for training a neural network that ceases training at or near the optimally trained point is presented. a neural network having an input layer, a hidden layer, and an output layer with each layer having one or more nodes is presented. each node in the input layer is connected to each node in the hidden layer and each node in the hidden layer is connected to each node in the output layer. each connection between nodes has an associated weight. all nodes in the input layer are connected to a different historical datum from the set of historical data. the neural network being operative by outputting a prediction or classification, the output of the output layer nodes, when presented with input data. the weights associated with the connections of the neural network are first adjusted by a training device. the training device then iteratively applies a training set to the neural network, the training set consisting of historical data. after each iteration the weights associated with the connections are adjusted according to the difference between the prediction or classification produced by the neural network given the training data and the known prediction or classification of the historical data. additionally, after each iteration, a test set, consisting of different historical data from that in the training set, is presented to the neural network. the training device then determines the difference between the known result from the test set and the result from the presentation of the test set to the neural network. this difference, herein referred to as the variance, is then recorded along with the weights in the neural network. the variance is monitored at each iteration to determine if it is monotonically, within a given margin of error, decreasing. that is the prediction or classification resulting from the test set being presented to the neural network is getting successively closer to matching the known result from the test set. when the variance hits the inflection point where it begins to increase, training is ceased. at this point the neural network is no longer learning the pattern underlying the input data, but is instead over fitting the input data.",2000-09-12,"The title of the patent is optimum cessation of training in neural networks and its abstract is a system and method for training a neural network that ceases training at or near the optimally trained point is presented. a neural network having an input layer, a hidden layer, and an output layer with each layer having one or more nodes is presented. each node in the input layer is connected to each node in the hidden layer and each node in the hidden layer is connected to each node in the output layer. each connection between nodes has an associated weight. all nodes in the input layer are connected to a different historical datum from the set of historical data. the neural network being operative by outputting a prediction or classification, the output of the output layer nodes, when presented with input data. the weights associated with the connections of the neural network are first adjusted by a training device. the training device then iteratively applies a training set to the neural network, the training set consisting of historical data. after each iteration the weights associated with the connections are adjusted according to the difference between the prediction or classification produced by the neural network given the training data and the known prediction or classification of the historical data. additionally, after each iteration, a test set, consisting of different historical data from that in the training set, is presented to the neural network. the training device then determines the difference between the known result from the test set and the result from the presentation of the test set to the neural network. this difference, herein referred to as the variance, is then recorded along with the weights in the neural network. the variance is monitored at each iteration to determine if it is monotonically, within a given margin of error, decreasing. that is the prediction or classification resulting from the test set being presented to the neural network is getting successively closer to matching the known result from the test set. when the variance hits the inflection point where it begins to increase, training is ceased. at this point the neural network is no longer learning the pattern underlying the input data, but is instead over fitting the input data. dated 2000-09-12"
6119529,fluid flow meter and corresponding flow measuring methods,"a fluid flow meter is of the type including a heated probe sensor of known electric resistance dipped into or swept by a fluid stream having a predetermined velocity. the sensor is capable of converting each flow velocity value to a voltage value, and is connected to a processor operating using fuzzy logic for producing the flow measurements. the sensor may be an ntc thermistor. the thermistor may be powered from a current generator, and the processor may include a neural network. the sensor may include at least two discrete thermistors, one being a hot thermistor and the other being a cold thermistor.",2000-09-19,"The title of the patent is fluid flow meter and corresponding flow measuring methods and its abstract is a fluid flow meter is of the type including a heated probe sensor of known electric resistance dipped into or swept by a fluid stream having a predetermined velocity. the sensor is capable of converting each flow velocity value to a voltage value, and is connected to a processor operating using fuzzy logic for producing the flow measurements. the sensor may be an ntc thermistor. the thermistor may be powered from a current generator, and the processor may include a neural network. the sensor may include at least two discrete thermistors, one being a hot thermistor and the other being a cold thermistor. dated 2000-09-19"
6125105,method and apparatus for forecasting future values of a time series,a method of predicting at least one future value of a time series of data using a neural network comprising the steps of: pa1 (i) inputting a plurality of values of the time series into the neural network; pa1 (ii) inputting information about a time into the neural network; and pa1 (iii) obtaining outputs from the neural network said outputs comprising predicted future value(s) of the time series.,2000-09-26,The title of the patent is method and apparatus for forecasting future values of a time series and its abstract is a method of predicting at least one future value of a time series of data using a neural network comprising the steps of: pa1 (i) inputting a plurality of values of the time series into the neural network; pa1 (ii) inputting information about a time into the neural network; and pa1 (iii) obtaining outputs from the neural network said outputs comprising predicted future value(s) of the time series. dated 2000-09-26
6125194,"method and system for re-screening nodules in radiological images using multi-resolution processing, neural network, and image processing","an automated detection method and system improve the diagnostic procedures of radiological images containing abnormalities, such as lung cancer nodules. the detection method and system use a multi-resolution approach to enable the efficient detection of nodules of different sizes, and to further enable the use of a single nodule phantom for correlation and matching in order to detect all or most nodule sizes. the detection method and system use spherical parameters to characterize the nodules, thus enabling a more accurate detection of non-conspicuous nodules. a robust pixel threshold generation technique is applied in order to increase the sensitivity of the system. in addition, the detection method and system increase the sensitivity of true nodule detection by analyzing only the negative cases, and by recommending further re-assessment only of cases determined by the detection method and system to be positive. the detection method and system use multiple classifiers including back propagation neural network, data fusion, decision based pruned neural network, and convolution neural network architecture to generate the classification score for the classification of lung nodules. such multiple neural network architectures enable the learning of subtle characteristics of nodules to differentiate the nodules from the corresponding anatomic background. a final decision making then selects a portion of films with highly suspicious nodules for further reviewing.",2000-09-26,"The title of the patent is method and system for re-screening nodules in radiological images using multi-resolution processing, neural network, and image processing and its abstract is an automated detection method and system improve the diagnostic procedures of radiological images containing abnormalities, such as lung cancer nodules. the detection method and system use a multi-resolution approach to enable the efficient detection of nodules of different sizes, and to further enable the use of a single nodule phantom for correlation and matching in order to detect all or most nodule sizes. the detection method and system use spherical parameters to characterize the nodules, thus enabling a more accurate detection of non-conspicuous nodules. a robust pixel threshold generation technique is applied in order to increase the sensitivity of the system. in addition, the detection method and system increase the sensitivity of true nodule detection by analyzing only the negative cases, and by recommending further re-assessment only of cases determined by the detection method and system to be positive. the detection method and system use multiple classifiers including back propagation neural network, data fusion, decision based pruned neural network, and convolution neural network architecture to generate the classification score for the classification of lung nodules. such multiple neural network architectures enable the learning of subtle characteristics of nodules to differentiate the nodules from the corresponding anatomic background. a final decision making then selects a portion of films with highly suspicious nodules for further reviewing. dated 2000-09-26"
6125311,railway operation monitoring and diagnosing systems,"to enhance the safety and security of the operation of a railway network, a railway operation monitoring and diagnosing system is disclosed that monitors and diagnoses the entire railway network as an integrated system. the railway operation monitoring and diagnosing system comprises a railway operation predictor and a diagnosing means. the railway operation predictor generates anticipated values of selected railway operation state (ros) variables. ros variables may discrete or continuous. if there are continuous ros variables selected, the railway operation predictor also determines the safety intervals of these continuous ros variables. the diagnosing means examines the measured values of the selected ros variables versus their anticipated values and/or safety intervals to detect and diagnose their discrepancies. a heuristics, statistics, fuzzy logic, artificial intelligence, neural network, or/and expert system is included in the diagnosing means for diagnosing the records of such discrepancies. if necessary, the railway operation predictor generates pessimistically anticipated values of a plurality of selected ros and possibly other variables for further diagnosing the railway operation. the diagnosing means issues a diagnosis report and/or a recommendation, whenever the diagnosing means decides that such an issuance is appropriate.",2000-09-26,"The title of the patent is railway operation monitoring and diagnosing systems and its abstract is to enhance the safety and security of the operation of a railway network, a railway operation monitoring and diagnosing system is disclosed that monitors and diagnoses the entire railway network as an integrated system. the railway operation monitoring and diagnosing system comprises a railway operation predictor and a diagnosing means. the railway operation predictor generates anticipated values of selected railway operation state (ros) variables. ros variables may discrete or continuous. if there are continuous ros variables selected, the railway operation predictor also determines the safety intervals of these continuous ros variables. the diagnosing means examines the measured values of the selected ros variables versus their anticipated values and/or safety intervals to detect and diagnose their discrepancies. a heuristics, statistics, fuzzy logic, artificial intelligence, neural network, or/and expert system is included in the diagnosing means for diagnosing the records of such discrepancies. if necessary, the railway operation predictor generates pessimistically anticipated values of a plurality of selected ros and possibly other variables for further diagnosing the railway operation. the diagnosing means issues a diagnosis report and/or a recommendation, whenever the diagnosing means decides that such an issuance is appropriate. dated 2000-09-26"
6125441,predicting a sequence of variable instruction lengths from previously identified length pattern indexed by an instruction fetch address,"an instruction cache having a pattern detector for use in predicting the length of variable length instructions in a microprocessor. the instruction cache comprises an instruction length calculation unit and the pattern detector. the pattern detector is configured with a content addressable memory and update logic. the content addressable memory stores fetch addresses and instruction lengths calculated by the calculation unit. the content addressable memory compares particular fetch addresses that it receives with fetch addresses already stored and outputs corresponding predicted instruction length sequences. the content addressable memory may receive, compare, and store instruction lengths or instruction bytes in addition to, or in lieu of, fetch addresses. a neural network or other type of memory configuration may be used in place of the content addressable memory.",2000-09-26,"The title of the patent is predicting a sequence of variable instruction lengths from previously identified length pattern indexed by an instruction fetch address and its abstract is an instruction cache having a pattern detector for use in predicting the length of variable length instructions in a microprocessor. the instruction cache comprises an instruction length calculation unit and the pattern detector. the pattern detector is configured with a content addressable memory and update logic. the content addressable memory stores fetch addresses and instruction lengths calculated by the calculation unit. the content addressable memory compares particular fetch addresses that it receives with fetch addresses already stored and outputs corresponding predicted instruction length sequences. the content addressable memory may receive, compare, and store instruction lengths or instruction bytes in addition to, or in lieu of, fetch addresses. a neural network or other type of memory configuration may be used in place of the content addressable memory. dated 2000-09-26"
6128606,module for constructing trainable modular network in which each module inputs and outputs data structured as a graph,"a machine learning paradigm called graph transformer networks extends the applicability of gradient-based learning algorithms to systems composed of modules that take graphs as inputs and produce graphs as output. training is performed by computing gradients of a global objective function with respect to all the parameters in the system using a kind of back-propagation procedure. a complete check reading system based on these concept is described. the system uses convolutional neural network character recognizers, combined with global training techniques to provides record accuracy on business and personal checks.",2000-10-03,"The title of the patent is module for constructing trainable modular network in which each module inputs and outputs data structured as a graph and its abstract is a machine learning paradigm called graph transformer networks extends the applicability of gradient-based learning algorithms to systems composed of modules that take graphs as inputs and produce graphs as output. training is performed by computing gradients of a global objective function with respect to all the parameters in the system using a kind of back-propagation procedure. a complete check reading system based on these concept is described. the system uses convolutional neural network character recognizers, combined with global training techniques to provides record accuracy on business and personal checks. dated 2000-10-03"
6128609,training a neural network using differential input,a neural network is trained using a training neural network having the same topology as the original network but having a differential network output and accepting also differential network inputs. this new training method enables deeper neural networks to be successfully trained by avoiding a problem occuring in conventional training methods in which errors vanish as they are propagated in the reverse direction through deep networks. an acceleration in convergence rate is achieved by adjusting the error used in training to compensate for the linkage between multiple training data points.,2000-10-03,The title of the patent is training a neural network using differential input and its abstract is a neural network is trained using a training neural network having the same topology as the original network but having a differential network output and accepting also differential network inputs. this new training method enables deeper neural networks to be successfully trained by avoiding a problem occuring in conventional training methods in which errors vanish as they are propagated in the reverse direction through deep networks. an acceleration in convergence rate is achieved by adjusting the error used in training to compensate for the linkage between multiple training data points. dated 2000-10-03
6129681,apparatus and method for analyzing information relating to physical and mental condition,"an apparatus and method are provided for analyzing information relating to the physiological and psychological conditions of a driver. psychological conditions such as comfortableness or degree of alertness are estimated on the basis of physical data such as fluctuation in brain waves. this apparatus comprises a first neural network having a pre-processed 1/f fluctuation signal for brain waves as an input and for estimating a degree of alertness of the driver, and a second neural network receiving the estimated degree of alertness and the pre-processed 1/f fluctuation signal, for estimating and outputting driving comfortableness. by employing a neural network, which has a mapping ability as well as flexible adaptability even for non-linear data, based on the learning function, more accurate estimation of mental conditions can be achieved in comparison with conventional statistical analysis.",2000-10-10,"The title of the patent is apparatus and method for analyzing information relating to physical and mental condition and its abstract is an apparatus and method are provided for analyzing information relating to the physiological and psychological conditions of a driver. psychological conditions such as comfortableness or degree of alertness are estimated on the basis of physical data such as fluctuation in brain waves. this apparatus comprises a first neural network having a pre-processed 1/f fluctuation signal for brain waves as an input and for estimating a degree of alertness of the driver, and a second neural network receiving the estimated degree of alertness and the pre-processed 1/f fluctuation signal, for estimating and outputting driving comfortableness. by employing a neural network, which has a mapping ability as well as flexible adaptability even for non-linear data, based on the learning function, more accurate estimation of mental conditions can be achieved in comparison with conventional statistical analysis. dated 2000-10-10"
6131444,misfire detection using a dynamic neural network with output feedback,a misfire detection apparatus and method for detecting misfire of an engine. the engine has an engine speed and a manifold absolute pressure. an engine signal preprocessing system is provided for determining: a fluctuation associated with the engine speed; an average associated with the engine speed; and an average associated with the manifold absolute pressure. a neural network determines a firing event signal based upon: the determined engine speed fluctuation; the determined engine speed average; and the determined manifold absolute pressure average. a misfire decision determinator detects a misfire of the engine based upon the determined firing event signal. a preferred embodiment also includes a dynamic neural network system with global output feedback.,2000-10-17,The title of the patent is misfire detection using a dynamic neural network with output feedback and its abstract is a misfire detection apparatus and method for detecting misfire of an engine. the engine has an engine speed and a manifold absolute pressure. an engine signal preprocessing system is provided for determining: a fluctuation associated with the engine speed; an average associated with the engine speed; and an average associated with the manifold absolute pressure. a neural network determines a firing event signal based upon: the determined engine speed fluctuation; the determined engine speed average; and the determined manifold absolute pressure average. a misfire decision determinator detects a misfire of the engine based upon the determined firing event signal. a preferred embodiment also includes a dynamic neural network system with global output feedback. dated 2000-10-17
6134528,method device and article of manufacture for neural-network based generation of postlexical pronunciations from lexical pronunciations,"a method (2000), device (2200) and article of manufacture (2300) provide, in response to lexical pronunciation information, efficient generation of postlexical pronunciation information. a method is presented for providing, in response to a lexical pronunciation, efficient generation of a postlexical pronunciation, including the steps of: determining lexical phones, lexical features, and boundary information for a predetermined portion of text; and utilizing a pretrained neural network that was pretrained using lexical phones, postlexical phones, lexical features, and boundary information to generate a neural network hypothesis for a postlexical pronunciation of the predetermined portion of text.",2000-10-17,"The title of the patent is method device and article of manufacture for neural-network based generation of postlexical pronunciations from lexical pronunciations and its abstract is a method (2000), device (2200) and article of manufacture (2300) provide, in response to lexical pronunciation information, efficient generation of postlexical pronunciation information. a method is presented for providing, in response to a lexical pronunciation, efficient generation of a postlexical pronunciation, including the steps of: determining lexical phones, lexical features, and boundary information for a predetermined portion of text; and utilizing a pretrained neural network that was pretrained using lexical phones, postlexical phones, lexical features, and boundary information to generate a neural network hypothesis for a postlexical pronunciation of the predetermined portion of text. dated 2000-10-17"
6134537,visualization and self organization of multidimensional data through equalized orthogonal mapping,""" the subject system provides reduced-dimension mapping of pattern data. mapping is applied through conventional single-hidden-layer feed-forward neural network with non-linear neurons. according to one aspect of the present invention, the system functions to equalize and orthogonalize lower dimensional output signals by reducing the covariance matrix of the output signals to the form of a diagonal matrix or constant times the identity matrix. the present invention allows for visualization of large bodies of complex multidimensional data in a relatively """"topologically correct"""" low-dimension approximation, to reduce randomness associated with other methods of similar purposes, and to keep the mapping computationally efficient at the same time. """,2000-10-17,"The title of the patent is visualization and self organization of multidimensional data through equalized orthogonal mapping and its abstract is "" the subject system provides reduced-dimension mapping of pattern data. mapping is applied through conventional single-hidden-layer feed-forward neural network with non-linear neurons. according to one aspect of the present invention, the system functions to equalize and orthogonalize lower dimensional output signals by reducing the covariance matrix of the output signals to the form of a diagonal matrix or constant times the identity matrix. the present invention allows for visualization of large bodies of complex multidimensional data in a relatively """"topologically correct"""" low-dimension approximation, to reduce randomness associated with other methods of similar purposes, and to keep the mapping computationally efficient at the same time. "" dated 2000-10-17"
6134538,procedure for equalizing distorted data signals,"a neural network is provided for equalizing distorted data signals. the data signal to be equalized is coupled via time-delay elements to a group of networks for weighting. the output signals of the networks for weighting are coupled to the input terminals of a plurality of neurons whose outputs are coupled, via a respective amplifier, to input terminals of a further neuron having an output terminals where the equalized data signal can be tapped.",2000-10-17,"The title of the patent is procedure for equalizing distorted data signals and its abstract is a neural network is provided for equalizing distorted data signals. the data signal to be equalized is coupled via time-delay elements to a group of networks for weighting. the output signals of the networks for weighting are coupled to the input terminals of a plurality of neurons whose outputs are coupled, via a respective amplifier, to input terminals of a further neuron having an output terminals where the equalized data signal can be tapped. dated 2000-10-17"
6135966,method and apparatus for non-invasive diagnosis of cardiovascular and related disorders,"apparatus and method for non-invasive diagnosis of cardiovascular and related disorders. the system establishes a correspondence between the dynamics of the wave contour of the arterial pressure pulse and the associated disease states. the system comprises an input module, a contour signal receiver, and a processing module. the input module utilizes a pressure transducer for taking a non-invasive measurement of the arterial pulse. the contour signal receiver amplifies, digitizes and normalizes the arterial pressure pulse signal. in the processing module, the normalized arterial pressure contour is subjected to wavelet analysis, which transforms the dynamics of the time series of arterial blood pressure contour into multi-resolution wavelet coefficients or signatures. the processing module includes a neural network which is trained to associate the diagnostic features of the transformed arterial pressure contour embedded in the coefficients with a disease condition. after the learning phase, the system is capable of diagnosing known cardiovascular conditions in patients.",2000-10-24,"The title of the patent is method and apparatus for non-invasive diagnosis of cardiovascular and related disorders and its abstract is apparatus and method for non-invasive diagnosis of cardiovascular and related disorders. the system establishes a correspondence between the dynamics of the wave contour of the arterial pressure pulse and the associated disease states. the system comprises an input module, a contour signal receiver, and a processing module. the input module utilizes a pressure transducer for taking a non-invasive measurement of the arterial pulse. the contour signal receiver amplifies, digitizes and normalizes the arterial pressure pulse signal. in the processing module, the normalized arterial pressure contour is subjected to wavelet analysis, which transforms the dynamics of the time series of arterial blood pressure contour into multi-resolution wavelet coefficients or signatures. the processing module includes a neural network which is trained to associate the diagnostic features of the transformed arterial pressure contour embedded in the coefficients with a disease condition. after the learning phase, the system is capable of diagnosing known cardiovascular conditions in patients. dated 2000-10-24"
6137886,active vibration control method and apparatus,"an improved active vibration control system using feedback and pseudo-feedforward sensor inputs is provided for solving the problem of random and repetitive active vibration control and noise cancellation in a system. in a first embodiment of the invention, an artificial neural network is used for learning the dynamics of a structure and for providing output signals that follow the state variables of the structure. in one implementation of the neural network, a plurality of neurons obtain biasing inputs derived from sensor inputs, as well as inputs from the other neurons in the network. further, each neuron obtains a feedback input from itself. each input to a neuron is weighted using a weighting function derived on-line. the neural network supplies structure parameters and state variables to an optimal controller which derives and provides a control signal to the actuators so as to counteract vibrations and/or noise sensed in the system. in a second embodiment an optimal controller utilizing a modified generalized predictive control algorithm is used to to consider the limitations on the physical characteristics of the actuator(s), on-line, in terms of the output level and the rate of change of the output in the system. additional embodiments wherein an optimized control signal is sent to the actuator(s) to minimize vibration incident to the structure are provided.",2000-10-24,"The title of the patent is active vibration control method and apparatus and its abstract is an improved active vibration control system using feedback and pseudo-feedforward sensor inputs is provided for solving the problem of random and repetitive active vibration control and noise cancellation in a system. in a first embodiment of the invention, an artificial neural network is used for learning the dynamics of a structure and for providing output signals that follow the state variables of the structure. in one implementation of the neural network, a plurality of neurons obtain biasing inputs derived from sensor inputs, as well as inputs from the other neurons in the network. further, each neuron obtains a feedback input from itself. each input to a neuron is weighted using a weighting function derived on-line. the neural network supplies structure parameters and state variables to an optimal controller which derives and provides a control signal to the actuators so as to counteract vibrations and/or noise sensed in the system. in a second embodiment an optimal controller utilizing a modified generalized predictive control algorithm is used to to consider the limitations on the physical characteristics of the actuator(s), on-line, in terms of the output level and the rate of change of the output in the system. additional embodiments wherein an optimized control signal is sent to the actuator(s) to minimize vibration incident to the structure are provided. dated 2000-10-24"
6137898,gabor filtering for improved microcalcification detection in digital mammograms,"a method and system for detecting and displaying clustered microcalcifications in a digital mammogram, wherein a single digital mammogram is first automatically cropped to a breast area sub-image which is then processed by means of an optimized difference of gaussians filter to enhance the appearance of potential microcalcifications in the sub-image. the potential microcalcifications are thresholded, clusters are detected, features are computed for the detected clusters, and the clusters are classified as either suspicious or not suspicious by means of a neural network. thresholding is preferably by sloping local thresholding, but may also be performed by global and dual-local thresholding. the locations in the original digital mammogram of the suspicious detected clustered microcalcifications are indicated. parameters for use in the detection and thresholding portions of the system are computer-optimized by means of a genetic algorithm. the results of the system are optimally combined with a radiologist's observation of the original mammogram by combining the observations with the results, after the radiologist has first accepted or rejected individual detections reported by the system. an alternative embodiment reduces false positive detections by means of gabor filtering the cropped mammogram image to identify elongated structures such as milk ducts and veins. individual microcalcifications coincident with the elongated structures are removed and the remaining detections grouped into clusters.",2000-10-24,"The title of the patent is gabor filtering for improved microcalcification detection in digital mammograms and its abstract is a method and system for detecting and displaying clustered microcalcifications in a digital mammogram, wherein a single digital mammogram is first automatically cropped to a breast area sub-image which is then processed by means of an optimized difference of gaussians filter to enhance the appearance of potential microcalcifications in the sub-image. the potential microcalcifications are thresholded, clusters are detected, features are computed for the detected clusters, and the clusters are classified as either suspicious or not suspicious by means of a neural network. thresholding is preferably by sloping local thresholding, but may also be performed by global and dual-local thresholding. the locations in the original digital mammogram of the suspicious detected clustered microcalcifications are indicated. parameters for use in the detection and thresholding portions of the system are computer-optimized by means of a genetic algorithm. the results of the system are optimally combined with a radiologist's observation of the original mammogram by combining the observations with the results, after the radiologist has first accepted or rejected individual detections reported by the system. an alternative embodiment reduces false positive detections by means of gabor filtering the cropped mammogram image to identify elongated structures such as milk ducts and veins. individual microcalcifications coincident with the elongated structures are removed and the remaining detections grouped into clusters. dated 2000-10-24"
6138109,neural network diagnostic classification of complex binary systems,a malfunction diagnostic and repair guidance system and method wherein a matrix of numbers indicating the state of a complex binary system is used as an input vector for a neural network pattern processing capability that is focused to distinguish malfunction types of patterns. the neural network capability provides two complementary network types to classify and generalize the binary matrix. an interactive operator interface is updated with each repair after the root cause and is proposed repair of a malfunction is identified.,2000-10-24,The title of the patent is neural network diagnostic classification of complex binary systems and its abstract is a malfunction diagnostic and repair guidance system and method wherein a matrix of numbers indicating the state of a complex binary system is used as an input vector for a neural network pattern processing capability that is focused to distinguish malfunction types of patterns. the neural network capability provides two complementary network types to classify and generalize the binary matrix. an interactive operator interface is updated with each repair after the root cause and is proposed repair of a malfunction is identified. dated 2000-10-24
6138945,neural network controller for a pulsed rocket motor tactical missile system,"a neural network controller for a pulsed rocket motor tactical missile. the missile includes a fuselage or body, with a propulsion system. the pulsed propulsion system has a need for a logical control of the application of propulsion energy throughout the missile's flight. the controller is trained to provide optimal initiation of individual rocket motor thrust pulses based on tactical information available at various points/times in the missile's flight. the controller training is through use of training cases, in which the network learns to output a specific target value(s) when specific values are input. when trained with a large sample of training cases selected from the multidimensional population of interest, the neural network effectively learns the correlations between inputs and outputs and can predict input/output relationships not previously seen in any training case.",2000-10-31,"The title of the patent is neural network controller for a pulsed rocket motor tactical missile system and its abstract is a neural network controller for a pulsed rocket motor tactical missile. the missile includes a fuselage or body, with a propulsion system. the pulsed propulsion system has a need for a logical control of the application of propulsion energy throughout the missile's flight. the controller is trained to provide optimal initiation of individual rocket motor thrust pulses based on tactical information available at various points/times in the missile's flight. the controller training is through use of training cases, in which the network learns to output a specific target value(s) when specific values are input. when trained with a large sample of training cases selected from the multidimensional population of interest, the neural network effectively learns the correlations between inputs and outputs and can predict input/output relationships not previously seen in any training case. dated 2000-10-31"
6140964,wireless communication system and method and system for detection of position of radio mobile station,"a method of detecting a position of a radio mobile station in radiocommunications, which is capable of accurately and simply finding the position of the mobile station. at a measuring point the mobile station measures the reception radio strength levels from a plurality of base stations and conveys the measurement results through the base station to a control station. the control station learns, through a neural network, the correlation between the reception radio strength levels and the position of the mobile station on the basis of the measurement results at a plurality of measuring points and the positions of the measuring points. subsequently, when the mobile station communicates to the control station the reception radio strength levels measured at an arbitrary point, the control station estimates the position of the mobile station, causing those measurement results, on the basis of the correlation obtained through the learning.",2000-10-31,"The title of the patent is wireless communication system and method and system for detection of position of radio mobile station and its abstract is a method of detecting a position of a radio mobile station in radiocommunications, which is capable of accurately and simply finding the position of the mobile station. at a measuring point the mobile station measures the reception radio strength levels from a plurality of base stations and conveys the measurement results through the base station to a control station. the control station learns, through a neural network, the correlation between the reception radio strength levels and the position of the mobile station on the basis of the measurement results at a plurality of measuring points and the positions of the measuring points. subsequently, when the mobile station communicates to the control station the reception radio strength levels measured at an arbitrary point, the control station estimates the position of the mobile station, causing those measurement results, on the basis of the correlation obtained through the learning. dated 2000-10-31"
6141437,"cad method, computer and storage medium for automated detection of lung nodules in digital chest images","a computer-aided diagnosis (cad) method for the automated detection of lung nodules in a digital chest image, a computer programmed to implement the method, and a storage medium which stores a program for implementing the method, wherein nodule candidates are first automatically selected by thresholding the difference image and then classified in six groups. a large number of false positives are eliminated by adaptive rule-based tests applied to the original chest image and in the difference image and an artificial neural network (ann) applied to remaining candidate nodule locations in the original chest image. using two hundred pa chest radiographs, 100 normal and 100 abnormal, as the database, the presence of nodules in the 100 abnormal cases was confirmed by two experienced radiologists on the basis of ct scans or radiographic follow-up. the cad method achieves, on average, the sensitivity of 70% at 1.7 false positives per chest image.",2000-10-31,"The title of the patent is cad method, computer and storage medium for automated detection of lung nodules in digital chest images and its abstract is a computer-aided diagnosis (cad) method for the automated detection of lung nodules in a digital chest image, a computer programmed to implement the method, and a storage medium which stores a program for implementing the method, wherein nodule candidates are first automatically selected by thresholding the difference image and then classified in six groups. a large number of false positives are eliminated by adaptive rule-based tests applied to the original chest image and in the difference image and an artificial neural network (ann) applied to remaining candidate nodule locations in the original chest image. using two hundred pa chest radiographs, 100 normal and 100 abnormal, as the database, the presence of nodules in the 100 abnormal cases was confirmed by two experienced radiologists on the basis of ct scans or radiographic follow-up. the cad method achieves, on average, the sensitivity of 70% at 1.7 false positives per chest image. dated 2000-10-31"
6145381,real-time adaptive control of rotationally-induced vibration,"disturbances of a periodic nature, such as those occurring in a rotating device can be attenuated by a cancellation function. a cancellation signal is produced by extracting one or more frequency components (harmonics) from a disturbance. a set of weighting coefficients are generated by an artificial neural network based on information derived from the selected frequency components of the periodic disturbance signal. the artificial neural network algorithm adapts in real time to shifts in the magnitude and phase of the disturbance frequencies selected for attenuation at the sensor location. in turn, these coefficients are applied to a function of a similar number of components and applied to the rotating device. over a period of time, feedback and adaptation will attenuate the disturbance at the selected frequencies.",2000-11-14,"The title of the patent is real-time adaptive control of rotationally-induced vibration and its abstract is disturbances of a periodic nature, such as those occurring in a rotating device can be attenuated by a cancellation function. a cancellation signal is produced by extracting one or more frequency components (harmonics) from a disturbance. a set of weighting coefficients are generated by an artificial neural network based on information derived from the selected frequency components of the periodic disturbance signal. the artificial neural network algorithm adapts in real time to shifts in the magnitude and phase of the disturbance frequencies selected for attenuation at the sensor location. in turn, these coefficients are applied to a function of a similar number of components and applied to the rotating device. over a period of time, feedback and adaptation will attenuate the disturbance at the selected frequencies. dated 2000-11-14"
6145751,method and apparatus for determining a thermal setpoint in a hvac system,"the present invention discloses methods for determining setpoint information in a hvac system. setpoint values are determined using occupant feedback provided by individual occupants over at least one of an internet or intranet communications network. according to a first aspect of the invention a setpoint is determined using fuzzy logic. according to a second aspect, historical setpoint data determined using occupant feedback is used to develop a neural network for predicting setpoint values.",2000-11-14,"The title of the patent is method and apparatus for determining a thermal setpoint in a hvac system and its abstract is the present invention discloses methods for determining setpoint information in a hvac system. setpoint values are determined using occupant feedback provided by individual occupants over at least one of an internet or intranet communications network. according to a first aspect of the invention a setpoint is determined using fuzzy logic. according to a second aspect, historical setpoint data determined using occupant feedback is used to develop a neural network for predicting setpoint values. dated 2000-11-14"
6148101,digital image processor,"taking into consideration the disadvantage that a large-scale analog neural network cannot be constructed as an lsi and, even if this were possible, the cost would be prohibitive and the network would lack universality, a digital image processor for processing input image data based upon a cellular neural network is provided with a first multiply-and-accumulate arithmetic unit for digitally processing multiplication and accumulation of input image data of a plurality of pixels and input weighting values in a predetermined area, a second multiply-and-accumulate arithmetic unit for digitally processing multiplication and accumulation of output image data of a plurality of pixels and output weighting values in a predetermined area, and a non-linear acting unit for deciding output image data in accordance with results of calculation from the first and second multiply-and-accumulate arithmetic unit and non-linear characteristic parameters. this makes it possible to realize an image processor which excels in universality, ease of control and ease of integration.",2000-11-14,"The title of the patent is digital image processor and its abstract is taking into consideration the disadvantage that a large-scale analog neural network cannot be constructed as an lsi and, even if this were possible, the cost would be prohibitive and the network would lack universality, a digital image processor for processing input image data based upon a cellular neural network is provided with a first multiply-and-accumulate arithmetic unit for digitally processing multiplication and accumulation of input image data of a plurality of pixels and input weighting values in a predetermined area, a second multiply-and-accumulate arithmetic unit for digitally processing multiplication and accumulation of output image data of a plurality of pixels and output weighting values in a predetermined area, and a non-linear acting unit for deciding output image data in accordance with results of calculation from the first and second multiply-and-accumulate arithmetic unit and non-linear characteristic parameters. this makes it possible to realize an image processor which excels in universality, ease of control and ease of integration. dated 2000-11-14"
6151424,system for identifying objects and features in an image,"the present invention features the use of the fundamental concept of color perception and multi-level resolution to perform scene segmentation and object/feature extraction in the context of self-determining and self-calibration modes. the technique uses only a single image, instead of multiple images as the input to generate segmented images. moreover, a flexible and arbitrary scheme is incorporated, rather than a fixed scheme of segmentation analysis. the process allows users to perform digital analysis using any appropriate means for object extraction after an image is segmented. first, an image is retrieved. the image is then transformed into at least two distinct bands. each transformed image is then projected into a color domain or a multi-level resolution setting. a segmented image is then created from all of the transformed images. the segmented image is analyzed to identify objects. object identification is achieved by matching a segmented region against an image library. a featureless library contains full shape, partial shape and real-world images in a dual library system. the depth contours and height-above-ground structural components constitute a dual library. also provided is a mathematical model called a parzen window-based statistical/neural network classifier, which forms an integral part of this featureless dual library object identification system. all images are considered three-dimensional. laser radar based 3-d images represent a special case.",2000-11-21,"The title of the patent is system for identifying objects and features in an image and its abstract is the present invention features the use of the fundamental concept of color perception and multi-level resolution to perform scene segmentation and object/feature extraction in the context of self-determining and self-calibration modes. the technique uses only a single image, instead of multiple images as the input to generate segmented images. moreover, a flexible and arbitrary scheme is incorporated, rather than a fixed scheme of segmentation analysis. the process allows users to perform digital analysis using any appropriate means for object extraction after an image is segmented. first, an image is retrieved. the image is then transformed into at least two distinct bands. each transformed image is then projected into a color domain or a multi-level resolution setting. a segmented image is then created from all of the transformed images. the segmented image is analyzed to identify objects. object identification is achieved by matching a segmented region against an image library. a featureless library contains full shape, partial shape and real-world images in a dual library system. the depth contours and height-above-ground structural components constitute a dual library. also provided is a mathematical model called a parzen window-based statistical/neural network classifier, which forms an integral part of this featureless dual library object identification system. all images are considered three-dimensional. laser radar based 3-d images represent a special case. dated 2000-11-21"
6151592,"recognition apparatus using neural network, and learning method therefor","a recognition apparatus and method using a neural network is provided. a neuron-like element stores a value of its inner condition. the neuron-like element also updates a values of its internal status on the basis of an output from the neuron-like element itself, outputs from other neuron-like elements and an external input, and an output value generator a value of its internal status into an external output. accordingly, the neuron-like element itself can retain the history of input data. this enables the time series data, such as speech, to be processed without providing any special devices in the neural network.",2000-11-21,"The title of the patent is recognition apparatus using neural network, and learning method therefor and its abstract is a recognition apparatus and method using a neural network is provided. a neuron-like element stores a value of its inner condition. the neuron-like element also updates a values of its internal status on the basis of an output from the neuron-like element itself, outputs from other neuron-like elements and an external input, and an output value generator a value of its internal status into an external output. accordingly, the neuron-like element itself can retain the history of input data. this enables the time series data, such as speech, to be processed without providing any special devices in the neural network. dated 2000-11-21"
6151593,apparatus for authenticating an individual based on a typing pattern by using a neural network system,"a user authentication apparatus for use in controlling access to a system inputs an owner's login name and password and then extracts the owner's timing vectors from keystroke characteristics with which the owner repeatedly types the owner's password to thereby form a training set. a neural network is trained by using each of the owner's timing vectors in the training set as an input. thereafter, when a user inputs the owner's login name and password, it is checked if the user's password is identical to the owner's password. the user's timing vector is extracted from a keystroke characteristic to type the user's password if the checked result is affirmative, and the user is prohibited from accessing the system if otherwise. the user's timing vector is applied to the trained neural network as an input and a difference between the input and an output of the neural network is compared with a predetermined threshold. the user will be permitted to access the system if the difference is not greater than the threshold and prohibited from accessing the system if otherwise.",2000-11-21,"The title of the patent is apparatus for authenticating an individual based on a typing pattern by using a neural network system and its abstract is a user authentication apparatus for use in controlling access to a system inputs an owner's login name and password and then extracts the owner's timing vectors from keystroke characteristics with which the owner repeatedly types the owner's password to thereby form a training set. a neural network is trained by using each of the owner's timing vectors in the training set as an input. thereafter, when a user inputs the owner's login name and password, it is checked if the user's password is identical to the owner's password. the user's timing vector is extracted from a keystroke characteristic to type the user's password if the checked result is affirmative, and the user is prohibited from accessing the system if otherwise. the user's timing vector is applied to the trained neural network as an input and a difference between the input and an output of the neural network is compared with a predetermined threshold. the user will be permitted to access the system if the difference is not greater than the threshold and prohibited from accessing the system if otherwise. dated 2000-11-21"
6154140,intelligent personal underwater monitoring device,the present invention relates to an intelligent personal underwater monitng system which recognizes when a swimmer is in trouble and transmits a warning signal. the system comprises a device worn by the swimmer which senses water pressure and transmits a first signal when a predetermined depth is passed by the swimmer. the system further comprises a device for receiving the signals generated by the device worn by the swimmer and a neural network processor for processing the signals and generating an output signal representative of whether the swimmer is in trouble and an alarm signal generation device.,2000-11-28,The title of the patent is intelligent personal underwater monitoring device and its abstract is the present invention relates to an intelligent personal underwater monitng system which recognizes when a swimmer is in trouble and transmits a warning signal. the system comprises a device worn by the swimmer which senses water pressure and transmits a first signal when a predetermined depth is passed by the swimmer. the system further comprises a device for receiving the signals generated by the device worn by the swimmer and a neural network processor for processing the signals and generating an output signal representative of whether the swimmer is in trouble and an alarm signal generation device. dated 2000-11-28
6157877,apparatus and method for testing automotive electronic control units and batteries and related equipment,a method and apparatus for testing automotive electronic control units and batteries and other equipment for identification and performance purposes utilizes neural networks to effect waveform analysis on a digitized signal. identification of electronic control units is by means of correlation of resultant waveform data with corresponding data on known units. battery testing is by waveform analysis of the battery current during transient connection of a load by a transistorized switching circuit. in both cases the method of testing includes a network learning stage and an ensuing recognition test routine for characteristic waveforms.,2000-12-05,The title of the patent is apparatus and method for testing automotive electronic control units and batteries and related equipment and its abstract is a method and apparatus for testing automotive electronic control units and batteries and other equipment for identification and performance purposes utilizes neural networks to effect waveform analysis on a digitized signal. identification of electronic control units is by means of correlation of resultant waveform data with corresponding data on known units. battery testing is by waveform analysis of the battery current during transient connection of a load by a transistorized switching circuit. in both cases the method of testing includes a network learning stage and an ensuing recognition test routine for characteristic waveforms. dated 2000-12-05
6163773,data storage system with trained predictive cache management engine,"in a data storage system, a cache is managed by a predictive cache management engine that evaluates cache contents and purges entries unlikely to receive sufficient future cache hits. the engine includes a single output back propagation neural network that is trained in response to various event triggers. accesses to stored datasets are logged in a data access log; conversely, log entries are removed according to a predefined expiration criteria. in response to access of a cached dataset or expiration of its log entry, the cache management engine prepares training data. this is achieved by determining characteristics of the dataset at various past times between the time of the access/expiration and a time of last access, and providing these characteristics and the times of access as input to train the neural network. as another part of training, the cache management engine provides the neural network with output representing the expiration or access of the dataset. according to a predefined schedule, the cache management engine operates the trained neural network to generate scores for cached datasets, these scores ranking the datasets relative to each other. according to this or a different schedule, the cache management engine reviews the scores, identifies one or more datasets with the least scores, and purges the identified datasets from the cache.",2000-12-19,"The title of the patent is data storage system with trained predictive cache management engine and its abstract is in a data storage system, a cache is managed by a predictive cache management engine that evaluates cache contents and purges entries unlikely to receive sufficient future cache hits. the engine includes a single output back propagation neural network that is trained in response to various event triggers. accesses to stored datasets are logged in a data access log; conversely, log entries are removed according to a predefined expiration criteria. in response to access of a cached dataset or expiration of its log entry, the cache management engine prepares training data. this is achieved by determining characteristics of the dataset at various past times between the time of the access/expiration and a time of last access, and providing these characteristics and the times of access as input to train the neural network. as another part of training, the cache management engine provides the neural network with output representing the expiration or access of the dataset. according to a predefined schedule, the cache management engine operates the trained neural network to generate scores for cached datasets, these scores ranking the datasets relative to each other. according to this or a different schedule, the cache management engine reviews the scores, identifies one or more datasets with the least scores, and purges the identified datasets from the cache. dated 2000-12-19"
6167117,voice-dialing system using model of calling behavior,"a method and apparatus for assisting voice-dialing using a model of an individual's calling behavior to improve recognition of an input name corresponding a desired telephone number. when the individual picks up a telephone, activity is initiated in a neural network model of the individual's calling behavior that predicts the likelihood that different numbers will be called, given such predictors as the day of the week and the time of day. the model is constructed by training the neural network with data from the user's history of making and receiving telephone calls. the auditory output from an automatic speech recognition system and the output from the user model are integrated together so as to select the number that is most likely to be the number desired by the speaker. the system can also provide automatic directory assistance, by speaking the number aloud rather than dialing it. in one version, the system is a personal directory for an individual maintained on that individual's personal computer. in another version, the system serves as a directory for a given physical or virtual site, with information about the institutional organization at the site in addition to individual calling histories used to track calling patterns and make predictions about the likelihood of calls within the site.",2000-12-26,"The title of the patent is voice-dialing system using model of calling behavior and its abstract is a method and apparatus for assisting voice-dialing using a model of an individual's calling behavior to improve recognition of an input name corresponding a desired telephone number. when the individual picks up a telephone, activity is initiated in a neural network model of the individual's calling behavior that predicts the likelihood that different numbers will be called, given such predictors as the day of the week and the time of day. the model is constructed by training the neural network with data from the user's history of making and receiving telephone calls. the auditory output from an automatic speech recognition system and the output from the user model are integrated together so as to select the number that is most likely to be the number desired by the speaker. the system can also provide automatic directory assistance, by speaking the number aloud rather than dialing it. in one version, the system is a personal directory for an individual maintained on that individual's personal computer. in another version, the system serves as a directory for a given physical or virtual site, with information about the institutional organization at the site in addition to individual calling histories used to track calling patterns and make predictions about the likelihood of calls within the site. dated 2000-12-26"
6167146,method and system for segmentation and detection of microcalcifications from digital mammograms,"a method and system for detecting and displaying clustered microcalcifications in a digital mammogram, wherein a single digital mammogram is first automatically cropped to a breast area sub-image which is then processed by means of an optimized difference of gaussians filter to enhance the appearance of potential microcalcifications in the sub-image. the potential microcalcifications are thresholded, clusters are detected, features are computed for the detected clusters, and the clusters are classified as either suspicious or not suspicious by means of a neural network. thresholding is preferably by sloping local thresholding, but may also be performed by global and dual-local thresholding. the locations in the original digital mammogram of the suspicious detected clustered microcalcifications are indicated. parameters for use in the detection and thresholding portions of the system are computer-optimized by means of a genetic algorithm. the results of the system are optimally combined with a radiologist's observation of the original mammogram by combining the observations with the results, after the radiologist has first accepted or rejected individual detections reported by the system.",2000-12-26,"The title of the patent is method and system for segmentation and detection of microcalcifications from digital mammograms and its abstract is a method and system for detecting and displaying clustered microcalcifications in a digital mammogram, wherein a single digital mammogram is first automatically cropped to a breast area sub-image which is then processed by means of an optimized difference of gaussians filter to enhance the appearance of potential microcalcifications in the sub-image. the potential microcalcifications are thresholded, clusters are detected, features are computed for the detected clusters, and the clusters are classified as either suspicious or not suspicious by means of a neural network. thresholding is preferably by sloping local thresholding, but may also be performed by global and dual-local thresholding. the locations in the original digital mammogram of the suspicious detected clustered microcalcifications are indicated. parameters for use in the detection and thresholding portions of the system are computer-optimized by means of a genetic algorithm. the results of the system are optimally combined with a radiologist's observation of the original mammogram by combining the observations with the results, after the radiologist has first accepted or rejected individual detections reported by the system. dated 2000-12-26"
6167390,facet classification neural network,"a classification neural network for piecewise linearly separating an input space to classify input patterns is described. the multilayered neural network comprises an input node, a plurality of difference nodes in a first layer, a minimum node, a plurality of perceptron nodes in a second layer and an output node. in operation, the input node broadcasts the input pattern to all of the difference nodes. the difference nodes, along with the minimum node, identify in which vornoi cell of the piecewise linear separation the input pattern lies. the difference node defining the vornoi cell localizes input pattern to a local coordinate space and sends it to a corresponding perceptron, which produces a class designator for the input pattern.",2000-12-26,"The title of the patent is facet classification neural network and its abstract is a classification neural network for piecewise linearly separating an input space to classify input patterns is described. the multilayered neural network comprises an input node, a plurality of difference nodes in a first layer, a minimum node, a plurality of perceptron nodes in a second layer and an output node. in operation, the input node broadcasts the input pattern to all of the difference nodes. the difference nodes, along with the minimum node, identify in which vornoi cell of the piecewise linear separation the input pattern lies. the difference node defining the vornoi cell localizes input pattern to a local coordinate space and sends it to a corresponding perceptron, which produces a class designator for the input pattern. dated 2000-12-26"
6169980,method for operating a neural network with missing and/or incomplete data,"a neural network system is provided that models the system in a system model (12) with the output thereof providing a predicted output. this predicted output is modified or controlled by an output control (14). input data is processed in a data preprocess step (10) to reconcile the data for input to the system model (12). additionally, the error resulted from the reconciliation is input to an uncertainty model to predict the uncertainty in the predicted output. this is input to a decision processor (20) which is utilized to control the output control (14). the output control (14) is controlled to either vary the predicted output or to inhibit the predicted output whenever the output of the uncertainty model (18) exceeds a predetermined decision threshold, input by a decision threshold block (22). additionally, a validity model (16) is also provided which represents the reliability or validity of the output as a function of the number of data points in a given data region during training of the system model (12). this predicts the confidence in the predicted output which is also input to the decision processor (20). the decision processor (20) therefore bases its decision on the predicted confidence and the predicted uncertainty. additionally, the uncertainty output by the data preprocess block (10) can be utilized to train the system model (12).",2001-01-02,"The title of the patent is method for operating a neural network with missing and/or incomplete data and its abstract is a neural network system is provided that models the system in a system model (12) with the output thereof providing a predicted output. this predicted output is modified or controlled by an output control (14). input data is processed in a data preprocess step (10) to reconcile the data for input to the system model (12). additionally, the error resulted from the reconciliation is input to an uncertainty model to predict the uncertainty in the predicted output. this is input to a decision processor (20) which is utilized to control the output control (14). the output control (14) is controlled to either vary the predicted output or to inhibit the predicted output whenever the output of the uncertainty model (18) exceeds a predetermined decision threshold, input by a decision threshold block (22). additionally, a validity model (16) is also provided which represents the reliability or validity of the output as a function of the number of data points in a given data region during training of the system model (12). this predicts the confidence in the predicted output which is also input to the decision processor (20). the decision processor (20) therefore bases its decision on the predicted confidence and the predicted uncertainty. additionally, the uncertainty output by the data preprocess block (10) can be utilized to train the system model (12). dated 2001-01-02"
6169981,3-brain architecture for an intelligent decision and control system,"a method and system for intelligent control of external devices using a mammalian brain-like structure having three parts. the method and system include a computer-implemented neural network system which is an extension of the model-based adaptive critic design and is applicable to real-time control (e.g., robotic control) and real-time distributed control. additional uses include data visualization, data mining, and other tasks requiring complex analysis of inter-relationships between data.",2001-01-02,"The title of the patent is 3-brain architecture for an intelligent decision and control system and its abstract is a method and system for intelligent control of external devices using a mammalian brain-like structure having three parts. the method and system include a computer-implemented neural network system which is an extension of the model-based adaptive critic design and is applicable to real-time control (e.g., robotic control) and real-time distributed control. additional uses include data visualization, data mining, and other tasks requiring complex analysis of inter-relationships between data. dated 2001-01-02"
6173218,neurocomputing control distribution system,"a method and system for controlling a propulsion system of a vehicle. the propulsion system includes at least one propulsion effector. the system also includes sensors for sensing vehicle position and motions, a control device for generating control signals, and a generator for generating desired vehicle forces/moments from the sensed vehicle position and motions and the generated control signals based on predefined vehicle compensation and control laws. also included is a neural network controller for generating propulsion commands for the at least one propulsion effector based on the generated desire forces/moments, wherein said neural network controller was trained based on pregenerated vehicle control distribution data.",2001-01-09,"The title of the patent is neurocomputing control distribution system and its abstract is a method and system for controlling a propulsion system of a vehicle. the propulsion system includes at least one propulsion effector. the system also includes sensors for sensing vehicle position and motions, a control device for generating control signals, and a generator for generating desired vehicle forces/moments from the sensed vehicle position and motions and the generated control signals based on predefined vehicle compensation and control laws. also included is a neural network controller for generating propulsion commands for the at least one propulsion effector based on the generated desire forces/moments, wherein said neural network controller was trained based on pregenerated vehicle control distribution data. dated 2001-01-09"
6173275,representation and retrieval of images using context vectors derived from image information elements,""" image features are generated by performing wavelet transformations at sample points on images stored in electronic form. multiple wavelet transformations at a point are combined to form an image feature vector. a prototypical set of feature vectors, or atoms, is derived from the set of feature vectors to form an """"atomic vocabulary."""" the prototypical feature vectors are derived using a vector quantization method (e.g., using neural network self-organization techniques) in which a vector quantization network is also generated. the atomic vocabulary is used to define new images. meaning is established between atoms in the atomic vocabulary. high-dimensional context vectors are assigned to each atom. the context vectors are then trained as a function of the proximity and co-occurrence of each atom to other atoms in the image. after training, the context vectors associated with the atoms that comprise an image are combined to form a summary vector for the image. images are retrieved using a number of query methods (e.g., images, image portions, vocabulary atoms, index terms). the user's query is converted into a query context vector. a dot product is calculated between the query vector and the summary vectors to locate images having the closest meaning. the invention is also applicable to video or temporally related images, and can also be used in conjunction with other context vector data domains such as text or audio, thereby linking images to such data domains. """,2001-01-09,"The title of the patent is representation and retrieval of images using context vectors derived from image information elements and its abstract is "" image features are generated by performing wavelet transformations at sample points on images stored in electronic form. multiple wavelet transformations at a point are combined to form an image feature vector. a prototypical set of feature vectors, or atoms, is derived from the set of feature vectors to form an """"atomic vocabulary."""" the prototypical feature vectors are derived using a vector quantization method (e.g., using neural network self-organization techniques) in which a vector quantization network is also generated. the atomic vocabulary is used to define new images. meaning is established between atoms in the atomic vocabulary. high-dimensional context vectors are assigned to each atom. the context vectors are then trained as a function of the proximity and co-occurrence of each atom to other atoms in the image. after training, the context vectors associated with the atoms that comprise an image are combined to form a summary vector for the image. images are retrieved using a number of query methods (e.g., images, image portions, vocabulary atoms, index terms). the user's query is converted into a query context vector. a dot product is calculated between the query vector and the summary vectors to locate images having the closest meaning. the invention is also applicable to video or temporally related images, and can also be used in conjunction with other context vector data domains such as text or audio, thereby linking images to such data domains. "" dated 2001-01-09"
6175643,neural network based auto-windowing system for mr images,"an adaptive hierarchical neural network based system with online adaptation capabilities has been developed to automatically adjust the display window width and center for mr images. our windowing system possesses the online training capabilities that make the adaptation of the optimal display parameters to personal preference as well as different viewing conditions possible. the online adaptation capabilities are primarily due to the use of the hierarchical neural networks and the development of a new width/center mapping system. the large training image set is hierarchically organized for efficient user interaction and effective re-mapping of the width/center settings in the training data set. the width/center values are modified in the training data through a width/center mapping function, which is estimated from the new width/center values of some representative images adjusted by the user. the width/center mapping process consists of a global spline mapping for the entire training images as well as a first-order polynomial sequence mapping for the image sequences selected in the user's new adjustment procedure.",2001-01-16,"The title of the patent is neural network based auto-windowing system for mr images and its abstract is an adaptive hierarchical neural network based system with online adaptation capabilities has been developed to automatically adjust the display window width and center for mr images. our windowing system possesses the online training capabilities that make the adaptation of the optimal display parameters to personal preference as well as different viewing conditions possible. the online adaptation capabilities are primarily due to the use of the hierarchical neural networks and the development of a new width/center mapping system. the large training image set is hierarchically organized for efficient user interaction and effective re-mapping of the width/center settings in the training data set. the width/center values are modified in the training data through a width/center mapping function, which is estimated from the new width/center values of some representative images adjusted by the user. the width/center mapping process consists of a global spline mapping for the entire training images as well as a first-order polynomial sequence mapping for the image sequences selected in the user's new adjustment procedure. dated 2001-01-16"
6176325,moling apparatus and a ground sensing system therefor,"the invention provides a ground sensing system (10) comprising: sensing means (19) located, in use, on a projectile being driven through ground by means of apparatus having a self adjustment between a vibration mode and a vibro-impact mode according to encountered ground resistance, the sensing means sensing the dynamic resistance of the ground that the projectile is passing through; signal processing means for processing the output of said sensing means to provide a dynamic resistance waveform (106); and waveform recognition means (108) for correlating said dynamic resistance waveform with stored dynamic waveforms for identifying a ground characteristic. the waveform recognition means may comprise a neural network system.",2001-01-23,"The title of the patent is moling apparatus and a ground sensing system therefor and its abstract is the invention provides a ground sensing system (10) comprising: sensing means (19) located, in use, on a projectile being driven through ground by means of apparatus having a self adjustment between a vibration mode and a vibro-impact mode according to encountered ground resistance, the sensing means sensing the dynamic resistance of the ground that the projectile is passing through; signal processing means for processing the output of said sensing means to provide a dynamic resistance waveform (106); and waveform recognition means (108) for correlating said dynamic resistance waveform with stored dynamic waveforms for identifying a ground characteristic. the waveform recognition means may comprise a neural network system. dated 2001-01-23"
6178402,"method, apparatus and system for generating acoustic parameters in a text-to-speech system using a neural network","the present invention provides a method, device and system to generate acoustic parameters in a text-to-speech system utilizing a neural network to generate a representation of a trajectory in an acoustic parameter space across a phonetic segment.",2001-01-23,"The title of the patent is method, apparatus and system for generating acoustic parameters in a text-to-speech system using a neural network and its abstract is the present invention provides a method, device and system to generate acoustic parameters in a text-to-speech system utilizing a neural network to generate a representation of a trajectory in an acoustic parameter space across a phonetic segment. dated 2001-01-23"
6182028,"method, device and system for part-of-speech disambiguation","a method (300), device (408), and system (400) provide part-of-speech disambiguation for words based on hybrid neural-network and stochastic processing. the method disambiguates the part-of-speech tags of text tokens by obtaining a set of probabilistically annotated tags for each text token, determining a locally predicted tag for each text token based on the local context of the text token, determining an alternative tag for each text token based on the expanded context of the text token, and choosing between the locally predicted tag and the alternative tag when the locally predicted tag and the alternative tag are different.",2001-01-30,"The title of the patent is method, device and system for part-of-speech disambiguation and its abstract is a method (300), device (408), and system (400) provide part-of-speech disambiguation for words based on hybrid neural-network and stochastic processing. the method disambiguates the part-of-speech tags of text tokens by obtaining a set of probabilistically annotated tags for each text token, determining a locally predicted tag for each text token based on the local context of the text token, determining an alternative tag for each text token based on the expanded context of the text token, and choosing between the locally predicted tag and the alternative tag when the locally predicted tag and the alternative tag are different. dated 2001-01-30"
6185171,system for accommodating vibrations resulting from rotating a data storage medium,"a control system in a data storage apparatus and associated methods for attempting to accommodate the vibrations resulting from rotating a data storage medium. the control system comprises a neural network which utilizes detected vibrations resulting from the rotation of data storage media to learn the characteristics of the rotational imbalance of rotating data storage media. thereafter, the rotation of a data storage medium and/or movement of a data head is controlled based on the characteristics learned.",2001-02-06,"The title of the patent is system for accommodating vibrations resulting from rotating a data storage medium and its abstract is a control system in a data storage apparatus and associated methods for attempting to accommodate the vibrations resulting from rotating a data storage medium. the control system comprises a neural network which utilizes detected vibrations resulting from the rotation of data storage media to learn the characteristics of the rotational imbalance of rotating data storage media. thereafter, the rotation of a data storage medium and/or movement of a data head is controlled based on the characteristics learned. dated 2001-02-06"
6185470,neural network predictive control method and system,"a method and system for controlling a dynamic nonlinear plant. an input signal controls the plant and an output signal represents a state of the plant in response to the received input signal. a memory stores input and output signals corresponding to m consecutive past states of the plant. a computer neural network predicts a set of future output states representative of the output signal corresponding to the next n consecutive future states of the plant in response to a set of trial control inputs. the trial control inputs represent the input signal corresponding to the next n consecutive future states of the plant. the neural network predicts the future output states based on the past input and output signals and the future trial control inputs. a processor generates the trial control inputs and determines a performance index, indicative of plant performance over time in response to the trial control inputs, as a function of the future output states. the processor generates the input signal for controlling the plant and modifies it as a function of the trial control inputs so that the performance index reaches a desired value.",2001-02-06,"The title of the patent is neural network predictive control method and system and its abstract is a method and system for controlling a dynamic nonlinear plant. an input signal controls the plant and an output signal represents a state of the plant in response to the received input signal. a memory stores input and output signals corresponding to m consecutive past states of the plant. a computer neural network predicts a set of future output states representative of the output signal corresponding to the next n consecutive future states of the plant in response to a set of trial control inputs. the trial control inputs represent the input signal corresponding to the next n consecutive future states of the plant. the neural network predicts the future output states based on the past input and output signals and the future trial control inputs. a processor generates the trial control inputs and determines a performance index, indicative of plant performance over time in response to the trial control inputs, as a function of the future output states. the processor generates the input signal for controlling the plant and modifies it as a function of the trial control inputs so that the performance index reaches a desired value. dated 2001-02-06"
6185528,method of and a device for speech recognition employing neural network and markov model recognition techniques,"a method and a device for recognition of isolated words in large vocabularies are described, wherein recognition is performed through two sequential steps using neural networks and markov models techniques, respectively, and the results of both techniques are adequately combined so as to improve recognition accuracy. the devices performing the combination also provide an evaluation of recognition reliability.",2001-02-06,"The title of the patent is method of and a device for speech recognition employing neural network and markov model recognition techniques and its abstract is a method and a device for recognition of isolated words in large vocabularies are described, wherein recognition is performed through two sequential steps using neural networks and markov models techniques, respectively, and the results of both techniques are adequately combined so as to improve recognition accuracy. the devices performing the combination also provide an evaluation of recognition reliability. dated 2001-02-06"
6185548,neural network methods to predict enzyme inhibitor or receptor ligand potency,"a new method to analyze and predict the binding energy for enzyme-transition state inhibitor interactions is presented. computational neural networks are employed to discovery quantum mechanical features of transition states and putative inhibitors necessary for binding. the method is able to generate its own relationship between the quantum mechanical structure of the inhibitor and the strength of binding. feed-forward neural networks with back propagation of error can be trained to recognize the quantum mechanical electrostatic potential at the entire van der waals surface, rather than a collapsed representation, of a group of training inhibitors and to predict the strength of interactions between the enzyme and a group of novel inhibitors. the experimental results show that the neural networks can predict with quantitative accuracy the binding strength of new inhibitors. the method is in fact able to predict the large binding free energy of the transition state, when trained with less tightly bound inhibitors. the present method is also applicable to prediction of the binding free energy of a ligand to a receptor. the application of this approach to the study of transition state inhibitors and ligands would permit evaluation of chemical libraries of potential inhibitory, agonistic, or antagonistic agents. the method is amenable to incorporation in a computer-readable medium accessible by general-purpose computers.",2001-02-06,"The title of the patent is neural network methods to predict enzyme inhibitor or receptor ligand potency and its abstract is a new method to analyze and predict the binding energy for enzyme-transition state inhibitor interactions is presented. computational neural networks are employed to discovery quantum mechanical features of transition states and putative inhibitors necessary for binding. the method is able to generate its own relationship between the quantum mechanical structure of the inhibitor and the strength of binding. feed-forward neural networks with back propagation of error can be trained to recognize the quantum mechanical electrostatic potential at the entire van der waals surface, rather than a collapsed representation, of a group of training inhibitors and to predict the strength of interactions between the enzyme and a group of novel inhibitors. the experimental results show that the neural networks can predict with quantitative accuracy the binding strength of new inhibitors. the method is in fact able to predict the large binding free energy of the transition state, when trained with less tightly bound inhibitors. the present method is also applicable to prediction of the binding free energy of a ligand to a receptor. the application of this approach to the study of transition state inhibitors and ligands would permit evaluation of chemical libraries of potential inhibitory, agonistic, or antagonistic agents. the method is amenable to incorporation in a computer-readable medium accessible by general-purpose computers. dated 2001-02-06"
6186953,non-invasive and continuous blood-pressure estimation apparatus,"an apparatus for iteratively estimating an intra-arterial blood-pressure value of a living subject, including a first information obtaining device which non-invasively and iteratively obtains information relating to propagation of a pulse wave through an arterial vessel of the subject, a second information obtaining device which non-invasively and iteratively obtains information relating to heartbeat of the subject and/or information relating to area of a heartbeat-synchronous pulse of a volumetric pulse wave obtained from a peripheral body portion of the subject, and an estimating device including a neural network which learns sets of information each set of which includes a blood-pressure value measured using a cuff, pulse-wave-propagation-relating information obtained when the blood-pressure value is measured using the cuff, and heartbeat-relating information and/or pulse-wave-area-relating information obtained when the blood-pressure value is measured using the cuff, the neural network iteratively estimating an intra-arterial blood-pressure value of the subject, based on each piece of pulse-wave-propagation-relating information iteratively obtained by the first information obtaining device, and each piece of heartbeat-relating information and/or each piece of pulse-wave-area-relating information which are or is iteratively obtained by the second information obtaining device.",2001-02-13,"The title of the patent is non-invasive and continuous blood-pressure estimation apparatus and its abstract is an apparatus for iteratively estimating an intra-arterial blood-pressure value of a living subject, including a first information obtaining device which non-invasively and iteratively obtains information relating to propagation of a pulse wave through an arterial vessel of the subject, a second information obtaining device which non-invasively and iteratively obtains information relating to heartbeat of the subject and/or information relating to area of a heartbeat-synchronous pulse of a volumetric pulse wave obtained from a peripheral body portion of the subject, and an estimating device including a neural network which learns sets of information each set of which includes a blood-pressure value measured using a cuff, pulse-wave-propagation-relating information obtained when the blood-pressure value is measured using the cuff, and heartbeat-relating information and/or pulse-wave-area-relating information obtained when the blood-pressure value is measured using the cuff, the neural network iteratively estimating an intra-arterial blood-pressure value of the subject, based on each piece of pulse-wave-propagation-relating information iteratively obtained by the first information obtaining device, and each piece of heartbeat-relating information and/or each piece of pulse-wave-area-relating information which are or is iteratively obtained by the second information obtaining device. dated 2001-02-13"
6186955,noninvasive continuous cardiac output monitor,"method and apparatus for continuous, non-invasive determination of cardiac output which processes a sequence of non-invasive cardiography signals which are quantitatively dependent upon cardiac output within a computer system and associated neural network capable of generating a single output signal for the combined input signals, wherein the neural network applies weighting factors determined during a training phase to force the output signal to match the known value of cardiac output determined by invasive means and reports the single output signal as the determined value of cardiac output.",2001-02-13,"The title of the patent is noninvasive continuous cardiac output monitor and its abstract is method and apparatus for continuous, non-invasive determination of cardiac output which processes a sequence of non-invasive cardiography signals which are quantitatively dependent upon cardiac output within a computer system and associated neural network capable of generating a single output signal for the combined input signals, wherein the neural network applies weighting factors determined during a training phase to force the output signal to match the known value of cardiac output determined by invasive means and reports the single output signal as the determined value of cardiac output. dated 2001-02-13"
6187145,method for process management in paper and cardboard manufacture,"a method and apparatus for using a neural network to control a paper or paper board production machine. by using spectrum measuring devices, the characteristics of the starting materials for the paper and board production and/or their intermediate or final products are registered and the values fed to a neural network. the network provides statements concerning the paper and board quality, from which signals for the feedback and/or feedforward control of the production process may be derived.",2001-02-13,"The title of the patent is method for process management in paper and cardboard manufacture and its abstract is a method and apparatus for using a neural network to control a paper or paper board production machine. by using spectrum measuring devices, the characteristics of the starting materials for the paper and board production and/or their intermediate or final products are registered and the values fed to a neural network. the network provides statements concerning the paper and board quality, from which signals for the feedback and/or feedforward control of the production process may be derived. dated 2001-02-13"
6189002,process and system for retrieval of documents using context-relevant semantic profiles,"a process and system for database storage and retrieval are described along with methods for obtaining semantic profiles from a training text corpus, i.e., text of known relevance, a method for using the training to guide context-relevant document retrieval, and a method for limiting the range of documents that need to be searched after a query. a neural network is used to extract semantic profiles from text corpus. a new set of documents, such as world wide web pages obtained from the internet, is then submitted for processing to the same neural network, which computes a semantic profile representation for these pages using the semantic relations learned from profiling the training documents. these semantic profiles are then organized into clusters in order to minimize the time required to answer a query. when a user queries the database, i.e., the set of documents, his or her query is similarly transformed into a semantic profile and compared with the semantic profiles of each cluster of documents. the query profile is then compared with each of the documents in that cluster. documents with the closest weighted match to the query are returned as search results.",2001-02-13,"The title of the patent is process and system for retrieval of documents using context-relevant semantic profiles and its abstract is a process and system for database storage and retrieval are described along with methods for obtaining semantic profiles from a training text corpus, i.e., text of known relevance, a method for using the training to guide context-relevant document retrieval, and a method for limiting the range of documents that need to be searched after a query. a neural network is used to extract semantic profiles from text corpus. a new set of documents, such as world wide web pages obtained from the internet, is then submitted for processing to the same neural network, which computes a semantic profile representation for these pages using the semantic relations learned from profiling the training documents. these semantic profiles are then organized into clusters in order to minimize the time required to answer a query. when a user queries the database, i.e., the set of documents, his or her query is similarly transformed into a semantic profile and compared with the semantic profiles of each cluster of documents. the query profile is then compared with each of the documents in that cluster. documents with the closest weighted match to the query are returned as search results. dated 2001-02-13"
6192273,non-programmable automated heart rhythm classifier,"a nonprogrammable automated heart rhythm classifier that may be used alone or in conjunction with a therapy system for delivering shock treatment or therapeutic drugs to a patient, a monitoring or recording system, a paging or alarm system, or other rhythm classifying device. the nonprogrammable heart rhythm classifier is used to determine whether a patient's heart rhythm is normal, monomorphic tachycardia or polymorphic tachycardia from extracted features of the cardiac signal of a patent's heart. the extracted features are cycle length and regularity, and preferably with the addition of morphology. prior to feature extraction, the cardiac electrical signal is conditioned with a signal conditioning system. the classifier may comprise a trained neural network or a trained discriminant function, which has been previously trained by a known set of classified heart rhythm data. morphology can be estimated by kurtosis or from the probability density function. regularity can be determined from approximate entropy, information dimension, correlation dimension or from lyapunov's exponents of the patient's cardiac electrical signal. in yet another embodiment of the invention, adaptive sampling may be utilized to selectively digitize the cardiac electrical signal before classification occurs.",2001-02-20,"The title of the patent is non-programmable automated heart rhythm classifier and its abstract is a nonprogrammable automated heart rhythm classifier that may be used alone or in conjunction with a therapy system for delivering shock treatment or therapeutic drugs to a patient, a monitoring or recording system, a paging or alarm system, or other rhythm classifying device. the nonprogrammable heart rhythm classifier is used to determine whether a patient's heart rhythm is normal, monomorphic tachycardia or polymorphic tachycardia from extracted features of the cardiac signal of a patent's heart. the extracted features are cycle length and regularity, and preferably with the addition of morphology. prior to feature extraction, the cardiac electrical signal is conditioned with a signal conditioning system. the classifier may comprise a trained neural network or a trained discriminant function, which has been previously trained by a known set of classified heart rhythm data. morphology can be estimated by kurtosis or from the probability density function. regularity can be determined from approximate entropy, information dimension, correlation dimension or from lyapunov's exponents of the patient's cardiac electrical signal. in yet another embodiment of the invention, adaptive sampling may be utilized to selectively digitize the cardiac electrical signal before classification occurs. dated 2001-02-20"
6192351,fuzzy neural networks,"there is disclosed a pattern identifying neural network comprising at least an input and an output layer, the output layer having a plurality of principal nodes, each principal node trained to recognize a different class of patterns, and at least one fuzzy node trained to recognize all classes of patterns recognized by the principal nodes but with outputs set out at levels lower than the corresponding outputs of the principal nodes.",2001-02-20,"The title of the patent is fuzzy neural networks and its abstract is there is disclosed a pattern identifying neural network comprising at least an input and an output layer, the output layer having a plurality of principal nodes, each principal node trained to recognize a different class of patterns, and at least one fuzzy node trained to recognize all classes of patterns recognized by the principal nodes but with outputs set out at levels lower than the corresponding outputs of the principal nodes. dated 2001-02-20"
6192352,artificial neural network and fuzzy logic based boiler tube leak detection systems,"power industry boiler tube failures are a major cause of utility forced outages in the united states, with approximately 41,000 tube failures occurring every year at a cost of $5 billion a year. accordingly, early tube leak detection and isolation is highly desirable. early detection allows scheduling of a repair rather than suffering a forced outage, and significantly increases the chance of preventing damage to adjacent tubes. the instant detection scheme starts with identification of boiler tube leak process variables which are divided into universal sensitive variables, local leak sensitive variables, group leak sensitive variables, and subgroup leak sensitive variables, and which may be automatically be obtained using a data driven approach and a leak sensitivity function. one embodiment uses artificial neural networks (ann) to learn the map between appropriate leak sensitive variables and the leak behavior. the second design philosophy integrates anns with approximate reasoning using fuzzy logic and fuzzy sets. in the second design, anns are used for learning, while approximate reasoning and inference engines are used for decision making. advantages include use of already monitored process variables, no additional hardware and/or maintenance requirements, systematic processing does not require an expert system and/or a skilled operator, and the systems are portable and can be easily tailored for use on a variety of different boilers.",2001-02-20,"The title of the patent is artificial neural network and fuzzy logic based boiler tube leak detection systems and its abstract is power industry boiler tube failures are a major cause of utility forced outages in the united states, with approximately 41,000 tube failures occurring every year at a cost of $5 billion a year. accordingly, early tube leak detection and isolation is highly desirable. early detection allows scheduling of a repair rather than suffering a forced outage, and significantly increases the chance of preventing damage to adjacent tubes. the instant detection scheme starts with identification of boiler tube leak process variables which are divided into universal sensitive variables, local leak sensitive variables, group leak sensitive variables, and subgroup leak sensitive variables, and which may be automatically be obtained using a data driven approach and a leak sensitivity function. one embodiment uses artificial neural networks (ann) to learn the map between appropriate leak sensitive variables and the leak behavior. the second design philosophy integrates anns with approximate reasoning using fuzzy logic and fuzzy sets. in the second design, anns are used for learning, while approximate reasoning and inference engines are used for decision making. advantages include use of already monitored process variables, no additional hardware and/or maintenance requirements, systematic processing does not require an expert system and/or a skilled operator, and the systems are portable and can be easily tailored for use on a variety of different boilers. dated 2001-02-20"
6198843,method and apparatus for color gamut mapping,"colorimetric values such as l*a*b* are input from an input section to a color gamut mapping portion. a filtering process is performed on an image of the difference between the image sent by the input section and an image obtained as a result of bi-directional conversion at a converter section, and the input to the converter section is controlled so as to satisfy a requirement of minimizing an evaluation function which is defined by the sum of square norm of the filtered image as a whole and a value obtained by performing a thresholding process on norm of the difference in each pixel between images before and after the conversion at the converter section output by a gamut extracting section, obtained for the image as a whole. device values satisfying predetermined conditions which are output of a first neural network of a first converter section to an image output section.",2001-03-06,"The title of the patent is method and apparatus for color gamut mapping and its abstract is colorimetric values such as l*a*b* are input from an input section to a color gamut mapping portion. a filtering process is performed on an image of the difference between the image sent by the input section and an image obtained as a result of bi-directional conversion at a converter section, and the input to the converter section is controlled so as to satisfy a requirement of minimizing an evaluation function which is defined by the sum of square norm of the filtered image as a whole and a value obtained by performing a thresholding process on norm of the difference in each pixel between images before and after the conversion at the converter section output by a gamut extracting section, obtained for the image as a whole. device values satisfying predetermined conditions which are output of a first neural network of a first converter section to an image output section. dated 2001-03-06"
6199057,bit-serial neuroprocessor architecture,"a neuroprocessor architecture employs a combination of bit-serial and serial-parallel techniques for implementing the neurons of the neuroprocessor. the neuroprocessor architecture includes a neural module containing a pool of neurons, a global controller, a sigmoid activation rom look-up-table, a plurality of neuron state registers, and a synaptic weight ram. the neuroprocessor reduces the number of neurons required to perform the task by time multiplexing groups of neurons from a fixed pool of neurons to achieve the successive hidden layers of a recurrent network topology.",2001-03-06,"The title of the patent is bit-serial neuroprocessor architecture and its abstract is a neuroprocessor architecture employs a combination of bit-serial and serial-parallel techniques for implementing the neurons of the neuroprocessor. the neuroprocessor architecture includes a neural module containing a pool of neurons, a global controller, a sigmoid activation rom look-up-table, a plurality of neuron state registers, and a synaptic weight ram. the neuroprocessor reduces the number of neurons required to perform the task by time multiplexing groups of neurons from a fixed pool of neurons to achieve the successive hidden layers of a recurrent network topology. dated 2001-03-06"
6205236,method and system for automated detection of clustered microcalcifications from digital mammograms,"a method and system for detecting and displaying clustered microcalcifications in a digital mammogram, wherein a single digital mammogram is first automatically cropped to a breast area sub-image which is then processed by means of an optimized difference of gaussians filter to enhance the appearance of potential microcalcifications in the sub-image. the potential microcalcifications are thresholded clusters are detected, features are computed for the detected clusters, and the clusters are classified as either suspicious or not suspicious by means of a neural network. thresholding is preferably by sloping local thresholding but may also be performed by global and dual-local thresholding. the locations in the original digital mammogram of the suspicious detected clustered microcalcifications are indicated. parameters for use in the detection and thresholding portions of the system are computer-optimized by means of a genetic algorithm. the results of the system are optimally combined with a radiologist's observation of the original mammogram by combining the observations with the results, after the radiologist has first accepted or rejected individual detections reported by the system.",2001-03-20,"The title of the patent is method and system for automated detection of clustered microcalcifications from digital mammograms and its abstract is a method and system for detecting and displaying clustered microcalcifications in a digital mammogram, wherein a single digital mammogram is first automatically cropped to a breast area sub-image which is then processed by means of an optimized difference of gaussians filter to enhance the appearance of potential microcalcifications in the sub-image. the potential microcalcifications are thresholded clusters are detected, features are computed for the detected clusters, and the clusters are classified as either suspicious or not suspicious by means of a neural network. thresholding is preferably by sloping local thresholding but may also be performed by global and dual-local thresholding. the locations in the original digital mammogram of the suspicious detected clustered microcalcifications are indicated. parameters for use in the detection and thresholding portions of the system are computer-optimized by means of a genetic algorithm. the results of the system are optimally combined with a radiologist's observation of the original mammogram by combining the observations with the results, after the radiologist has first accepted or rejected individual detections reported by the system. dated 2001-03-20"
6205556,semiconductor integrated circuit device comprising a memory array and a processing circuit,"herein disclosed is a data processing system having a memory packaged therein for realizing a large-scale and high-speed parallel distributed processing and, especially, a data processing system for the neural network processing. the neural network processing system according to the present invention comprises: a memory circuit for storing neuron output values, connection weights, the desired values of outputs, and data necessary for learning; an input/output circuit for writing or reading data in or out of said memory circuit; a processing circuit for performing a processing for determining the neuron outputs such as the product, sum and nonlinear conversion of the data stored in said memory circuit, a comparison of the output value and its desired value, and a processing necessary for learning; and a control circuit for controlling the operations of said memory circuit, said input/output circuit and said processing circuit. the processing circuit is constructed to include at least one of an adder, a multiplier, a nonlinear transfer function circuit and a comparator so that at least a portion of the processing necessary for determining the neutron output values such as the product or sum may be accomplished in parallel. moreover, these circuits are shared among a plurality of neutrons and are operated in a time sharing manner to determine the plural neuron output values. still moreover, the aforementioned comparator compares the neuron output value determined and the desired value of the output in parallel.",2001-03-20,"The title of the patent is semiconductor integrated circuit device comprising a memory array and a processing circuit and its abstract is herein disclosed is a data processing system having a memory packaged therein for realizing a large-scale and high-speed parallel distributed processing and, especially, a data processing system for the neural network processing. the neural network processing system according to the present invention comprises: a memory circuit for storing neuron output values, connection weights, the desired values of outputs, and data necessary for learning; an input/output circuit for writing or reading data in or out of said memory circuit; a processing circuit for performing a processing for determining the neuron outputs such as the product, sum and nonlinear conversion of the data stored in said memory circuit, a comparison of the output value and its desired value, and a processing necessary for learning; and a control circuit for controlling the operations of said memory circuit, said input/output circuit and said processing circuit. the processing circuit is constructed to include at least one of an adder, a multiplier, a nonlinear transfer function circuit and a comparator so that at least a portion of the processing necessary for determining the neutron output values such as the product or sum may be accomplished in parallel. moreover, these circuits are shared among a plurality of neutrons and are operated in a time sharing manner to determine the plural neuron output values. still moreover, the aforementioned comparator compares the neuron output value determined and the desired value of the output in parallel. dated 2001-03-20"
6206324,"wing-drive mechanism, vehicle employing same, and method for controlling the wing-drive mechanism and vehicle employing same","a wing-drive mechanism is described that permits, with proper control, movement of a wing about multiple wing trajectories. the wing-drive is capable of independent movement about three rotational degrees of movement; movement about a flap axis is independent of movement about a yaw axis, and both are independent of changes in the pitch of the wing. methods of controlling the wing-drive mechanism to affect a desired wing trajectory include the use of a non-linear automated controller that generates input signals to the wing-drive mechanism by comparing actual and desired wing trajectories in real time. specification of wing trajectories is preferably also accomplished in real time using an automated trajectory specification system, which can include a fuzzy logic processor or a neural network. a vehicle that derives controlled motion as a whole from the wing-drive mechanism is also disclosed.",2001-03-27,"The title of the patent is wing-drive mechanism, vehicle employing same, and method for controlling the wing-drive mechanism and vehicle employing same and its abstract is a wing-drive mechanism is described that permits, with proper control, movement of a wing about multiple wing trajectories. the wing-drive is capable of independent movement about three rotational degrees of movement; movement about a flap axis is independent of movement about a yaw axis, and both are independent of changes in the pitch of the wing. methods of controlling the wing-drive mechanism to affect a desired wing trajectory include the use of a non-linear automated controller that generates input signals to the wing-drive mechanism by comparing actual and desired wing trajectories in real time. specification of wing trajectories is preferably also accomplished in real time using an automated trajectory specification system, which can include a fuzzy logic processor or a neural network. a vehicle that derives controlled motion as a whole from the wing-drive mechanism is also disclosed. dated 2001-03-27"
6207936,model-based predictive control of thermal processing,"a nonlinear model-based predictive temperature control system is described for use in thermal process reactors. a multivariable temperature response is predicted using a nonlinear parameterized model of a thermal process reactor. the nonlinear parameterized model is implemented using a neural network. predictions are made in an auto-regressive moving average fashion with a receding prediction horizon. model predictions are incorporated into a control law for estimating the optimum future control strategy. the high-speed, predictive nature of the controller renders it advantageous in multivariable rapid thermal processing reactors where fast response and high temperature uniformity are needed.",2001-03-27,"The title of the patent is model-based predictive control of thermal processing and its abstract is a nonlinear model-based predictive temperature control system is described for use in thermal process reactors. a multivariable temperature response is predicted using a nonlinear parameterized model of a thermal process reactor. the nonlinear parameterized model is implemented using a neural network. predictions are made in an auto-regressive moving average fashion with a receding prediction horizon. model predictions are incorporated into a control law for estimating the optimum future control strategy. the high-speed, predictive nature of the controller renders it advantageous in multivariable rapid thermal processing reactors where fast response and high temperature uniformity are needed. dated 2001-03-27"
6208758,method for learning by a neural network including extracting a target object image for which learning operations are to be carried out,"a method for recognizing an object image comprises the steps of extracting a candidate for a predetermined object image from an image, and making a judgment as to whether the extracted candidate for the predetermined object image is or is not the predetermined object image. the candidate for the predetermined object image is extracted by causing the center point of a view window, which has a predetermined size, to travel to the position of the candidate for the predetermined object image, and determining an extraction area in accordance with the size and/or the shape of the candidate for the predetermined object image, the center point of the view window being taken as a reference during the determination of the extraction area. a learning method for a neural network comprises the steps of extracting a target object image, for which learning operations are to be carried out, from an image, feeding a signal, which represents the extracted target object image, into a neural network, and carrying out the learning operations of the neural network in accordance with the input target object image.",2001-03-27,"The title of the patent is method for learning by a neural network including extracting a target object image for which learning operations are to be carried out and its abstract is a method for recognizing an object image comprises the steps of extracting a candidate for a predetermined object image from an image, and making a judgment as to whether the extracted candidate for the predetermined object image is or is not the predetermined object image. the candidate for the predetermined object image is extracted by causing the center point of a view window, which has a predetermined size, to travel to the position of the candidate for the predetermined object image, and determining an extraction area in accordance with the size and/or the shape of the candidate for the predetermined object image, the center point of the view window being taken as a reference during the determination of the extraction area. a learning method for a neural network comprises the steps of extracting a target object image, for which learning operations are to be carried out, from an image, feeding a signal, which represents the extracted target object image, into a neural network, and carrying out the learning operations of the neural network in accordance with the input target object image. dated 2001-03-27"
6208981,circuit configuration for controlling a running-gear or drive system in a motor vehicle,"motor vehicle sensor signals are evaluated by a fuzzy system, which generates control signals for a system device of the motor vehicle--for example an automatic transmission, active suspension, speed stabilization, power-steering assistance, or traction control. the fuzzy system is connected to a neural network, which evaluates the sensor signals and reference data from a recording of driving data of the motor vehicle. the neural network optimizes the rule base of the fuzzy system. during a driving operation, the fuzzy system generates on-line signals categorizing the respective driving situation, and thus makes possible intelligent, time-adaptive, driving-situation-dependent control. the fuzzy system and the neural network each contain a classification system which can be reciprocally converted by a correspondence-maintaining bidirectional transformation.",2001-03-27,"The title of the patent is circuit configuration for controlling a running-gear or drive system in a motor vehicle and its abstract is motor vehicle sensor signals are evaluated by a fuzzy system, which generates control signals for a system device of the motor vehicle--for example an automatic transmission, active suspension, speed stabilization, power-steering assistance, or traction control. the fuzzy system is connected to a neural network, which evaluates the sensor signals and reference data from a recording of driving data of the motor vehicle. the neural network optimizes the rule base of the fuzzy system. during a driving operation, the fuzzy system generates on-line signals categorizing the respective driving situation, and thus makes possible intelligent, time-adaptive, driving-situation-dependent control. the fuzzy system and the neural network each contain a classification system which can be reciprocally converted by a correspondence-maintaining bidirectional transformation. dated 2001-03-27"
6208982,method and apparatus for solving complex and computationally intensive inverse problems in real-time,""" the system of the present invention may """"solve"""" a variety of inverse physical problem types by using neural network techniques. in operation, the present invention may generate data sets characterizing a particular starting condition of a physical process (such as data sets characterizing the parameters of an initial metal die), based upon an ending condition of the physical process (such as the parameters of the metal part to be stamped by the die). in one embodiment, the system of the present invention may generate a plurality of training data sets, each training data set characterizing a sample ending condition, the physical process that results in the sample ending condition, and a sample starting condition of the physical process. the training data sets may then be applied to a neural network so as to train the network. a network definition associated with the trained neural network may be stored, and an ending data set characterizing an ending condition of the physical process may be generated. a starting data set characterizing a starting condition of the physical process may thereafter be generated based upon the stored network definition and the ending data set. """,2001-03-27,"The title of the patent is method and apparatus for solving complex and computationally intensive inverse problems in real-time and its abstract is "" the system of the present invention may """"solve"""" a variety of inverse physical problem types by using neural network techniques. in operation, the present invention may generate data sets characterizing a particular starting condition of a physical process (such as data sets characterizing the parameters of an initial metal die), based upon an ending condition of the physical process (such as the parameters of the metal part to be stamped by the die). in one embodiment, the system of the present invention may generate a plurality of training data sets, each training data set characterizing a sample ending condition, the physical process that results in the sample ending condition, and a sample starting condition of the physical process. the training data sets may then be applied to a neural network so as to train the network. a network definition associated with the trained neural network may be stored, and an ending data set characterizing an ending condition of the physical process may be generated. a starting data set characterizing a starting condition of the physical process may thereafter be generated based upon the stored network definition and the ending data set. "" dated 2001-03-27"
6208983,method and apparatus for training and operating a neural network for detecting breast cancer,""" a method and apparatus for training and operating a neural network using gated data. the neural network is a mixture of experts that performs """"soft"""" partitioning of a network of experts. in a specific embodiment, the technique is used to detect malignancy by analyzing skin surface potential data. in particular, the invention uses certain patient information, such as menstrual cycle information, to """"gate"""" the expert output data into particular populations, i.e., the network is soft partitioned into the populations. an expectation-maximization (em) routine is used to train the neural network using known patient information, known measured skin potential data and correct diagnosis for the particular training data and patient information. once trained, the neural network parameters are used in a classifier for predicting breast cancer malignancy when given the patient information and skin potentials of other patients. """,2001-03-27,"The title of the patent is method and apparatus for training and operating a neural network for detecting breast cancer and its abstract is "" a method and apparatus for training and operating a neural network using gated data. the neural network is a mixture of experts that performs """"soft"""" partitioning of a network of experts. in a specific embodiment, the technique is used to detect malignancy by analyzing skin surface potential data. in particular, the invention uses certain patient information, such as menstrual cycle information, to """"gate"""" the expert output data into particular populations, i.e., the network is soft partitioned into the populations. an expectation-maximization (em) routine is used to train the neural network using known patient information, known measured skin potential data and correct diagnosis for the particular training data and patient information. once trained, the neural network parameters are used in a classifier for predicting breast cancer malignancy when given the patient information and skin potentials of other patients. "" dated 2001-03-27"
6212438,method and apparatus for generating a model of an industrial production,"a process model of an industrial process or system is generated. the model correlates a first number m of process parameters forming input values with a second number l of quality characteristics forming output values, which are processed to form feedback control signals for the process or system. a third number n of training data sets of the industrial process are first gathered and processed during a learning phase of the model with the help of a central processing unit, whereby a preliminary approximately model is used including a neural network with local approximation characteristics. the neural network is connected in parallel with a linear network. both networks are connected to the same inputs. the neural network initially has a number n of neural cells corresponding to the number of training data sets. a weighted linear combination of the m process parameters is performed. the linear network and the neural network are connected with their outputs through weighting circuits to a common summing point. a stepwise regression is performed to reduce the number of neural cells from n to k and of linear paths from m to m-r. closed loop feedback signals control the industrial process.",2001-04-03,"The title of the patent is method and apparatus for generating a model of an industrial production and its abstract is a process model of an industrial process or system is generated. the model correlates a first number m of process parameters forming input values with a second number l of quality characteristics forming output values, which are processed to form feedback control signals for the process or system. a third number n of training data sets of the industrial process are first gathered and processed during a learning phase of the model with the help of a central processing unit, whereby a preliminary approximately model is used including a neural network with local approximation characteristics. the neural network is connected in parallel with a linear network. both networks are connected to the same inputs. the neural network initially has a number n of neural cells corresponding to the number of training data sets. a weighted linear combination of the m process parameters is performed. the linear network and the neural network are connected with their outputs through weighting circuits to a common summing point. a stepwise regression is performed to reduce the number of neural cells from n to k and of linear paths from m to m-r. closed loop feedback signals control the industrial process. dated 2001-04-03"
6212466,optimization control method for shock absorber,a control system for optimizing the performance of a vehicle suspension system by controlling the damping factor of one or more shock absorbers is described. the control system uses a fitness (performance) function that is based on the physical laws of minimum entropy. the control system uses a fuzzy neural network that is trained by a genetic analyzer. the genetic analyzer uses a fitness function that maximizes information while minimizing entropy production. the fitness function uses a difference between the time differential of entropy from a control signal produced in a learning control module and the time differential of the entropy calculated by a model of the suspension system that uses the control signal as an input. the entropy calculation is based on a dynamic model of an equation of motion for the suspension system such that the suspension system is treated as an open dynamic system.,2001-04-03,The title of the patent is optimization control method for shock absorber and its abstract is a control system for optimizing the performance of a vehicle suspension system by controlling the damping factor of one or more shock absorbers is described. the control system uses a fitness (performance) function that is based on the physical laws of minimum entropy. the control system uses a fuzzy neural network that is trained by a genetic analyzer. the genetic analyzer uses a fitness function that maximizes information while minimizing entropy production. the fitness function uses a difference between the time differential of entropy from a control signal produced in a learning control module and the time differential of the entropy calculated by a model of the suspension system that uses the control signal as an input. the entropy calculation is based on a dynamic model of an equation of motion for the suspension system such that the suspension system is treated as an open dynamic system. dated 2001-04-03
6212508,process and arrangement for conditioning an input variable of a neural network,"a process and an arrangement for conditioning input variables of a neural network are described by the invention. from the input variables of the network, time series are formed and these are subdivided into intervals whose length depends on how far the interval and the measured variables contained therein lie back in the past. in this case, the interval length is selected to be larger the further the interval lies back in the past. by means of convolution using a bell-shaped function, a representative input value for the neural network is obtained from all these measured variables contained in the interval. all the input variables which are obtained in this way are fed to the network simultaneously during training and during operation. a memory is thus realized in a simple way for a forwardly directed neural network. potential applications include, in particular, chemical processes having very different time constants.",2001-04-03,"The title of the patent is process and arrangement for conditioning an input variable of a neural network and its abstract is a process and an arrangement for conditioning input variables of a neural network are described by the invention. from the input variables of the network, time series are formed and these are subdivided into intervals whose length depends on how far the interval and the measured variables contained therein lie back in the past. in this case, the interval length is selected to be larger the further the interval lies back in the past. by means of convolution using a bell-shaped function, a representative input value for the neural network is obtained from all these measured variables contained in the interval. all the input variables which are obtained in this way are fed to the network simultaneously during training and during operation. a memory is thus realized in a simple way for a forwardly directed neural network. potential applications include, in particular, chemical processes having very different time constants. dated 2001-04-03"
6212509,visualization and self-organization of multidimensional data through equalized orthogonal mapping,""" the subject system provides reduced-dimension mapping of pattern data. mapping is applied through conventional single-hidden-layer feed-forward neural network with non-linear neurons. according to one aspect of the present invention, the system functions to equalize and orthogonalize lower dimensional output signals by reducing the covariance matrix of the output signals to the form of a diagonal matrix or constant times the identity matrix. the present invention allows for visualization of large bodies of complex multidimensional data in a relatively """"topologically correct"""" low-dimension approximation, to reduce randomness associated with other methods of similar purposes, and to keep the mapping computationally efficient at the same time. """,2001-04-03,"The title of the patent is visualization and self-organization of multidimensional data through equalized orthogonal mapping and its abstract is "" the subject system provides reduced-dimension mapping of pattern data. mapping is applied through conventional single-hidden-layer feed-forward neural network with non-linear neurons. according to one aspect of the present invention, the system functions to equalize and orthogonalize lower dimensional output signals by reducing the covariance matrix of the output signals to the form of a diagonal matrix or constant times the identity matrix. the present invention allows for visualization of large bodies of complex multidimensional data in a relatively """"topologically correct"""" low-dimension approximation, to reduce randomness associated with other methods of similar purposes, and to keep the mapping computationally efficient at the same time. "" dated 2001-04-03"
6213934,electromagnetic bone-assessment and treatment: apparatus and method,"non-invasive quantitative in-vivo electromagnetic evaluation of bone is performed by subjecting bone to an electrical excitation waveform supplied to a single magnetic field coil near the skin of a bony member, and involving a repetitive finite duration signal consisting of plural frequencies that are in the range 0 hz-200 mhz. signal-processing of a bone-current response signal and a bone-voltage response signal is operative to sequentially average the most recently received given number of successive bone-current and bone-voltage response signals to obtain an averaged per-pulse bone-current signal and an averaged per-pulse bone-voltage signal, and to produce their associated fourier transforms. these fourier transforms are further processed to obtain the inductively determined frequency-dependent bone-admittance function. a neural network, configured to generate an estimate of one or more of the desired bone-related quantities, is connected for response to the bone-admittance function, whereby to generate the indicated estimates of bone status, namely, bone-density, bone-architecture, bone-strength and bone-fracture risk. in another embodiment, a stochastic electromagnetic field generated by a single magnetic field coil is used to therapeutically treat living tissue in vivo.",2001-04-10,"The title of the patent is electromagnetic bone-assessment and treatment: apparatus and method and its abstract is non-invasive quantitative in-vivo electromagnetic evaluation of bone is performed by subjecting bone to an electrical excitation waveform supplied to a single magnetic field coil near the skin of a bony member, and involving a repetitive finite duration signal consisting of plural frequencies that are in the range 0 hz-200 mhz. signal-processing of a bone-current response signal and a bone-voltage response signal is operative to sequentially average the most recently received given number of successive bone-current and bone-voltage response signals to obtain an averaged per-pulse bone-current signal and an averaged per-pulse bone-voltage signal, and to produce their associated fourier transforms. these fourier transforms are further processed to obtain the inductively determined frequency-dependent bone-admittance function. a neural network, configured to generate an estimate of one or more of the desired bone-related quantities, is connected for response to the bone-admittance function, whereby to generate the indicated estimates of bone status, namely, bone-density, bone-architecture, bone-strength and bone-fracture risk. in another embodiment, a stochastic electromagnetic field generated by a single magnetic field coil is used to therapeutically treat living tissue in vivo. dated 2001-04-10"
6213958,"method and apparatus for the acoustic emission monitoring detection, localization, and classification of metabolic bone disease","a non-invasive bone condition data acquisition system performs sensitive and reliable clinical data acquisition, localization and classification of bone disease, particularly osteoporosis. the bone condition data acquisition system measures a correlation between a wideband ae signature and a spatially localized bone microarchitecture, which is used to determine fracture risk. the bone condition data acquisition system includes processors and memory for analyzing ae signals from bone tissue to generate information-bearing attributes, for extracting a set of times-of-arrival and a feature vector from the attributes, for utilizing the set of times-of-arrival to derive the locations of the ae events, and for responding to the feature vector to classify the bone using a neural network and a nearest neighbor rule processor.",2001-04-10,"The title of the patent is method and apparatus for the acoustic emission monitoring detection, localization, and classification of metabolic bone disease and its abstract is a non-invasive bone condition data acquisition system performs sensitive and reliable clinical data acquisition, localization and classification of bone disease, particularly osteoporosis. the bone condition data acquisition system measures a correlation between a wideband ae signature and a spatially localized bone microarchitecture, which is used to determine fracture risk. the bone condition data acquisition system includes processors and memory for analyzing ae signals from bone tissue to generate information-bearing attributes, for extracting a set of times-of-arrival and a feature vector from the attributes, for utilizing the set of times-of-arrival to derive the locations of the ae events, and for responding to the feature vector to classify the bone using a neural network and a nearest neighbor rule processor. dated 2001-04-10"
6215271,charging system having a controlled rectifier bridge and a single voltage sensor,"a charging system has an alternator, a battery, and a controlled rectifier bridge between the alternator and the battery. the controlled rectifier bridge has controllable elements. a controller is provided wherein the controller controls (a) the duty cycle of the alternator field winding and (b) the switching in the controllable elements to control the phase advance angle. by utilizing a half-wave controlled rectifier bridge and three controllable elements, power output can be increased at a low overall cost. in another embodiment, the controller determines switching points for the controllable elements through a third harmonic extraction device which employs a neural network for determining the zero crossings of the third harmonic voltage obtained from a single phase voltage.",2001-04-10,"The title of the patent is charging system having a controlled rectifier bridge and a single voltage sensor and its abstract is a charging system has an alternator, a battery, and a controlled rectifier bridge between the alternator and the battery. the controlled rectifier bridge has controllable elements. a controller is provided wherein the controller controls (a) the duty cycle of the alternator field winding and (b) the switching in the controllable elements to control the phase advance angle. by utilizing a half-wave controlled rectifier bridge and three controllable elements, power output can be increased at a low overall cost. in another embodiment, the controller determines switching points for the controllable elements through a third harmonic extraction device which employs a neural network for determining the zero crossings of the third harmonic voltage obtained from a single phase voltage. dated 2001-04-10"
6216048,method and apparatus for determining the sensitivity of inputs to a neural network on output parameters,"a distributed control system (14) receives on the input thereof the control inputs and then outputs control signals to a plant (10) for the operation thereof. the measured variables of the plant and the control inputs are input to a predictive model (34) that operates in conjunction with an inverse model (36) to generate predicted control inputs. the predicted control inputs are processed through a filter (46) to apply hard constraints and sensitivity modifiers, the values of which are received from a control parameter block (22). during operation, the sensitivity of output variables on various input variables is determined. this information can be displayed and then the user allowed to select which of the input variables constitute the most sensitive input variables. these can then be utilized with a control network (470) to modify the predicted values of the input variables. additionally, a neural network (406) can be trained on only the selected input variables that are determined to be the most sensitive. in this operation, the network is first configured and trained with all input nodes and with all training data. this provides a learned representation of the output wherein the combined effects of all other input variables are taken into account in the determination of the effect of each of the input variables thereon. the network (406) is then reconfigured with only the selected inputs and then the network (406) again trained on only the input/output pairs associated with the select input variables.",2001-04-10,"The title of the patent is method and apparatus for determining the sensitivity of inputs to a neural network on output parameters and its abstract is a distributed control system (14) receives on the input thereof the control inputs and then outputs control signals to a plant (10) for the operation thereof. the measured variables of the plant and the control inputs are input to a predictive model (34) that operates in conjunction with an inverse model (36) to generate predicted control inputs. the predicted control inputs are processed through a filter (46) to apply hard constraints and sensitivity modifiers, the values of which are received from a control parameter block (22). during operation, the sensitivity of output variables on various input variables is determined. this information can be displayed and then the user allowed to select which of the input variables constitute the most sensitive input variables. these can then be utilized with a control network (470) to modify the predicted values of the input variables. additionally, a neural network (406) can be trained on only the selected input variables that are determined to be the most sensitive. in this operation, the network is first configured and trained with all input nodes and with all training data. this provides a learned representation of the output wherein the combined effects of all other input variables are taken into account in the determination of the effect of each of the input variables thereon. the network (406) is then reconfigured with only the selected inputs and then the network (406) again trained on only the input/output pairs associated with the select input variables. dated 2001-04-10"
6216083,system for intelligent control of an engine based on soft computing,a reduced control system suitable for control of an engine as a nonlinear plant is described. the reduced control system is configured to use a reduced sensor set for controlling the plant without significant loss of control quality (accuracy) as compared to an optimal control system with an optimum sensor set. the control system calculates the information content provided by the reduced sensor set as compared to the information content provided by the optimum set. the control system also calculates the difference between the entropy production rate of the plant and the entropy production rate of the controller. a genetic optimizer is used to tune a fuzzy neural network in the reduced controller. a fitness function for the genetic optimizer provides optimum control accuracy in the reduced control system by minimizing the difference in entropy production while maximizing the sensor information content.,2001-04-10,The title of the patent is system for intelligent control of an engine based on soft computing and its abstract is a reduced control system suitable for control of an engine as a nonlinear plant is described. the reduced control system is configured to use a reduced sensor set for controlling the plant without significant loss of control quality (accuracy) as compared to an optimal control system with an optimum sensor set. the control system calculates the information content provided by the reduced sensor set as compared to the information content provided by the optimum set. the control system also calculates the difference between the entropy production rate of the plant and the entropy production rate of the controller. a genetic optimizer is used to tune a fuzzy neural network in the reduced controller. a fitness function for the genetic optimizer provides optimum control accuracy in the reduced control system by minimizing the difference in entropy production while maximizing the sensor information content. dated 2001-04-10
6216119,"multi-kernel neural network concurrent learning, monitoring, and forecasting system","a multi-kernel neural network computing architecture configured to learn correlations among feature values 34, 38 as the network monitors and imputes measured input values 30 and also predicts future output values 46. this computing architecture, referred to as a concurrent-learning information processor (cip 10), includes a multi-kernel neural network array 14 with the capability to learn and predict in real time. the cip 10 also includes a manager 16 and an input-output transducer 12 that may be used for input-output refinement. these components allow the computing capacity of the multi-kernel array 14 to be reassigned in response to measured performance or other factors. the output feature values 46 computed by the multi-kernel array 14 and processed by an output processor 44 of the transducer 12 are supplied to a response unit 18 that may be configured to perform a variety of monitoring, forecasting, and control operations in response to the computed output values. important characteristics of the cip 10, such as feature function specifications 35 and 49, connection specifications 42, learning weight schedules 55, and the like may be set by a technician through a graphical user interface 20. refinement processes also allow the cip 10 be reconfigured in accordance with user commands for application to different physical applications.",2001-04-10,"The title of the patent is multi-kernel neural network concurrent learning, monitoring, and forecasting system and its abstract is a multi-kernel neural network computing architecture configured to learn correlations among feature values 34, 38 as the network monitors and imputes measured input values 30 and also predicts future output values 46. this computing architecture, referred to as a concurrent-learning information processor (cip 10), includes a multi-kernel neural network array 14 with the capability to learn and predict in real time. the cip 10 also includes a manager 16 and an input-output transducer 12 that may be used for input-output refinement. these components allow the computing capacity of the multi-kernel array 14 to be reassigned in response to measured performance or other factors. the output feature values 46 computed by the multi-kernel array 14 and processed by an output processor 44 of the transducer 12 are supplied to a response unit 18 that may be configured to perform a variety of monitoring, forecasting, and control operations in response to the computed output values. important characteristics of the cip 10, such as feature function specifications 35 and 49, connection specifications 42, learning weight schedules 55, and the like may be set by a technician through a graphical user interface 20. refinement processes also allow the cip 10 be reconfigured in accordance with user commands for application to different physical applications. dated 2001-04-10"
6216267,"media capture and compression communication system using holographic optical classification, voice recognition and neural network decision processing",a system and method for capturing and processing audio and video images for optimized transmission over a lower bandwidth communications channel are disclosed. a digital processing system receives video and audio signals from a media source such that the video and audio signals are representative of captured images and sounds as a multimedia signal. an optical processing system is coupled to the digital processing system and performs signal processing algorithms on the multimedia signal. a transceiver transmits the multimedia signal over a lower bandwidth communications channel wherein the multimedia signal is optimized for transmission over the lower bandwidth communications channel by the signal processing algorithms performed by the optical processing system.,2001-04-10,"The title of the patent is media capture and compression communication system using holographic optical classification, voice recognition and neural network decision processing and its abstract is a system and method for capturing and processing audio and video images for optimized transmission over a lower bandwidth communications channel are disclosed. a digital processing system receives video and audio signals from a media source such that the video and audio signals are representative of captured images and sounds as a multimedia signal. an optical processing system is coupled to the digital processing system and performs signal processing algorithms on the multimedia signal. a transceiver transmits the multimedia signal over a lower bandwidth communications channel wherein the multimedia signal is optimized for transmission over the lower bandwidth communications channel by the signal processing algorithms performed by the optical processing system. dated 2001-04-10"
6219642,quantization using frequency and mean compensated frequency input data for robust speech recognition,""" a speech recognition system utilizes multiple quantizers to process frequency parameters and mean compensated frequency parameters derived from an input signal. the quantizers may be matrix and vector quantizer pairs, and such quantizer pairs may also function as front ends to a second stage speech classifiers such as hidden markov models (hmms) and/or utilizes neural network postprocessing to, for example, improve speech recognition performance. mean compensating the frequency parameters can remove noise frequency components that remain approximately constant during the duration of the input signal. hmm initial state and state transition probabilities derived from common quantizer types and the same input signal may be consolidated to improve recognition system performance and efficiency. matrix quantization exploits the """"evolution"""" of the speech short-term spectral envelopes as well as frequency domain information, and vector quantization (vq) primarily operates on frequency domain information. time domain information may be substantially limited which may introduce error into the matrix quantization, and the vq may provide error compensation. the matrix and vector quantizers may split spectral subbands to target selected frequencies for enhanced processing and may use fuzzy associations to develop fuzzy observation sequence data. a mixer may provide a variety of input data to the neural network for classification determination. fuzzy operators may be utilized to reduce quantization error. multiple codebooks may also be combined to form single respective codebooks for split matrix and split vector quantization to reduce processing resources demand. """,2001-04-17,"The title of the patent is quantization using frequency and mean compensated frequency input data for robust speech recognition and its abstract is "" a speech recognition system utilizes multiple quantizers to process frequency parameters and mean compensated frequency parameters derived from an input signal. the quantizers may be matrix and vector quantizer pairs, and such quantizer pairs may also function as front ends to a second stage speech classifiers such as hidden markov models (hmms) and/or utilizes neural network postprocessing to, for example, improve speech recognition performance. mean compensating the frequency parameters can remove noise frequency components that remain approximately constant during the duration of the input signal. hmm initial state and state transition probabilities derived from common quantizer types and the same input signal may be consolidated to improve recognition system performance and efficiency. matrix quantization exploits the """"evolution"""" of the speech short-term spectral envelopes as well as frequency domain information, and vector quantization (vq) primarily operates on frequency domain information. time domain information may be substantially limited which may introduce error into the matrix quantization, and the vq may provide error compensation. the matrix and vector quantizers may split spectral subbands to target selected frequencies for enhanced processing and may use fuzzy associations to develop fuzzy observation sequence data. a mixer may provide a variety of input data to the neural network for classification determination. fuzzy operators may be utilized to reduce quantization error. multiple codebooks may also be combined to form single respective codebooks for split matrix and split vector quantization to reduce processing resources demand. "" dated 2001-04-17"
6219657,device and method for creation of emotions,"a device and a method for creation of emotions are provided for an interface of information, such as an artificial agent and a personified agent, intervened between a human being (i.e., user) and an electronic apparatus. for instance, an emotion creating device is configured by a neural network, a behavior determination engine and a feature determination engine. the neural network inputs user information, representing conditions of the user, and apparatus information, representing conditions of the apparatus, so as to produce emotional states. herein, a present set of emotional states are produced in consideration of a previous set of emotional states. the emotional states represent prescribed emotions such as pleasure, anger, sadness and surprise. the behavior determination engine refers to a behavior determination database using the user information and the emotional states of the neural network so as to determine a behavior of the interface. the feature determination engine refers to a database using the emotional states of the neural network to determine a feature of the interface, which corresponds to a facial feature.",2001-04-17,"The title of the patent is device and method for creation of emotions and its abstract is a device and a method for creation of emotions are provided for an interface of information, such as an artificial agent and a personified agent, intervened between a human being (i.e., user) and an electronic apparatus. for instance, an emotion creating device is configured by a neural network, a behavior determination engine and a feature determination engine. the neural network inputs user information, representing conditions of the user, and apparatus information, representing conditions of the apparatus, so as to produce emotional states. herein, a present set of emotional states are produced in consideration of a previous set of emotional states. the emotional states represent prescribed emotions such as pleasure, anger, sadness and surprise. the behavior determination engine refers to a behavior determination database using the user information and the emotional states of the neural network so as to determine a behavior of the interface. the feature determination engine refers to a database using the emotional states of the neural network to determine a feature of the interface, which corresponds to a facial feature. dated 2001-04-17"
6219658,method for learning data classification in two separate classes separated by a region divider of order 1 or 2,"a method for learning to classify data according to two distinct classes (c11, c12) separated by a separating surface (s), by means of a neurone of the binary type comprising a parameter describing the separating surface and whose inputs are weighted by a weight (w.sub.i), and including the following steps: pa1 a) defining a cost function c: ##equ1## pa1 b) initializing the weights (w.sub.i), the radii (r.sup.i), the parameters (.sigma., t+, t-), the learning rate (.epsilon.) and speeds of the temperature decreasing (.delta.t+, .delta.t-); pa1 c) minimizing the cost function c by successive iterations; pa1 d) obtaining the values of the weights of the connections and radii of the neurone. application to the classification and recognition of shapes by a neural network.",2001-04-17,"The title of the patent is method for learning data classification in two separate classes separated by a region divider of order 1 or 2 and its abstract is a method for learning to classify data according to two distinct classes (c11, c12) separated by a separating surface (s), by means of a neurone of the binary type comprising a parameter describing the separating surface and whose inputs are weighted by a weight (w.sub.i), and including the following steps: pa1 a) defining a cost function c: ##equ1## pa1 b) initializing the weights (w.sub.i), the radii (r.sup.i), the parameters (.sigma., t+, t-), the learning rate (.epsilon.) and speeds of the temperature decreasing (.delta.t+, .delta.t-); pa1 c) minimizing the cost function c by successive iterations; pa1 d) obtaining the values of the weights of the connections and radii of the neurone. application to the classification and recognition of shapes by a neural network. dated 2001-04-17"
6220517,air-conditioning device,"temperatures in a dr side air-conditioning zone and a pa side air-conditioning zone are controlled highly independently of each other without temperature interference between each zone. a room internal air temperature sensor and a room external air temperature sensor are provided. dr side and pa side temperature setters separately set room setpoint temperatures (tset(dr), tset(pa)) in each zone. first and second target blow-out temperature calculating portions, which include neural network, input the room setpoint temperatures and the temperature data. then it calculates dr side and pa side target blow-out temperatures (tao(dr), tao(pa)) relative to each air-conditioning zones by using a neural network. air-mixing doors separately adjusts the temperatures of conditioned air blown out from dr side air passage and pa side air passage to be the first and second target blow-out temperatures. here, the neural network has the learning function, which adjusts its output to be desired data (teacher signal). therefore, the output at a specific input condition can be adjusted without temperature interference between each zone.",2001-04-24,"The title of the patent is air-conditioning device and its abstract is temperatures in a dr side air-conditioning zone and a pa side air-conditioning zone are controlled highly independently of each other without temperature interference between each zone. a room internal air temperature sensor and a room external air temperature sensor are provided. dr side and pa side temperature setters separately set room setpoint temperatures (tset(dr), tset(pa)) in each zone. first and second target blow-out temperature calculating portions, which include neural network, input the room setpoint temperatures and the temperature data. then it calculates dr side and pa side target blow-out temperatures (tao(dr), tao(pa)) relative to each air-conditioning zones by using a neural network. air-mixing doors separately adjusts the temperatures of conditioned air blown out from dr side air passage and pa side air passage to be the first and second target blow-out temperatures. here, the neural network has the learning function, which adjusts its output to be desired data (teacher signal). therefore, the output at a specific input condition can be adjusted without temperature interference between each zone. dated 2001-04-24"
6226544,living body internal active source estimation apparatus,obtaining fast dipole size estimation with less influence of noise by using a regional dipole model. an artificial neural network section 30 executes regional dipole size estimation using a neural network having coupling coefficients representing coupling states among as plurality of units and thresholds thereof.,2001-05-01,The title of the patent is living body internal active source estimation apparatus and its abstract is obtaining fast dipole size estimation with less influence of noise by using a regional dipole model. an artificial neural network section 30 executes regional dipole size estimation using a neural network having coupling coefficients representing coupling states among as plurality of units and thresholds thereof. dated 2001-05-01
6227842,automatically optimized combustion control,"systems and methods are disclosed that optimize the combustion process in various reactors, furnaces, and internal combustion engines. video cameras are used to evaluate the combustion flame grade. depending on the desired form, standard or special video devices, or beam scanning devices, are used to image the combustion flame and by-products. the video device generates and outputs image signals during various phases of, and at various locations in, the combustion process. other forms of sensors monitor and generate data signals defining selected parameters of the combustion process, such as air flow, fuel flow, turbulence, exhaust and inlet valve openings, etc. in a preferred form, a neural networks initially processes the image data and characterizes the combustion flame. a fuzzy logic controller and associated fuzzy logic rule base analyzes the image data from the neural network, along with other sensor information. the fuzzy logic controller determines and generates control signals defining adjustments necessary to optimize the combustion process.",2001-05-08,"The title of the patent is automatically optimized combustion control and its abstract is systems and methods are disclosed that optimize the combustion process in various reactors, furnaces, and internal combustion engines. video cameras are used to evaluate the combustion flame grade. depending on the desired form, standard or special video devices, or beam scanning devices, are used to image the combustion flame and by-products. the video device generates and outputs image signals during various phases of, and at various locations in, the combustion process. other forms of sensors monitor and generate data signals defining selected parameters of the combustion process, such as air flow, fuel flow, turbulence, exhaust and inlet valve openings, etc. in a preferred form, a neural networks initially processes the image data and characterizes the combustion flame. a fuzzy logic controller and associated fuzzy logic rule base analyzes the image data from the neural network, along with other sensor information. the fuzzy logic controller determines and generates control signals defining adjustments necessary to optimize the combustion process. dated 2001-05-08"
6233230,neural network is-95 rate determination,"an apparatus and the method of its operation for decoding a received, encoded data signal having a plurality of data rates, such as a cdma signal, by means of a deinterleaver for deinterleaving the recorded, encoded signal and outputting a frame of deinterleaved symbols, a feature calculation circuit for applying a plurality of different algorithms to the frame of deinterleaved symbols to produce a corresponding plurality of output feature values which are indicative of a degree of repetition of the deinterleaved symbols, a neural network for processing the plurality of output feature values according to a predetermined set of weights to produce a plurality of output rate determination values y.sub.1, y.sub.2, . . . y.sub.n, each of which corresponds to a different data rate, where m.ltoreq.y.ltoreq.m, n is an integer and m and m are predetermined minimum and maximum values, respectively, a rate detection circuit for comparing the plurality of output rate determination values y.sub.1, y.sub.2, . . . y.sub.n and selecting a rate corresponding to the largest output rate determination value y.sub.1, y.sub.2, . . . y.sub.n, and a decoder for decoding the encoded data signal at the data rate selected by the rate detection circuit.",2001-05-15,"The title of the patent is neural network is-95 rate determination and its abstract is an apparatus and the method of its operation for decoding a received, encoded data signal having a plurality of data rates, such as a cdma signal, by means of a deinterleaver for deinterleaving the recorded, encoded signal and outputting a frame of deinterleaved symbols, a feature calculation circuit for applying a plurality of different algorithms to the frame of deinterleaved symbols to produce a corresponding plurality of output feature values which are indicative of a degree of repetition of the deinterleaved symbols, a neural network for processing the plurality of output feature values according to a predetermined set of weights to produce a plurality of output rate determination values y.sub.1, y.sub.2, . . . y.sub.n, each of which corresponds to a different data rate, where m.ltoreq.y.ltoreq.m, n is an integer and m and m are predetermined minimum and maximum values, respectively, a rate detection circuit for comparing the plurality of output rate determination values y.sub.1, y.sub.2, . . . y.sub.n and selecting a rate corresponding to the largest output rate determination value y.sub.1, y.sub.2, . . . y.sub.n, and a decoder for decoding the encoded data signal at the data rate selected by the rate detection circuit. dated 2001-05-15"
6233365,image-processing method,"an image-processing method is provided with the first step of dividing an input image having n(n>1) gray scales into a plurality of matrixes, the second step of carrying out at least either a resolution-converting process or a variable magnification process for each of the divided matrixes, by using a hierarchical neural network that can execute a learning process for each input image, and the third step of outputting the image processed in the second step as an output image having n gray scales. thus, the weights adjustment of the network can be carried out on each input image whatever image is inputted thereto; therefore, it is possible to always provide an optimal converting process.",2001-05-15,"The title of the patent is image-processing method and its abstract is an image-processing method is provided with the first step of dividing an input image having n(n>1) gray scales into a plurality of matrixes, the second step of carrying out at least either a resolution-converting process or a variable magnification process for each of the divided matrixes, by using a hierarchical neural network that can execute a learning process for each input image, and the third step of outputting the image processed in the second step as an output image having n gray scales. thus, the weights adjustment of the network can be carried out on each input image whatever image is inputted thereto; therefore, it is possible to always provide an optimal converting process. dated 2001-05-15"
6236749,image recognition method,"the method provides object recognition procedure and a neural network by using the discrete-cosine transform (dct) (4) and histogram adaptive quantization (5). the method employs the dct transform with the added advantage of having a computationally-efficient and data-independent matrix as an alternative to the karhunen-loeve transform or principal component analysis which requires data-independent eigenvectors as a priori information. since the set of learning samples (1) may be small, we employ a mixture model of prior distributions for accurate estimation of local distribution of feature patterns obtained from several two dimensional images. the model selection method based on the mixture classes is presented to optimize the mixture number and local metric parameters. this method also provides image synthesis to generate a set of image databases to be used for training a neural network.",2001-05-22,"The title of the patent is image recognition method and its abstract is the method provides object recognition procedure and a neural network by using the discrete-cosine transform (dct) (4) and histogram adaptive quantization (5). the method employs the dct transform with the added advantage of having a computationally-efficient and data-independent matrix as an alternative to the karhunen-loeve transform or principal component analysis which requires data-independent eigenvectors as a priori information. since the set of learning samples (1) may be small, we employ a mixture model of prior distributions for accurate estimation of local distribution of feature patterns obtained from several two dimensional images. the model selection method based on the mixture classes is presented to optimize the mixture number and local metric parameters. this method also provides image synthesis to generate a set of image databases to be used for training a neural network. dated 2001-05-22"
6236908,virtual vehicle sensors based on neural networks trained using data generated by simulation models,"a virtual vehicle sensor includes a neural network which produces a sensor output based on a linear combination of non-linear physical signals generated by conventional physical sensors. instead of determining an output directly, the neural network determines the polynomial coefficients as functions of the physical signals indicative of other engine operating parameters. the sensor is manufactured using relatively limited data collection to calibrate a simulation model. the output of the simulation model is used for model-based mapping to generate more comprehensive maps used for training the neural network. the trained neural network is embedded in a controller and acts as the virtual sensor to monitor engine parameters which are difficult to measure or for which conventional physical sensors do not currently exist. the virtual sensor may be used to sense parameters such as in-cylinder residual mass fraction, emission levels, in-cylinder pressure rise during combustion, and exhaust gas temperature.",2001-05-22,"The title of the patent is virtual vehicle sensors based on neural networks trained using data generated by simulation models and its abstract is a virtual vehicle sensor includes a neural network which produces a sensor output based on a linear combination of non-linear physical signals generated by conventional physical sensors. instead of determining an output directly, the neural network determines the polynomial coefficients as functions of the physical signals indicative of other engine operating parameters. the sensor is manufactured using relatively limited data collection to calibrate a simulation model. the output of the simulation model is used for model-based mapping to generate more comprehensive maps used for training the neural network. the trained neural network is embedded in a controller and acts as the virtual sensor to monitor engine parameters which are difficult to measure or for which conventional physical sensors do not currently exist. the virtual sensor may be used to sense parameters such as in-cylinder residual mass fraction, emission levels, in-cylinder pressure rise during combustion, and exhaust gas temperature. dated 2001-05-22"
6236942,"system and method for delineating spatially dependent objects, such as hydrocarbon accumulations from seismic data","a neural network based system, method, and process for the automated delineation of spatially dependent objects is disclosed. the method is applicable to objects such as hydrocarbon accumulations, aeromagnetic profiles, astronomical clusters, weather clusters, objects from radar, sonar, seismic and infrared returns, etc. one of the novelties in the present invention is that the method can be utilized whether or not known data is available to provide traditional training sets. the output consists of a classification of the input data into clearly delineated accumulations, clusters, objects, etc. that have various types and properties. a preferred but non-exclusive application of the present invention is the automated delineation of hydrocarbon accumulations and sub-regions within the accumulations with various properties, in an oil and gas field, prior to the commencement of drilling operations.",2001-05-22,"The title of the patent is system and method for delineating spatially dependent objects, such as hydrocarbon accumulations from seismic data and its abstract is a neural network based system, method, and process for the automated delineation of spatially dependent objects is disclosed. the method is applicable to objects such as hydrocarbon accumulations, aeromagnetic profiles, astronomical clusters, weather clusters, objects from radar, sonar, seismic and infrared returns, etc. one of the novelties in the present invention is that the method can be utilized whether or not known data is available to provide traditional training sets. the output consists of a classification of the input data into clearly delineated accumulations, clusters, objects, etc. that have various types and properties. a preferred but non-exclusive application of the present invention is the automated delineation of hydrocarbon accumulations and sub-regions within the accumulations with various properties, in an oil and gas field, prior to the commencement of drilling operations. dated 2001-05-22"
6236965,method for automatically generating pronunciation dictionary in speech recognition system,"a method for automatically generating a pronunciation dictionary in a speech recognition system is disclosed. pronunciation patterns of a large scale pronunciation dictionary are learned through a neural network without resorting to a phonetic knowledge, and the pronunciation sequences for input words are accurately formed by utilizing an exception grapheme pronunciation dictionary, an exception word pronunciation dictionary for graphemes and words prohibiting the formation of an accurate pronunciation dictionary through the learning neural network, thereby reducing the size of the memory and the amount of calculations. a multi-layer perceptron for directly mapping phonemes relevant to respective graphemes is taught by utilizing a neural network, so as to form an exception word pronunciation dictionary data base, an exception grapheme pronunciation dictionary data base, and a phoneme output multi-layer perceptron parameter data base for respective graphemes. the exception word pronunciation dictionary data base, the exception grapheme pronunciation dictionary data base, and the phoneme output multi-layer perceptron parameter data base for the input word thus pre-processed is inspected, to post-process the pronunciation sequences for the relevant word.",2001-05-22,"The title of the patent is method for automatically generating pronunciation dictionary in speech recognition system and its abstract is a method for automatically generating a pronunciation dictionary in a speech recognition system is disclosed. pronunciation patterns of a large scale pronunciation dictionary are learned through a neural network without resorting to a phonetic knowledge, and the pronunciation sequences for input words are accurately formed by utilizing an exception grapheme pronunciation dictionary, an exception word pronunciation dictionary for graphemes and words prohibiting the formation of an accurate pronunciation dictionary through the learning neural network, thereby reducing the size of the memory and the amount of calculations. a multi-layer perceptron for directly mapping phonemes relevant to respective graphemes is taught by utilizing a neural network, so as to form an exception word pronunciation dictionary data base, an exception grapheme pronunciation dictionary data base, and a phoneme output multi-layer perceptron parameter data base for respective graphemes. the exception word pronunciation dictionary data base, the exception grapheme pronunciation dictionary data base, and the phoneme output multi-layer perceptron parameter data base for the input word thus pre-processed is inspected, to post-process the pronunciation sequences for the relevant word. dated 2001-05-22"
6240291,method for handoff in wireless communication systems using pattern recognition,"a handoff technique for wireless communication systems uses pattern recognition of signal strength data to anticipate handoffs and reduce the total number of handoffs in the system. a criterion for system performance is used in determining the necessity for handoff. a window of signal samples from nearby base stations constitutes a pattern vector which is classified using a probabilistic neural network or other learning machine. the use of averaged signals and the sequencing of classes allow for a small number of training vectors for the pattern classifier. substantially increased performance requires only one training vector per class. simulation results indicate that, for a given probability of failure, the pattern recognition based handoff technique yields fewer handoffs than the conventional hysteresis rule.",2001-05-29,"The title of the patent is method for handoff in wireless communication systems using pattern recognition and its abstract is a handoff technique for wireless communication systems uses pattern recognition of signal strength data to anticipate handoffs and reduce the total number of handoffs in the system. a criterion for system performance is used in determining the necessity for handoff. a window of signal samples from nearby base stations constitutes a pattern vector which is classified using a probabilistic neural network or other learning machine. the use of averaged signals and the sequencing of classes allow for a small number of training vectors for the pattern classifier. substantially increased performance requires only one training vector per class. simulation results indicate that, for a given probability of failure, the pattern recognition based handoff technique yields fewer handoffs than the conventional hysteresis rule. dated 2001-05-29"
6240343,apparatus and method for diagnosing an engine using computer based models in combination with a neural network,"an apparatus and method for diagnosing an engine using computer based models in combination with a neural network which reduces the neural network training and updating required is disclosed. the method and apparatus determine modeled values for a plurality of engine parameters as a function of sensed values in a plurality of initial values; sense a plurality of actual values for the engine parameters; and determine a difference between the actual values and the modeled values for the respective engine parameters. the differences are inputted into a neural network which generates an output pattern as a function of the differences and a plurality of weight values. the weight values and the initial values are then updated as a function of a comparison between the output patterns and a desired pattern, and the engine is responsively diagnosed as a function of the output pattern.",2001-05-29,"The title of the patent is apparatus and method for diagnosing an engine using computer based models in combination with a neural network and its abstract is an apparatus and method for diagnosing an engine using computer based models in combination with a neural network which reduces the neural network training and updating required is disclosed. the method and apparatus determine modeled values for a plurality of engine parameters as a function of sensed values in a plurality of initial values; sense a plurality of actual values for the engine parameters; and determine a difference between the actual values and the modeled values for the respective engine parameters. the differences are inputted into a neural network which generates an output pattern as a function of the differences and a plurality of weight values. the weight values and the initial values are then updated as a function of a comparison between the output patterns and a desired pattern, and the engine is responsively diagnosed as a function of the output pattern. dated 2001-05-29"
6243489,method for a neural network for representing imaging functions,"in a method for a self-organizing neural network for representing multidimensional, nonlinear imaging functions onto simpler imaging functions use divider-membranes are employed for achieving an error free representation of the imaging function via the learning sample, allowing for a high level of generalization. kohonen cell borders coincide with a required imaging function. the neural network can independently determine a number of neurons necessary for an error-free solution of a problem. a readout of the neural network can occur through the calculation of the minimum of the squares of the distances.",2001-06-05,"The title of the patent is method for a neural network for representing imaging functions and its abstract is in a method for a self-organizing neural network for representing multidimensional, nonlinear imaging functions onto simpler imaging functions use divider-membranes are employed for achieving an error free representation of the imaging function via the learning sample, allowing for a high level of generalization. kohonen cell borders coincide with a required imaging function. the neural network can independently determine a number of neurons necessary for an error-free solution of a problem. a readout of the neural network can occur through the calculation of the minimum of the squares of the distances. dated 2001-06-05"
6243490,data processing using neural networks having conversion tables in an intermediate layer,"in a neural network which includes one input layer, one or more intermediate layers and one output layer, neural elements in the input layer and neural elements in the intermediate layer are divided into groups. arithmetic operations representing the coupling between the neural elements of the input layer and the neural elements of the intermediate layer are put into table form.",2001-06-05,"The title of the patent is data processing using neural networks having conversion tables in an intermediate layer and its abstract is in a neural network which includes one input layer, one or more intermediate layers and one output layer, neural elements in the input layer and neural elements in the intermediate layer are divided into groups. arithmetic operations representing the coupling between the neural elements of the input layer and the neural elements of the intermediate layer are put into table form. dated 2001-06-05"
6246962,method and apparatus for adaptively filtering noise to detect downhole events,"adaptive neural networks can be used to effectively enhance signal detection in the inherently noisy environment of an oil well. the neural network can be either non-recurrent or recurrent in nature. the system is implemented with a computer that accepts input from at least one sensor mounted to the wellhead or near the wellhead. the detected contaminated signal is a combination of the event signal and the noise from the environment. the event signal can be, for example, the detonation of a perforation gun. the noise can be either random or periodic. the use of adaptive filtering allows the noise to be more precisely predicted and then subtracted from the contaminated signal to produce a cleaner representation of the event signal. once the predicted noise is subtracted, the remaining event signal can be analyzed using voice or sound recognition systems to produce an output describing what event occurred.",2001-06-12,"The title of the patent is method and apparatus for adaptively filtering noise to detect downhole events and its abstract is adaptive neural networks can be used to effectively enhance signal detection in the inherently noisy environment of an oil well. the neural network can be either non-recurrent or recurrent in nature. the system is implemented with a computer that accepts input from at least one sensor mounted to the wellhead or near the wellhead. the detected contaminated signal is a combination of the event signal and the noise from the environment. the event signal can be, for example, the detonation of a perforation gun. the noise can be either random or periodic. the use of adaptive filtering allows the noise to be more precisely predicted and then subtracted from the contaminated signal to produce a cleaner representation of the event signal. once the predicted noise is subtracted, the remaining event signal can be analyzed using voice or sound recognition systems to produce an output describing what event occurred. dated 2001-06-12"
6247001,method of training a neural network,"a state vector (sv.sub.t) is determined with elements that characterize a financial market (101). taking into account predetermined evaluation variables, an evaluation (v.sub.t) is determined (102) for the state vector (sv.sub.t). in addition, a chronologically following state vector (sv.sub.t+1) is determined (103) and evaluated (v.sub.t+1). on the basis of the two evaluations (v.sub.t, v.sub.t+1), weights (w.sub.i) of the neural network (nn) are adapted (104) using a reinforcement learning method (.delta.w.sub.i).",2001-06-12,"The title of the patent is method of training a neural network and its abstract is a state vector (sv.sub.t) is determined with elements that characterize a financial market (101). taking into account predetermined evaluation variables, an evaluation (v.sub.t) is determined (102) for the state vector (sv.sub.t). in addition, a chronologically following state vector (sv.sub.t+1) is determined (103) and evaluated (v.sub.t+1). on the basis of the two evaluations (v.sub.t, v.sub.t+1), weights (w.sub.i) of the neural network (nn) are adapted (104) using a reinforcement learning method (.delta.w.sub.i). dated 2001-06-12"
6247003,current transformer saturation correction using artificial neural networks,"a method and apparatus of correcting for saturation in a current transformer, which outputs a current measurement, is provided. a switching algorithm receives a value of the current measurement from the current transformer and determines within which of three ranges the value falls. if the value falls in a first range, the current measurement is provided to a protective device such as a relay. if the value falls in a second range, the current measurement is provided to an artificial neural network that produces an output that accounts for saturation of the current transformer. if the value falls in a third range, the current measurement is provided to another artificial neural network that produces an output that accounts for saturation of the current transformer.",2001-06-12,"The title of the patent is current transformer saturation correction using artificial neural networks and its abstract is a method and apparatus of correcting for saturation in a current transformer, which outputs a current measurement, is provided. a switching algorithm receives a value of the current measurement from the current transformer and determines within which of three ranges the value falls. if the value falls in a first range, the current measurement is provided to a protective device such as a relay. if the value falls in a second range, the current measurement is provided to an artificial neural network that produces an output that accounts for saturation of the current transformer. if the value falls in a third range, the current measurement is provided to another artificial neural network that produces an output that accounts for saturation of the current transformer. dated 2001-06-12"
6248063,computer assisted methods for diagnosing diseases,"the simultaneous multi access reasoning technology system of the present invention utilizes both existing knowledge and implicit information that can be numerically extracted from training data to provide a method and apparatus for diagnosing disease and treating a patient. this technology further comprises a system for receiving patient data from another location, analyzing the data in a trained neural network, producing a diagnostic value, and optionally transmitting the diagnostic value to another location.",2001-06-19,"The title of the patent is computer assisted methods for diagnosing diseases and its abstract is the simultaneous multi access reasoning technology system of the present invention utilizes both existing knowledge and implicit information that can be numerically extracted from training data to provide a method and apparatus for diagnosing disease and treating a patient. this technology further comprises a system for receiving patient data from another location, analyzing the data in a trained neural network, producing a diagnostic value, and optionally transmitting the diagnostic value to another location. dated 2001-06-19"
6249605,key character extraction and lexicon reduction for cursive text recognition,""" a method, apparatus, and article of manufacturing employing lexicon reduction using key characters and a neural network, for recognizing a line of cursive text. unambiguous parts of a cursive image, referred to as """"key characters,"""" are identified. if the level of confidence that a segment of a line of cursive text is a particular character is higher than a threshold, and is also sufficiently higher than the level of confidence of neighboring segments, then the character is designated as a key character candidate. key character candidates are then screened using geometric information. the key character candidates that pass the screening are designated key characters. two-stages of lexicon reduction are employed. the first stage of lexicon reduction uses a neural network to estimate a lower bound and an upper bound of the number of characters in a line of cursive text. lexicon entries having a total number of characters outside of the bounds are eliminated. for the second stage of lexicon reduction, the lexicon is further reduced by comparing character strings using the key characters, with lexicon entries. for each of the key characters in the character strings, it is determined whether there is a mismatch between the key character and characters in a corresponding search range in the lexicon entry. if the number of mismatches for all of the key characters in a search string is greater than (1+(the number of key characters in the search string/4)), then the lexicon entry is eliminated. accordingly, the invention advantageously accomplishes lexicon reduction, thereby decreasing the time required to recognize a line of cursive text, without reducing accuracy. """,2001-06-19,"The title of the patent is key character extraction and lexicon reduction for cursive text recognition and its abstract is "" a method, apparatus, and article of manufacturing employing lexicon reduction using key characters and a neural network, for recognizing a line of cursive text. unambiguous parts of a cursive image, referred to as """"key characters,"""" are identified. if the level of confidence that a segment of a line of cursive text is a particular character is higher than a threshold, and is also sufficiently higher than the level of confidence of neighboring segments, then the character is designated as a key character candidate. key character candidates are then screened using geometric information. the key character candidates that pass the screening are designated key characters. two-stages of lexicon reduction are employed. the first stage of lexicon reduction uses a neural network to estimate a lower bound and an upper bound of the number of characters in a line of cursive text. lexicon entries having a total number of characters outside of the bounds are eliminated. for the second stage of lexicon reduction, the lexicon is further reduced by comparing character strings using the key characters, with lexicon entries. for each of the key characters in the character strings, it is determined whether there is a mismatch between the key character and characters in a corresponding search range in the lexicon entry. if the number of mismatches for all of the key characters in a search string is greater than (1+(the number of key characters in the search string/4)), then the lexicon entry is eliminated. accordingly, the invention advantageously accomplishes lexicon reduction, thereby decreasing the time required to recognize a line of cursive text, without reducing accuracy. "" dated 2001-06-19"
6249606,method and system for gesture category recognition and training using a feature vector,"a computer implemented method and system for gesture category recognition and training. generally, a gesture is a hand or body initiated movement of a cursor directing device to outline a particular pattern in particular directions done in particular periods of time. the present invention allows a computer system to accept input data, originating from a user, in the form gesture data that are made using the cursor directing device. in one embodiment, a mouse device is used, but the present invention is equally well suited for use with other cursor directing devices (e.g., a track ball, a finger pad, an electronic stylus, etc.). in one embodiment, gesture data is accepted by pressing a key on the keyboard and then moving the mouse (with mouse button pressed) to trace out the gesture. mouse position information and time stamps are recorded. the present invention then determines a multi-dimensional feature vector based on the gesture data. the feature vector is then passed through a gesture category recognition engine that, in one implementation, uses a radial basis function neural network to associate the feature vector to a pre-existing gesture category. once identified, a set of user commands that are associated with the gesture category are applied to the computer system. the user commands can originate from an automatic process that extracts commands that are associated with the menu items of a particular application program. the present invention also allows user training so that user-defined gestures, and the computer commands associated therewith, can be programmed into the computer system.",2001-06-19,"The title of the patent is method and system for gesture category recognition and training using a feature vector and its abstract is a computer implemented method and system for gesture category recognition and training. generally, a gesture is a hand or body initiated movement of a cursor directing device to outline a particular pattern in particular directions done in particular periods of time. the present invention allows a computer system to accept input data, originating from a user, in the form gesture data that are made using the cursor directing device. in one embodiment, a mouse device is used, but the present invention is equally well suited for use with other cursor directing devices (e.g., a track ball, a finger pad, an electronic stylus, etc.). in one embodiment, gesture data is accepted by pressing a key on the keyboard and then moving the mouse (with mouse button pressed) to trace out the gesture. mouse position information and time stamps are recorded. the present invention then determines a multi-dimensional feature vector based on the gesture data. the feature vector is then passed through a gesture category recognition engine that, in one implementation, uses a radial basis function neural network to associate the feature vector to a pre-existing gesture category. once identified, a set of user commands that are associated with the gesture category are applied to the computer system. the user commands can originate from an automatic process that extracts commands that are associated with the menu items of a particular application program. the present invention also allows user training so that user-defined gestures, and the computer commands associated therewith, can be programmed into the computer system. dated 2001-06-19"
6253186,method and apparatus for detecting fraud,"a computerized arrangement for detecting potentially fraudulent suppliers or providers of goods or services includes a processor, a storage device, an input device for communicating data to the processor and storage device, and an output device for communicating data from the processor and storage device. the storage device includes a claims data file for storing information relating to a plurality of claims submitted for payment by a selected supplier or provider, one or more encoding lookup tables for use with the claims data file to produce an encoded claims data file, and a neural network program for analyzing the encoded data to produce an indicator of potentially fraudulent activity. the indicator may be compared to a predetermined threshold value by the apparatus or method to identify fraudulent suppliers. in addition to the neural network, at least one expert system may be used in the identification process.",2001-06-26,"The title of the patent is method and apparatus for detecting fraud and its abstract is a computerized arrangement for detecting potentially fraudulent suppliers or providers of goods or services includes a processor, a storage device, an input device for communicating data to the processor and storage device, and an output device for communicating data from the processor and storage device. the storage device includes a claims data file for storing information relating to a plurality of claims submitted for payment by a selected supplier or provider, one or more encoding lookup tables for use with the claims data file to produce an encoded claims data file, and a neural network program for analyzing the encoded data to produce an indicator of potentially fraudulent activity. the indicator may be compared to a predetermined threshold value by the apparatus or method to identify fraudulent suppliers. in addition to the neural network, at least one expert system may be used in the identification process. dated 2001-06-26"
6256619,self optimizing neural network analog data processing system,"a analog data neural network processing system is provided in which there is a self optimization capability that varies the signal processing factors in response to a detected contrast in the system output patterns. the system provides feedback type guidance in varying such processing factors as sampling rate, frame length, signal transformation, neural network vigilance and architecture, each in a direction that will maximize or minimize the contracts with patterns used to train the network. the processing system is useful in all signal classification tasks.",2001-07-03,"The title of the patent is self optimizing neural network analog data processing system and its abstract is a analog data neural network processing system is provided in which there is a self optimization capability that varies the signal processing factors in response to a detected contrast in the system output patterns. the system provides feedback type guidance in varying such processing factors as sampling rate, frame length, signal transformation, neural network vigilance and architecture, each in a direction that will maximize or minimize the contracts with patterns used to train the network. the processing system is useful in all signal classification tasks. dated 2001-07-03"
6259406,wireless communication system and method and system for detection of position of radio mobile station,"a method of detecting a position of a radio mobile station in radiocommunications, which is capable of accurately and simply finding the position of the mobile station. at a measuring point the mobile station measures the reception radio strength levels from a plurality of base stations and conveys the measurement results through the base station to a control station. the control station learns, through a neural network, the correlation between the reception radio strength levels and the position of the mobile station on the basis of the measurement results at a plurality of measuring points and the positions of the measuring points. subsequently, when the mobile station communicates to the control station the reception radio strength levels measured at an arbitrary point, the control station estimates the position of the mobile station, causing those measurement results, on the basis of the correlation obtained through the learning.",2001-07-10,"The title of the patent is wireless communication system and method and system for detection of position of radio mobile station and its abstract is a method of detecting a position of a radio mobile station in radiocommunications, which is capable of accurately and simply finding the position of the mobile station. at a measuring point the mobile station measures the reception radio strength levels from a plurality of base stations and conveys the measurement results through the base station to a control station. the control station learns, through a neural network, the correlation between the reception radio strength levels and the position of the mobile station on the basis of the measurement results at a plurality of measuring points and the positions of the measuring points. subsequently, when the mobile station communicates to the control station the reception radio strength levels measured at an arbitrary point, the control station estimates the position of the mobile station, causing those measurement results, on the basis of the correlation obtained through the learning. dated 2001-07-10"
6259812,key character extraction and lexicon reduction cursive text recognition,""" a method, apparatus, and article of manufacture employing lexicon reduction using key characters and a neural network, for recognizing a line of cursive text. unambiguous parts of a cursive image, referred to as """"key characters,"""" are identified. if the level of confidence that a segment of a line of cursive text is a particular character is higher than a threshold, and is also sufficiently higher than the level of confidence of neighboring segments, then the character is designated as a key character candidate. key character candidates are then screened using geometric information. the key character candidates that pass the screening are designated key characters. two-stages of lexicon reduction are employed. the first stage of lexicon reduction uses a neural network to estimate a lower bound and an upper bound of the number of characters in a line of cursive text. lexicon entries having a total number of characters outside of the bounds are eliminated. for the second stage of lexicon reduction, the lexicon is further reduced by comparing character strings using the key characters, with lexicon entries. for each of the key characters in the character strings, it is determined whether there is a mismatch between the key character and characters in a corresponding search range in the lexicon entry. if the number of mismatches for all of the key characters in a search string is greater than (1+(the number of key characters in the search string/4)), then the lexicon entry is eliminated. accordingly, the invention advantageously accomplishes lexicon reduction, thereby decreasing the time required to recognize a line of cursive text, without reducing accuracy. """,2001-07-10,"The title of the patent is key character extraction and lexicon reduction cursive text recognition and its abstract is "" a method, apparatus, and article of manufacture employing lexicon reduction using key characters and a neural network, for recognizing a line of cursive text. unambiguous parts of a cursive image, referred to as """"key characters,"""" are identified. if the level of confidence that a segment of a line of cursive text is a particular character is higher than a threshold, and is also sufficiently higher than the level of confidence of neighboring segments, then the character is designated as a key character candidate. key character candidates are then screened using geometric information. the key character candidates that pass the screening are designated key characters. two-stages of lexicon reduction are employed. the first stage of lexicon reduction uses a neural network to estimate a lower bound and an upper bound of the number of characters in a line of cursive text. lexicon entries having a total number of characters outside of the bounds are eliminated. for the second stage of lexicon reduction, the lexicon is further reduced by comparing character strings using the key characters, with lexicon entries. for each of the key characters in the character strings, it is determined whether there is a mismatch between the key character and characters in a corresponding search range in the lexicon entry. if the number of mismatches for all of the key characters in a search string is greater than (1+(the number of key characters in the search string/4)), then the lexicon entry is eliminated. accordingly, the invention advantageously accomplishes lexicon reduction, thereby decreasing the time required to recognize a line of cursive text, without reducing accuracy. "" dated 2001-07-10"
6259824,image processing apparatus utilizing a neural network to improve printed image quality,"an information processing apparatus including a switch for manually requesting change of output image quality, detecting unit for detecting a condition of the apparatus, setting unit of setting an image forming condition in accordance with the detected condition and the requested output image quality and control unit for changing the image forming condition by learning the request previously requested by the user. image forming conditions are adjusted so as to satisfy the user manually based upon the degree of satisfaction determined by the user.",2001-07-10,"The title of the patent is image processing apparatus utilizing a neural network to improve printed image quality and its abstract is an information processing apparatus including a switch for manually requesting change of output image quality, detecting unit for detecting a condition of the apparatus, setting unit of setting an image forming condition in accordance with the detected condition and the requested output image quality and control unit for changing the image forming condition by learning the request previously requested by the user. image forming conditions are adjusted so as to satisfy the user manually based upon the degree of satisfaction determined by the user. dated 2001-07-10"
6260034,method and a system for nucleic acid sequence analysis,"a method and a system for identifying mutations within nucleic acid sequences. using raw data signals from conventional nucleic acid sequencing equipment, a method to create input signals that enables a properly trained neural network to output a mutation/no mutation signal is provided. further, an instrument system to perform the method is provided.",2001-07-10,"The title of the patent is method and a system for nucleic acid sequence analysis and its abstract is a method and a system for identifying mutations within nucleic acid sequences. using raw data signals from conventional nucleic acid sequencing equipment, a method to create input signals that enables a properly trained neural network to output a mutation/no mutation signal is provided. further, an instrument system to perform the method is provided. dated 2001-07-10"
6263238,automatic external defibrillator having a ventricular fibrillation detector,"in an automatic external defibrillator (aed) having a ventricular fibrillation detector, the ventricular fibrillation detector may generally be defined as a filter containing both an adaptive non-linear section and a linear section. the non-linear section is preferably a complex-domain neural network that can be trained to differentiate between various rhythm patterns and produce linear data for input to the linear section. the linear section is preferably an ongoing, continuous operation based on a sliding window of a predetermined time period, e.g., a tapped time-delay filter. in combination the non-linear section and linear section of the filter operate to detect and extract artifacts from a patient's ecg signal in a substantially accurate fashion so that the determination to deliver a defibrillation pulse may be accurately made.",2001-07-17,"The title of the patent is automatic external defibrillator having a ventricular fibrillation detector and its abstract is in an automatic external defibrillator (aed) having a ventricular fibrillation detector, the ventricular fibrillation detector may generally be defined as a filter containing both an adaptive non-linear section and a linear section. the non-linear section is preferably a complex-domain neural network that can be trained to differentiate between various rhythm patterns and produce linear data for input to the linear section. the linear section is preferably an ongoing, continuous operation based on a sliding window of a predetermined time period, e.g., a tapped time-delay filter. in combination the non-linear section and linear section of the filter operate to detect and extract artifacts from a patient's ecg signal in a substantially accurate fashion so that the determination to deliver a defibrillation pulse may be accurately made. dated 2001-07-17"
6263257,method and device for determining relevant variables in the processing of textile structures,"a device for the simulation of spinning machines with a view to their optimum economic use and operation. starting from material properties of a preproduct (1) determined by measurement, as well as from configuration and setting parameters (3) of such a spinning machine, material properties (2) of the intermediate or output products which are being produced, which properties can be determined by measurement, are predicted using a process model describing the behavior of the spinning machine, by a simulation device. the process model on which a method being based is presented by a neural network. the coefficients determining the actual behavior of this neural network area calculated from a set of sample data in a training phase. this sample data and/or from otherwise conclusively predicted properties of the behavior of such a spinning machine.",2001-07-17,"The title of the patent is method and device for determining relevant variables in the processing of textile structures and its abstract is a device for the simulation of spinning machines with a view to their optimum economic use and operation. starting from material properties of a preproduct (1) determined by measurement, as well as from configuration and setting parameters (3) of such a spinning machine, material properties (2) of the intermediate or output products which are being produced, which properties can be determined by measurement, are predicted using a process model describing the behavior of the spinning machine, by a simulation device. the process model on which a method being based is presented by a neural network. the coefficients determining the actual behavior of this neural network area calculated from a set of sample data in a training phase. this sample data and/or from otherwise conclusively predicted properties of the behavior of such a spinning machine. dated 2001-07-17"
6266664,"method for scanning, analyzing and rating digital information content",""" computer-implemented methods are described for, first, characterizing a specific category of information content--pornography, for example--and then accurately identifying instances of that category of content within a real-time media stream, such as a web page, e-mail or other digital dataset. this content-recognition technology enables a new class of highly scalable applications to manage such content, including filtering, classifying, prioritizing, tracking, etc. an illustrative application of the invention is a software product for use in conjunction with web-browser client software for screening access to web pages that contain pornography or other potentially harmful or offensive content. a target attribute set of regular expression, such as natural language words and/or phrases, is formed by statistical analysis of a number of samples of datasets characterized as """"containing,"""" and another set of samples characterized as """"not containing,"""" the selected category of information content. this list of expressions is refined by applying correlation analysis to the samples or """"training data."""" neural-network feed-forward techniques are then applied, again using a substantial training dataset, for adaptively assigning relative weights to each of the expressions in the target attribute set, thereby forming an awaited list that is highly predictive of the information content category of interest. """,2001-07-24,"The title of the patent is method for scanning, analyzing and rating digital information content and its abstract is "" computer-implemented methods are described for, first, characterizing a specific category of information content--pornography, for example--and then accurately identifying instances of that category of content within a real-time media stream, such as a web page, e-mail or other digital dataset. this content-recognition technology enables a new class of highly scalable applications to manage such content, including filtering, classifying, prioritizing, tracking, etc. an illustrative application of the invention is a software product for use in conjunction with web-browser client software for screening access to web pages that contain pornography or other potentially harmful or offensive content. a target attribute set of regular expression, such as natural language words and/or phrases, is formed by statistical analysis of a number of samples of datasets characterized as """"containing,"""" and another set of samples characterized as """"not containing,"""" the selected category of information content. this list of expressions is refined by applying correlation analysis to the samples or """"training data."""" neural-network feed-forward techniques are then applied, again using a substantial training dataset, for adaptively assigning relative weights to each of the expressions in the target attribute set, thereby forming an awaited list that is highly predictive of the information content category of interest. "" dated 2001-07-24"
6269351,method and system for training an artificial neural network,""" a method and system for training an artificial neural network (""""ann"""") are disclosed. one embodiment of the method of the present invention initializes an artificial neural network by assigning values to one or more weights. an adaptive learning rate is set to an initial starting value and training patterns for an input layer and an output layer are stored. the input layer training pattern is processed in the ann to obtain an output pattern. an error is calculated between the output layer training pattern and the output pattern and used to calculate an error ratio, which is used to adjust the value of the adaptive learning rate. if the error ratio is less than a threshold value, the adaptive learning rate can be multiplied by a step-up factor to increase the learning rate. if the error ratio is greater than the threshold value, the adaptive learning rate can be multiplied by a step-down factor to reduce the learning rate. the value of the weights used to initialize the ann are adjusted based on the calculated error and the adaptive learning rate. the training method of the present invention is repeated until ann achieves a final trained state. """,2001-07-31,"The title of the patent is method and system for training an artificial neural network and its abstract is "" a method and system for training an artificial neural network (""""ann"""") are disclosed. one embodiment of the method of the present invention initializes an artificial neural network by assigning values to one or more weights. an adaptive learning rate is set to an initial starting value and training patterns for an input layer and an output layer are stored. the input layer training pattern is processed in the ann to obtain an output pattern. an error is calculated between the output layer training pattern and the output pattern and used to calculate an error ratio, which is used to adjust the value of the adaptive learning rate. if the error ratio is less than a threshold value, the adaptive learning rate can be multiplied by a step-up factor to increase the learning rate. if the error ratio is greater than the threshold value, the adaptive learning rate can be multiplied by a step-down factor to reduce the learning rate. the value of the weights used to initialize the ann are adjusted based on the calculated error and the adaptive learning rate. the training method of the present invention is repeated until ann achieves a final trained state. "" dated 2001-07-31"
6269352,"low-voltage, very-low-power conductance mode neuron","a neural network including a number of synaptic weighting elements, and a neuron stage; each of the synaptic weighting elements having a respective synaptic input connection supplied with a respective input signal; and the neuron stage having inputs connected to the synaptic weighting elements, and being connected to an output of the neural network supplying a digital output signal. the accumulated weighted inputs are represented as conductances, and a conductance-mode neuron is used to apply nonlinearity and produce an output. the synaptic weighting elements are formed by memory cells programmable to different threshold voltage levels, so that each presents a respective programmable conductance; and the neuron stage provides for measuring conductance on the basis of the current through the memory cells, and for generating a binary output signal on the basis of the total conductance of the synaptic elements.",2001-07-31,"The title of the patent is low-voltage, very-low-power conductance mode neuron and its abstract is a neural network including a number of synaptic weighting elements, and a neuron stage; each of the synaptic weighting elements having a respective synaptic input connection supplied with a respective input signal; and the neuron stage having inputs connected to the synaptic weighting elements, and being connected to an output of the neural network supplying a digital output signal. the accumulated weighted inputs are represented as conductances, and a conductance-mode neuron is used to apply nonlinearity and produce an output. the synaptic weighting elements are formed by memory cells programmable to different threshold voltage levels, so that each presents a respective programmable conductance; and the neuron stage provides for measuring conductance on the basis of the current through the memory cells, and for generating a binary output signal on the basis of the total conductance of the synaptic elements. dated 2001-07-31"
6272261,image processing device,"the present invention provides an image processing device capable of performing high-resolution conversion, enlargement processing, etc. which, even with images including both text images and photographic images, satisfies both sharpness of text areas and smoothness of photograph areas. the image processing device according to the present invention performs frequency conversion, using a frequency conversion device, of partial images which are the objects of processing, resulting in a matrix of frequency-converted coefficients; divides into a plurality of domains, using a plurality of patterns, the matrix of frequency-converted coefficients; and quantifies each such domain by finding a mean value of the coefficients of that domain as a feature quantity for that domain. the mean coefficient value of each domain is inputted into a conversion filter selecting device, where a hierarchical neural network calculates and outputs the suitability of each conversion filter for the partial image in question, and the conversion filter selecting device selects the conversion filter with the highest suitability. then an interpolation processing device performs interpolation processing of each partial image using the conversion filter most suited thereto.",2001-08-07,"The title of the patent is image processing device and its abstract is the present invention provides an image processing device capable of performing high-resolution conversion, enlargement processing, etc. which, even with images including both text images and photographic images, satisfies both sharpness of text areas and smoothness of photograph areas. the image processing device according to the present invention performs frequency conversion, using a frequency conversion device, of partial images which are the objects of processing, resulting in a matrix of frequency-converted coefficients; divides into a plurality of domains, using a plurality of patterns, the matrix of frequency-converted coefficients; and quantifies each such domain by finding a mean value of the coefficients of that domain as a feature quantity for that domain. the mean coefficient value of each domain is inputted into a conversion filter selecting device, where a hierarchical neural network calculates and outputs the suitability of each conversion filter for the partial image in question, and the conversion filter selecting device selects the conversion filter with the highest suitability. then an interpolation processing device performs interpolation processing of each partial image using the conversion filter most suited thereto. dated 2001-08-07"
6272392,methodology for extracting effective lens aberrations using a neural network,a method (250) of extracting effective imaging system aberrations from test data collected from test structures (220) constructed from a lithography system having an imaging system associated therewith includes inputting (264) experimental critical dimension data corresponding to fabricated features (220) on a substrate (212) to a neural network (208). the method (250) also includes inputting (266) nominal critical dimension data corresponding to the fabricated features on the substrate (212) to the neural network (208) and determining (268) the effective aberrations of the imaging system associated with the lithography system used to fabricate the features (220) using the neural network (208).,2001-08-07,The title of the patent is methodology for extracting effective lens aberrations using a neural network and its abstract is a method (250) of extracting effective imaging system aberrations from test data collected from test structures (220) constructed from a lithography system having an imaging system associated therewith includes inputting (264) experimental critical dimension data corresponding to fabricated features (220) on a substrate (212) to a neural network (208). the method (250) also includes inputting (266) nominal critical dimension data corresponding to the fabricated features on the substrate (212) to the neural network (208) and determining (268) the effective aberrations of the imaging system associated with the lithography system used to fabricate the features (220) using the neural network (208). dated 2001-08-07
6272426,predicting cylinder pressure for on-vehicle control,"an apparatus for predicting cylinder pressure for on-vehicle control of an internal combustion engine includes a piston sensor, a cylinder pressure sensor, and a controller. the piston sensor is coupled to a piston located in the engine. the piston sensor detects the piston position and generates a piston position signal. the cylinder pressure sensor is coupled to a cylinder located in the engine. the cylinder pressure sensor detects the cylinder pressure and generates a cylinder pressure signal. the controller receives both the piston position signal and the cylinder pressure signal. a neural network, located in the controller, uses this data to predict an undesirable cylinder pressure during a future combustion event. the controller then modifies the future combustion event in response to the predicted undesirable cylinder pressure.",2001-08-07,"The title of the patent is predicting cylinder pressure for on-vehicle control and its abstract is an apparatus for predicting cylinder pressure for on-vehicle control of an internal combustion engine includes a piston sensor, a cylinder pressure sensor, and a controller. the piston sensor is coupled to a piston located in the engine. the piston sensor detects the piston position and generates a piston position signal. the cylinder pressure sensor is coupled to a cylinder located in the engine. the cylinder pressure sensor detects the cylinder pressure and generates a cylinder pressure signal. the controller receives both the piston position signal and the cylinder pressure signal. a neural network, located in the controller, uses this data to predict an undesirable cylinder pressure during a future combustion event. the controller then modifies the future combustion event in response to the predicted undesirable cylinder pressure. dated 2001-08-07"
6272480,method and arrangement for the neural modelling of a dynamic system with non-linear stochastic behavior,"in a method and arrangement for the neural modelling of a dynamic system with non-linear stochastic behavior wherein only a few measured values of the influencing variable are available and the remaining values of the time series are modelled, a combination of a non-linear computerized recurrent neural predictive network and a linear error model are employed to produce a prediction with the application of maximum likelihood adaption rules. the computerized recurrent neural network can be trained with the assistance of the real-time recurrent learning rule, and the linear error model is trained with the assistance of the error model adaption rule that is implemented on the basis of forward-backward kalman equations. this model is utilized in order to predict values of the glucose-insulin metabolism of a diabetes patient.",2001-08-07,"The title of the patent is method and arrangement for the neural modelling of a dynamic system with non-linear stochastic behavior and its abstract is in a method and arrangement for the neural modelling of a dynamic system with non-linear stochastic behavior wherein only a few measured values of the influencing variable are available and the remaining values of the time series are modelled, a combination of a non-linear computerized recurrent neural predictive network and a linear error model are employed to produce a prediction with the application of maximum likelihood adaption rules. the computerized recurrent neural network can be trained with the assistance of the real-time recurrent learning rule, and the linear error model is trained with the assistance of the error model adaption rule that is implemented on the basis of forward-backward kalman equations. this model is utilized in order to predict values of the glucose-insulin metabolism of a diabetes patient. dated 2001-08-07"
6275190,wireless communication system and method and system for detection of position of radio mobile station,"a method of detecting a position of a radio mobile station in radiocommunications, which is capable of accurately and simply finding the position of the mobile station. at a measuring point the mobile station measures the reception radio strength levels from a plurality of base stations and conveys the measurement results through the base station to a control station. the control station learns, through a neural network, the correlation between the reception radio strength levels and the position of the mobile station on the basis of the measurement results at a plurality of measuring points and the positions of the measuring points. subsequently, when the mobile station communicates to the control station the reception radio strength levels measured at an arbitrary point, the control station estimates the position of the mobile station, causing those measurement results, on the basis of the correlation obtained through the learning.",2001-08-14,"The title of the patent is wireless communication system and method and system for detection of position of radio mobile station and its abstract is a method of detecting a position of a radio mobile station in radiocommunications, which is capable of accurately and simply finding the position of the mobile station. at a measuring point the mobile station measures the reception radio strength levels from a plurality of base stations and conveys the measurement results through the base station to a control station. the control station learns, through a neural network, the correlation between the reception radio strength levels and the position of the mobile station on the basis of the measurement results at a plurality of measuring points and the positions of the measuring points. subsequently, when the mobile station communicates to the control station the reception radio strength levels measured at an arbitrary point, the control station estimates the position of the mobile station, causing those measurement results, on the basis of the correlation obtained through the learning. dated 2001-08-14"
6275761,neural network-based virtual sensor for automatic transmission slip,"a method is disclosed for computer-based control of the timing and level of gear shifts in a multi-speed automatic transmission operated in combination with an internal combustion engine and interposed fluid torque converter. the computer containing power control module signals gear shifts in response to its repeated cyclic processing of engine and transmission operation parameters including torque converter slippage. here, such slippage is estimated using a neural network with suitable such parameters as input data. in preferred modes of operation, different neural networks are available for selection and use by the computer in different modes of engine-transmission operation.",2001-08-14,"The title of the patent is neural network-based virtual sensor for automatic transmission slip and its abstract is a method is disclosed for computer-based control of the timing and level of gear shifts in a multi-speed automatic transmission operated in combination with an internal combustion engine and interposed fluid torque converter. the computer containing power control module signals gear shifts in response to its repeated cyclic processing of engine and transmission operation parameters including torque converter slippage. here, such slippage is estimated using a neural network with suitable such parameters as input data. in preferred modes of operation, different neural networks are available for selection and use by the computer in different modes of engine-transmission operation. dated 2001-08-14"
6277080,method and apparatus for measuring exertion endurance,"the present invention relates to a method and an apparatus for measuring exercise condition, especially a method for measuring an exertion endurance indicator representing exercise condition of a subject to be measured, such as maximal oxygen uptake or any such exertion endurance indicator representing exercise condition. the method is characterized in that, in the method a predetermined calculation formula is used preferably by means of a neural network, to which formula physiological parameters representing the subject to be measured are supplied. the input parameters comprise at least one or more of the following physiological parameters, such as sex, age, height, weight. one or more output parameters representing the exertion endurance indicator representing the exercise condition of the subject to be measured are obtained as a result from the calculation formula. in addition to the physiological parameters, one or more resting heartbeat parameters measured specifically from resting heartbeat are used as input parameters of the calculation formula. in the preferred embodiment of the invention, the calculation formula is formed by means of a neural network construction.",2001-08-21,"The title of the patent is method and apparatus for measuring exertion endurance and its abstract is the present invention relates to a method and an apparatus for measuring exercise condition, especially a method for measuring an exertion endurance indicator representing exercise condition of a subject to be measured, such as maximal oxygen uptake or any such exertion endurance indicator representing exercise condition. the method is characterized in that, in the method a predetermined calculation formula is used preferably by means of a neural network, to which formula physiological parameters representing the subject to be measured are supplied. the input parameters comprise at least one or more of the following physiological parameters, such as sex, age, height, weight. one or more output parameters representing the exertion endurance indicator representing the exercise condition of the subject to be measured are obtained as a result from the calculation formula. in addition to the physiological parameters, one or more resting heartbeat parameters measured specifically from resting heartbeat are used as input parameters of the calculation formula. in the preferred embodiment of the invention, the calculation formula is formed by means of a neural network construction. dated 2001-08-21"
6278799,hierarchical data matrix pattern recognition system,""" the present invention relates to a hierarchical artificial neural network (hann) for automating the recognition and identification of patterns in data matrices. it has particular, although not exclusive, application to the identification of severe storm events (sses) from spatial precipitation patterns, derived from conventional volumetric radar imagery. to identify characteristic features a data matrix, the data matrix is processed with a self organizing network to produce a self organizing feature space mapping. the self organizing feature space mapping is processed to produce a density characterization of the feature space mapping. the self organizing network is preferably completely unsupervised. it may, under some circumstances include a supervised layer, but it must include at least an unsupervised component for the purposes of the invention. the """"self organizing feature space"""" is intended to include any map with the self organizing characteristics of the kohonen self organizing feature map. the frequency vector of a cappi image that has been derived is a data abstraction that can be displayed directly for examination. in preferred embodiments, it is presented to a classification network, e. g. the standard cpn network, for classifying the density vector representation of the three dimensional data and displaying a representation of classified features in the three dimensional data. a novel methodology is preferably used for incorporating vigilance and conscience mechanisms in the forward counterpropagation network during training. """,2001-08-21,"The title of the patent is hierarchical data matrix pattern recognition system and its abstract is "" the present invention relates to a hierarchical artificial neural network (hann) for automating the recognition and identification of patterns in data matrices. it has particular, although not exclusive, application to the identification of severe storm events (sses) from spatial precipitation patterns, derived from conventional volumetric radar imagery. to identify characteristic features a data matrix, the data matrix is processed with a self organizing network to produce a self organizing feature space mapping. the self organizing feature space mapping is processed to produce a density characterization of the feature space mapping. the self organizing network is preferably completely unsupervised. it may, under some circumstances include a supervised layer, but it must include at least an unsupervised component for the purposes of the invention. the """"self organizing feature space"""" is intended to include any map with the self organizing characteristics of the kohonen self organizing feature map. the frequency vector of a cappi image that has been derived is a data abstraction that can be displayed directly for examination. in preferred embodiments, it is presented to a classification network, e. g. the standard cpn network, for classifying the density vector representation of the three dimensional data and displaying a representation of classified features in the three dimensional data. a novel methodology is preferably used for incorporating vigilance and conscience mechanisms in the forward counterpropagation network during training. "" dated 2001-08-21"
6278937,method and apparatus for controlling the position of floating rig,"disclosed is a system for controlling the position of a floating rig that permits holding the rig at a position optimum to an excavation riser even if a position signal of the floating rig is not received, provided that the angles of inclination at the upper and lower ends of the riser are detected. in the method of controlling the position of a floating rig, the floating rig 10 is joined to a well head 14 at the sea bottom by an excavation riser 16, and the rig 10 is driven to a corrected position by thrusters or a combination of thrusters and a propulsion system. a neural network is allowed to learn in advance the position information of the floating rig accompanying the behaving characteristics of the excavation riser. the angles of inclination at the upper and lower ends of the excavation riser are detected and a signal represent of the detected angles is supplied to the neural network so as to permit the neural network to output the information on the correction of the present position of the floating rig. based on the position information, the correcting information that permits diminishing the angles of inclination at the upper and lower ends of the riser is calculated so as to automatically control the position of the floating rig. where the position information of the floating rig has ceased to be received, the angles of inclination at the upper and lower ends of the excavation riser that are to be detected are supplied to the position estimating section of the rig based on the algorithm of kalman filter so as to estimate the rig position and, thus, to perform the position control.",2001-08-21,"The title of the patent is method and apparatus for controlling the position of floating rig and its abstract is disclosed is a system for controlling the position of a floating rig that permits holding the rig at a position optimum to an excavation riser even if a position signal of the floating rig is not received, provided that the angles of inclination at the upper and lower ends of the riser are detected. in the method of controlling the position of a floating rig, the floating rig 10 is joined to a well head 14 at the sea bottom by an excavation riser 16, and the rig 10 is driven to a corrected position by thrusters or a combination of thrusters and a propulsion system. a neural network is allowed to learn in advance the position information of the floating rig accompanying the behaving characteristics of the excavation riser. the angles of inclination at the upper and lower ends of the excavation riser are detected and a signal represent of the detected angles is supplied to the neural network so as to permit the neural network to output the information on the correction of the present position of the floating rig. based on the position information, the correcting information that permits diminishing the angles of inclination at the upper and lower ends of the riser is calculated so as to automatically control the position of the floating rig. where the position information of the floating rig has ceased to be received, the angles of inclination at the upper and lower ends of the excavation riser that are to be detected are supplied to the position estimating section of the rig based on the algorithm of kalman filter so as to estimate the rig position and, thus, to perform the position control. dated 2001-08-21"
6278962,hybrid linear-neural network process control,"a hybrid analyzer having a data derived primary analyzer and an error correction analyzer connected in parallel is disclosed. the primary analyzer, preferably a data derived linear model such as a partial least squares model, is trained using training data to generate major predictions of defined output variables. the error correction analyzer, preferably a neural network model is trained to capture the residuals between the primary analyzer outputs and the target process variables. the residuals generated by the error correction analyzer is summed with the output of the primary analyzer to compensate for the error residuals of the primary analyzer to arrive at a more accurate overall model of the target process. additionally, an adaptive filter can be applied to the output of the primary analyzer to further capture the process dynamics. the data derived hybrid analyzer provides a readily adaptable framework to build the process model without requiring up-front knowledge. additionally, the primary analyzer, which incorporates the pls model, is well accepted by process control engineers. further, the hybrid analyzer also addresses the reliability of the process model output over the operating range since the primary analyzer can extrapolate data in a predictable way beyond the data used to train the model. together, the primary and the error correction analyzers provide a more accurate hybrid process analyzer which mitigates the disadvantages, and enhances the advantages, of each modeling methodology when used alone.",2001-08-21,"The title of the patent is hybrid linear-neural network process control and its abstract is a hybrid analyzer having a data derived primary analyzer and an error correction analyzer connected in parallel is disclosed. the primary analyzer, preferably a data derived linear model such as a partial least squares model, is trained using training data to generate major predictions of defined output variables. the error correction analyzer, preferably a neural network model is trained to capture the residuals between the primary analyzer outputs and the target process variables. the residuals generated by the error correction analyzer is summed with the output of the primary analyzer to compensate for the error residuals of the primary analyzer to arrive at a more accurate overall model of the target process. additionally, an adaptive filter can be applied to the output of the primary analyzer to further capture the process dynamics. the data derived hybrid analyzer provides a readily adaptable framework to build the process model without requiring up-front knowledge. additionally, the primary analyzer, which incorporates the pls model, is well accepted by process control engineers. further, the hybrid analyzer also addresses the reliability of the process model output over the operating range since the primary analyzer can extrapolate data in a predictable way beyond the data used to train the model. together, the primary and the error correction analyzers provide a more accurate hybrid process analyzer which mitigates the disadvantages, and enhances the advantages, of each modeling methodology when used alone. dated 2001-08-21"
6278985,virtual element selection device and method using a neural network,"in a virtual element selection device for use in growing, in an electronic device, virtual life composed of various parts or elements, an indication is given from a user and is converted by a neural network into a new element of the virtual life. the new element is substituted for an old one to be displayed on a display device. using the neural network makes it possible to select the new element which reflects the user's intention.",2001-08-21,"The title of the patent is virtual element selection device and method using a neural network and its abstract is in a virtual element selection device for use in growing, in an electronic device, virtual life composed of various parts or elements, an indication is given from a user and is converted by a neural network into a new element of the virtual life. the new element is substituted for an old one to be displayed on a display device. using the neural network makes it possible to select the new element which reflects the user's intention. dated 2001-08-21"
6281928,positional detector device for a vehicular license plate,"in a positional detector device for a license plate of a motor vehicle, a camera is provided to photograph a front and rear portion of a motor vehicle so as to produce an image signal. an a/d converter converts the image signal into a digital image. a positional detector detects a position of a lisence plate of a motor vehicle based on the digital image. a cut-off treatment device cuts off a specified region from an original image photographed by the camera while scanning the original image when the motor vehicle approaches the camera member within a predetermined distance. an edge refining treatment device emphasizes a contour of a pattern represented by the specified region cut off from the original image. a contraction treatment device contracts an image size of the specified region which was treated with the edge refining treatment device. a calculation treatment device feeds the contracted pattern to a learned neural network so as to calculate an output value for a positional detection neural network. a coordinate transformation device transforms the output value of the positional detection neural network to such a scale as to meet the original image. an addition treatment device adds data at specified points which were multiplied by gaussian window so as to obtain a projective addition value after carrying out the coordinate transformation. a distinction treatment device recognizes a position of the license plate based on a maximum value on which the projective addition value falls.",2001-08-28,"The title of the patent is positional detector device for a vehicular license plate and its abstract is in a positional detector device for a license plate of a motor vehicle, a camera is provided to photograph a front and rear portion of a motor vehicle so as to produce an image signal. an a/d converter converts the image signal into a digital image. a positional detector detects a position of a lisence plate of a motor vehicle based on the digital image. a cut-off treatment device cuts off a specified region from an original image photographed by the camera while scanning the original image when the motor vehicle approaches the camera member within a predetermined distance. an edge refining treatment device emphasizes a contour of a pattern represented by the specified region cut off from the original image. a contraction treatment device contracts an image size of the specified region which was treated with the edge refining treatment device. a calculation treatment device feeds the contracted pattern to a learned neural network so as to calculate an output value for a positional detection neural network. a coordinate transformation device transforms the output value of the positional detection neural network to such a scale as to meet the original image. an addition treatment device adds data at specified points which were multiplied by gaussian window so as to obtain a projective addition value after carrying out the coordinate transformation. a distinction treatment device recognizes a position of the license plate based on a maximum value on which the projective addition value falls. dated 2001-08-28"
6282323,image processing method and apparatus,"in an image processing method and apparatus, image data having multi-value levels for one pixel is input, and the input image data is quantized such that an output area of one pixel is adapted to an output device in which an output area of one pixel changes depending on the position of the pixel. a quantizing process executes an arithmetic operation based on an algorithm of a neural network on the basis of a value obtained by multiplying an output value by a weight corresponding to an area of each pixel. therefore, even if pixels have different maximum luminances, the different numbers of bits, and different color expression capabilities, an optimum half-tone process can be performed by an algorithm based on a cellular neural network, and a high-quality image can be obtained.",2001-08-28,"The title of the patent is image processing method and apparatus and its abstract is in an image processing method and apparatus, image data having multi-value levels for one pixel is input, and the input image data is quantized such that an output area of one pixel is adapted to an output device in which an output area of one pixel changes depending on the position of the pixel. a quantizing process executes an arithmetic operation based on an algorithm of a neural network on the basis of a value obtained by multiplying an output value by a weight corresponding to an area of each pixel. therefore, even if pixels have different maximum luminances, the different numbers of bits, and different color expression capabilities, an optimum half-tone process can be performed by an algorithm based on a cellular neural network, and a high-quality image can be obtained. dated 2001-08-28"
6282529,method and apparatus for computer-supported generation of at least one artificial training data vector for a neural network,"a method and apparatus for computer-supported generation of at least one artificial training data vector for a neural network is provided wherein a residual error is determined after a training of a neural network has occurred. a backward error is then determined from the residual error. artificial training data vectors are generated from a statistical random process that is based on a statistical distribution, such that the respective backward error for an input of the neural network is taken into consideration.",2001-08-28,"The title of the patent is method and apparatus for computer-supported generation of at least one artificial training data vector for a neural network and its abstract is a method and apparatus for computer-supported generation of at least one artificial training data vector for a neural network is provided wherein a residual error is determined after a training of a neural network has occurred. a backward error is then determined from the residual error. artificial training data vectors are generated from a statistical random process that is based on a statistical distribution, such that the respective backward error for an input of the neural network is taken into consideration. dated 2001-08-28"
6285992,neural network based methods and systems for analyzing complex data,"fully automated methods and systems for processing complex data sets to identify abnormalities are described. in one embodiment, the system includes wavelet processing, recursive processing to determine prominent features, and then utilizing feed forward neural networks (ffnns) to classify feature vectors generated in the wavelet and recursive processing. with respect to wavelet processing, multiresolution (five-level) and multidirection (two-dimensional) wavelet analysis with quadratic spline wavelets is performed to transform each image. the wavelets are a first-order derivative of a smoothing function and enhance the edges of image objects. because two-dimensional wavelet transforms quantize an image in terms of space and spatial frequency and can be ordered linearly, the data is processed recursively to determine prominent features. a neural network approach derived from sequential recursive auto-associative memory is then used to parse the wavelet coefficients and hierarchy data. since the wavelet coefficients are continuous, linear output instead of sigmoidal output is used. this variation is therefore referred to as linear output sequential recursive auto-associative memory, or losraam. the objective of training the losraam network is to have the output exactly match the input. context units arising from serial evaluation of the wavelet coefficient triplets may be collected as vectors. these vectors are subjected to cluster analysis. this analysis yields a number of identifiable and discrete states. from these states, feature vectors are created. each element in the feature vector represents the number of times the corresponding state from the above cluster analysis is found. then, feed forward neural networks (ffnns) are trained to classify feature vectors.",2001-09-04,"The title of the patent is neural network based methods and systems for analyzing complex data and its abstract is fully automated methods and systems for processing complex data sets to identify abnormalities are described. in one embodiment, the system includes wavelet processing, recursive processing to determine prominent features, and then utilizing feed forward neural networks (ffnns) to classify feature vectors generated in the wavelet and recursive processing. with respect to wavelet processing, multiresolution (five-level) and multidirection (two-dimensional) wavelet analysis with quadratic spline wavelets is performed to transform each image. the wavelets are a first-order derivative of a smoothing function and enhance the edges of image objects. because two-dimensional wavelet transforms quantize an image in terms of space and spatial frequency and can be ordered linearly, the data is processed recursively to determine prominent features. a neural network approach derived from sequential recursive auto-associative memory is then used to parse the wavelet coefficients and hierarchy data. since the wavelet coefficients are continuous, linear output instead of sigmoidal output is used. this variation is therefore referred to as linear output sequential recursive auto-associative memory, or losraam. the objective of training the losraam network is to have the output exactly match the input. context units arising from serial evaluation of the wavelet coefficient triplets may be collected as vectors. these vectors are subjected to cluster analysis. this analysis yields a number of identifiable and discrete states. from these states, feature vectors are created. each element in the feature vector represents the number of times the corresponding state from the above cluster analysis is found. then, feed forward neural networks (ffnns) are trained to classify feature vectors. dated 2001-09-04"
6289275,neural network based transient fuel control method,"a system and method for use in a motor vehicles is disclosed for calculating a fuel multiplier during transient engine operation. the fuel multiplier modifies the amount of fuel released from a fuel actuator into an engine. the fuel control system uses neural network logic to establish the fuel multiplier. the neural network logic involves taking inputs from engine sensors, processing the inputs through an input layer, a hidden layer and an output layer resulting in a fuel multiplier.",2001-09-11,"The title of the patent is neural network based transient fuel control method and its abstract is a system and method for use in a motor vehicles is disclosed for calculating a fuel multiplier during transient engine operation. the fuel multiplier modifies the amount of fuel released from a fuel actuator into an engine. the fuel control system uses neural network logic to establish the fuel multiplier. the neural network logic involves taking inputs from engine sensors, processing the inputs through an input layer, a hidden layer and an output layer resulting in a fuel multiplier. dated 2001-09-11"
6289328,chemical sensor pattern recognition system and method using a self-training neural network classifier with automated outlier detection,"a device and method for a pattern recognition system using a self-training neural network classifier with automated outlier detection for use in chemical sensor array systems. the pattern recognition system uses a probabilistic neural network (pnn) training computer system to develop automated classification algorithms for field-portable chemical sensor array systems. the pnn training computer system uses a pattern extraction unit to determine pattern vectors for chemical analytes. these pattern vectors form the initial hidden layer of the pnn. the hidden layer of the pnn is reduced in size by a learning vector quantization (lvq) classifier unit. the hidden layer neurons are further reduced in number by checking them against the pattern vectors and further eliminating dead neurons using a dead neuron elimination device. using the remaining neurons in the hidden layer of the pnn, a global, .sigma. value is calculated and a threshold rejection value is determined. the hidden layer, .sigma. value and the threshold value are then downloaded into a pnn module for use in a chemical sensor field unit. based on the threshold value, outliers seen in the real world environment may be rejected and a predicted chemical analyte identification with a measure of uncertainty will be provided to the user.",2001-09-11,"The title of the patent is chemical sensor pattern recognition system and method using a self-training neural network classifier with automated outlier detection and its abstract is a device and method for a pattern recognition system using a self-training neural network classifier with automated outlier detection for use in chemical sensor array systems. the pattern recognition system uses a probabilistic neural network (pnn) training computer system to develop automated classification algorithms for field-portable chemical sensor array systems. the pnn training computer system uses a pattern extraction unit to determine pattern vectors for chemical analytes. these pattern vectors form the initial hidden layer of the pnn. the hidden layer of the pnn is reduced in size by a learning vector quantization (lvq) classifier unit. the hidden layer neurons are further reduced in number by checking them against the pattern vectors and further eliminating dead neurons using a dead neuron elimination device. using the remaining neurons in the hidden layer of the pnn, a global, .sigma. value is calculated and a threshold rejection value is determined. the hidden layer, .sigma. value and the threshold value are then downloaded into a pnn module for use in a chemical sensor field unit. based on the threshold value, outliers seen in the real world environment may be rejected and a predicted chemical analyte identification with a measure of uncertainty will be provided to the user. dated 2001-09-11"
6289329,system for converting neural network to rule-based expert system using multiple-valued logic representation of neurons in feedforward network,a computer-implemented apparatus and method for generating a rule-based expert system from a trained neural network which is expressed as network data stored in a computer-readable medium. the rule-based expert system represents an interconnected network of neurons with associated weights data and threshold data. a network configuration extractor is provided for accessing the network data and for ascertaining the interconnection structure of the trained neural network by examining the network data. a transformation system is utilized to alter the algebraic sign of at least a portion of the weights data to eliminate differences in the algebraic sign among the weights data while selectively adjusting the threshold data to preserve the logical relationships defined by the neural network. a symbolic representation generator applies a sum-of-products search upon each neuron in the network to generate a multivalued logic representation for each neuron. a propagation mechanism combines the multivalued logic representation of each neuron through network propagation to yield a final logical expression corresponding to a rule-based expert system of the trained neural network. the resulting apparatus permits the knowledge incorporated in the connection strengths of neurons to be expressed as rule-based expert system.,2001-09-11,The title of the patent is system for converting neural network to rule-based expert system using multiple-valued logic representation of neurons in feedforward network and its abstract is a computer-implemented apparatus and method for generating a rule-based expert system from a trained neural network which is expressed as network data stored in a computer-readable medium. the rule-based expert system represents an interconnected network of neurons with associated weights data and threshold data. a network configuration extractor is provided for accessing the network data and for ascertaining the interconnection structure of the trained neural network by examining the network data. a transformation system is utilized to alter the algebraic sign of at least a portion of the weights data to eliminate differences in the algebraic sign among the weights data while selectively adjusting the threshold data to preserve the logical relationships defined by the neural network. a symbolic representation generator applies a sum-of-products search upon each neuron in the network to generate a multivalued logic representation for each neuron. a propagation mechanism combines the multivalued logic representation of each neuron through network propagation to yield a final logical expression corresponding to a rule-based expert system of the trained neural network. the resulting apparatus permits the knowledge incorporated in the connection strengths of neurons to be expressed as rule-based expert system. dated 2001-09-11
6290654,obstructive sleep apnea detection apparatus and method using pattern recognition,apparatus for detecting a breath pattern of a breathing patient having lungs and a nose and mouth in communication with the lungs and breathing through the nose and/or mouth and creating an airflow into and out of the lungs. the apparatus comprises a sensor in close proximity to the face of the patient for detecting said airflow to provide a first channel of airflow information in an analog format. an analog-to-digital converter is provided for converting the first channel of airflow information in an analog format to a first channel of airflow information in a digital format. a filter is provided for filtering the airflow information in a digital format in the first channel of information to improve the signal-to-noise ratio of the signal to provide filtered airflow information. an estimated volume airflow estimator operates on the filtered airflow information for estimating the amount of air volume inhaled and exhaled by the patient to provide a signal representing the estimated volume of air. a wavelet transform feature extractor is provided for obtaining a continuous-time wavelet transform of the estimated volume of air for ascertaining whether a breathing pattern has been recognized and providing a breathing pattern signal. a neural network pattern recognizer operates on the breathing pattern signal to ascertain when disordered breathing is occurring and provides a disordered breathing signal. a pattern classifier operates on the disordered breathing signal to separate the disordered breathing into apnea and hypopnea categories.,2001-09-18,The title of the patent is obstructive sleep apnea detection apparatus and method using pattern recognition and its abstract is apparatus for detecting a breath pattern of a breathing patient having lungs and a nose and mouth in communication with the lungs and breathing through the nose and/or mouth and creating an airflow into and out of the lungs. the apparatus comprises a sensor in close proximity to the face of the patient for detecting said airflow to provide a first channel of airflow information in an analog format. an analog-to-digital converter is provided for converting the first channel of airflow information in an analog format to a first channel of airflow information in a digital format. a filter is provided for filtering the airflow information in a digital format in the first channel of information to improve the signal-to-noise ratio of the signal to provide filtered airflow information. an estimated volume airflow estimator operates on the filtered airflow information for estimating the amount of air volume inhaled and exhaled by the patient to provide a signal representing the estimated volume of air. a wavelet transform feature extractor is provided for obtaining a continuous-time wavelet transform of the estimated volume of air for ascertaining whether a breathing pattern has been recognized and providing a breathing pattern signal. a neural network pattern recognizer operates on the breathing pattern signal to ascertain when disordered breathing is occurring and provides a disordered breathing signal. a pattern classifier operates on the disordered breathing signal to separate the disordered breathing into apnea and hypopnea categories. dated 2001-09-18
6292791,method and apparatus of synthesizing plucked string instruments using recurrent neural networks,""" a """"virtual string"""" is generated for synthesizing sound produced by plucked-string instruments using recurrent neural networks. the disclosed recurrent neural network, called a scattering recurrent network (srn), is based on the physics of waves traveling in the string. vibration measured from a plucked string is used as the training data for the srn. the trained srn is a virtual model capable of generating tones similar to the tones generated by the physical string. as with a real string, the """"virtual string"""" corresponding to the srn responds differently to different types of string """"plucking"""" motions. """,2001-09-18,"The title of the patent is method and apparatus of synthesizing plucked string instruments using recurrent neural networks and its abstract is "" a """"virtual string"""" is generated for synthesizing sound produced by plucked-string instruments using recurrent neural networks. the disclosed recurrent neural network, called a scattering recurrent network (srn), is based on the physics of waves traveling in the string. vibration measured from a plucked string is used as the training data for the srn. the trained srn is a virtual model capable of generating tones similar to the tones generated by the physical string. as with a real string, the """"virtual string"""" corresponding to the srn responds differently to different types of string """"plucking"""" motions. "" dated 2001-09-18"
6297439,system and method for automatic music generation using a neural network architecture,"a system and method are disclosed for automatically generating music on the basis of an initial sequence of input notes, and in particular to such a system and method utilizing a recursive artificial neural network (rann) architecture. the aforementioned system includes a score interpreter (2) interpreting an initial input sequence, a rhythm production rann (4) for generating a subsequent note duration, a note generation rann (6) for generating a subsequent note, and feedback means for feeding the pitch and duration of the subsequent note back to the rhythm generation (4) and note generation (6) ranns, the subsequent note thereby becoming the current note for a following iteration.",2001-10-02,"The title of the patent is system and method for automatic music generation using a neural network architecture and its abstract is a system and method are disclosed for automatically generating music on the basis of an initial sequence of input notes, and in particular to such a system and method utilizing a recursive artificial neural network (rann) architecture. the aforementioned system includes a score interpreter (2) interpreting an initial input sequence, a rhythm production rann (4) for generating a subsequent note duration, a note generation rann (6) for generating a subsequent note, and feedback means for feeding the pitch and duration of the subsequent note back to the rhythm generation (4) and note generation (6) ranns, the subsequent note thereby becoming the current note for a following iteration. dated 2001-10-02"
6298470,method for efficient manufacturing of integrated circuits,"this invention pertains to a method for the systematic development of integrated chip technology. the method may include obtaining empirical data of parameters for an existing integrated circuit manufacturing process and extrapolating the known data to a new technology to assess potential yields of the new technology from the known process. further, process variables of the new process may be adjusted based upon the empirical data in order to optimize the yields of the new technology. a logic based computing system such as a fuzzy logic or neural-network system may be utilized. the computing system may also be utilized to improve the yields of an existing manufacturing process by adjust process variables within downstream process tools based upon data collected in upstream process for a particular semiconductor substrate or lot.",2001-10-02,"The title of the patent is method for efficient manufacturing of integrated circuits and its abstract is this invention pertains to a method for the systematic development of integrated chip technology. the method may include obtaining empirical data of parameters for an existing integrated circuit manufacturing process and extrapolating the known data to a new technology to assess potential yields of the new technology from the known process. further, process variables of the new process may be adjusted based upon the empirical data in order to optimize the yields of the new technology. a logic based computing system such as a fuzzy logic or neural-network system may be utilized. the computing system may also be utilized to improve the yields of an existing manufacturing process by adjust process variables within downstream process tools based upon data collected in upstream process for a particular semiconductor substrate or lot. dated 2001-10-02"
6299536,"card dispensing shoe with scanner apparatus, system and method therefor",""" the present invention is directed to a playing card dispensing shoe apparatus, system and method wherein the shoe has a card scanner which scans the indicia on a playing card as the card moves along and out of a chute of the shoe by operation of the dealer. the scanner comprises an optical-sensor used in combination with a neural network which is trained using error back-propagation to recognize the card suits and card values of the playing cards as they are moved past the scanner. the scanning process in combination with a central processing unit (cpu) determines the progress of the play of the game and, by identifying card counting systems or basic playing strategies in use by the players of the game, provides means to limit or prevent casino losses and calculate the theoretical win of the casino, thus also providing an accurate quality method of the amount of comps to be given a particular player. the shoe is also provided with additional devices which make it simple and easy to access, record and display other data relevant to the play of the game. these include means for acconunodating a """"customer-tracking card"""" which reads each player's account information from a magnetic stripe on the card, thus providing access to the player's customer data file stored on the casino's computer system, and one or more alpha-numeric keyboards and lcd displays used to enter and retrieve player and game information. also included are keyboards on the game table so that each player can individually select various playing or wagering options using their own keyboard. """,2001-10-09,"The title of the patent is card dispensing shoe with scanner apparatus, system and method therefor and its abstract is "" the present invention is directed to a playing card dispensing shoe apparatus, system and method wherein the shoe has a card scanner which scans the indicia on a playing card as the card moves along and out of a chute of the shoe by operation of the dealer. the scanner comprises an optical-sensor used in combination with a neural network which is trained using error back-propagation to recognize the card suits and card values of the playing cards as they are moved past the scanner. the scanning process in combination with a central processing unit (cpu) determines the progress of the play of the game and, by identifying card counting systems or basic playing strategies in use by the players of the game, provides means to limit or prevent casino losses and calculate the theoretical win of the casino, thus also providing an accurate quality method of the amount of comps to be given a particular player. the shoe is also provided with additional devices which make it simple and easy to access, record and display other data relevant to the play of the game. these include means for acconunodating a """"customer-tracking card"""" which reads each player's account information from a magnetic stripe on the card, thus providing access to the player's customer data file stored on the casino's computer system, and one or more alpha-numeric keyboards and lcd displays used to enter and retrieve player and game information. also included are keyboards on the game table so that each player can individually select various playing or wagering options using their own keyboard. "" dated 2001-10-09"
6301381,"neurofilter, and method of training same to operate on image data such as to discriminate between text and picture regions of an image which is expressed by image data","a neurofilter is implemented as a neural network in which the weighting coefficients have previously been set, by an appropriate training procedure, such as to provide a desired form of filter response. the neurofilter is applicable to filtering of image data or serial data signals. also, by training a neurofilter to produce output data based on amounts of error that occur in the output data from a conventional filter, a filter apparatus can be provided whereby the neurofilter compensates for errors in output data from the conventional filter. the design and manufacturing constraints on the conventional filter can thereby be substantially relaxed.",2001-10-09,"The title of the patent is neurofilter, and method of training same to operate on image data such as to discriminate between text and picture regions of an image which is expressed by image data and its abstract is a neurofilter is implemented as a neural network in which the weighting coefficients have previously been set, by an appropriate training procedure, such as to provide a desired form of filter response. the neurofilter is applicable to filtering of image data or serial data signals. also, by training a neurofilter to produce output data based on amounts of error that occur in the output data from a conventional filter, a filter apparatus can be provided whereby the neurofilter compensates for errors in output data from the conventional filter. the design and manufacturing constraints on the conventional filter can thereby be substantially relaxed. dated 2001-10-09"
6301385,method and apparatus for segmenting images prior to coding,"to segment moving foreground from background, where the moving foreground is of most interest to the viewer, this method uses three detection algorithms as the input to a neural network. the multiple cues used are focus, intensity, and motion. the neural network consists of a two-layered neural network. focus and motion measurements are taken from high frequency data, edges; whereas, intensity measurements are taken from low frequency data, object interiors. combined, these measurements are used to segment a complete object. results indicate that moving foreground can be segmented from stationary foreground and moving or stationary background. the neural network segments the entire object, both interior and exterior, in this integrated approach. results also demonstrate that combining cues allows flexibility in both type and complexity of scenes. integration of cues improves accuracy in segmenting complex scenes containing both moving foreground and background. good segmentation yields bit rate savings when coding the object of interest, also called the video object in mpeg4. this method combines simple measurements to increase segmentation robustness.",2001-10-09,"The title of the patent is method and apparatus for segmenting images prior to coding and its abstract is to segment moving foreground from background, where the moving foreground is of most interest to the viewer, this method uses three detection algorithms as the input to a neural network. the multiple cues used are focus, intensity, and motion. the neural network consists of a two-layered neural network. focus and motion measurements are taken from high frequency data, edges; whereas, intensity measurements are taken from low frequency data, object interiors. combined, these measurements are used to segment a complete object. results indicate that moving foreground can be segmented from stationary foreground and moving or stationary background. the neural network segments the entire object, both interior and exterior, in this integrated approach. results also demonstrate that combining cues allows flexibility in both type and complexity of scenes. integration of cues improves accuracy in segmenting complex scenes containing both moving foreground and background. good segmentation yields bit rate savings when coding the object of interest, also called the video object in mpeg4. this method combines simple measurements to increase segmentation robustness. dated 2001-10-09"
6301572,neural network based analysis system for vibration analysis and condition monitoring,"a system and a method for tracking long term performance of a vibrating body such as a gas turbine, includes a vibration sensor who time domain outputs are transformed to the frequency domain, using a fast fourier transform processing. frequency domain outputs are provided as inputs to a fuzzy adaptive resonance theory neural network. outputs from the network can be coupled to an expert system for analysis, to display devices for presentation to an operator or are available for other control and information purposes.",2001-10-09,"The title of the patent is neural network based analysis system for vibration analysis and condition monitoring and its abstract is a system and a method for tracking long term performance of a vibrating body such as a gas turbine, includes a vibration sensor who time domain outputs are transformed to the frequency domain, using a fast fourier transform processing. frequency domain outputs are provided as inputs to a fuzzy adaptive resonance theory neural network. outputs from the network can be coupled to an expert system for analysis, to display devices for presentation to an operator or are available for other control and information purposes. dated 2001-10-09"
6304218,wireless communication system and method and system for detection of position of radio mobile station,"a method of detecting a position of a radio mobile station in radiocommunications, which is capable of accurately and simply finding the position of the mobile station. at a measuring point the mobile station measures the reception radio strength levels from a plurality of base stations and conveys the measurement results through the base station to a control station. the control station learns, through a neural network, the correlation between the reception radio strength levels and the position of the mobile station on the basis of the measurement results at a plurality of measuring points and the positions of the measuring points. subsequently, when the mobile station communicates to the control station the reception radio strength levels measured at an arbitrary point, the control station estimates the position of the mobile station, causing those measurement results, on the basis of the correlation obtained through the learning.",2001-10-16,"The title of the patent is wireless communication system and method and system for detection of position of radio mobile station and its abstract is a method of detecting a position of a radio mobile station in radiocommunications, which is capable of accurately and simply finding the position of the mobile station. at a measuring point the mobile station measures the reception radio strength levels from a plurality of base stations and conveys the measurement results through the base station to a control station. the control station learns, through a neural network, the correlation between the reception radio strength levels and the position of the mobile station on the basis of the measurement results at a plurality of measuring points and the positions of the measuring points. subsequently, when the mobile station communicates to the control station the reception radio strength levels measured at an arbitrary point, the control station estimates the position of the mobile station, causing those measurement results, on the basis of the correlation obtained through the learning. dated 2001-10-16"
6304539,multilayer perception neural network scheme disk memory device and signal processing device,"a disk memory device comprises a head which reads a data recorded on a disk and outputs a read-out signal, an a/d converter which converts the read-out signal into a digital signal, and a signal processing circuit of a multilayer perceptron type neural network scheme which receives the digital signal converted by the a/d converter, wherein the signal processing circuit has an input layer which inputs the digital signal, a hidden layer having a plurality of hidden nodes, and an output layer which outputs an output signal, and the input layer has a weighting arithmetic circuitry which is shared by the each hidden node, calculates a coupling weighting coefficient between the input layer and the plurality of hidden nodes, and outputs a result of multiplying a digital signal by the weighting coefficient to the plurality of hidden nodes.",2001-10-16,"The title of the patent is multilayer perception neural network scheme disk memory device and signal processing device and its abstract is a disk memory device comprises a head which reads a data recorded on a disk and outputs a read-out signal, an a/d converter which converts the read-out signal into a digital signal, and a signal processing circuit of a multilayer perceptron type neural network scheme which receives the digital signal converted by the a/d converter, wherein the signal processing circuit has an input layer which inputs the digital signal, a hidden layer having a plurality of hidden nodes, and an output layer which outputs an output signal, and the input layer has a weighting arithmetic circuitry which is shared by the each hidden node, calculates a coupling weighting coefficient between the input layer and the plurality of hidden nodes, and outputs a result of multiplying a digital signal by the weighting coefficient to the plurality of hidden nodes. dated 2001-10-16"
6304845,method of transmitting voice data,"in a transmission of voice data, the stream of voice data is first decomposed into phonemes. for each phoneme a code symbol which is assigned to that specific phoneme in a selectable voice-specific and/or speaker-specific phoneme catalog is transmitted to a voice synthesizer at the transmission destination. the amount of data which has to be transmitted is generally greatly reduced. the decomposition of the stream of voice data into phonemes is carried out by a neural network which is trained to detect the phonemes stored in the selected voice-specific and/or speaker-specific phoneme catalog. the voice synthesizer reconverts the stream of received code symbols into a sequence of phonemes and outputs it.",2001-10-16,"The title of the patent is method of transmitting voice data and its abstract is in a transmission of voice data, the stream of voice data is first decomposed into phonemes. for each phoneme a code symbol which is assigned to that specific phoneme in a selectable voice-specific and/or speaker-specific phoneme catalog is transmitted to a voice synthesizer at the transmission destination. the amount of data which has to be transmitted is generally greatly reduced. the decomposition of the stream of voice data into phonemes is carried out by a neural network which is trained to detect the phonemes stored in the selected voice-specific and/or speaker-specific phoneme catalog. the voice synthesizer reconverts the stream of received code symbols into a sequence of phonemes and outputs it. dated 2001-10-16"
6304864,system for retrieving multimedia information from the internet using multiple evolving intelligent agents,"a system for retrieving multimedia information is provided using a computer coupled to a computer-based network, such as the internet, and particularly the world wide web (www). the system includes a web browser, a graphic user interface enabled through the web browser to allow a user to input a query representing the information the user wishes to retrieve, and an agent server for producing, training, and evolving first agents and second agents. each of the first agents retrieves documents (web page) from the network at a different first network address and at other addresses linked from the document at the first network address. each of the second agents executes a search on different search engines on the network in accordance with the query to retrieve documents at network addresses provided by the search engine. the system includes a natural language processor which determines the subject categories and important terms of the query, and of the text of each agent retrieved document. the agent server generates and trains an artificial neural network in accordance with the natural language processed query, and embeds the trained artificial neural network in each of the first and second agents. during the search, the first and second agents process through their artificial neural network the subject categories and important terms of each document they retrieve to determine a retrieval value for the document. the graphic user interface displays to the user the addresses of the retrieved documents which are above a threshold retrieval value. the user manually, or the agent server automatically, selects which of the retrieved documents are relevant. periodically, the artificial neural network of the first and second agents is expanded and retrained by the agent server in accordance with the selected relevant documents to improve their ability to retrieve documents which may be relevant to the query. further, the agent server can evolve an artificial neural network based on the current artificial neural network, the retrieved documents, and their selected relevancy, by iteratively producing, training, and testing several generations of neural networks to produce an evolved agent. the artificial neural network of the evolved agent then replaces the current artificial neural network used by the agents to search the internet. one or more concurrent search of the internet may be provided.",2001-10-16,"The title of the patent is system for retrieving multimedia information from the internet using multiple evolving intelligent agents and its abstract is a system for retrieving multimedia information is provided using a computer coupled to a computer-based network, such as the internet, and particularly the world wide web (www). the system includes a web browser, a graphic user interface enabled through the web browser to allow a user to input a query representing the information the user wishes to retrieve, and an agent server for producing, training, and evolving first agents and second agents. each of the first agents retrieves documents (web page) from the network at a different first network address and at other addresses linked from the document at the first network address. each of the second agents executes a search on different search engines on the network in accordance with the query to retrieve documents at network addresses provided by the search engine. the system includes a natural language processor which determines the subject categories and important terms of the query, and of the text of each agent retrieved document. the agent server generates and trains an artificial neural network in accordance with the natural language processed query, and embeds the trained artificial neural network in each of the first and second agents. during the search, the first and second agents process through their artificial neural network the subject categories and important terms of each document they retrieve to determine a retrieval value for the document. the graphic user interface displays to the user the addresses of the retrieved documents which are above a threshold retrieval value. the user manually, or the agent server automatically, selects which of the retrieved documents are relevant. periodically, the artificial neural network of the first and second agents is expanded and retrained by the agent server in accordance with the selected relevant documents to improve their ability to retrieve documents which may be relevant to the query. further, the agent server can evolve an artificial neural network based on the current artificial neural network, the retrieved documents, and their selected relevancy, by iteratively producing, training, and testing several generations of neural networks to produce an evolved agent. the artificial neural network of the evolved agent then replaces the current artificial neural network used by the agents to search the internet. one or more concurrent search of the internet may be provided. dated 2001-10-16"
6304865,audio diagnostic system and method using frequency spectrum and neural network,"a method for testing an audio device with a trained neural network includes a loopback connector connecting the output port of the audio device to the input port of the audio device. a test signal is transmitted through the audio port and received at the input port. the test signal is converted into a frequency spectrum for analysis. the frequency spectrum is provided as input to a trained neural network, the neural network being previously trained to recognize the frequency spectrum pattern created by a properly working, or ideal, audio device. the neural network is trained by connecting the input port to the output port of an audio device from which the training is to occur. prior to converting signals to a frequency spectrum, the waveform characteristics of the signal may be iteratively evaluated and recording levels adjusted so that the signal received has characteristics that can be tested by the neural network. the analyzed signal may be a portion of the signal received after analog to digital converters in the audio device have stabilized. the neural network generates a confidence level based on comparing the pattern of the tested audio device's frequency spectrum to the frequency spectrum of a working audio device. a pass value may be predetermined so that the tested audio device is reported as passing or failing the test by comparing the confidence level value generated by the system with the predetermined pass value.",2001-10-16,"The title of the patent is audio diagnostic system and method using frequency spectrum and neural network and its abstract is a method for testing an audio device with a trained neural network includes a loopback connector connecting the output port of the audio device to the input port of the audio device. a test signal is transmitted through the audio port and received at the input port. the test signal is converted into a frequency spectrum for analysis. the frequency spectrum is provided as input to a trained neural network, the neural network being previously trained to recognize the frequency spectrum pattern created by a properly working, or ideal, audio device. the neural network is trained by connecting the input port to the output port of an audio device from which the training is to occur. prior to converting signals to a frequency spectrum, the waveform characteristics of the signal may be iteratively evaluated and recording levels adjusted so that the signal received has characteristics that can be tested by the neural network. the analyzed signal may be a portion of the signal received after analog to digital converters in the audio device have stabilized. the neural network generates a confidence level based on comparing the pattern of the tested audio device's frequency spectrum to the frequency spectrum of a working audio device. a pass value may be predetermined so that the tested audio device is reported as passing or failing the test by comparing the confidence level value generated by the system with the predetermined pass value. dated 2001-10-16"
6305232,method of dry-calibrating vortex flow sensors,"to achieve accuracies of the order of 0.75% of the measured value, a digitized, two-dimensional overall image of a bluff body (7), of the internal surface of a measuring tube (2) in the area of the bluff body, of the two fixing zones (71, 72) of the bluff body, and of contour line (51) of the inlet end (5) of the measuring tube is generated by a high-resolution electronic camera (9) located in front of the measuring tube (2) on the axis (3) of this tube. the overall image is divided into three partial images. the first partial image contains only information about the inlet end (5) and the internal surface (4) of the measuring tube, the second contains only information about the bluff body (7) without the fixing zones (71, 72), and the third contains only information about the fixing zones. from shape information about the fixing zones (71, 72) and ideal information characteristic of the ideal shapes of the fixing zones, cross-correlation information is formed. in a neural network (22), all information and corresponding standard information derived from a plurality of wet calibrations are processed into calibration factor information and/or dimension information about the geometrical dimensions of the calibrated vortex sensor.",2001-10-23,"The title of the patent is method of dry-calibrating vortex flow sensors and its abstract is to achieve accuracies of the order of 0.75% of the measured value, a digitized, two-dimensional overall image of a bluff body (7), of the internal surface of a measuring tube (2) in the area of the bluff body, of the two fixing zones (71, 72) of the bluff body, and of contour line (51) of the inlet end (5) of the measuring tube is generated by a high-resolution electronic camera (9) located in front of the measuring tube (2) on the axis (3) of this tube. the overall image is divided into three partial images. the first partial image contains only information about the inlet end (5) and the internal surface (4) of the measuring tube, the second contains only information about the bluff body (7) without the fixing zones (71, 72), and the third contains only information about the fixing zones. from shape information about the fixing zones (71, 72) and ideal information characteristic of the ideal shapes of the fixing zones, cross-correlation information is formed. in a neural network (22), all information and corresponding standard information derived from a plurality of wet calibrations are processed into calibration factor information and/or dimension information about the geometrical dimensions of the calibrated vortex sensor. dated 2001-10-23"
6306087,computer assisted methods for diagnosing diseases,"the simultaneous multi access reasoning technology system of the present invention utilizes both existing knowledge and implicit information that can be numerically extracted from training data to provide a method and apparatus for diagnosing disease and treating a patient. this technology further comprises a system for receiving patient data from another location, analyzing the data in a trained neural network, producing a diagnostic value, and optionally transmitting the diagnostic value to another location.",2001-10-23,"The title of the patent is computer assisted methods for diagnosing diseases and its abstract is the simultaneous multi access reasoning technology system of the present invention utilizes both existing knowledge and implicit information that can be numerically extracted from training data to provide a method and apparatus for diagnosing disease and treating a patient. this technology further comprises a system for receiving patient data from another location, analyzing the data in a trained neural network, producing a diagnostic value, and optionally transmitting the diagnostic value to another location. dated 2001-10-23"
6307854,telecommunications switch,"a self-routing switch such as a banyan switch has a controller which recognizes incoming routing requests which would give rise to blocking in the switch and makes and optimum selection of queued requests which can be handled without blocking. the controller is implemented by means of an optical neural network having a light source array to illuminate a photodetector array through a mask, there being a light source array element and a photodetector element for each possible path through the switch. the assignment of paths to array locations is such that, for the path corresponding to any light source array element, the photodetector array element positions corresponding to the paths blocked thereby form a pattern which is a shifted version of the pattern formed by the photodetector array elements corresponding to the paths blocked by a path corresponding to any other light source array element, whereby a single mask may be employed.",2001-10-23,"The title of the patent is telecommunications switch and its abstract is a self-routing switch such as a banyan switch has a controller which recognizes incoming routing requests which would give rise to blocking in the switch and makes and optimum selection of queued requests which can be handled without blocking. the controller is implemented by means of an optical neural network having a light source array to illuminate a photodetector array through a mask, there being a light source array element and a photodetector element for each possible path through the switch. the assignment of paths to array locations is such that, for the path corresponding to any light source array element, the photodetector array element positions corresponding to the paths blocked thereby form a pattern which is a shifted version of the pattern formed by the photodetector array elements corresponding to the paths blocked by a path corresponding to any other light source array element, whereby a single mask may be employed. dated 2001-10-23"
6308649,sailboat and crew performance optimization system,"a sailboat and crew performance optimization system includes a modular system of sensors, data acquisition, computational analysis, graphical display and optional feedback control for optimizing sailboat and crew performance. the system acquires data relating to external factors (e.g. wind speed, wind direction, variations in wind speed, variations in wind direction, sea state, and wave conditions), performance parameters (e.g. boat speed, time to reach a specified destination and velocity made good, safety parameters and sailboat comfort parameters), dependent variable setpoints (e.g. sail shape, sail pressure distribution, etc.), and control variables (e.g. line tensions, rudder angle, sail plan, etc.) and correlates or analyzes the data to determine or predict the optimum setpoint targets and control variables. the system displays information and relationships to the sailboat crew in order to optimize sailboat performance and crew performance. the system also provides benchmark measures of sailboat performance and crew performance to compare performance at different times or under different conditions or to measure progress or improvement in performance. optionally, the system can be used for automatic feedback control of sailboat operation. the system may utilize a computer or an artificial intelligence system, such as a neural network system, a fuzzy logic system, a genetic algorithm system or an expert system, to analyze and predict optimum setpoint targets and control variables.",2001-10-30,"The title of the patent is sailboat and crew performance optimization system and its abstract is a sailboat and crew performance optimization system includes a modular system of sensors, data acquisition, computational analysis, graphical display and optional feedback control for optimizing sailboat and crew performance. the system acquires data relating to external factors (e.g. wind speed, wind direction, variations in wind speed, variations in wind direction, sea state, and wave conditions), performance parameters (e.g. boat speed, time to reach a specified destination and velocity made good, safety parameters and sailboat comfort parameters), dependent variable setpoints (e.g. sail shape, sail pressure distribution, etc.), and control variables (e.g. line tensions, rudder angle, sail plan, etc.) and correlates or analyzes the data to determine or predict the optimum setpoint targets and control variables. the system displays information and relationships to the sailboat crew in order to optimize sailboat performance and crew performance. the system also provides benchmark measures of sailboat performance and crew performance to compare performance at different times or under different conditions or to measure progress or improvement in performance. optionally, the system can be used for automatic feedback control of sailboat operation. the system may utilize a computer or an artificial intelligence system, such as a neural network system, a fuzzy logic system, a genetic algorithm system or an expert system, to analyze and predict optimum setpoint targets and control variables. dated 2001-10-30"
6309342,management of physiological and psychological state of an individual using images biometric analyzer,"a method of determining the physiological reactivity of an individual with respect to images, comprising: measuring at least one physiological parameter of an individual during a baseline calm period of time; measuring the said at least one physiological parameter of said individual during a stress period of time and a succeeding rest period of time; dividing said measured data for each of said periods of time into a plurality of predetermined time segments; computing a histogram or fourier analysis as appropriate on said predetermined time segments for each said period of time; standardizing said computed data; conducting a principal component analysis, a canonical discriminant analysis, or a neural network, on said data to establish a baseline calm or rest period set of scores and a stress period set of scores; measuring the said at least one physiological parameter of said individual during an image presentation period of time; repeating said time segmenting, said histogram or fourier analysis, and said standardizing steps on said measured physiological data for said image presentation period of time; applying said vectors from said conducting step to said image data to produce an image set of scores which are compared with said baseline and stress sets of scores; and determining whether said presented image is activating, deactivating, or neutral based on said comparison.",2001-10-30,"The title of the patent is management of physiological and psychological state of an individual using images biometric analyzer and its abstract is a method of determining the physiological reactivity of an individual with respect to images, comprising: measuring at least one physiological parameter of an individual during a baseline calm period of time; measuring the said at least one physiological parameter of said individual during a stress period of time and a succeeding rest period of time; dividing said measured data for each of said periods of time into a plurality of predetermined time segments; computing a histogram or fourier analysis as appropriate on said predetermined time segments for each said period of time; standardizing said computed data; conducting a principal component analysis, a canonical discriminant analysis, or a neural network, on said data to establish a baseline calm or rest period set of scores and a stress period set of scores; measuring the said at least one physiological parameter of said individual during an image presentation period of time; repeating said time segmenting, said histogram or fourier analysis, and said standardizing steps on said measured physiological data for said image presentation period of time; applying said vectors from said conducting step to said image data to produce an image set of scores which are compared with said baseline and stress sets of scores; and determining whether said presented image is activating, deactivating, or neutral based on said comparison. dated 2001-10-30"
6311172,"method for determination of weights, suitable for elimination, of a neural network using a computer","the training phase of a neural network nn is stopped before an error function, which is to be minimized in the training phase, reaches a minimum (301). a first variable (eg) is defined using, for example, the optimal brain damage method or the optimal brain surgeon method, on the assumption that the error function is at the minimum. furthermore, a second variable (zg) is determined which provides an indication of the manner in which the value of the error function varies when a weight (w.sub.i) is removed from the neural network (nn). the first variable (eg) and the second variable (zg) are used to classify the weight (w.sub.i) as being suitable or unsuitable for removal from the neural network (nn).",2001-10-30,"The title of the patent is method for determination of weights, suitable for elimination, of a neural network using a computer and its abstract is the training phase of a neural network nn is stopped before an error function, which is to be minimized in the training phase, reaches a minimum (301). a first variable (eg) is defined using, for example, the optimal brain damage method or the optimal brain surgeon method, on the assumption that the error function is at the minimum. furthermore, a second variable (zg) is determined which provides an indication of the manner in which the value of the error function varies when a weight (w.sub.i) is removed from the neural network (nn). the first variable (eg) and the second variable (zg) are used to classify the weight (w.sub.i) as being suitable or unsuitable for removal from the neural network (nn). dated 2001-10-30"
6311174,problem solving apparatus having learning function,"a problem solving unit for obtaining a solution in a symbol process in response to a given problem is provided with a learning control unit for making a neural network learn the solution output from the problem solving unit. the output in response to the given problem from the learned neural network is transmitted to the problem solving unit as a (first) hint on obtaining the solution to the problem. furthermore, a second neural network for outputting a second hint is provided, selects either the first hint or the second hint, and provides the selection result to the problem solving unit.",2001-10-30,"The title of the patent is problem solving apparatus having learning function and its abstract is a problem solving unit for obtaining a solution in a symbol process in response to a given problem is provided with a learning control unit for making a neural network learn the solution output from the problem solving unit. the output in response to the given problem from the learned neural network is transmitted to the problem solving unit as a (first) hint on obtaining the solution to the problem. furthermore, a second neural network for outputting a second hint is provided, selects either the first hint or the second hint, and provides the selection result to the problem solving unit. dated 2001-10-30"
6314413,method for controlling process events using neural network,"the invention relates to a method for controlling process events of a technical plant. in order to permit a simultaneous and coherent assessment of relevant process variables of the plant, it is proposed to use a neural analysis on the basis of self-organizing neural maps to evaluate the relevant process variables in relation to one another by realizing a topology-maintaining nonlinear projection of data from the relevant process variables onto a multidimensional neural map.",2001-11-06,"The title of the patent is method for controlling process events using neural network and its abstract is the invention relates to a method for controlling process events of a technical plant. in order to permit a simultaneous and coherent assessment of relevant process variables of the plant, it is proposed to use a neural analysis on the basis of self-organizing neural maps to evaluate the relevant process variables in relation to one another by realizing a topology-maintaining nonlinear projection of data from the relevant process variables onto a multidimensional neural map. dated 2001-11-06"
6314414,method for training and/or testing a neural network with missing and/or incomplete data,"a neural network system is provided that models the system in a system model (12) with the output thereof providing a predicted output. this predicted output is modified or controlled by an output control (14). input data is processed in a data preprocess step (10) to reconcile the data for input to the system model (12). additionally, the error resulted from the reconciliation is input to an uncertainty model to predict the uncertainty in the predicted output. this is input to a decision processor (20) which is utilized to control the output control (14). the output control (14) is controlled to either vary the predicted output or to inhibit the predicted output whenever the output of the uncertainty model (18) exceeds a predetermined decision threshold, input by a decision threshold block (22). additionally, a validity model (16) is also provided which represents the reliability or validity of the output as a function of the number of data points in a given data region during training of the system model (12). this predicts the confidence in the predicted output which is also input to the decision processor (20). the decision processor (20) therefore bases its decision on the predicted confidence and the predicted uncertainty. additionally, the uncertainty output by the data preprocess block (10) can be utilized to train the system model (12).",2001-11-06,"The title of the patent is method for training and/or testing a neural network with missing and/or incomplete data and its abstract is a neural network system is provided that models the system in a system model (12) with the output thereof providing a predicted output. this predicted output is modified or controlled by an output control (14). input data is processed in a data preprocess step (10) to reconcile the data for input to the system model (12). additionally, the error resulted from the reconciliation is input to an uncertainty model to predict the uncertainty in the predicted output. this is input to a decision processor (20) which is utilized to control the output control (14). the output control (14) is controlled to either vary the predicted output or to inhibit the predicted output whenever the output of the uncertainty model (18) exceeds a predetermined decision threshold, input by a decision threshold block (22). additionally, a validity model (16) is also provided which represents the reliability or validity of the output as a function of the number of data points in a given data region during training of the system model (12). this predicts the confidence in the predicted output which is also input to the decision processor (20). the decision processor (20) therefore bases its decision on the predicted confidence and the predicted uncertainty. additionally, the uncertainty output by the data preprocess block (10) can be utilized to train the system model (12). dated 2001-11-06"
6317658,neurocomputing control distribution system,"a method, system and computer-readable medium for controlling a control subsystem of a vehicle. the control subsystem includes at least one propulsion, aerodynamic, or other control effector. the system also includes sensors for sensing vehicle position and motions and operating conditions, a control input device for generating control signals, and a generator for generating desired vehicle forces/moments from the sensed vehicle position and motions and operating conditions, and the generated control input signals based on predefined vehicle compensation and control laws. also included is a neural network controller for generating control subsystem commands for the at least one propulsion, aerodynamic, or other control effector based on the generated desired forces/moments, wherein said neural network controller was trained based on pregenerated vehicle control distribution data. the neural network controller is also trained to compensate for one or more failed control effectors.",2001-11-13,"The title of the patent is neurocomputing control distribution system and its abstract is a method, system and computer-readable medium for controlling a control subsystem of a vehicle. the control subsystem includes at least one propulsion, aerodynamic, or other control effector. the system also includes sensors for sensing vehicle position and motions and operating conditions, a control input device for generating control signals, and a generator for generating desired vehicle forces/moments from the sensed vehicle position and motions and operating conditions, and the generated control input signals based on predefined vehicle compensation and control laws. also included is a neural network controller for generating control subsystem commands for the at least one propulsion, aerodynamic, or other control effector based on the generated desired forces/moments, wherein said neural network controller was trained based on pregenerated vehicle control distribution data. the neural network controller is also trained to compensate for one or more failed control effectors. dated 2001-11-13"
6317730,method for optimizing a set of fuzzy rules using a computer,"a set of fuzzy rules (fr) is mapped onto a neural network (nn) (501). the neural network (nn) is trained (502), and weights (w.sub.i) and/or neurons (ne) of the neural network (nn) are pruned or grown (503). a new neural network (nnn) formed in this way is mapped onto a new fuzzy rule set (nfr) (504).",2001-11-13,"The title of the patent is method for optimizing a set of fuzzy rules using a computer and its abstract is a set of fuzzy rules (fr) is mapped onto a neural network (nn) (501). the neural network (nn) is trained (502), and weights (w.sub.i) and/or neurons (ne) of the neural network (nn) are pruned or grown (503). a new neural network (nnn) formed in this way is mapped onto a new fuzzy rule set (nfr) (504). dated 2001-11-13"
6324531,system and method for identifying the geographic origin of a fresh commodity,"the detection method includes generating a plurality of neural network models. each model has as a training set a data set from a plurality of samples of a commodity of known origins. each sample has been analyzed for a plurality of elemental concentrations. each neural network model is presented for classification a test data set from a plurality of samples of a commodity of unknown origins. as with the training set, the samples have been analyzed for the same plurality of elemental concentrations. next a bootstrap aggregating strategy is employed to combine the results of the classifications for each sample in the test data set made by each neural network model. finally, a determination is made from the bootstrap aggregating strategy as to a final classification of each sample in the test data set. this final classification is indicative of the geographical origin of the commodity. the system includes software for generating the neural network models and a software routine for performing the bootstrap aggregating strategy.",2001-11-27,"The title of the patent is system and method for identifying the geographic origin of a fresh commodity and its abstract is the detection method includes generating a plurality of neural network models. each model has as a training set a data set from a plurality of samples of a commodity of known origins. each sample has been analyzed for a plurality of elemental concentrations. each neural network model is presented for classification a test data set from a plurality of samples of a commodity of unknown origins. as with the training set, the samples have been analyzed for the same plurality of elemental concentrations. next a bootstrap aggregating strategy is employed to combine the results of the classifications for each sample in the test data set made by each neural network model. finally, a determination is made from the bootstrap aggregating strategy as to a final classification of each sample in the test data set. this final classification is indicative of the geographical origin of the commodity. the system includes software for generating the neural network models and a software routine for performing the bootstrap aggregating strategy. dated 2001-11-27"
6324532,method and apparatus for training a neural network to detect objects in an image,""" a signal processing apparatus and concomitant method for learning and integrating features from multiple resolutions for detecting and/or classifying objects. the signal processing apparatus comprises a hierarchical pyramid of neural networks (hpnn) having a """"fine-to-coarse"""" structure or a combination of the """"fine-to-coarse"""" and the """"coarse-to-fine"""" structures. """,2001-11-27,"The title of the patent is method and apparatus for training a neural network to detect objects in an image and its abstract is "" a signal processing apparatus and concomitant method for learning and integrating features from multiple resolutions for detecting and/or classifying objects. the signal processing apparatus comprises a hierarchical pyramid of neural networks (hpnn) having a """"fine-to-coarse"""" structure or a combination of the """"fine-to-coarse"""" and the """"coarse-to-fine"""" structures. "" dated 2001-11-27"
6325178,elevator group managing system with selective performance prediction,"a rule base storing control rule sets predicts elevator group management performance, such as waiting time distribution, obtained when applying each rule set stored in the rule base to the current traffic situation, and selects a rule set in accordance with a performance prediction. in addition, a weight database stores weighting parameters of a neural network corresponding to the rule sets and performance learning measures for correcting the weighting parameters in accordance with learning by the neural network. as a result, the optimal rule set is applied at all times for group management control of the elevators to provide passengers with excellent service and to enhance prediction accuracy in correspondence with the actual operational situation of the elevators.",2001-12-04,"The title of the patent is elevator group managing system with selective performance prediction and its abstract is a rule base storing control rule sets predicts elevator group management performance, such as waiting time distribution, obtained when applying each rule set stored in the rule base to the current traffic situation, and selects a rule set in accordance with a performance prediction. in addition, a weight database stores weighting parameters of a neural network corresponding to the rule sets and performance learning measures for correcting the weighting parameters in accordance with learning by the neural network. as a result, the optimal rule set is applied at all times for group management control of the elevators to provide passengers with excellent service and to enhance prediction accuracy in correspondence with the actual operational situation of the elevators. dated 2001-12-04"
6327377,automated cytological specimen classification system and method,an automated screening system and method for cytological specimen classification in which a neural network is utilized in performance of the classification function. also included is an automated microscope and associated image processing circuitry.,2001-12-04,The title of the patent is automated cytological specimen classification system and method and its abstract is an automated screening system and method for cytological specimen classification in which a neural network is utilized in performance of the classification function. also included is an automated microscope and associated image processing circuitry. dated 2001-12-04
6327386,key character extraction and lexicon reduction for cursive text recognition,""" a method, apparatus, and article of manufacture employing lexicon reduction using key characters and a neural network, for recognizing a line of cursive text. unambiguous parts of a cursive image, referred to as """"key characters,"""" are identified. if the level of confidence that a segment of a line of cursive text is a particular character is higher than a threshold, and is also sufficiently higher than the level of confidence of neighboring segments, then the character is designated as a key character candidate. key character candidates are then screened using geometric information. the key character candidates that pass the screening are designated key characters. two-stages of lexicon reduction are employed. the first stage of lexicon reduction uses a neural network to estimate a lower bound and an upper bound of the number of characters in a line of cursive text. lexicon entries having a total number of characters outside of the bounds are eliminated. for the second stage of lexicon reduction, the lexicon is fitter reduced by comparing character strings using the key characters, with lexicon entries. for each of the key characters in the character strings, it is determined whether there is a mismatch between the key character and characters in a corresponding search range in the lexicon entry. if the number of mismatches for all of the key characters in a search string is greater than (1+(the number of key characters in the search string/4)), then the lexicon entry is eliminated. accordingly, the invention advantageously accomplishes lexicon reduction, thereby decreasing the time required to recognize a line of cursive text, without reducing accuracy. """,2001-12-04,"The title of the patent is key character extraction and lexicon reduction for cursive text recognition and its abstract is "" a method, apparatus, and article of manufacture employing lexicon reduction using key characters and a neural network, for recognizing a line of cursive text. unambiguous parts of a cursive image, referred to as """"key characters,"""" are identified. if the level of confidence that a segment of a line of cursive text is a particular character is higher than a threshold, and is also sufficiently higher than the level of confidence of neighboring segments, then the character is designated as a key character candidate. key character candidates are then screened using geometric information. the key character candidates that pass the screening are designated key characters. two-stages of lexicon reduction are employed. the first stage of lexicon reduction uses a neural network to estimate a lower bound and an upper bound of the number of characters in a line of cursive text. lexicon entries having a total number of characters outside of the bounds are eliminated. for the second stage of lexicon reduction, the lexicon is fitter reduced by comparing character strings using the key characters, with lexicon entries. for each of the key characters in the character strings, it is determined whether there is a mismatch between the key character and characters in a corresponding search range in the lexicon entry. if the number of mismatches for all of the key characters in a search string is greater than (1+(the number of key characters in the search string/4)), then the lexicon entry is eliminated. accordingly, the invention advantageously accomplishes lexicon reduction, thereby decreasing the time required to recognize a line of cursive text, without reducing accuracy. "" dated 2001-12-04"
6330546,risk determination and management using predictive modeling and transaction profiles for individual transacting entities,an automated system and method detects fraudulent transactions using a predictive model such as a neural network to evaluate individual customer accounts and identify potentially fraudulent transactions based on learned relationships among known variables. the system may also output reason codes indicating relative contributions of various variables to a particular result. the system periodically monitors its performance and redevelops the model when performance drops below a predetermined level.,2001-12-11,The title of the patent is risk determination and management using predictive modeling and transaction profiles for individual transacting entities and its abstract is an automated system and method detects fraudulent transactions using a predictive model such as a neural network to evaluate individual customer accounts and identify potentially fraudulent transactions based on learned relationships among known variables. the system may also output reason codes indicating relative contributions of various variables to a particular result. the system periodically monitors its performance and redevelops the model when performance drops below a predetermined level. dated 2001-12-11
6330553,autonomic system for updating fuzzy neural network and control system using the fuzzy neural network,"an autonomic system for updating a fuzzy neural network includes a process of calculating an estimated value based on fuzzy inference by using a neural network structure, wherein a parameter to be adjusted or identified by fuzzy inference and outputted from the neural network is made to correspond to coupling loads which are updated by learning, i.e., fuzzy rules and membership functions are adjusted by learning. this system is characterized in that the addition and deletion of fuzzy rules are conducted based on changes in output errors in an autonomic manner, thereby effectively obtaining appropriate numbers of fuzzy rules optimal for an object such as a vehicle engine having strong non-linearity. fuzzy rules are formed by a combination of membership functions representing variables such as an engine speed and a throttle angle.",2001-12-11,"The title of the patent is autonomic system for updating fuzzy neural network and control system using the fuzzy neural network and its abstract is an autonomic system for updating a fuzzy neural network includes a process of calculating an estimated value based on fuzzy inference by using a neural network structure, wherein a parameter to be adjusted or identified by fuzzy inference and outputted from the neural network is made to correspond to coupling loads which are updated by learning, i.e., fuzzy rules and membership functions are adjusted by learning. this system is characterized in that the addition and deletion of fuzzy rules are conducted based on changes in output errors in an autonomic manner, thereby effectively obtaining appropriate numbers of fuzzy rules optimal for an object such as a vehicle engine having strong non-linearity. fuzzy rules are formed by a combination of membership functions representing variables such as an engine speed and a throttle angle. dated 2001-12-11"
6332105,neural network based automatic limit prediction and avoidance system and method,"a method for performance envelope boundary cueing for a vehicle control system comprises the steps of formulating a prediction system for a neural network and training the neural network to predict values of limited parameters as a function of current control positions and current vehicle operating conditions. the method further comprises the steps of applying the neural network to the control system of the vehicle, where the vehicle has capability for measuring current control positions and current vehicle operating conditions. the neural network generates a map of current control positions and vehicle operating conditions versus the limited parameters in a pre-determined vehicle operating condition. the method estimates critical control deflections from the current control positions required to drive the vehicle to a performance envelope boundary. finally, the method comprises the steps of communicating the critical control deflection to the vehicle control system; and driving the vehicle control system to provide a tactile cue to an operator of the vehicle as the control positions approach the critical control deflections.",2001-12-18,"The title of the patent is neural network based automatic limit prediction and avoidance system and method and its abstract is a method for performance envelope boundary cueing for a vehicle control system comprises the steps of formulating a prediction system for a neural network and training the neural network to predict values of limited parameters as a function of current control positions and current vehicle operating conditions. the method further comprises the steps of applying the neural network to the control system of the vehicle, where the vehicle has capability for measuring current control positions and current vehicle operating conditions. the neural network generates a map of current control positions and vehicle operating conditions versus the limited parameters in a pre-determined vehicle operating condition. the method estimates critical control deflections from the current control positions required to drive the vehicle to a performance envelope boundary. finally, the method comprises the steps of communicating the critical control deflection to the vehicle control system; and driving the vehicle control system to provide a tactile cue to an operator of the vehicle as the control positions approach the critical control deflections. dated 2001-12-18"
6332107,efficient high density train operations,""" the present invention provides methods for preventing low train voltages and managing interference, thereby improving the efficiency, reliability, and passenger comfort associated with commuter trains. an algorithm implementing neural network technology is used to predict low voltages before they occur. once voltages are predicted, then multiple trains can be controlled to prevent low voltage events. further, algorithms for managing inference are presented in the present invention. different types of interference problems are addressed in the present invention such as """"interference. during acceleration"""", """"interference near station stops"""", and """"interference during delay recovery."""" managing such interference avoids unnecessary brake/acceleration cycles during acceleration, immediately before station stops, and after substantial delays. algorithms are demonstrated to avoid oscillatory brake/acceleration cycles due to interference and to smooth the trajectories of closely following trains. this is achieved by maintaining sufficient following distances to avoid unnecessary braking/accelerating. these methods generate smooth train trajectories, making for a more comfortable ride, and improve train motor reliability by avoiding unnecessary mode-changes between propulsion and braking. these algorithms can also have a favorable impact on traction power system requirements and energy consumption. """,2001-12-18,"The title of the patent is efficient high density train operations and its abstract is "" the present invention provides methods for preventing low train voltages and managing interference, thereby improving the efficiency, reliability, and passenger comfort associated with commuter trains. an algorithm implementing neural network technology is used to predict low voltages before they occur. once voltages are predicted, then multiple trains can be controlled to prevent low voltage events. further, algorithms for managing inference are presented in the present invention. different types of interference problems are addressed in the present invention such as """"interference. during acceleration"""", """"interference near station stops"""", and """"interference during delay recovery."""" managing such interference avoids unnecessary brake/acceleration cycles during acceleration, immediately before station stops, and after substantial delays. algorithms are demonstrated to avoid oscillatory brake/acceleration cycles due to interference and to smooth the trajectories of closely following trains. this is achieved by maintaining sufficient following distances to avoid unnecessary braking/accelerating. these methods generate smooth train trajectories, making for a more comfortable ride, and improve train motor reliability by avoiding unnecessary mode-changes between propulsion and braking. these algorithms can also have a favorable impact on traction power system requirements and energy consumption. "" dated 2001-12-18"
6332137,parallel associative learning memory for a standalone hardwired recognition system,a recognition system comprises at least two field-programmable logic array devices connected to a common vector-input port of an array of a zero-instruction-set computers. each field-programmable logic array device is configured to preprocess data from different respective media inputs and provide feature extraction vectors to the common vector-input port. neural networks within the zero-instruction-set computer recognize the input patterns by comparing in parallel their vectors with those stored in each neural network cell. a variety of recognition jobs are made possible by changing the programming on-the-fly of the field-programmable logic array devices to suit each new job.,2001-12-18,The title of the patent is parallel associative learning memory for a standalone hardwired recognition system and its abstract is a recognition system comprises at least two field-programmable logic array devices connected to a common vector-input port of an array of a zero-instruction-set computers. each field-programmable logic array device is configured to preprocess data from different respective media inputs and provide feature extraction vectors to the common vector-input port. neural networks within the zero-instruction-set computer recognize the input patterns by comparing in parallel their vectors with those stored in each neural network cell. a variety of recognition jobs are made possible by changing the programming on-the-fly of the field-programmable logic array devices to suit each new job. dated 2001-12-18
6334121,usage pattern based user authenticator,"a usage based pattern authenticator for monitoring and reporting on user usage patterns in an operating system using a set of security rules and user usage patterns. this computer system security tool authenticates users at the operating system level in multi-user operating systems. it supports system administrators in limiting the ability of unauthorized users to disrupt system operations using a neural network and set of rules to track usage patterns and flag suspicious activity on the system. the data collection mode collects and stores usage patterns of authenticated users. the training mode trains an artificial neural network and sets the interconnection weights of the network. the production mode monitors and reports on usage patterns, and optionally performs automatic responses when confronted with non-authenticated users.",2001-12-25,"The title of the patent is usage pattern based user authenticator and its abstract is a usage based pattern authenticator for monitoring and reporting on user usage patterns in an operating system using a set of security rules and user usage patterns. this computer system security tool authenticates users at the operating system level in multi-user operating systems. it supports system administrators in limiting the ability of unauthorized users to disrupt system operations using a neural network and set of rules to track usage patterns and flag suspicious activity on the system. the data collection mode collects and stores usage patterns of authenticated users. the training mode trains an artificial neural network and sets the interconnection weights of the network. the production mode monitors and reports on usage patterns, and optionally performs automatic responses when confronted with non-authenticated users. dated 2001-12-25"
6335913,disk memory device and disk read-out signal processor,"a disk memory device comprises a head which reads a read-out signal from a disk, an amplifier which amplifies an analog signal waveform of the read-out signal read from the head, a filter which decreases a noise of the read-out signal output from the amplifier, an a/d converter which converts the read-out signal of which noise is decreased by the filter into a digital signal including a waveform distortion component, and a neural network type signal processing circuit which detects a binarized data from the digital signal.",2002-01-01,"The title of the patent is disk memory device and disk read-out signal processor and its abstract is a disk memory device comprises a head which reads a read-out signal from a disk, an amplifier which amplifies an analog signal waveform of the read-out signal read from the head, a filter which decreases a noise of the read-out signal output from the amplifier, an a/d converter which converts the read-out signal of which noise is decreased by the filter into a digital signal including a waveform distortion component, and a neural network type signal processing circuit which detects a binarized data from the digital signal. dated 2002-01-01"
6337568,system and method for enhanced vertical resolution magnetic resonance imaging logs,"an interpretation method and system for nmr echo-train data in thinly laminated sequences. the invention uses geological information obtained at higher vertical resolution, such as using electric micro imaging, to enhance the vertical resolution of echo-train data, and thus avoids log interpretations in which the hydrocarbon potential of the formation can be misread because low resolution logs tend to provide an average description of the formation. such averaging is especially problematic in thinly laminated sequences that consist of highly permeable and porous sand layers and less permeable silt or essentially impermeable shale layers. in a preferred embodiment, using the additional high-resolution formation information one can estimate the typical t2-spectra of lithological laminae, and significantly enhance the permeability estimate in the laminated sequences. in another aspect the system and method of the preferred embodiment use neural network(s) to further enhance the resolution of a particular log measurement. the method and system are applicable to any temporal data from other logging tools, such as the thermal neutron decay log and others. the system and method enable proper evaluation of the high potential of thinly laminated formations, which may otherwise be overlooked as low permeable formations.",2002-01-08,"The title of the patent is system and method for enhanced vertical resolution magnetic resonance imaging logs and its abstract is an interpretation method and system for nmr echo-train data in thinly laminated sequences. the invention uses geological information obtained at higher vertical resolution, such as using electric micro imaging, to enhance the vertical resolution of echo-train data, and thus avoids log interpretations in which the hydrocarbon potential of the formation can be misread because low resolution logs tend to provide an average description of the formation. such averaging is especially problematic in thinly laminated sequences that consist of highly permeable and porous sand layers and less permeable silt or essentially impermeable shale layers. in a preferred embodiment, using the additional high-resolution formation information one can estimate the typical t2-spectra of lithological laminae, and significantly enhance the permeability estimate in the laminated sequences. in another aspect the system and method of the preferred embodiment use neural network(s) to further enhance the resolution of a particular log measurement. the method and system are applicable to any temporal data from other logging tools, such as the thermal neutron decay log and others. the system and method enable proper evaluation of the high potential of thinly laminated formations, which may otherwise be overlooked as low permeable formations. dated 2002-01-08"
6337654,a-scan isar classification system and method therefor,"a target recognition system and method wherein only target amplitude data, i.e., coherent a-scan data, is interrogated for target recognition. target aspect angle is ignored within the angular segmentation of the feature library without degrading classification performance. observed signature characteristics are collected at various aspect angles and through unknown and arbitrary roll, pitch and yaw motions of each anticipated target and provided to a neural network as training sets. the neural network forms feature vectors for each target class which are useful for valid classification comparisons in all sea states, especially in calm and littoral waters. these feature vectors are useful for valid classification comparisons over at least 30 degrees of target aspect angle.",2002-01-08,"The title of the patent is a-scan isar classification system and method therefor and its abstract is a target recognition system and method wherein only target amplitude data, i.e., coherent a-scan data, is interrogated for target recognition. target aspect angle is ignored within the angular segmentation of the feature library without degrading classification performance. observed signature characteristics are collected at various aspect angles and through unknown and arbitrary roll, pitch and yaw motions of each anticipated target and provided to a neural network as training sets. the neural network forms feature vectors for each target class which are useful for valid classification comparisons in all sea states, especially in calm and littoral waters. these feature vectors are useful for valid classification comparisons over at least 30 degrees of target aspect angle. dated 2002-01-08"
6338052,method for optimizing matching network of semiconductor process apparatus,"a method for optimizing matching network between an output impedance and an input impedance in a semiconductor process apparatus is disclosed. the method includes the steps of: providing a neural network capable of being trained through repeated learning; training the neural network from previously performed process conditions; setting up an initial value; comparing the initial value with a theoretically calculated value, to obtain error between the values; and repeating the training, setting, and comparing steps until the error becomes zero.",2002-01-08,"The title of the patent is method for optimizing matching network of semiconductor process apparatus and its abstract is a method for optimizing matching network between an output impedance and an input impedance in a semiconductor process apparatus is disclosed. the method includes the steps of: providing a neural network capable of being trained through repeated learning; training the neural network from previously performed process conditions; setting up an initial value; comparing the initial value with a theoretically calculated value, to obtain error between the values; and repeating the training, setting, and comparing steps until the error becomes zero. dated 2002-01-08"
6341275,programmable and expandable hamming neural network circuit,a hamming neural network circuit which can be programmed and expanded is disclosed. the hamming neural network includes an i/o circuit for inputting and outputting a plurality of standard patterns. a bi-directional transmission gate array is connected to the i/o circuit and controlled by a programming signal for transmitting the standard patterns. a plurality of standard pattern memory units is connected to the bi-directional transmission gate array for storing the plurality of standard patterns respectively. an address decoder is connected to the plurality of standard pattern memory units for addressing one of the plurality of standard pattern memory units. a plurality of pattern matching calculation circuit units are respectively connected to the plurality of standard pattern memory units for generating a plurality of matching rates between a to-be-recognized pattern and the plurality of standard patterns. an expandable matching rate comparing circuit is provided for comparing and sorting said plurality of matching rates.,2002-01-22,The title of the patent is programmable and expandable hamming neural network circuit and its abstract is a hamming neural network circuit which can be programmed and expanded is disclosed. the hamming neural network includes an i/o circuit for inputting and outputting a plurality of standard patterns. a bi-directional transmission gate array is connected to the i/o circuit and controlled by a programming signal for transmitting the standard patterns. a plurality of standard pattern memory units is connected to the bi-directional transmission gate array for storing the plurality of standard patterns respectively. an address decoder is connected to the plurality of standard pattern memory units for addressing one of the plurality of standard pattern memory units. a plurality of pattern matching calculation circuit units are respectively connected to the plurality of standard pattern memory units for generating a plurality of matching rates between a to-be-recognized pattern and the plurality of standard patterns. an expandable matching rate comparing circuit is provided for comparing and sorting said plurality of matching rates. dated 2002-01-22
6347297,matrix quantization with vector quantization error compensation and neural network postprocessing for robust speech recognition,"a speech recognition system utilizes both matrix and vector quantizers as front ends to a second stage speech classifier such as hidden markov models (hmms) and utilizes neural network postprocessing to, for example, improve speech recognition performance. matrix quantization exploits the &#8220;evolution&#8221; of the speech short-term spectral envelopes as well as frequency domain information, and vector quantization (vq) primarily operates on frequency domain information. time domain information may be substantially limited which may introduce error into the matrix quantization, and the vq may provide error compensation. the matrix and vector quantizers may split spectral subbands to target selected frequencies for enhanced processing and may use fuzzy associations to develop fuzzy observation sequence data. a mixer provides a variety of input data to the neural network for classification determination. the neural network's ability to analyze the input data generally enhances recognition accuracy. fuzzy operators may be utilized to reduce quantization error. multiple codebooks may also be combined to form single respective codebooks for split matrix and split vector quantization to reduce processing resources demand.",2002-02-12,"The title of the patent is matrix quantization with vector quantization error compensation and neural network postprocessing for robust speech recognition and its abstract is a speech recognition system utilizes both matrix and vector quantizers as front ends to a second stage speech classifier such as hidden markov models (hmms) and utilizes neural network postprocessing to, for example, improve speech recognition performance. matrix quantization exploits the &#8220;evolution&#8221; of the speech short-term spectral envelopes as well as frequency domain information, and vector quantization (vq) primarily operates on frequency domain information. time domain information may be substantially limited which may introduce error into the matrix quantization, and the vq may provide error compensation. the matrix and vector quantizers may split spectral subbands to target selected frequencies for enhanced processing and may use fuzzy associations to develop fuzzy observation sequence data. a mixer provides a variety of input data to the neural network for classification determination. the neural network's ability to analyze the input data generally enhances recognition accuracy. fuzzy operators may be utilized to reduce quantization error. multiple codebooks may also be combined to form single respective codebooks for split matrix and split vector quantization to reduce processing resources demand. dated 2002-02-12"
6347309,circuits and method for shaping the influence field of neurons and neural networks resulting therefrom,"the improved neural network of the present invention results from the combination of a dedicated logic block with a conventional neural network based upon a mapping of the input space usually employed to classify an input data by computing the distance between said input data and prototypes memorized therein. the improved neural network is able to classify an input data, for instance, represented by a vector a even when some of its components are noisy or unknown during either the learning or the recognition phase. to that end, influence fields of various and different shapes are created for each neuron of the conventional neural network. the logic block transforms at least some of the n components (a1, . . . , an) of the input vector a into the m components (v1, . . . , vm) of a network input vector v according to a linear or non-linear transform function f. in turn, vector v is applied as the input data to said conventional neural network. the transform function f is such that certain components of vector v are not modified, e.g. vk=aj, while other components are transformed as mentioned above, e.g. vi=fi(a1, . . . , an). in addition, one (or more) component of vector v can be used to compensate an offset that is present in the distance evaluation of vector v. because, the logic block is placed in front of the said conventional neural network any modification thereof is avoided.",2002-02-12,"The title of the patent is circuits and method for shaping the influence field of neurons and neural networks resulting therefrom and its abstract is the improved neural network of the present invention results from the combination of a dedicated logic block with a conventional neural network based upon a mapping of the input space usually employed to classify an input data by computing the distance between said input data and prototypes memorized therein. the improved neural network is able to classify an input data, for instance, represented by a vector a even when some of its components are noisy or unknown during either the learning or the recognition phase. to that end, influence fields of various and different shapes are created for each neuron of the conventional neural network. the logic block transforms at least some of the n components (a1, . . . , an) of the input vector a into the m components (v1, . . . , vm) of a network input vector v according to a linear or non-linear transform function f. in turn, vector v is applied as the input data to said conventional neural network. the transform function f is such that certain components of vector v are not modified, e.g. vk=aj, while other components are transformed as mentioned above, e.g. vi=fi(a1, . . . , an). in addition, one (or more) component of vector v can be used to compensate an offset that is present in the distance evaluation of vector v. because, the logic block is placed in front of the said conventional neural network any modification thereof is avoided. dated 2002-02-12"
6349231,method and apparatus for will determination and bio-signal control,"an apparatus and a method for automatically determining the present will of a human subject. the characteristic values of the subject are detected and output signals corresponding to the detected characteristic values are produced, amplified and digitized. a set of state variables for each selected frequency sub-band of a selected frequency band for each of the output signals is determined. sets of reference weights and sets of reference biases for a neural network from sets of state reference variables corresponding to known wills are formed. each of the sets of state variables, the sets of reference weights and the sets of reference biases are applied to the neural network to determine present will of the subject. the present will of the subject can be displayed or used to control an external device, such as a robot.",2002-02-19,"The title of the patent is method and apparatus for will determination and bio-signal control and its abstract is an apparatus and a method for automatically determining the present will of a human subject. the characteristic values of the subject are detected and output signals corresponding to the detected characteristic values are produced, amplified and digitized. a set of state variables for each selected frequency sub-band of a selected frequency band for each of the output signals is determined. sets of reference weights and sets of reference biases for a neural network from sets of state reference variables corresponding to known wills are formed. each of the sets of state variables, the sets of reference weights and the sets of reference biases are applied to the neural network to determine present will of the subject. the present will of the subject can be displayed or used to control an external device, such as a robot. dated 2002-02-19"
6349293,method for optimization of a fuzzy neural network,"optimization of a fnn (fnn)-based controller is described. the optimization includes selecting which input signals will be used by the fnn to compute a desired control output. output parameters are identified and computed by fuzzy reasoning using a neural network. adjustment of fuzzy rules and/or membership functions for the fnn is provided by a learning process. the learning process includes selecting candidate input data signals (e.g. selecting candidate sensor signals) as inputs for the fnn. the input data is categorized and coded into a chromosome structure for use by a genetic algorithm. the genetic algorithm is used to select an optimum chromosome (individual). the optimum chromosome specifies the number(s) and type(s) of input data signals for the fnn so as to optimize the operation of the fnn-based control system. the optimized fnn-based control system can be used in many control environments, including control of an internal combustion engine.",2002-02-19,"The title of the patent is method for optimization of a fuzzy neural network and its abstract is optimization of a fnn (fnn)-based controller is described. the optimization includes selecting which input signals will be used by the fnn to compute a desired control output. output parameters are identified and computed by fuzzy reasoning using a neural network. adjustment of fuzzy rules and/or membership functions for the fnn is provided by a learning process. the learning process includes selecting candidate input data signals (e.g. selecting candidate sensor signals) as inputs for the fnn. the input data is categorized and coded into a chromosome structure for use by a genetic algorithm. the genetic algorithm is used to select an optimum chromosome (individual). the optimum chromosome specifies the number(s) and type(s) of input data signals for the fnn so as to optimize the operation of the fnn-based control system. the optimized fnn-based control system can be used in many control environments, including control of an internal combustion engine. dated 2002-02-19"
6351273,system and methods for controlling automatic scrolling of information on a display or screen,"a system for controlling the automatic scrolling of information on a computer display. the system includes a computer display, a computer gimbaled sensor for tracking the position of the user's head and user's eye, and a scroll activating interface algorithm using a neural network to find screen gaze coordinates implemented by the computer. a scrolling function is performed based upon the screen gaze coordinates of the user's eye relative t activation area(s) on the display. the gimbaled sensor system contains a platform mounted at the top of the display. the gimbaled sensor system tracks the user's head and eye allowing the user to be free from attachments while the gimbaled sensor system is tracking, still allowing the user to freely move his head. a method of controlling automatic scrolling of information on a display includes the steps of finding a screen gaze coordinate on the display of the user determining whether the screen gaze coordinate is within at least one activated control region, and activating scrolling to provide a desired display of information when the gaze direction is within at least one activated control region. in one embodiment, the control regions are defined as upper control region, lower region, right region and left region for controlling the scrolling respectively in downward, upward, leftward and rightward directions. in another embodiment, control regions are defined by concentric rings for maintaining the stationary position of the information or controlling the scrolling of the information towards the center of the display or screen.",2002-02-26,"The title of the patent is system and methods for controlling automatic scrolling of information on a display or screen and its abstract is a system for controlling the automatic scrolling of information on a computer display. the system includes a computer display, a computer gimbaled sensor for tracking the position of the user's head and user's eye, and a scroll activating interface algorithm using a neural network to find screen gaze coordinates implemented by the computer. a scrolling function is performed based upon the screen gaze coordinates of the user's eye relative t activation area(s) on the display. the gimbaled sensor system contains a platform mounted at the top of the display. the gimbaled sensor system tracks the user's head and eye allowing the user to be free from attachments while the gimbaled sensor system is tracking, still allowing the user to freely move his head. a method of controlling automatic scrolling of information on a display includes the steps of finding a screen gaze coordinate on the display of the user determining whether the screen gaze coordinate is within at least one activated control region, and activating scrolling to provide a desired display of information when the gaze direction is within at least one activated control region. in one embodiment, the control regions are defined as upper control region, lower region, right region and left region for controlling the scrolling respectively in downward, upward, leftward and rightward directions. in another embodiment, control regions are defined by concentric rings for maintaining the stationary position of the information or controlling the scrolling of the information towards the center of the display or screen. dated 2002-02-26"
6351711,gps navigation system using neural networks,a gps receiver includes a satellite receiver/processor having an input that receives input signals from at least one gps satellite. the output of the receiver/processor provides satellite-related navigation information. a neural network receives the satellite-related information to obtain an output signal representative of receiver-related navigation information. the neural network includes a first node layer connected to a second node layer through a first connection layer and a third node layer connected to the second node layer through a second connection layer. each of the node layers comprises a plurality of neurons.,2002-02-26,The title of the patent is gps navigation system using neural networks and its abstract is a gps receiver includes a satellite receiver/processor having an input that receives input signals from at least one gps satellite. the output of the receiver/processor provides satellite-related navigation information. a neural network receives the satellite-related information to obtain an output signal representative of receiver-related navigation information. the neural network includes a first node layer connected to a second node layer through a first connection layer and a third node layer connected to the second node layer through a second connection layer. each of the node layers comprises a plurality of neurons. dated 2002-02-26
6351713,distributed stress wave analysis system,"a distributed stress wave analysis system is disclosed for detecting structure borne sounds cause by friction. the detected information is processed using feature extraction and neural network artificial intelligence software. the system consists of stress wave sensors, interconnect cables, and preferably three modules: (1) distributed processing units, (2) maintenance advisory panel, and (3) laptop computer. a derived stress wave pulse train which is independent of background levels of vibration and audible noise is used to extract signature features, which when processed by neural networks of polynomial equations, characterize the mechanical health of the monitored components. the system includes an adjustable data fusion architecture to optimize indication thresholds, maximize fault detection probability, and minimize false alarms.",2002-02-26,"The title of the patent is distributed stress wave analysis system and its abstract is a distributed stress wave analysis system is disclosed for detecting structure borne sounds cause by friction. the detected information is processed using feature extraction and neural network artificial intelligence software. the system consists of stress wave sensors, interconnect cables, and preferably three modules: (1) distributed processing units, (2) maintenance advisory panel, and (3) laptop computer. a derived stress wave pulse train which is independent of background levels of vibration and audible noise is used to extract signature features, which when processed by neural networks of polynomial equations, characterize the mechanical health of the monitored components. the system includes an adjustable data fusion architecture to optimize indication thresholds, maximize fault detection probability, and minimize false alarms. dated 2002-02-26"
6353766,method for generating control parameters from a response signal of a controlled system and system for adaptive setting of a pid controller,setting parameters of a pid controller are obtained by feeding a step signal or another input signal to an assigned controlled system. the response signal emitted by the controlled system is sampled and the characteristics of the bode diagram are generated from the input signal and the response signal by using a smoothing method and elementary correspondences. the characteristics are normalized and input values are derived therefrom for a neural network which is trained on the properties of the controlled systems. the neural network directly generates the setting parameters for the controller.,2002-03-05,The title of the patent is method for generating control parameters from a response signal of a controlled system and system for adaptive setting of a pid controller and its abstract is setting parameters of a pid controller are obtained by feeding a step signal or another input signal to an assigned controlled system. the response signal emitted by the controlled system is sampled and the characteristics of the bode diagram are generated from the input signal and the response signal by using a smoothing method and elementary correspondences. the characteristics are normalized and input values are derived therefrom for a neural network which is trained on the properties of the controlled systems. the neural network directly generates the setting parameters for the controller. dated 2002-03-05
6353815,statistically qualified neuro-analytic failure detection method and system,"an apparatus and method for monitoring a process involve development and application of a statistically qualified neuro-analytic (sqna) model to accurately and reliably identify process change. the development of the sqna model is accomplished in two stages: deterministic model adaption and stochastic model modification of the deterministic model adaptation. deterministic model adaption involves formulating an analytic model of the process representing known process characteristics, augmenting the analytic model with a neural network that captures unknown process characteristics, and training the resulting neuro-analytic model by adjusting the neural network weights according to a unique scaled equation error minimization technique. stochastic model modification involves qualifying any remaining uncertainty in the trained neuro-analytic model by formulating a likelihood function, given an error propagation equation, for computing the probability that the neuro-analytic model generates measured process output. preferably, the developed sqna model is validated using known sequential probability ratio tests and applied to the process as an on-line monitoring system. illustrative of the method and apparatus, the method is applied to a peristaltic pump system.",2002-03-05,"The title of the patent is statistically qualified neuro-analytic failure detection method and system and its abstract is an apparatus and method for monitoring a process involve development and application of a statistically qualified neuro-analytic (sqna) model to accurately and reliably identify process change. the development of the sqna model is accomplished in two stages: deterministic model adaption and stochastic model modification of the deterministic model adaptation. deterministic model adaption involves formulating an analytic model of the process representing known process characteristics, augmenting the analytic model with a neural network that captures unknown process characteristics, and training the resulting neuro-analytic model by adjusting the neural network weights according to a unique scaled equation error minimization technique. stochastic model modification involves qualifying any remaining uncertainty in the trained neuro-analytic model by formulating a likelihood function, given an error propagation equation, for computing the probability that the neuro-analytic model generates measured process output. preferably, the developed sqna model is validated using known sequential probability ratio tests and applied to the process as an on-line monitoring system. illustrative of the method and apparatus, the method is applied to a peristaltic pump system. dated 2002-03-05"
6353816,"method, apparatus and storage medium configured to analyze predictive accuracy of a trained neural network","a neural network analysis apparatus has an input means 1 for inputting multilinear functions that represent the various units of a trained neural network, a storage means 2 therefor, term generators 3 that successively generate terms of a boolean function used for approximation from the coefficients of each term of the multilinear function input for each unit, a judgment condition storage section 22 into which is stored the conditions for generating each term of the boolean function to be used in the approximation, an apparatus 5 that links the generated boolean function terms, and a boolean function synthesizing apparatus 6 for representing the output unit by input variables. the term generators 3 each have four sub-apparatuses: a data-limiting apparatus 31, a minimum value calculation apparatus 32, a non-important, attribute processing apparatus 33, and a function judgment apparatus 34.",2002-03-05,"The title of the patent is method, apparatus and storage medium configured to analyze predictive accuracy of a trained neural network and its abstract is a neural network analysis apparatus has an input means 1 for inputting multilinear functions that represent the various units of a trained neural network, a storage means 2 therefor, term generators 3 that successively generate terms of a boolean function used for approximation from the coefficients of each term of the multilinear function input for each unit, a judgment condition storage section 22 into which is stored the conditions for generating each term of the boolean function to be used in the approximation, an apparatus 5 that links the generated boolean function terms, and a boolean function synthesizing apparatus 6 for representing the output unit by input variables. the term generators 3 each have four sub-apparatuses: a data-limiting apparatus 31, a minimum value calculation apparatus 32, a non-important, attribute processing apparatus 33, and a function judgment apparatus 34. dated 2002-03-05"
6356884,device system for the autonomous generation of useful information,"an artificial neural network-based system and method for determining desired concepts and relationships within a predefined field of endeavor, including a neural network portion, which neural network portion includes an artificial neural network that has been previously trained in accordance with a set of given training exemplars, a monitor portion associated with the neural network portion to observe the data outputs produced by the previously trained artificial neural network, and a perturbation portion for perturbing the neural network portion to effect changes, subject to design constraints of the artificial neural network that remain unperturbed, in the outputs produced by the neural network portion, the perturbation portion operable such that production of an output by the neural network portion thereafter effects a perturbation of the neural network portion by the perturbation portion, the monitor portion responsive to detection of the data outputs being produced by the previously trained neural network, whereby the system is operable to derive over a period of time a plurality of input/perturbation/output mapping relationships that differ from the input/perturbation/mapping relationships of the training exemplars.",2002-03-12,"The title of the patent is device system for the autonomous generation of useful information and its abstract is an artificial neural network-based system and method for determining desired concepts and relationships within a predefined field of endeavor, including a neural network portion, which neural network portion includes an artificial neural network that has been previously trained in accordance with a set of given training exemplars, a monitor portion associated with the neural network portion to observe the data outputs produced by the previously trained artificial neural network, and a perturbation portion for perturbing the neural network portion to effect changes, subject to design constraints of the artificial neural network that remain unperturbed, in the outputs produced by the neural network portion, the perturbation portion operable such that production of an output by the neural network portion thereafter effects a perturbation of the neural network portion by the perturbation portion, the monitor portion responsive to detection of the data outputs being produced by the previously trained neural network, whereby the system is operable to derive over a period of time a plurality of input/perturbation/output mapping relationships that differ from the input/perturbation/mapping relationships of the training exemplars. dated 2002-03-12"
6359587,wireless communication system and method and system for detection of position of radio mobile station,"a method of detecting a position of a radio mobile station in radiocommiunications, which is capable of accurately and simply finding the position of the mobile station. at a measuring point the mobile station measures the reception radio strength levels from a plurality of base stations and conveys the measurement results through the base station to a control station. the control station learns, through a neural network, the correlation between the reception radio strength levels and the position of the mobile station on the basis of the measurement results at a plurality of measuring points and the positions of the measuring points. subsequently, when the mobile station communicates to the control station the reception radio strength levels measured at an arbitrary point, the control station estimates the position of the mobile station, causing those measurement results, on the basis of the correlation obtained through the learning.",2002-03-19,"The title of the patent is wireless communication system and method and system for detection of position of radio mobile station and its abstract is a method of detecting a position of a radio mobile station in radiocommiunications, which is capable of accurately and simply finding the position of the mobile station. at a measuring point the mobile station measures the reception radio strength levels from a plurality of base stations and conveys the measurement results through the base station to a control station. the control station learns, through a neural network, the correlation between the reception radio strength levels and the position of the mobile station on the basis of the measurement results at a plurality of measuring points and the positions of the measuring points. subsequently, when the mobile station communicates to the control station the reception radio strength levels measured at an arbitrary point, the control station estimates the position of the mobile station, causing those measurement results, on the basis of the correlation obtained through the learning. dated 2002-03-19"
6362783,wireless communication system and method and system for detection of position of radio mobile station,"a method of detecting a position of a radio mobile station in radiocommunications, which is capable of accurately and simply finding the position of the mobile station. at a measuring point the mobile station measures the reception radio strength levels from a plurality of base stations and conveys the measurement results through the base station to a control station. the control station learns, through a neural network, the correlation between the reception radio strength levels and the position of the mobile station on the basis of the measurement results at a plurality of measuring points and the positions of the measuring points. subsequently, when the mobile station communicates to the control station the reception radio strength levels measured at an arbitrary point, the control station estimates the position of the mobile station, causing those measurement results, on the basis of the correlation obtained through the learning.",2002-03-26,"The title of the patent is wireless communication system and method and system for detection of position of radio mobile station and its abstract is a method of detecting a position of a radio mobile station in radiocommunications, which is capable of accurately and simply finding the position of the mobile station. at a measuring point the mobile station measures the reception radio strength levels from a plurality of base stations and conveys the measurement results through the base station to a control station. the control station learns, through a neural network, the correlation between the reception radio strength levels and the position of the mobile station on the basis of the measurement results at a plurality of measuring points and the positions of the measuring points. subsequently, when the mobile station communicates to the control station the reception radio strength levels measured at an arbitrary point, the control station estimates the position of the mobile station, causing those measurement results, on the basis of the correlation obtained through the learning. dated 2002-03-26"
6363289,residual activation neural network,"a plant (72) is operable to receive control inputs c(t) and provide an output y(t). the plant (72) has associated therewith state variables s(t) that are not variable. a control network (74) is provided that accurately models the plant (72). the output of the control network (74) provides a predicted output which is combined with a desired output to generate an error. this error is back propagated through an inverse control network (76), which is the inverse of the control network (74) to generate a control error signal that is input to a distributed control system (73) to vary the control inputs to the plant (72) in order to change the output y(t) to meet the desired output. the control network (74) is comprised of a first network net 1 that is operable to store a representation of the dependency of the control variables on the state variables. the predicted result is subtracted from the actual state variable input and stored as a residual in a residual layer (102). the output of the residual layer (102) is input to a hidden layer (108) which also receives the control inputs to generate a predicted output in an output layer (106). during back propagation of error, the residual values in the residual layer (102) are latched and only the control inputs allowed to vary.",2002-03-26,"The title of the patent is residual activation neural network and its abstract is a plant (72) is operable to receive control inputs c(t) and provide an output y(t). the plant (72) has associated therewith state variables s(t) that are not variable. a control network (74) is provided that accurately models the plant (72). the output of the control network (74) provides a predicted output which is combined with a desired output to generate an error. this error is back propagated through an inverse control network (76), which is the inverse of the control network (74) to generate a control error signal that is input to a distributed control system (73) to vary the control inputs to the plant (72) in order to change the output y(t) to meet the desired output. the control network (74) is comprised of a first network net 1 that is operable to store a representation of the dependency of the control variables on the state variables. the predicted result is subtracted from the actual state variable input and stored as a residual in a residual layer (102). the output of the residual layer (102) is input to a hidden layer (108) which also receives the control inputs to generate a predicted output in an output layer (106). during back propagation of error, the residual values in the residual layer (102) are latched and only the control inputs allowed to vary. dated 2002-03-26"
6363328,software controlled meat probe for use in determining meat tenderness,"a data processor used in the overall process of determining meat tenderness which receives, analyses and graphically displays in a dynamic format collected fluorescence emitted by connective tissue as a meat probe passes by such tissue during either insertion or removal of the meat probe from the meat. the data processor also collects and calculates feature variables based on the data collected during the insertion and removal of the meat probe, and through the use of artificial intelligence and artificial neural network processing can be taught to recognize patterns in the meat probe data indicative of tenderness. the data processor also performs an analysis of probe data to make a prediction of meat tenderness.",2002-03-26,"The title of the patent is software controlled meat probe for use in determining meat tenderness and its abstract is a data processor used in the overall process of determining meat tenderness which receives, analyses and graphically displays in a dynamic format collected fluorescence emitted by connective tissue as a meat probe passes by such tissue during either insertion or removal of the meat probe from the meat. the data processor also collects and calculates feature variables based on the data collected during the insertion and removal of the meat probe, and through the use of artificial intelligence and artificial neural network processing can be taught to recognize patterns in the meat probe data indicative of tenderness. the data processor also performs an analysis of probe data to make a prediction of meat tenderness. dated 2002-03-26"
6363333,method of classifying statistical dependency of a measurable series of statistical values,"a time series that is established by a measured signal of a dynamic system, for example a quotation curve on the stock market, is modelled according to its probability density in order to be able to make a prediction of future values. a non-linear markov process of the order m is suited for describing the conditioned probability densities. a neural network is trained according to the probabilities of the markov process using the maximum likelihood principle, which is a training rule for maximizing the product of probabilities. the neural network predicts a value in the future for a prescribable number of values m from the past of the signal to be predicted. a number of steps in the future can be predicted by iteration. the order m of the non-linear markov process, which corresponds to the number of values from the past that are important in the modelling of the conditioned probability densities, serves as parameter for improving the probability of the prediction.",2002-03-26,"The title of the patent is method of classifying statistical dependency of a measurable series of statistical values and its abstract is a time series that is established by a measured signal of a dynamic system, for example a quotation curve on the stock market, is modelled according to its probability density in order to be able to make a prediction of future values. a non-linear markov process of the order m is suited for describing the conditioned probability densities. a neural network is trained according to the probabilities of the markov process using the maximum likelihood principle, which is a training rule for maximizing the product of probabilities. the neural network predicts a value in the future for a prescribable number of values m from the past of the signal to be predicted. a number of steps in the future can be predicted by iteration. the order m of the non-linear markov process, which corresponds to the number of values from the past that are important in the modelling of the conditioned probability densities, serves as parameter for improving the probability of the prediction. dated 2002-03-26"
6366236,neural network radar processor,"a neural network radar processor (10) comprises a multilayer perceptron neural network (100.1) comprising an input layer (102), a second layer (122), and at least a third layer (124), wherein each layer has a plurality of nodes (108), and respective subsets of nodes (108) of the second (122) and third (124) layers are interconnected so as to form mutually exclusive subnetworks (120). in-phase and quadrature phase time series from a sampled down-converted fmcw radar signal (19) are applied to the input layer, and the neural network (100) is trained so that the nodes of the output layer (106) are responsive to targets in corresponding range cells, and different subnetworks (120) are responsive to respectively different non-overlapping sets of target ranges. the neural network is trained with signals that are germane to an fmcw radar, including a wide range of target scenarios as well as leakage signals, dc bias signals, and background clutter signals.",2002-04-02,"The title of the patent is neural network radar processor and its abstract is a neural network radar processor (10) comprises a multilayer perceptron neural network (100.1) comprising an input layer (102), a second layer (122), and at least a third layer (124), wherein each layer has a plurality of nodes (108), and respective subsets of nodes (108) of the second (122) and third (124) layers are interconnected so as to form mutually exclusive subnetworks (120). in-phase and quadrature phase time series from a sampled down-converted fmcw radar signal (19) are applied to the input layer, and the neural network (100) is trained so that the nodes of the output layer (106) are responsive to targets in corresponding range cells, and different subnetworks (120) are responsive to respectively different non-overlapping sets of target ranges. the neural network is trained with signals that are germane to an fmcw radar, including a wide range of target scenarios as well as leakage signals, dc bias signals, and background clutter signals. dated 2002-04-02"
6366885,speech driven lip synthesis using viseme based hidden markov models,"a method of speech driven lip synthesis which applies viseme based training models to units of visual speech. the audio data is grouped into a smaller number of visually distinct visemes rather than the larger number of phonemes. these visemes then form the basis for a hidden markov model (hmm) state sequence or the output nodes of a neural network. during the training phase, audio and visual features are extracted from input speech, which is then aligned according to the apparent viseme sequence with the corresponding audio features being used to calculate the hmm state output probabilities or the output of the neutral network. during the synthesis phase, the acoustic input is aligned with the most likely viseme hmm sequence (in the case of an hmm based model) or with the nodes of the network (in the case of a neural network based system), which is then used for animation.",2002-04-02,"The title of the patent is speech driven lip synthesis using viseme based hidden markov models and its abstract is a method of speech driven lip synthesis which applies viseme based training models to units of visual speech. the audio data is grouped into a smaller number of visually distinct visemes rather than the larger number of phonemes. these visemes then form the basis for a hidden markov model (hmm) state sequence or the output nodes of a neural network. during the training phase, audio and visual features are extracted from input speech, which is then aligned according to the apparent viseme sequence with the corresponding audio features being used to calculate the hmm state output probabilities or the output of the neutral network. during the synthesis phase, the acoustic input is aligned with the most likely viseme hmm sequence (in the case of an hmm based model) or with the nodes of the network (in the case of a neural network based system), which is then used for animation. dated 2002-04-02"
6366896,adaptive agent using neural network,an adaptive agent including an artificial neural network having a plurality of input nodes for receiving input signals and a plurality of output nodes generating responses. a situation value unit receives a plurality of the responses and generating a situation value signal. a change sensor coupled to receive the situation value signal generates an output signal representing a change of the situation value signal from a prior time to a current time. a connection coupling the change sensor output to one of the input nodes of the artificial neural network.,2002-04-02,The title of the patent is adaptive agent using neural network and its abstract is an adaptive agent including an artificial neural network having a plurality of input nodes for receiving input signals and a plurality of output nodes generating responses. a situation value unit receives a plurality of the responses and generating a situation value signal. a change sensor coupled to receive the situation value signal generates an output signal representing a change of the situation value signal from a prior time to a current time. a connection coupling the change sensor output to one of the input nodes of the artificial neural network. dated 2002-04-02
6366897,cortronic neural networks with distributed processing,"a cortronic neural network defines connections between neurons in a number of regions using target lists, which identify the output connections of each neuron and the connection strength. neurons are preferably sparsely interconnected between regions. training of connection weights employs a three stage process, which involves computation of the contribution to the input intensity of each neuron by every currently active neuron, a competition process that determines the next set of active neurons based on their current input intensity, and a weight adjustment process that updates and normalizes the connection weights based on which neurons won the competition process, and their connectivity with other winning neurons.",2002-04-02,"The title of the patent is cortronic neural networks with distributed processing and its abstract is a cortronic neural network defines connections between neurons in a number of regions using target lists, which identify the output connections of each neuron and the connection strength. neurons are preferably sparsely interconnected between regions. training of connection weights employs a three stage process, which involves computation of the contribution to the input intensity of each neuron by every currently active neuron, a competition process that determines the next set of active neurons based on their current input intensity, and a weight adjustment process that updates and normalizes the connection weights based on which neurons won the competition process, and their connectivity with other winning neurons. dated 2002-04-02"
6369545,neural network controlled power distribution element,"the power distribution control element significantly improves the efficiency by which solar energy is distributed and controlled to large phased array antenna assemblies by providing current directly from photovoltaic cells to lithium-ion battery cells through a neural-network based charge controller. the small current required to operate each transmit/receive module is provided from an adjacent battery cell rather than a large centralized battery assembly located in the spacecraft bus. in the preferred embodiment, the charge control is regulated by a back-propagation neural network.",2002-04-09,"The title of the patent is neural network controlled power distribution element and its abstract is the power distribution control element significantly improves the efficiency by which solar energy is distributed and controlled to large phased array antenna assemblies by providing current directly from photovoltaic cells to lithium-ion battery cells through a neural-network based charge controller. the small current required to operate each transmit/receive module is provided from an adjacent battery cell rather than a large centralized battery assembly located in the spacecraft bus. in the preferred embodiment, the charge control is regulated by a back-propagation neural network. dated 2002-04-09"
6373033,model-based predictive control of thermal processing,"a nonlinear model-based predictive temperature control system is described for use in thermal process reactors. a multivariable temperature response is predicted using a nonlinear parameterized model of a thermal process reactor. the nonlinear parameterized model is implemented using a neural network. predictions are made in an auto-regressive moving average fashion with a receding prediction horizon. model predictions are incorporated into a control law for estimating the optimum future control strategy. the high-speed, predictive nature of the controller renders it advantageous in multivariable rapid thermal processing reactors where fast response and high temperature uniformity are needed.",2002-04-16,"The title of the patent is model-based predictive control of thermal processing and its abstract is a nonlinear model-based predictive temperature control system is described for use in thermal process reactors. a multivariable temperature response is predicted using a nonlinear parameterized model of a thermal process reactor. the nonlinear parameterized model is implemented using a neural network. predictions are made in an auto-regressive moving average fashion with a receding prediction horizon. model predictions are incorporated into a control law for estimating the optimum future control strategy. the high-speed, predictive nature of the controller renders it advantageous in multivariable rapid thermal processing reactors where fast response and high temperature uniformity are needed. dated 2002-04-16"
6373962,license plate information reader device for motor vehicles,"in a license plate information reader device (a) for motor vehicles, a ccd camera (1) is provided to produce video image data (11) involving a license plate obtained by photographing a front and rear portion of a motor vehicle. an a/d converter (3) produces a digital multivalue image data (31) by a/d converting the video image data (11). a license plate extracting device (4) is provided to produce a digital multivalue image data (41) corresponding to an area in which the license plate occupies. a literal region extracting device (5) extracts a literal positional region of a letter sequence of the license plate based on the image obtained from the license plate extracting device (4). a literal recognition device (6) is provided to recognize a letter from a literal image (571) of the literal positional region obtained from the literal region extracting device (5). an image emphasis device is provided to emphasize the literal image (571) of the literal positional region by replacing a part of the literal region extracting device (5) with a filter net which serves as a neural network.",2002-04-16,"The title of the patent is license plate information reader device for motor vehicles and its abstract is in a license plate information reader device (a) for motor vehicles, a ccd camera (1) is provided to produce video image data (11) involving a license plate obtained by photographing a front and rear portion of a motor vehicle. an a/d converter (3) produces a digital multivalue image data (31) by a/d converting the video image data (11). a license plate extracting device (4) is provided to produce a digital multivalue image data (41) corresponding to an area in which the license plate occupies. a literal region extracting device (5) extracts a literal positional region of a letter sequence of the license plate based on the image obtained from the license plate extracting device (4). a literal recognition device (6) is provided to recognize a letter from a literal image (571) of the literal positional region obtained from the literal region extracting device (5). an image emphasis device is provided to emphasize the literal image (571) of the literal positional region by replacing a part of the literal region extracting device (5) with a filter net which serves as a neural network. dated 2002-04-16"
6374385,method and arrangement for implementing convolutional decoding,"the invention relates to a method and an arrangement for decoding a convolutionally encoded signal which comprises code words and the arrangement comprises a neural network which comprises a set of neurons comprising a set of inputs and an output. the received code words are applied to the inputs of the neurons, and the arrangement combines some of the inputs of the neuron in the neuron. in order to enable efficient decoding of the convolutional encoding, some of the output signals of the neural network neurons are fed back to the inputs of the neuron and the neuron multiplies at least some of the inputs of the neuron with one another before combining. the output signal of a predetermined neuron comprises an estimate of a decoded symbol.",2002-04-16,"The title of the patent is method and arrangement for implementing convolutional decoding and its abstract is the invention relates to a method and an arrangement for decoding a convolutionally encoded signal which comprises code words and the arrangement comprises a neural network which comprises a set of neurons comprising a set of inputs and an output. the received code words are applied to the inputs of the neurons, and the arrangement combines some of the inputs of the neuron in the neuron. in order to enable efficient decoding of the convolutional encoding, some of the output signals of the neural network neurons are fed back to the inputs of the neuron and the neuron multiplies at least some of the inputs of the neuron with one another before combining. the output signal of a predetermined neuron comprises an estimate of a decoded symbol. dated 2002-04-16"
6376831,neural network system for estimating conditions on submerged surfaces of seawater vessels,"an algorithm method of predicting estimated sea energy, wave directions and other seastate data with respect to submerged sea-going vessels, based on inputs derived from signal processed measurements of keel depth, pitch, roll and forward speed applied to three neural networks for heading detection and to a fourth neural network for seastate estimations.",2002-04-23,"The title of the patent is neural network system for estimating conditions on submerged surfaces of seawater vessels and its abstract is an algorithm method of predicting estimated sea energy, wave directions and other seastate data with respect to submerged sea-going vessels, based on inputs derived from signal processed measurements of keel depth, pitch, roll and forward speed applied to three neural networks for heading detection and to a fourth neural network for seastate estimations. dated 2002-04-23"
6377545,open loop adaptive access control of atm networks using a neural network,"a pyramidal pram neural network is used in an access control scheme for an atm node, or other switch handling bursty traffic. specifically, the neural network is a configured in a teacher forcing mode, and is trained either off-line or on-line. cells arriving at the node, which are found to violate agreed quality of service parameters, may be dropped, depending on the expected state of the switch buffers, based on expected traffic arrival rates, but the control is adaptive, which means that violating cells may be allowed to gain access to the node, if the state of the switch allows it. control is open-loop, which means that the response can be quicker.",2002-04-23,"The title of the patent is open loop adaptive access control of atm networks using a neural network and its abstract is a pyramidal pram neural network is used in an access control scheme for an atm node, or other switch handling bursty traffic. specifically, the neural network is a configured in a teacher forcing mode, and is trained either off-line or on-line. cells arriving at the node, which are found to violate agreed quality of service parameters, may be dropped, depending on the expected state of the switch buffers, based on expected traffic arrival rates, but the control is adaptive, which means that violating cells may be allowed to gain access to the node, if the state of the switch allows it. control is open-loop, which means that the response can be quicker. dated 2002-04-23"
6377832,system and method for analyzing a medical image,"disclosed is a system and method for determining a severity of a stenosis in a blood vessel depicted in a magnetic resonance imaging (mri) data set. the system comprises a neural network configured to calculate the severity of the stenosis in the blood vessel based upon a number of input parameters, and the input parameters including at least one characteristic of a signal void associated with the stenosis in the mri data set. the input parameters may include, for example, a flow rate of blood through the blood vessel, a length of a longitudinal axis of the signal void, and an average image intensity along the longitudinal axis of the signal void as well as other input parameters.",2002-04-23,"The title of the patent is system and method for analyzing a medical image and its abstract is disclosed is a system and method for determining a severity of a stenosis in a blood vessel depicted in a magnetic resonance imaging (mri) data set. the system comprises a neural network configured to calculate the severity of the stenosis in the blood vessel based upon a number of input parameters, and the input parameters including at least one characteristic of a signal void associated with the stenosis in the mri data set. the input parameters may include, for example, a flow rate of blood through the blood vessel, a length of a longitudinal axis of the signal void, and an average image intensity along the longitudinal axis of the signal void as well as other input parameters. dated 2002-04-23"
6377941,implementing automatic learning according to the k nearest neighbor mode in artificial neural networks,"a method of achieving automatic learning of an input vector presented to an artificial neural network (ann) formed by a plurality of neurons, using the k nearest neighbor (knn) mode. upon providing an input vector to be learned to the ann, a write component operation is performed to store the input vector components in the first available free neuron of the ann. then, a write category operation is performed by assigning a category defined by the user to the input vector. next, a test is performed to determine whether this category matches the categories of the nearest prototypes, i.e. which are located at the minimum distance. if it matches, this first free neuron is not engaged. otherwise, it is engaged by assigning the matching category to it. as a result, the input vector becomes the new prototype with the matching category associated thereto. further described is a circuit which automatically retains the first free neuron of the ann for learning.",2002-04-23,"The title of the patent is implementing automatic learning according to the k nearest neighbor mode in artificial neural networks and its abstract is a method of achieving automatic learning of an input vector presented to an artificial neural network (ann) formed by a plurality of neurons, using the k nearest neighbor (knn) mode. upon providing an input vector to be learned to the ann, a write component operation is performed to store the input vector components in the first available free neuron of the ann. then, a write category operation is performed by assigning a category defined by the user to the input vector. next, a test is performed to determine whether this category matches the categories of the nearest prototypes, i.e. which are located at the minimum distance. if it matches, this first free neuron is not engaged. otherwise, it is engaged by assigning the matching category to it. as a result, the input vector becomes the new prototype with the matching category associated thereto. further described is a circuit which automatically retains the first free neuron of the ann for learning. dated 2002-04-23"
6381083,multilevel signal equalization utilizing nonlinear maps,"in a recording/playback system, increased information is achieved by 4 level biased magnetic recording where the maximum amplitude 4 level recording signal drives the medium's magnetization into a nonlinear region of its transfer function. the bias does not eliminate distortion at the maximum signal input level, however the system's signal to noise ratio is improved due to an increase in the amplitude of the playback signal resulting from the increased recording level. the nonlinear mapping capability of a neural network provides equalization of playback signals distorted due to the record/playback nonlinearity. the 4 level recorded signals provide a factor of 2 in information storage compared to binary recording, and quadrature amplitude modulation (qam) combined with the 4 level recording technique provides an additional factor of 2, for a factor of 4 in the information content stored.",2002-04-30,"The title of the patent is multilevel signal equalization utilizing nonlinear maps and its abstract is in a recording/playback system, increased information is achieved by 4 level biased magnetic recording where the maximum amplitude 4 level recording signal drives the medium's magnetization into a nonlinear region of its transfer function. the bias does not eliminate distortion at the maximum signal input level, however the system's signal to noise ratio is improved due to an increase in the amplitude of the playback signal resulting from the increased recording level. the nonlinear mapping capability of a neural network provides equalization of playback signals distorted due to the record/playback nonlinearity. the 4 level recorded signals provide a factor of 2 in information storage compared to binary recording, and quadrature amplitude modulation (qam) combined with the 4 level recording technique provides an additional factor of 2, for a factor of 4 in the information content stored. dated 2002-04-30"
6381345,method and apparatus for detecting eye location in an image,"a simple method for segmenting eyes and extracting parameters enables further processing of the image to enable a person to appear to be making eye contact with another via a video conferencing system. this method is a first step for eye synthesis and gaze detection because it can automatically extract select eye parameters useful to these processes. its advantage is that no a priori information is necessary to segment eyes, unlike modeling and neural network methods. the method of the present invention first blurs the image to make it easier to determine the location of the two eye regions in the image. the eyebrows are then eliminated based on the located eye regions. the eyes are then segmented and the eye parameters are extracted from the resulting image according to the present invention, the process applies a gaussian filter, h(x, y), where g(x, y) is the resulting image and &#402;(x, y) is the original image.",2002-04-30,"The title of the patent is method and apparatus for detecting eye location in an image and its abstract is a simple method for segmenting eyes and extracting parameters enables further processing of the image to enable a person to appear to be making eye contact with another via a video conferencing system. this method is a first step for eye synthesis and gaze detection because it can automatically extract select eye parameters useful to these processes. its advantage is that no a priori information is necessary to segment eyes, unlike modeling and neural network methods. the method of the present invention first blurs the image to make it easier to determine the location of the two eye regions in the image. the eyebrows are then eliminated based on the located eye regions. the eyes are then segmented and the eye parameters are extracted from the resulting image according to the present invention, the process applies a gaussian filter, h(x, y), where g(x, y) is the resulting image and &#402;(x, y) is the original image. dated 2002-04-30"
6381591,"method for transformation of fuzzy logic, which is used to simulate a technical process, into a neural network","a method for transformation of fuzzy logic (fs) into a neural network (nn), in which, in order to form a defuzzified output value (y2) from normalized single-element functions (f1 . . . fm), the single-element functions (f1 . . . fm) are each assigned a singleton position (a1 . . . am) and at least one singleton weighting factor (r1 . . . rn), those singleton weighting factors (r1 . . . rn) which are assigned to the same single-element function (f1 . . . fm) are additively linked, and the singleton weighting factors (r1 . . . rn) and the additively linked singleton weighting factors (r1 . . . rn) are weighted via the corresponding singleton positions (a1 . . . am) and are additively linked in order to form the defuzzified output value (y2). one advantage of the method according to the invention is that the singleton positions (a1 . . . am) in the neural network (nn) can be varied, in order to optimize this network, such that their number before and after the optimization process remains constant and thus, in any case, subsequent reverse transformation of the neural network (nn) can be carried out to optimize fuzzy logic (fs). this advantageously allows the use of, in particular, standardized fuzzy system software to describe the optimized fuzzy logic (fs).",2002-04-30,"The title of the patent is method for transformation of fuzzy logic, which is used to simulate a technical process, into a neural network and its abstract is a method for transformation of fuzzy logic (fs) into a neural network (nn), in which, in order to form a defuzzified output value (y2) from normalized single-element functions (f1 . . . fm), the single-element functions (f1 . . . fm) are each assigned a singleton position (a1 . . . am) and at least one singleton weighting factor (r1 . . . rn), those singleton weighting factors (r1 . . . rn) which are assigned to the same single-element function (f1 . . . fm) are additively linked, and the singleton weighting factors (r1 . . . rn) and the additively linked singleton weighting factors (r1 . . . rn) are weighted via the corresponding singleton positions (a1 . . . am) and are additively linked in order to form the defuzzified output value (y2). one advantage of the method according to the invention is that the singleton positions (a1 . . . am) in the neural network (nn) can be varied, in order to optimize this network, such that their number before and after the optimization process remains constant and thus, in any case, subsequent reverse transformation of the neural network (nn) can be carried out to optimize fuzzy logic (fs). this advantageously allows the use of, in particular, standardized fuzzy system software to describe the optimized fuzzy logic (fs). dated 2002-04-30"
6382029,apparatus and method for utilizing electromagnetic acoustic transducers to non-destructively analyze in-service conductive materials,"the method of the invention identifies damage to an in-service conductor associated with the delivery (transmission and distribution) of electric power. electro-magnetic acoustic energy is generated in an in-service conductor associated with the delivery of electric power. corresponding return electro-magnetic acoustic energy is then measured. features are then extracted from the return electro-magnetic acoustic energy to characterize damage to the in-service conductor. the features may be extracted through a variety of signal processing techniques, such as wavelet signal processing. the extracted features may be classified using a neural network, fuzzy logic, or a combination of both.",2002-05-07,"The title of the patent is apparatus and method for utilizing electromagnetic acoustic transducers to non-destructively analyze in-service conductive materials and its abstract is the method of the invention identifies damage to an in-service conductor associated with the delivery (transmission and distribution) of electric power. electro-magnetic acoustic energy is generated in an in-service conductor associated with the delivery of electric power. corresponding return electro-magnetic acoustic energy is then measured. features are then extracted from the return electro-magnetic acoustic energy to characterize damage to the in-service conductor. the features may be extracted through a variety of signal processing techniques, such as wavelet signal processing. the extracted features may be classified using a neural network, fuzzy logic, or a combination of both. dated 2002-05-07"
6384995,apparatus and method for detecting defects in data storage devices,"a disk drive is tested during the manufacturing process, after the head/disk assembly is completely assembled and enclosed in its protective enclosure. a known data pattern is written to selected tracks on the disk surface, and the data is read back. during the read process, the analog read signal is sampled at first and third harmonic rates, and the logarithmic ratio of the two sampled signals used to derive a harmonic ratio flyheight (hrf) signal approximating the flyheight of the head. when a transducer head passes over a surface asperity, a collision occurs, causing the transducer to be lifted momentarily above its normal flyheight. if the amplitude of the hrf signal exceeds a predetermined clipping level, a possible disk defect is indicated. in order to characterize the possible defect, a window of the hrf signal samples in the vicinity of the suspected abnormality is digitized and used as the input to a neural network. the neural network is trained with actual hrf samples from previously detected disk drive abnormalities, which have been categorized by microscopic examination or other means. the neural network produces an output for the hrf signal samples from the drive being tested which indicates whether the defect is of a type which can be ignored, or of a type which requires scrapping or rework of the drive.",2002-05-07,"The title of the patent is apparatus and method for detecting defects in data storage devices and its abstract is a disk drive is tested during the manufacturing process, after the head/disk assembly is completely assembled and enclosed in its protective enclosure. a known data pattern is written to selected tracks on the disk surface, and the data is read back. during the read process, the analog read signal is sampled at first and third harmonic rates, and the logarithmic ratio of the two sampled signals used to derive a harmonic ratio flyheight (hrf) signal approximating the flyheight of the head. when a transducer head passes over a surface asperity, a collision occurs, causing the transducer to be lifted momentarily above its normal flyheight. if the amplitude of the hrf signal exceeds a predetermined clipping level, a possible disk defect is indicated. in order to characterize the possible defect, a window of the hrf signal samples in the vicinity of the suspected abnormality is digitized and used as the input to a neural network. the neural network is trained with actual hrf samples from previously detected disk drive abnormalities, which have been categorized by microscopic examination or other means. the neural network produces an output for the hrf signal samples from the drive being tested which indicates whether the defect is of a type which can be ignored, or of a type which requires scrapping or rework of the drive. dated 2002-05-07"
6386706,visual function testing with virtual retinal display,"a system for testing and quantifying visual field and other visual function information in a head-mounted virtual reality environment, utilizing a directed image formation device for scanning of an image onto the retina of the test subject. a method and an apparatus are also provided for utilizing a central neural network and a central data bank to perform automatic interpretation of the visual function test parameters obtained in a plurality of visual field testing systems, for a plurality of patients, with control and response signals being transmitted via the internet. the data produced by the testing systems are automatically analyzed and compared with patterns on which the neural network was previously trained, and clinical diagnoses for pathological conditions are thereby suggested to the respective clinician for each patient.",2002-05-14,"The title of the patent is visual function testing with virtual retinal display and its abstract is a system for testing and quantifying visual field and other visual function information in a head-mounted virtual reality environment, utilizing a directed image formation device for scanning of an image onto the retina of the test subject. a method and an apparatus are also provided for utilizing a central neural network and a central data bank to perform automatic interpretation of the visual function test parameters obtained in a plurality of visual field testing systems, for a plurality of patients, with control and response signals being transmitted via the internet. the data produced by the testing systems are automatically analyzed and compared with patterns on which the neural network was previously trained, and clinical diagnoses for pathological conditions are thereby suggested to the respective clinician for each patient. dated 2002-05-14"
6389157,joint optimization of parameters for the detection of clustered microcalcifications in digital mammograms,"a method and system for detecting and displaying clustered microcalcifications in a digital mammogram, wherein a single digital mammogram is first automatically cropped to a breast area sub-image which is then processed by means of an optimized difference of gaussians filter to enhance the appearance of potential microcalcifications in the sub-image. the potential microcalcifications are thresholded, clusters are detected, features are computed for the detected clusters, and the clusters are classified as either suspicious or not suspicious by means of a neural network. thresholding is preferably by sloping local thresholding, but may also be performed by global and dual-local thresholding. the locations in the original digital mammogram of the suspicious detected clustered microcalcifications are indicated. parameters for use in the detection and thresholding portions of the system are computer-optimized by means of a genetic algorithm. the results of the system are optimally combined with a radiologist's observation of the original mammogram by combining the observations with the results, after the radiologist has first accepted or rejected individual detections reported by the system.",2002-05-14,"The title of the patent is joint optimization of parameters for the detection of clustered microcalcifications in digital mammograms and its abstract is a method and system for detecting and displaying clustered microcalcifications in a digital mammogram, wherein a single digital mammogram is first automatically cropped to a breast area sub-image which is then processed by means of an optimized difference of gaussians filter to enhance the appearance of potential microcalcifications in the sub-image. the potential microcalcifications are thresholded, clusters are detected, features are computed for the detected clusters, and the clusters are classified as either suspicious or not suspicious by means of a neural network. thresholding is preferably by sloping local thresholding, but may also be performed by global and dual-local thresholding. the locations in the original digital mammogram of the suspicious detected clustered microcalcifications are indicated. parameters for use in the detection and thresholding portions of the system are computer-optimized by means of a genetic algorithm. the results of the system are optimally combined with a radiologist's observation of the original mammogram by combining the observations with the results, after the radiologist has first accepted or rejected individual detections reported by the system. dated 2002-05-14"
6389408,neural network systems for chemical and biological pattern recognition via the mueller matrix,"a neural network pattern recognition system for remotely sensing and identifying chemical and biological materials having a software component having an adaptive gradient descent training algorithm capable of performing backward-error-propagation and an input layer that is formatted to accept differential absorption mueller matrix spectroscopic data, a filtering weight matrix component capable of filtering pattern recognition from mueller data for specific predetermined materials and a processing component capable of receiving the pattern recognition from the filtering weight matrix component and determining the presence of specific predetermined materials. a method for sensing and identifying chemical and biological materials also is disclosed.",2002-05-14,"The title of the patent is neural network systems for chemical and biological pattern recognition via the mueller matrix and its abstract is a neural network pattern recognition system for remotely sensing and identifying chemical and biological materials having a software component having an adaptive gradient descent training algorithm capable of performing backward-error-propagation and an input layer that is formatted to accept differential absorption mueller matrix spectroscopic data, a filtering weight matrix component capable of filtering pattern recognition from mueller data for specific predetermined materials and a processing component capable of receiving the pattern recognition from the filtering weight matrix component and determining the presence of specific predetermined materials. a method for sensing and identifying chemical and biological materials also is disclosed. dated 2002-05-14"
6390380,air-conditioning device,"temperatures in a dr side air-conditioning zone and a pa side air-conditioning zone are controlled highly independently of each other without temperature interference between each zone. a room internal air temperature sensor and a room external air temperature sensor are provided. dr side and pa side temperature setters separately set room setpoint temperatures (tset(dr), tset(pa)) in each zone. first and second target blow-out temperature calculating portions, which include neural network, input the room setpoint temperatures and the temperature data. then it calculates dr side and pa side target blow-out temperatures (tao(dr), tao(pa)) relative to each air-conditioning zones by using a neural network. air-mixing doors separately adjusts the temperatures of conditioned air blown out from dr side air passage and pa side air passage to be the first and second target blow-out temperatures. here, the neural network has the learning function, which adjusts its output to be desired data (teacher signal). therefore, the output at a specific input condition can be adjusted without temperature interference between each zone.",2002-05-21,"The title of the patent is air-conditioning device and its abstract is temperatures in a dr side air-conditioning zone and a pa side air-conditioning zone are controlled highly independently of each other without temperature interference between each zone. a room internal air temperature sensor and a room external air temperature sensor are provided. dr side and pa side temperature setters separately set room setpoint temperatures (tset(dr), tset(pa)) in each zone. first and second target blow-out temperature calculating portions, which include neural network, input the room setpoint temperatures and the temperature data. then it calculates dr side and pa side target blow-out temperatures (tao(dr), tao(pa)) relative to each air-conditioning zones by using a neural network. air-mixing doors separately adjusts the temperatures of conditioned air blown out from dr side air passage and pa side air passage to be the first and second target blow-out temperatures. here, the neural network has the learning function, which adjusts its output to be desired data (teacher signal). therefore, the output at a specific input condition can be adjusted without temperature interference between each zone. dated 2002-05-21"
6392550,method and apparatus for monitoring driver alertness,a system for determining the alertness of a driver of a vehicle by monitoring the driver's seated posture. a pressure sensor array disposed in or on the seat produces output signals that indicate the pattern of pressure exerted by the driver's body at a plurality of points distributed over the seat surface. a microprocessor-based alertness evaluation module receives pressure pattern data from the sensor array and uses neural network processing techniques to determine a probable driver alertness level based upon changes in the pattern of pressure. the sensor array may also be used to supply data to an occupant restraint system control module. using the sensor for two purposes avoids the additional cost of providing a dedicated alertness sensor is avoided.,2002-05-21,The title of the patent is method and apparatus for monitoring driver alertness and its abstract is a system for determining the alertness of a driver of a vehicle by monitoring the driver's seated posture. a pressure sensor array disposed in or on the seat produces output signals that indicate the pattern of pressure exerted by the driver's body at a plurality of points distributed over the seat surface. a microprocessor-based alertness evaluation module receives pressure pattern data from the sensor array and uses neural network processing techniques to determine a probable driver alertness level based upon changes in the pattern of pressure. the sensor array may also be used to supply data to an occupant restraint system control module. using the sensor for two purposes avoids the additional cost of providing a dedicated alertness sensor is avoided. dated 2002-05-21
6393395,handwriting and speech recognizer using neural network with separate start and continuation output scores,"a method and system for recognizing user input information including cursive handwriting and spoken words. a time-delayed neural network having an improved architecture is trained at the word level with an improved method, which, along with preprocessing improvements, results in a recognizer with greater recognition accuracy. preprocessing is performed on the input data and, for example, may include resampling the data with sample points based on the second derivative to focus the recognizer on areas of the input data where the slope change per time is greatest. the input data is segmented, featurized and fed to the time-delayed neural network which outputs a matrix of character scores per segment. the neural network architecture outputs a separate score for the start and the continuation of a character. a dynamic time warp (dtw) is run against dictionary words to find the most probable path through the output matrix for that word, and each word is assigned a score based on the least costly path that can be traversed through the output matrix. the word (or words) with the overall lowest score (or scores) are returned. a dtw is similarly used in training, whereby the sample ink only need be labeled at the word level.",2002-05-21,"The title of the patent is handwriting and speech recognizer using neural network with separate start and continuation output scores and its abstract is a method and system for recognizing user input information including cursive handwriting and spoken words. a time-delayed neural network having an improved architecture is trained at the word level with an improved method, which, along with preprocessing improvements, results in a recognizer with greater recognition accuracy. preprocessing is performed on the input data and, for example, may include resampling the data with sample points based on the second derivative to focus the recognizer on areas of the input data where the slope change per time is greatest. the input data is segmented, featurized and fed to the time-delayed neural network which outputs a matrix of character scores per segment. the neural network architecture outputs a separate score for the start and the continuation of a character. a dynamic time warp (dtw) is run against dictionary words to find the most probable path through the output matrix for that word, and each word is assigned a score based on the least costly path that can be traversed through the output matrix. the word (or words) with the overall lowest score (or scores) are returned. a dtw is similarly used in training, whereby the sample ink only need be labeled at the word level. dated 2002-05-21"
6393413,n-tuple or ram based neural network classification system and method,"a method and system for training a computer classification system which can be defined by a network of a number of n-tuples or look up tables (luts), with each n-tuple or lut including a number of rows corresponding to at least a subset of possible classes and further including a number of columns being addressed by signals or elements of sampled training input data examples, each column being defined by a vector having cells with values, wherein the column vector cell values are determined based on one or more training sets of input data examples for different classes so that at least part of the cells comprise or point to information based on the number of times the corresponding cell address is sample from one or more sets of training input examples, and weight cell values are determined, corresponding to one or more column vector cells being addressed or sampled by the training examples.",2002-05-21,"The title of the patent is n-tuple or ram based neural network classification system and method and its abstract is a method and system for training a computer classification system which can be defined by a network of a number of n-tuples or look up tables (luts), with each n-tuple or lut including a number of rows corresponding to at least a subset of possible classes and further including a number of columns being addressed by signals or elements of sampled training input data examples, each column being defined by a vector having cells with values, wherein the column vector cell values are determined based on one or more training sets of input data examples for different classes so that at least part of the cells comprise or point to information based on the number of times the corresponding cell address is sample from one or more sets of training input examples, and weight cell values are determined, corresponding to one or more column vector cells being addressed or sampled by the training examples. dated 2002-05-21"
6397136,system for determining the occupancy state of a seat in a vehicle,"system for determining the occupancy of a seat in a vehicle using a variety of transducers and pattern recognition technologies and techniques that applies to any combination of transducers that provide information about seat occupancy. these include weight sensors, capacitive sensors, inductive sensors, ultrasonic, optical, electromagnetic, motion, infrared, and radar among others. the system includes a processor coupled to the transducers for receiving the data from the transducers and processing the data to obtain an output indicative of the current occupancy state of the seat. an algorithm is resident in the processor and is created from a plurality of data sets, each representing a different occupancy state of the seat and being formed from data from the transducers while the seat is in that occupancy state. the algorithm produces the output indicative of the current occupancy state of the seat upon inputting a data set representing the current occupancy state of the seat and being formed from data from the transducers. the algorithm may be a neural network or neural fuzzy algorithm generated by an appropriate algorithm-generating program.",2002-05-28,"The title of the patent is system for determining the occupancy state of a seat in a vehicle and its abstract is system for determining the occupancy of a seat in a vehicle using a variety of transducers and pattern recognition technologies and techniques that applies to any combination of transducers that provide information about seat occupancy. these include weight sensors, capacitive sensors, inductive sensors, ultrasonic, optical, electromagnetic, motion, infrared, and radar among others. the system includes a processor coupled to the transducers for receiving the data from the transducers and processing the data to obtain an output indicative of the current occupancy state of the seat. an algorithm is resident in the processor and is created from a plurality of data sets, each representing a different occupancy state of the seat and being formed from data from the transducers while the seat is in that occupancy state. the algorithm produces the output indicative of the current occupancy state of the seat upon inputting a data set representing the current occupancy state of the seat and being formed from data from the transducers. the algorithm may be a neural network or neural fuzzy algorithm generated by an appropriate algorithm-generating program. dated 2002-05-28"
6398914,method and device for process control in cellulose and paper manufacture,"in the production of de-inked pulp, measuring devices are used to register spectral and/or physical characteristic values of a starting material. these values are then fed to a neural network, by means of which correction variables are obtained for a regulating or controlling device which in provided. according to the invention, the measuring device is used to evaluate at least the starting materials of the production of pulp and/or paper. the evaluation of the characteristics of the raw material used in the production of pulp and paper is thereby possible.",2002-06-04,"The title of the patent is method and device for process control in cellulose and paper manufacture and its abstract is in the production of de-inked pulp, measuring devices are used to register spectral and/or physical characteristic values of a starting material. these values are then fed to a neural network, by means of which correction variables are obtained for a regulating or controlling device which in provided. according to the invention, the measuring device is used to evaluate at least the starting materials of the production of pulp and/or paper. the evaluation of the characteristics of the raw material used in the production of pulp and paper is thereby possible. dated 2002-06-04"
6401082,autoassociative-heteroassociative neural network,"an efficient neural network computing technique capable of synthesizing two sets of output signal data from a single input signal data set. the method and device of the invention involves a unique integration of autoassociative and heteroassociative neural network mappings, the autoassociative neural network mapping enabling a quality metric for assessing the generalization or prediction accuracy of the heteroassociative neural network mapping.",2002-06-04,"The title of the patent is autoassociative-heteroassociative neural network and its abstract is an efficient neural network computing technique capable of synthesizing two sets of output signal data from a single input signal data set. the method and device of the invention involves a unique integration of autoassociative and heteroassociative neural network mappings, the autoassociative neural network mapping enabling a quality metric for assessing the generalization or prediction accuracy of the heteroassociative neural network mapping. dated 2002-06-04"
6404750,sensor-assisted aloha for wireless networks,"the invention is made of an apparatus which incorporates a sensor-based means for stabilizing random access networks. in the preferred embodiment, a grid of sensors is used to gather energy measurements for analysis. in a preferred embodiment a neural network has been trained to estimate the number of colliding users in a given slot. this information is used to set parameters in a backoff algorithm so as to stabilize the network and minimize the delay experienced by users. the invention has the ability to locate users geographically within the network coverage area. this information can be used in conjunction with a steerable beam or an array of antennas to develop geographically-determined aloha subchannels, further increasing the capacity of the system.",2002-06-11,"The title of the patent is sensor-assisted aloha for wireless networks and its abstract is the invention is made of an apparatus which incorporates a sensor-based means for stabilizing random access networks. in the preferred embodiment, a grid of sensors is used to gather energy measurements for analysis. in a preferred embodiment a neural network has been trained to estimate the number of colliding users in a given slot. this information is used to set parameters in a backoff algorithm so as to stabilize the network and minimize the delay experienced by users. the invention has the ability to locate users geographically within the network coverage area. this information can be used in conjunction with a steerable beam or an array of antennas to develop geographically-determined aloha subchannels, further increasing the capacity of the system. dated 2002-06-11"
6404920,system for generalizing objects and features in an image,"the present invention features the use of the fundamental concept of color perception and multi-level resolution to perform scene segmentation and object/feature extraction in the context of self-determining and self-calibration modes. the technique uses only a single image, instead of multiple images as the input to generate segmented images. moreover, a flexible and arbitrary scheme is incorporated, rather than a fixed scheme of segmentation analysis. the process allows users to perform digital analysis using any appropriate means for object extraction after an image is segmented. first, an image is retrieved. the image is then transformed into at least two distinct bands. each transformed image is then projected into a color domain or a multi-level resolution setting. a segmented image is then created from all of the transformed images. the segmented image is analyzed to identify objects. object identification is achieved by matching a segmented region against an image library. a featureless library contains full shape, partial shape and real-world images in a dual library system. the depth contours and height-above-ground structural components constitute a dual library. also provided is a mathematical model called a parzen window-based statistical/neural network classifier, which forms an integral part of this featureless dual library object identification system. all images are considered three-dimensional. laser radar based 3-d images represent a special case.",2002-06-11,"The title of the patent is system for generalizing objects and features in an image and its abstract is the present invention features the use of the fundamental concept of color perception and multi-level resolution to perform scene segmentation and object/feature extraction in the context of self-determining and self-calibration modes. the technique uses only a single image, instead of multiple images as the input to generate segmented images. moreover, a flexible and arbitrary scheme is incorporated, rather than a fixed scheme of segmentation analysis. the process allows users to perform digital analysis using any appropriate means for object extraction after an image is segmented. first, an image is retrieved. the image is then transformed into at least two distinct bands. each transformed image is then projected into a color domain or a multi-level resolution setting. a segmented image is then created from all of the transformed images. the segmented image is analyzed to identify objects. object identification is achieved by matching a segmented region against an image library. a featureless library contains full shape, partial shape and real-world images in a dual library system. the depth contours and height-above-ground structural components constitute a dual library. also provided is a mathematical model called a parzen window-based statistical/neural network classifier, which forms an integral part of this featureless dual library object identification system. all images are considered three-dimensional. laser radar based 3-d images represent a special case. dated 2002-06-11"
6405184,process for producing fault classification signals,"a method for generating fault classification signals which identify faulty loops which develop in a multiphase energy supply network observed in the event of a fault from a protective device with a starting arrangement. to be able to generate such fault classification signals in a relatively simple manner, a neural network is used which is trained using input variables simulating faulty loops in the form of normalized resistance and reactance variables formed taking into consideration the starting characteristic of the starting arrangement. in the case of a fault, normalized resistance and reactance measured variables network for generating fault classification signals.",2002-06-11,"The title of the patent is process for producing fault classification signals and its abstract is a method for generating fault classification signals which identify faulty loops which develop in a multiphase energy supply network observed in the event of a fault from a protective device with a starting arrangement. to be able to generate such fault classification signals in a relatively simple manner, a neural network is used which is trained using input variables simulating faulty loops in the form of normalized resistance and reactance variables formed taking into consideration the starting characteristic of the starting arrangement. in the case of a fault, normalized resistance and reactance measured variables network for generating fault classification signals. dated 2002-06-11"
6411903,"system and method for delineating spatially dependent objects, such as hydrocarbon accumulations from seismic data","a neural network based system, method, and process for the automated delineation of spatially dependent objects is disclosed. the method is applicable to objects such as hydrocarbon accumulations, aeromagnetic profiles, astronomical clusters, weather clusters, objects from radar, sonar, seismic and infrared returns, etc. one of the novelties in the present invention is that the method can be utilized whether or not known data is available to provide traditional training sets. the output consists of a classification of the input data into clearly delineated accumulations, clusters, objects, etc. that have various types and properties. a preferred but non-exclusive application of the present invention is the automated delineation of hydrocarbon accumulations and sub-regions within the accumulations with various properties, in an oil and gas field, prior to the commencement of drilling operations.",2002-06-25,"The title of the patent is system and method for delineating spatially dependent objects, such as hydrocarbon accumulations from seismic data and its abstract is a neural network based system, method, and process for the automated delineation of spatially dependent objects is disclosed. the method is applicable to objects such as hydrocarbon accumulations, aeromagnetic profiles, astronomical clusters, weather clusters, objects from radar, sonar, seismic and infrared returns, etc. one of the novelties in the present invention is that the method can be utilized whether or not known data is available to provide traditional training sets. the output consists of a classification of the input data into clearly delineated accumulations, clusters, objects, etc. that have various types and properties. a preferred but non-exclusive application of the present invention is the automated delineation of hydrocarbon accumulations and sub-regions within the accumulations with various properties, in an oil and gas field, prior to the commencement of drilling operations. dated 2002-06-25"
6411905,method and apparatus for estimating odor concentration using an electronic nose,"a system and method for obtaining an estimate of the concentration of an odor in an air sample from data obtained by evaluating the sample with a sensor-array type electronic nose. principal components analysis is applied to a set of air sample data including sensor-array data obtained from evaluating the air sample with the sensor-array type electronic nose and measurements of the humidity of the air sample and clean reference air used by the electronic nose to obtain a predetermined number of principal components of the air sample data. the principal components obtained from the air sample are used as inputs to a neural network to obtain as output an estimate of the concentration of the odor in the air sample. the neural network uses parameters obtained by using an olfactometer to obtain discrete measurements of odor concentration from each of a plurality of calibration samples of air containing the odor, using the sensor-array type electronic nose to obtain a discrete set of calibration data from each of the calibration samples, each set including sensor-array data and measurements of the humidity of the calibration sample and the clean reference air used by the electronic nose, applying principal components analysis to each set of calibration data to obtain a discrete set of the predetermined number of principal components, and training a neural network using the sets of principal components as input data and the corresponding measured odor concentrations as expected output to obtain the parameters of a trained neural network.",2002-06-25,"The title of the patent is method and apparatus for estimating odor concentration using an electronic nose and its abstract is a system and method for obtaining an estimate of the concentration of an odor in an air sample from data obtained by evaluating the sample with a sensor-array type electronic nose. principal components analysis is applied to a set of air sample data including sensor-array data obtained from evaluating the air sample with the sensor-array type electronic nose and measurements of the humidity of the air sample and clean reference air used by the electronic nose to obtain a predetermined number of principal components of the air sample data. the principal components obtained from the air sample are used as inputs to a neural network to obtain as output an estimate of the concentration of the odor in the air sample. the neural network uses parameters obtained by using an olfactometer to obtain discrete measurements of odor concentration from each of a plurality of calibration samples of air containing the odor, using the sensor-array type electronic nose to obtain a discrete set of calibration data from each of the calibration samples, each set including sensor-array data and measurements of the humidity of the calibration sample and the clean reference air used by the electronic nose, applying principal components analysis to each set of calibration data to obtain a discrete set of the predetermined number of principal components, and training a neural network using the sets of principal components as input data and the corresponding measured odor concentrations as expected output to obtain the parameters of a trained neural network. dated 2002-06-25"
6411945,"method and apparatus for designing multi-component material, optimization analyzer and storage medium using learning process","in an optimization apparatus 30, a known compositional ratios and the like, and mechanical behaviors thereof are inputted by an experimental data input unit 40 and a learning is conducted in a non-linear calculation unit 32 in order to establish a corresponding relation between compositional ratios of multi-component materials and the like, and mechanical behaviors thereof as a conversion system based on a neural network. compositional ratios and the like are inputted in an optimization item input unit 42, and a mechanical behaviors are predicted in an optimization calculation unit 34 from compositional ratios and the like of the multi-component materials using the optimization item and the conversion system of the calculation unit 32, and an objective function is optimized until the objective function, expressing the mechanical behaviors are converged.",2002-06-25,"The title of the patent is method and apparatus for designing multi-component material, optimization analyzer and storage medium using learning process and its abstract is in an optimization apparatus 30, a known compositional ratios and the like, and mechanical behaviors thereof are inputted by an experimental data input unit 40 and a learning is conducted in a non-linear calculation unit 32 in order to establish a corresponding relation between compositional ratios of multi-component materials and the like, and mechanical behaviors thereof as a conversion system based on a neural network. compositional ratios and the like are inputted in an optimization item input unit 42, and a mechanical behaviors are predicted in an optimization calculation unit 34 from compositional ratios and the like of the multi-component materials using the optimization item and the conversion system of the calculation unit 32, and an objective function is optimized until the objective function, expressing the mechanical behaviors are converged. dated 2002-06-25"
6411946,route optimization and traffic management in an atm network using neural computing,"neural computing techniques are used to optimize route selection in a communication network, such as an atm network. output measurements of the network are used to provide optimal routing selection and traffic management. specifically, link data traffic is monitored in the network to obtain traffic history data. an autoregressive backpropagation neural network is trained using the traffic history data to obtain respective predicted traffic profiles for the links. particular links are then selected for carrying data based on the predicted traffic profiles. a cost function, limits on network parameters such as link cost and cell rate, and other quality of service factors are also considered in selecting the optimal route.",2002-06-25,"The title of the patent is route optimization and traffic management in an atm network using neural computing and its abstract is neural computing techniques are used to optimize route selection in a communication network, such as an atm network. output measurements of the network are used to provide optimal routing selection and traffic management. specifically, link data traffic is monitored in the network to obtain traffic history data. an autoregressive backpropagation neural network is trained using the traffic history data to obtain respective predicted traffic profiles for the links. particular links are then selected for carrying data based on the predicted traffic profiles. a cost function, limits on network parameters such as link cost and cell rate, and other quality of service factors are also considered in selecting the optimal route. dated 2002-06-25"
6415272,system for intelligent control based on soft computing,a reduced control system suitable for control of a nonlinear or unstable plant is described. the reduced control system is configured to use a reduced sensor set for controlling the plant without significant loss of control quality (accuracy) as compared to an optimal control system with an optimum sensor set. the control system calculates the information content provided by the reduced sensor set as compared to the information content provided by the optimum set. the control system also calculates the difference between the entropy production rate of the plant and the entropy production rate of the controller. a genetic optimizer is used to tune a fuzzy neural network in the reduced controller. a fitness function for the genetic optimizer provides optimum control accuracy in the reduced control system by minimizing the difference in entropy production while maximizing the sensor information content.,2002-07-02,The title of the patent is system for intelligent control based on soft computing and its abstract is a reduced control system suitable for control of a nonlinear or unstable plant is described. the reduced control system is configured to use a reduced sensor set for controlling the plant without significant loss of control quality (accuracy) as compared to an optimal control system with an optimum sensor set. the control system calculates the information content provided by the reduced sensor set as compared to the information content provided by the optimum set. the control system also calculates the difference between the entropy production rate of the plant and the entropy production rate of the controller. a genetic optimizer is used to tune a fuzzy neural network in the reduced controller. a fitness function for the genetic optimizer provides optimum control accuracy in the reduced control system by minimizing the difference in entropy production while maximizing the sensor information content. dated 2002-07-02
6415286,computer system and computerized method for partitioning data for parallel processing,"a computer system splits a data space to partition data between processors or processes. the data space may be split into sub-regions which need not be orthogonal to the axes defined the data space's parameters, using a decision tree. the decision tree can have neural networks in each of its non-terminal nodes that are trained on, and are used to partition, training data. each terminal, or leaf, node can have a hidden layer neural network trained on the training data that reaches the terminal node. the training of the non-terminal nodes' neural networks can be performed on one processor and the training of the leaf nodes' neural networks can be run on separate processors. different target values can be used for the training of the networks of different non-terminal nodes. the non-terminal node networks may be hidden layer neural networks. each non-terminal node automatically may send a desired ratio of the training records it receives to each of its child nodes, so the leaf node networks each receives approximately the same number of training records. the system may automatically configures the tree to have a number of leaf nodes equal to the number of separate processors available to train leaf node networks. after the non-terminal and leaf node networks have been trained, the records of a large data base can be passed through the tree for classification or for estimation of certain parameter values.",2002-07-02,"The title of the patent is computer system and computerized method for partitioning data for parallel processing and its abstract is a computer system splits a data space to partition data between processors or processes. the data space may be split into sub-regions which need not be orthogonal to the axes defined the data space's parameters, using a decision tree. the decision tree can have neural networks in each of its non-terminal nodes that are trained on, and are used to partition, training data. each terminal, or leaf, node can have a hidden layer neural network trained on the training data that reaches the terminal node. the training of the non-terminal nodes' neural networks can be performed on one processor and the training of the leaf nodes' neural networks can be run on separate processors. different target values can be used for the training of the networks of different non-terminal nodes. the non-terminal node networks may be hidden layer neural networks. each non-terminal node automatically may send a desired ratio of the training records it receives to each of its child nodes, so the leaf node networks each receives approximately the same number of training records. the system may automatically configures the tree to have a number of leaf nodes equal to the number of separate processors available to train leaf node networks. after the non-terminal and leaf node networks have been trained, the records of a large data base can be passed through the tree for classification or for estimation of certain parameter values. dated 2002-07-02"
6418378,neural net prediction of seismic streamer shape,"a neural network to predict seismic streamer shape during seismic operations having an input layer, an optional hidden layer, and an output layer, each layer having one or more nodes. the first layer comprises input nodes attached to seismic data acquisition operational parameters as follows: vessel coordinates, receiver coordinates, time, vessel velocity, current velocity, wind velocity, water temperature, salinity, tidal information, water depth, streamer density, and streamer dimensions. each node in the input layer is connected to each node in the hidden layer and each node in the hidden layer is connected to each node in the output layer, the output layer outputting a predicted cable shape. the hidden layer may be omitted. when the hidden lay is omitted, each node in the input layer is attached to each node in the output layer.each connection between nodes has an associated weight and a training process for determining the weights for each of the connections of the neural network. the trained neural network is responsive to the inputs and outputs to generate a predicted cable shape. the training process applies a plurality of training sets to the neural network. each training set comprises a set of inputs and a desired cable shape. with each training data set, the training process determines the difference between the cable shape predicted by the neural network and the desired or known cable shape. the training process then adjusts the weights of the neural network nodes based on the difference between the output predicted cable shape and the desired cable shape. the error assigned to each node in the neural network may be assigned by the training process via the use of back propagation or some other learning technique.",2002-07-09,"The title of the patent is neural net prediction of seismic streamer shape and its abstract is a neural network to predict seismic streamer shape during seismic operations having an input layer, an optional hidden layer, and an output layer, each layer having one or more nodes. the first layer comprises input nodes attached to seismic data acquisition operational parameters as follows: vessel coordinates, receiver coordinates, time, vessel velocity, current velocity, wind velocity, water temperature, salinity, tidal information, water depth, streamer density, and streamer dimensions. each node in the input layer is connected to each node in the hidden layer and each node in the hidden layer is connected to each node in the output layer, the output layer outputting a predicted cable shape. the hidden layer may be omitted. when the hidden lay is omitted, each node in the input layer is attached to each node in the output layer.each connection between nodes has an associated weight and a training process for determining the weights for each of the connections of the neural network. the trained neural network is responsive to the inputs and outputs to generate a predicted cable shape. the training process applies a plurality of training sets to the neural network. each training set comprises a set of inputs and a desired cable shape. with each training data set, the training process determines the difference between the cable shape predicted by the neural network and the desired or known cable shape. the training process then adjusts the weights of the neural network nodes based on the difference between the output predicted cable shape and the desired cable shape. the error assigned to each node in the neural network may be assigned by the training process via the use of back propagation or some other learning technique. dated 2002-07-09"
6418385,method for determining the location of a partial discharge,"in a high-voltage installation, for example a tube gas line, several partial discharge measurement probes are provided that detect electromagnetic signals of partial discharges. the corresponding measurement data is supplied to an evaluation device that includes a neural network that is trained in such a way that it determines the location of occurrence of the partial discharge from the measurement data of two sensors.",2002-07-09,"The title of the patent is method for determining the location of a partial discharge and its abstract is in a high-voltage installation, for example a tube gas line, several partial discharge measurement probes are provided that detect electromagnetic signals of partial discharges. the corresponding measurement data is supplied to an evaluation device that includes a neural network that is trained in such a way that it determines the location of occurrence of the partial discharge from the measurement data of two sensors. dated 2002-07-09"
6418412,quantization using frequency and mean compensated frequency input data for robust speech recognition,"a speech recognition system utilizes multiple quantizers to process frequency parameters and mean compensated frequency parameters derived from an input signal. the quantizers may be matrix and vector quantizer pairs, and such quantizer pairs may also function as front ends to a second stage speech classifiers such as hidden markov models (hmms) and/or utilizes neural network postprocessing to, for example, improve speech recognition performance. mean compensating the frequency parameters can remove noise frequency components that remain approximately constant during the duration of the input signal. hmm initial state and state transition probabilities derived from common quantizer types and the same input signal may be consolidated to improve recognition system performance and efficiency. matrix quantization exploits the &#8220;evolution&#8221; of the speech short-term spectral envelopes as well as frequency domain information, and vector quantization (vq) primarily operates on frequency domain information. time domain information may be substantially limited which may introduce error into the matrix quantization, and the vq may provide error compensation. the matrix and vector quantizers may split spectral subbands to target selected frequencies for enhanced processing and may use fuzzy associations to develop fuzzy observation sequence data. a mixer may provide a variety of input data to the neural network for classification determination. fuzzy operators may be utilized to reduce quantization error. multiple codebooks may also be combined to form single respective codebooks for split matrix and split vector quantization to reduce processing resources demand.",2002-07-09,"The title of the patent is quantization using frequency and mean compensated frequency input data for robust speech recognition and its abstract is a speech recognition system utilizes multiple quantizers to process frequency parameters and mean compensated frequency parameters derived from an input signal. the quantizers may be matrix and vector quantizer pairs, and such quantizer pairs may also function as front ends to a second stage speech classifiers such as hidden markov models (hmms) and/or utilizes neural network postprocessing to, for example, improve speech recognition performance. mean compensating the frequency parameters can remove noise frequency components that remain approximately constant during the duration of the input signal. hmm initial state and state transition probabilities derived from common quantizer types and the same input signal may be consolidated to improve recognition system performance and efficiency. matrix quantization exploits the &#8220;evolution&#8221; of the speech short-term spectral envelopes as well as frequency domain information, and vector quantization (vq) primarily operates on frequency domain information. time domain information may be substantially limited which may introduce error into the matrix quantization, and the vq may provide error compensation. the matrix and vector quantizers may split spectral subbands to target selected frequencies for enhanced processing and may use fuzzy associations to develop fuzzy observation sequence data. a mixer may provide a variety of input data to the neural network for classification determination. fuzzy operators may be utilized to reduce quantization error. multiple codebooks may also be combined to form single respective codebooks for split matrix and split vector quantization to reduce processing resources demand. dated 2002-07-09"
6418423,method and apparatus for executing neural network applications on a network of embedded devices,"disclosed is a system and a method for combining the computational resources of numerous embedded devices to enable any of them to perform complex tasks like speech recognition or natural language understanding. a distinguished master device communicates with a network of embedded devices, and organizes them as the nodes of a neural network. to each node (embedded device) in the neural network, the master device sends the activation function for that node and the connectivity pattern for that node. the master device sends the inputs for the network to the distinguished input nodes of the network. during computation, each node computes the activation function of all of its inputs and sends its activation to all the nodes to which it needs to send output to. the outputs of the neural network are sent to the master device. thus, the network of embedded devices can perform any computation (like speech recognition, natural language understanding, etc.) which can be mapped onto a neural network model.",2002-07-09,"The title of the patent is method and apparatus for executing neural network applications on a network of embedded devices and its abstract is disclosed is a system and a method for combining the computational resources of numerous embedded devices to enable any of them to perform complex tasks like speech recognition or natural language understanding. a distinguished master device communicates with a network of embedded devices, and organizes them as the nodes of a neural network. to each node (embedded device) in the neural network, the master device sends the activation function for that node and the connectivity pattern for that node. the master device sends the inputs for the network to the distinguished input nodes of the network. during computation, each node computes the activation function of all of its inputs and sends its activation to all the nodes to which it needs to send output to. the outputs of the neural network are sent to the master device. thus, the network of embedded devices can perform any computation (like speech recognition, natural language understanding, etc.) which can be mapped onto a neural network model. dated 2002-07-09"
6421064,system and methods for controlling automatic scrolling of information on a display screen,"a system for controlling the automatic scrolling of information includes a screen, a computer system, gimbaled sensor system for following and tracking the position and movement of the user's head and user's eye, and a scroll activating interface algorithm using a neural network to find screen gaze coordinates implemented by the computer system so that scrolling function is performance based upon the screen gaze coordinates of the user's eye relative to a certain activation area on the screen. a method of controlling scrolling includes the acts of finding a screen gaze coordinates on the screen, determining whether the screen gaze coordinate is within at least one activated control region, and activating scrolling to provide a display of information when the gaze direction is within at least one activated control region.",2002-07-16,"The title of the patent is system and methods for controlling automatic scrolling of information on a display screen and its abstract is a system for controlling the automatic scrolling of information includes a screen, a computer system, gimbaled sensor system for following and tracking the position and movement of the user's head and user's eye, and a scroll activating interface algorithm using a neural network to find screen gaze coordinates implemented by the computer system so that scrolling function is performance based upon the screen gaze coordinates of the user's eye relative to a certain activation area on the screen. a method of controlling scrolling includes the acts of finding a screen gaze coordinates on the screen, determining whether the screen gaze coordinate is within at least one activated control region, and activating scrolling to provide a display of information when the gaze direction is within at least one activated control region. dated 2002-07-16"
6421341,high speed packet switching controller for telephone switching system,"this invention relates to a high speed packet switching controller in a telephone switching system which can suitably be applied to a packet controller having large capacity using a neural network chip and maximize the system performance by the optimized switching operation. the high speed packet switching controller comprises a row address decoder for decoding a weight raw address which is inputted thereto, a column address decoder for decoding a weight column address which is inputted thereto, a matrix array for providing the neural network using address signals provided from the row address decoder and column address decoder and outputing varied voltage in accordance with an external weight value, a neural network for producing a final crossbar switching control signal, an external input/output bus for transmitting an output signal of the neural network, and an internal neural data bus for transmitting the address signal output from the row address decoder and column address decoder to the matrix array.",2002-07-16,"The title of the patent is high speed packet switching controller for telephone switching system and its abstract is this invention relates to a high speed packet switching controller in a telephone switching system which can suitably be applied to a packet controller having large capacity using a neural network chip and maximize the system performance by the optimized switching operation. the high speed packet switching controller comprises a row address decoder for decoding a weight raw address which is inputted thereto, a column address decoder for decoding a weight column address which is inputted thereto, a matrix array for providing the neural network using address signals provided from the row address decoder and column address decoder and outputing varied voltage in accordance with an external weight value, a neural network for producing a final crossbar switching control signal, an external input/output bus for transmitting an output signal of the neural network, and an internal neural data bus for transmitting the address signal output from the row address decoder and column address decoder to the matrix array. dated 2002-07-16"
6422079,ground anchorage testing apparatus,"a ground anchorage testing arrangement having an impulse imparting apparatus connectable to a ground anchorage tendon (20) or element thereof to be tested, the impulse imparting apparatus comprising an attachment means (22) for attachment to the ground anchorage tendon (20), a movable mass (31), a guide (28, 34) for guiding movement of the mass in the direction substantially aligned with the axis 0 of the ground anchorage to be tested and a drive means for imparting a driving force to move the mass in said direction (not shown). a method of assessing the integrity of ground anchorages, the method comprising the steps of (a) imparting a load impulse to a ground anchorage tendon to be tested, (b) monitoring the vibrational response signal of the anchorage to the imparted load impulse, (c) conditioning the vibrational response signal and (d) applying the conditioned vibrational response signal to an artificial neural network.",2002-07-23,"The title of the patent is ground anchorage testing apparatus and its abstract is a ground anchorage testing arrangement having an impulse imparting apparatus connectable to a ground anchorage tendon (20) or element thereof to be tested, the impulse imparting apparatus comprising an attachment means (22) for attachment to the ground anchorage tendon (20), a movable mass (31), a guide (28, 34) for guiding movement of the mass in the direction substantially aligned with the axis 0 of the ground anchorage to be tested and a drive means for imparting a driving force to move the mass in said direction (not shown). a method of assessing the integrity of ground anchorages, the method comprising the steps of (a) imparting a load impulse to a ground anchorage tendon to be tested, (b) monitoring the vibrational response signal of the anchorage to the imparted load impulse, (c) conditioning the vibrational response signal and (d) applying the conditioned vibrational response signal to an artificial neural network. dated 2002-07-23"
6422751,method and system for prediction of exposure and dose area product for radiographic x-ray imaging,"a neural network prediction has been provided for predicting radiation exposure and/or air-kerma at a predefined arbitrary distance during an x-ray exposure; and for predicting radiation exposure and/or air-kerma area product for a radiographic x-ray exposure. the air-kerma levels are predicted directly from the x-ray exposure parameters. the method or model is provided to predict the radiation exposure or air-kerma for an arbitrary radiographic x-ray exposure by providing input variables to identify the spectral characteristics of the x-ray beam, providing a neural net which has been trained to calculate the exposure or air-kerma value, and by scaling the neural net output by the calibrated tube efficiency, and the actual current through the x-ray tube and the duration of the exposure. the prediction for exposure/air-kerma further applies the actual source-to-object distance, and the prediction for exposure/air-kerma area product further applies the actual imaged field area at a source-to-image distance.",2002-07-23,"The title of the patent is method and system for prediction of exposure and dose area product for radiographic x-ray imaging and its abstract is a neural network prediction has been provided for predicting radiation exposure and/or air-kerma at a predefined arbitrary distance during an x-ray exposure; and for predicting radiation exposure and/or air-kerma area product for a radiographic x-ray exposure. the air-kerma levels are predicted directly from the x-ray exposure parameters. the method or model is provided to predict the radiation exposure or air-kerma for an arbitrary radiographic x-ray exposure by providing input variables to identify the spectral characteristics of the x-ray beam, providing a neural net which has been trained to calculate the exposure or air-kerma value, and by scaling the neural net output by the calibrated tube efficiency, and the actual current through the x-ray tube and the duration of the exposure. the prediction for exposure/air-kerma further applies the actual source-to-object distance, and the prediction for exposure/air-kerma area product further applies the actual imaged field area at a source-to-image distance. dated 2002-07-23"
6424737,method and apparatus of compressing images using localized radon transforms,a method and an apparatus of compressing data. the method and apparatus include constructing a neural network having a specific geometry using a finite and discrete radon transform. the data is then fed through the neural network to produce a transformed data stream. the transformed data stream is thresholded. a fixed input signal is fed back through the neural network to generate a decoding calculation of an average value. the thresholded data stream is entropy encoded.,2002-07-23,The title of the patent is method and apparatus of compressing images using localized radon transforms and its abstract is a method and an apparatus of compressing data. the method and apparatus include constructing a neural network having a specific geometry using a finite and discrete radon transform. the data is then fed through the neural network to produce a transformed data stream. the transformed data stream is thresholded. a fixed input signal is fed back through the neural network to generate a decoding calculation of an average value. the thresholded data stream is entropy encoded. dated 2002-07-23
6424919,"method for determining preferred drill bit design parameters and drilling parameters using a trained artificial neural network, and methods for training the artificial neural network","a method for selecting a design parameter for a drill bit is disclosed. the method includes entering a value of at least one property of an earth formation to be drilled into a trained neural network. the neural network is trained by selecting data from drilled wellbores. the data comprise values of the formation property for formations through which the drilled wellbores have penetrated. corresponding to the values of formation property are values of at least one drilling operating parameter, the drill bit design parameter, and values of a rate of penetration and a rate of wear of a drill bit used on each of the formations. data from the wellbores are entered into the neural network to train it, and the design parameter is then selected based on output of the trained neural network.",2002-07-23,"The title of the patent is method for determining preferred drill bit design parameters and drilling parameters using a trained artificial neural network, and methods for training the artificial neural network and its abstract is a method for selecting a design parameter for a drill bit is disclosed. the method includes entering a value of at least one property of an earth formation to be drilled into a trained neural network. the neural network is trained by selecting data from drilled wellbores. the data comprise values of the formation property for formations through which the drilled wellbores have penetrated. corresponding to the values of formation property are values of at least one drilling operating parameter, the drill bit design parameter, and values of a rate of penetration and a rate of wear of a drill bit used on each of the formations. data from the wellbores are entered into the neural network to train it, and the design parameter is then selected based on output of the trained neural network. dated 2002-07-23"
6424956,stochastic encoder/decoder/predictor,"an artificial intelligence system is provided which makes use of a dual subroutine to adapt weights. elastic fuzzy logic (&#8220;elf&#8221;) system is provided in which classical neural network learning techniques are combined with fuzzy logic techniques in order to accomplish artificial intelligence tasks such as pattern recognition, expert cloning and trajectory control. the system may be implemented in a computer provided with multiplier means and storage means for storing a vector of weights to be used as multiplier factors in an apparatus for fuzzy control.",2002-07-23,"The title of the patent is stochastic encoder/decoder/predictor and its abstract is an artificial intelligence system is provided which makes use of a dual subroutine to adapt weights. elastic fuzzy logic (&#8220;elf&#8221;) system is provided in which classical neural network learning techniques are combined with fuzzy logic techniques in order to accomplish artificial intelligence tasks such as pattern recognition, expert cloning and trajectory control. the system may be implemented in a computer provided with multiplier means and storage means for storing a vector of weights to be used as multiplier factors in an apparatus for fuzzy control. dated 2002-07-23"
6429699,generator of neuron transfer function and its derivative,"this invention relates to an artificial neural network (ann), particularly to a neuron circuit and its activation function including the derivative. the neuron circuit capable of generating an adjustable sigmoid-like function and a good approximation of its derivative, comprises: a current generator for generating a current; a current-controlled transistor for changing an output voltage according to the current from the current generator; and at least one differential pair of transistors for generating the adjustable sigmoid-like function output and the good approximation of its derivative by the changed output voltage.",2002-08-06,"The title of the patent is generator of neuron transfer function and its derivative and its abstract is this invention relates to an artificial neural network (ann), particularly to a neuron circuit and its activation function including the derivative. the neuron circuit capable of generating an adjustable sigmoid-like function and a good approximation of its derivative, comprises: a current generator for generating a current; a current-controlled transistor for changing an output voltage according to the current from the current generator; and at least one differential pair of transistors for generating the adjustable sigmoid-like function output and the good approximation of its derivative by the changed output voltage. dated 2002-08-06"
6434849,method for determining a lateral and/or angular offset between two rotatable parts,"data obtained by rotating a set of sensors attached to rotatable parts or the like, in order to perform an alignment of such rotatable parts, is organized in three dimensional vector form (triplets). two components of such a three dimensional vector will relate to observed apparent geometrical shifts and one component relates to an angle of rotation. in order to arrive at an improved &#8220;best fit&#8221; when evaluating the data, a three dimensional analysis is performed, based on the concept of an elliptical helix, and using spatial filtering or a neural network. for spatial filtering, three dimensional fourier transforms are used.",2002-08-20,"The title of the patent is method for determining a lateral and/or angular offset between two rotatable parts and its abstract is data obtained by rotating a set of sensors attached to rotatable parts or the like, in order to perform an alignment of such rotatable parts, is organized in three dimensional vector form (triplets). two components of such a three dimensional vector will relate to observed apparent geometrical shifts and one component relates to an angle of rotation. in order to arrive at an improved &#8220;best fit&#8221; when evaluating the data, a three dimensional analysis is performed, based on the concept of an elliptical helix, and using spatial filtering or a neural network. for spatial filtering, three dimensional fourier transforms are used. dated 2002-08-20"
6437728,a-scan isar target recognition system and method,"a target recognition system and method wherein only target amplitude data, i.e., coherent a-scan data, is interrogated at each of a plurality of range resolution cells along the radar line of sight path for target recognition. target aspect angle is ignored within the angular segmentation of the feature library without degrading classification performance. observed signature characteristics are collected at various aspect angles and through unknown roll, pitch and yaw motions of each anticipated target and provided to a neural network as training sets. the neural network forms feature vectors for each target class which are useful for valid classification comparisons in all sea states, especially in calm and littoral waters. these feature vectors are useful for valid classification comparisons over at least 30 degrees of target aspect angle.",2002-08-20,"The title of the patent is a-scan isar target recognition system and method and its abstract is a target recognition system and method wherein only target amplitude data, i.e., coherent a-scan data, is interrogated at each of a plurality of range resolution cells along the radar line of sight path for target recognition. target aspect angle is ignored within the angular segmentation of the feature library without degrading classification performance. observed signature characteristics are collected at various aspect angles and through unknown roll, pitch and yaw motions of each anticipated target and provided to a neural network as training sets. the neural network forms feature vectors for each target class which are useful for valid classification comparisons in all sea states, especially in calm and littoral waters. these feature vectors are useful for valid classification comparisons over at least 30 degrees of target aspect angle. dated 2002-08-20"
6438493,method for seismic facies interpretation using textural analysis and neural networks,"seismic facies are identified in a volume of seismic data, wherein, first, a plurality of initial textural attributes representative of the volume of seismic data are calculated. next, a probabilistic neural network is constructed from the calculated initial textural attributes. then, final textural attributes are calculated throughout the volume of seismic data. finally, the calculated final textural attributes are classified using the constructed probabilistic neural network.",2002-08-20,"The title of the patent is method for seismic facies interpretation using textural analysis and neural networks and its abstract is seismic facies are identified in a volume of seismic data, wherein, first, a plurality of initial textural attributes representative of the volume of seismic data are calculated. next, a probabilistic neural network is constructed from the calculated initial textural attributes. then, final textural attributes are calculated throughout the volume of seismic data. finally, the calculated final textural attributes are classified using the constructed probabilistic neural network. dated 2002-08-20"
6438534,"process and system for commissioning industrial plants, in particular in the primary industry","a method and system for commissioning industrial plants, in particular in the basic materials industry, having a plant control system which carries out both non-control functions and control functions and whose control system operates with process models, in particular control engineering models, for example in the form of mathematical models, neural network models, expert systems etc., in a control system computing unit. the commissioning is carried out in subdivided fashion into commissioning the non-control functions with extensive initialization of the control functions, by means of personnel located on site, and extensive commissioning of the control functions by means of remotely-transmitted data via data lines from at least one site remote from the plant, preferably from an engineering center.",2002-08-20,"The title of the patent is process and system for commissioning industrial plants, in particular in the primary industry and its abstract is a method and system for commissioning industrial plants, in particular in the basic materials industry, having a plant control system which carries out both non-control functions and control functions and whose control system operates with process models, in particular control engineering models, for example in the form of mathematical models, neural network models, expert systems etc., in a control system computing unit. the commissioning is carried out in subdivided fashion into commissioning the non-control functions with extensive initialization of the control functions, by means of personnel located on site, and extensive commissioning of the control functions by means of remotely-transmitted data via data lines from at least one site remote from the plant, preferably from an engineering center. dated 2002-08-20"
6442287,method and system for the computerized analysis of bone mass and structure,"an automated method, storage medium, and system for analyzing bone. digital image data corresponding to an image of the bone are obtained. next there is determined, based on the digital images, a measure of bone mineral density (bmd) and at least one of a measure of bone geometry, a minkowski dimension, and a trabecular orientation. the strength of the bone is estimated based upon the measure of bmd and at least one of the measure of bone geometry, the minkowski dimension, and the trabecular orientation. to improve bone texture analysis, the present invention also provides a novel automated method, storage medium, and system in which digital image data corresponding to an image of the bone is obtained, and a region of interest (roi) is selected within the bone. a fractal characteristic of the image data within the roi using an artificial neural network is extracted. the strength of the bone is estimated based at least in part on the extracted fractal characteristic. to perform bone analysis with an improved measure of bone mineral density, the present invention also provides a novel automated method, storage medium, and system in which digital image data corresponding to an image of the bone is obtained. a measure of normalized bone mineral density (bmd) corresponding to a volumetric bone mineral density of the bone is determined, and the strength of the bone based is estimated based at least in part on the normalized bmd.",2002-08-27,"The title of the patent is method and system for the computerized analysis of bone mass and structure and its abstract is an automated method, storage medium, and system for analyzing bone. digital image data corresponding to an image of the bone are obtained. next there is determined, based on the digital images, a measure of bone mineral density (bmd) and at least one of a measure of bone geometry, a minkowski dimension, and a trabecular orientation. the strength of the bone is estimated based upon the measure of bmd and at least one of the measure of bone geometry, the minkowski dimension, and the trabecular orientation. to improve bone texture analysis, the present invention also provides a novel automated method, storage medium, and system in which digital image data corresponding to an image of the bone is obtained, and a region of interest (roi) is selected within the bone. a fractal characteristic of the image data within the roi using an artificial neural network is extracted. the strength of the bone is estimated based at least in part on the extracted fractal characteristic. to perform bone analysis with an improved measure of bone mineral density, the present invention also provides a novel automated method, storage medium, and system in which digital image data corresponding to an image of the bone is obtained. a measure of normalized bone mineral density (bmd) corresponding to a volumetric bone mineral density of the bone is determined, and the strength of the bone based is estimated based at least in part on the normalized bmd. dated 2002-08-27"
6442535,"method and apparatus for implementing a low cost, intelligent controller for a switched reluctance machine","a controller for a switched reluctance machine utilizing a feedforward neural network in combination with either a fuzzy logic controller or a proportional-integral controller to provide output control signals (e.g., turn-on angle, turn-off angle and peak current) for controlling the energization of a switched reluctance machine. in an alternate embodiment, a fuzzy logic controller is utilized by itself to control a switched reluctance machine.",2002-08-27,"The title of the patent is method and apparatus for implementing a low cost, intelligent controller for a switched reluctance machine and its abstract is a controller for a switched reluctance machine utilizing a feedforward neural network in combination with either a fuzzy logic controller or a proportional-integral controller to provide output control signals (e.g., turn-on angle, turn-off angle and peak current) for controlling the energization of a switched reluctance machine. in an alternate embodiment, a fuzzy logic controller is utilized by itself to control a switched reluctance machine. dated 2002-08-27"
6442536,method for predicting flammability limits of complex mixtures,"this invention is directed to a method for predicting the flammability of complex mixtures using critical variables of structural groups comprising training data from the critical variables of each structural groups, the critical variables comprising compositional and thermochemical data from each of the structural groups to produce a neural network model; testing the trained data from the neural network model; and validating the trained and tested data from the neural network to accurately predict the flammability limit of an analogous complex mixture having similar structural groups.",2002-08-27,"The title of the patent is method for predicting flammability limits of complex mixtures and its abstract is this invention is directed to a method for predicting the flammability of complex mixtures using critical variables of structural groups comprising training data from the critical variables of each structural groups, the critical variables comprising compositional and thermochemical data from each of the structural groups to produce a neural network model; testing the trained data from the neural network model; and validating the trained and tested data from the neural network to accurately predict the flammability limit of an analogous complex mixture having similar structural groups. dated 2002-08-27"
6443889,provision of decision support for acute myocardial infarction,"the present invention provides methods and apparatuses, which make use of at least one trained and tuned artificial neural network (16) to generate decision regions (32) in the n-dimensional space of n input variables associated with ami. the set of measured variables (30) is related to the decision regions (32), in order to provide decision support. preferably, the decision regions (32) are graphically visualized as areas in a two-dimensional diagram. preferably, the artificial neural network (16) is trained by patient specific parameters. the variables associated with ami (30) are preferably selected as biochemical markers and/or quantities derived from continuous/intermittent ecg/vcg. the performance of the artificial neural network (16) is preferably optimally tuned to clinical requirements on predictive values of the artificial neural network output in given prevalence situations.",2002-09-03,"The title of the patent is provision of decision support for acute myocardial infarction and its abstract is the present invention provides methods and apparatuses, which make use of at least one trained and tuned artificial neural network (16) to generate decision regions (32) in the n-dimensional space of n input variables associated with ami. the set of measured variables (30) is related to the decision regions (32), in order to provide decision support. preferably, the decision regions (32) are graphically visualized as areas in a two-dimensional diagram. preferably, the artificial neural network (16) is trained by patient specific parameters. the variables associated with ami (30) are preferably selected as biochemical markers and/or quantities derived from continuous/intermittent ecg/vcg. the performance of the artificial neural network (16) is preferably optimally tuned to clinical requirements on predictive values of the artificial neural network output in given prevalence situations. dated 2002-09-03"
6445988,system for determining the occupancy state of a seat in a vehicle and controlling a component based thereon,"system for determining the occupancy of a seat in a vehicle using pattern recognition technologies and techniques that apply to any combination of transducers that provide information about seat occupancy, for example, weight sensors, capacitive sensors, inductive sensors, ultrasonic, optical, electromagnetic, motion, infrared and radar sensors. a processor is coupled to the transducers for receiving data therefrom and processes the data to obtain an output indicative of the seat's current occupancy state. a combination neural network is resident in the processor and is created from data sets, each representing a different occupancy state of the seat and being formed from data from the transducers while the seat is in that occupancy state. the combination neural network produces the output indicative of the current occupancy state of the seat upon inputting a data set representing the current occupancy state of the seat and being formed from data from the transducers.",2002-09-03,"The title of the patent is system for determining the occupancy state of a seat in a vehicle and controlling a component based thereon and its abstract is system for determining the occupancy of a seat in a vehicle using pattern recognition technologies and techniques that apply to any combination of transducers that provide information about seat occupancy, for example, weight sensors, capacitive sensors, inductive sensors, ultrasonic, optical, electromagnetic, motion, infrared and radar sensors. a processor is coupled to the transducers for receiving data therefrom and processes the data to obtain an output indicative of the seat's current occupancy state. a combination neural network is resident in the processor and is created from data sets, each representing a different occupancy state of the seat and being formed from data from the transducers while the seat is in that occupancy state. the combination neural network produces the output indicative of the current occupancy state of the seat upon inputting a data set representing the current occupancy state of the seat and being formed from data from the transducers. dated 2002-09-03"
6446038,method and system for objectively evaluating speech,a method and system for objectively evaluating the quality of speech in a voice communication system. a plurality of speech reference vectors is first obtained based on a plurality of clean speech samples. a corrupted speech signal is received and processed to determine a plurality of distortions derived from a plurality of distortion measures based on the plurality of speech reference vectors. the plurality of distortions are processed by a non-linear neural network model to generate a subjective score representing user acceptance of the corrupted speech signal. the non-linear neural network model is first trained on clean speech samples as well as corrupted speech samples through the use of backpropagation to obtain the weights and bias terms necessary to predict subjective scores from several objective measures.,2002-09-03,The title of the patent is method and system for objectively evaluating speech and its abstract is a method and system for objectively evaluating the quality of speech in a voice communication system. a plurality of speech reference vectors is first obtained based on a plurality of clean speech samples. a corrupted speech signal is received and processed to determine a plurality of distortions derived from a plurality of distortion measures based on the plurality of speech reference vectors. the plurality of distortions are processed by a non-linear neural network model to generate a subjective score representing user acceptance of the corrupted speech signal. the non-linear neural network model is first trained on clean speech samples as well as corrupted speech samples through the use of backpropagation to obtain the weights and bias terms necessary to predict subjective scores from several objective measures. dated 2002-09-03
6446055,process control,"a process control system is described using a plurality of autonomous process cells. each process cell has data inputs, data outputs, processing logic, state variable, and link data specifying the other cells that provides its inputs. a scheduler triggers the plurality of cells as a whole to update their state. cell may be recursive and contain child cells, at least some of which are linked to the parent cell. the cells can be subject to mutation and a non-brittle program language is provided for the logic within the cells in order to stop mutation causing a cell to cease to function. cells within a neural network are provided with both fast and slow feedback mechanisms to improve their responsiveness to action reinforcement cycles. a mediated peer-to-peer network is described for use to simulate a tessellated virtual environment.",2002-09-03,"The title of the patent is process control and its abstract is a process control system is described using a plurality of autonomous process cells. each process cell has data inputs, data outputs, processing logic, state variable, and link data specifying the other cells that provides its inputs. a scheduler triggers the plurality of cells as a whole to update their state. cell may be recursive and contain child cells, at least some of which are linked to the parent cell. the cells can be subject to mutation and a non-brittle program language is provided for the logic within the cells in order to stop mutation causing a cell to cease to function. cells within a neural network are provided with both fast and slow feedback mechanisms to improve their responsiveness to action reinforcement cycles. a mediated peer-to-peer network is described for use to simulate a tessellated virtual environment. dated 2002-09-03"
6447460,method for automated exclusion of deep venous thrombosis,"an automated screening tool for the exclusion of deep venous thrombosis generally comprises a sensor array for gathering thermal data from the lower limbs of a patient suspected of dvt; a processor for automated analysis of the gathered data; and a display device for reporting the exclusion or non-exclusion of dvt. in the preferred embodiment of the present invention, a microprocessor based system is utilized to control the gathering of thermal data and, thereafter, the reporting of the gathered data to the processor. according to the preferred method for use of the present invention, the gathered thermal data is utilized, alone or in combination with other indicators, as a factor for exclusion of dvt based upon an implemented algorithm.according to the preferred embodiment of the present invention, a neural network or genetic algorithm is implemented within the processor in order to make an entirely objective determination relative the presence of dvt. this feature makes the present invention particularly adapted to multiple risk factor analysis, wherein factors such as calf circumference; positive homan sign; colorimetry reading of the limbs; recent surgery or trauma; history of dvt, phlebitic syndrome or venous insufficiency; or presently uncannulated veins may be considered along with thermographic data.",2002-09-10,"The title of the patent is method for automated exclusion of deep venous thrombosis and its abstract is an automated screening tool for the exclusion of deep venous thrombosis generally comprises a sensor array for gathering thermal data from the lower limbs of a patient suspected of dvt; a processor for automated analysis of the gathered data; and a display device for reporting the exclusion or non-exclusion of dvt. in the preferred embodiment of the present invention, a microprocessor based system is utilized to control the gathering of thermal data and, thereafter, the reporting of the gathered data to the processor. according to the preferred method for use of the present invention, the gathered thermal data is utilized, alone or in combination with other indicators, as a factor for exclusion of dvt based upon an implemented algorithm.according to the preferred embodiment of the present invention, a neural network or genetic algorithm is implemented within the processor in order to make an entirely objective determination relative the presence of dvt. this feature makes the present invention particularly adapted to multiple risk factor analysis, wherein factors such as calf circumference; positive homan sign; colorimetry reading of the limbs; recent surgery or trauma; history of dvt, phlebitic syndrome or venous insufficiency; or presently uncannulated veins may be considered along with thermographic data. dated 2002-09-10"
6453206,neural network for predicting values in non-linear functional mappings,"a neural network for predicting values in non-linear functional mappings having a single hidden layer function generator (12) and an output layer (40). the single hidden layer function generator (12) is operable to receive one or more mapping inputs (x1) and generate a plurality of terms (14) from each mapping input. the plurality of terms generated by the single hidden layer function generator (12) includes at least one trigonometric term selected from the group comprising sin(x1), sin(2x1), sin(3x1), cos(x1), cos(2xl), cos(3xl), cosec(xl), cotan(xl), and being free of gaussian and sigmoidal terms.",2002-09-17,"The title of the patent is neural network for predicting values in non-linear functional mappings and its abstract is a neural network for predicting values in non-linear functional mappings having a single hidden layer function generator (12) and an output layer (40). the single hidden layer function generator (12) is operable to receive one or more mapping inputs (x1) and generate a plurality of terms (14) from each mapping input. the plurality of terms generated by the single hidden layer function generator (12) includes at least one trigonometric term selected from the group comprising sin(x1), sin(2x1), sin(3x1), cos(x1), cos(2xl), cos(3xl), cosec(xl), cotan(xl), and being free of gaussian and sigmoidal terms. dated 2002-09-17"
6453284,multiple voice tracking system and method,"for tracking multiple, simultaneous voices, predicted tracking is used to follow individual voices through time, even when the voices are very similar in fundamental frequency. an acoustic waveform comprised of a group of voices is submitted to a frequency estimator, which may employ an average magnitude difference function (amdf) calculation to determine the voice fundamental frequencies that are present for each voice. these frequency estimates are then used as input values to a recurrent neural network that tracks each of the frequencies by predicting the current fundamental frequency value for each voice present based on past fundamental frequency values in order to disambiguate any fundamental frequency trajectories that may be converging in frequency.",2002-09-17,"The title of the patent is multiple voice tracking system and method and its abstract is for tracking multiple, simultaneous voices, predicted tracking is used to follow individual voices through time, even when the voices are very similar in fundamental frequency. an acoustic waveform comprised of a group of voices is submitted to a frequency estimator, which may employ an average magnitude difference function (amdf) calculation to determine the voice fundamental frequencies that are present for each voice. these frequency estimates are then used as input values to a recurrent neural network that tracks each of the frequencies by predicting the current fundamental frequency value for each voice present based on past fundamental frequency values in order to disambiguate any fundamental frequency trajectories that may be converging in frequency. dated 2002-09-17"
6453309,"method for correcting errors in parallel a/d conversion, corrector and parallel a/d converter","the invention pertains to a method and corrector (ic6) for correcting an error in a parallel analog-to-digital conversion. such a correctable error is caused by uncertainties in the reading of the states of parallel comparing elements (ic1, ic2, ic3, ic4) in the converter, said uncertainties being brought about by nonideality, such as non-simultaneous state latching. this error is corrected using a nonlinear cellular neural network preferably such that the real level of the phenomenon compared by means of comparing elements (ic1, ic2, ic3, ic4) is estimated by estimating the states corresponding to correct reading of the comparing elements (ic1, ic2, ic3, ic4) read temporally or otherwise erroneously.",2002-09-17,"The title of the patent is method for correcting errors in parallel a/d conversion, corrector and parallel a/d converter and its abstract is the invention pertains to a method and corrector (ic6) for correcting an error in a parallel analog-to-digital conversion. such a correctable error is caused by uncertainties in the reading of the states of parallel comparing elements (ic1, ic2, ic3, ic4) in the converter, said uncertainties being brought about by nonideality, such as non-simultaneous state latching. this error is corrected using a nonlinear cellular neural network preferably such that the real level of the phenomenon compared by means of comparing elements (ic1, ic2, ic3, ic4) is estimated by estimating the states corresponding to correct reading of the comparing elements (ic1, ic2, ic3, ic4) read temporally or otherwise erroneously. dated 2002-09-17"
6456239,method and apparatus for locating mobile tags,"a method and apparatus for determining tag location is disclosed. tag reference data may be stored, e.g., in the form of a lookup table, as a trained neural network, and so on, and used to determine the location of tags. readings used to determine tag location and/or preliminary tag locations may be filtered to produce reliable tag location indications. packages of user configurable parameters can be provided and used for the filtering of the preliminary tag locations. confidence levels may also be generated for determined tag locations and used, for example, to indicate how well an asset location system can distinguish between different tag locations.",2002-09-24,"The title of the patent is method and apparatus for locating mobile tags and its abstract is a method and apparatus for determining tag location is disclosed. tag reference data may be stored, e.g., in the form of a lookup table, as a trained neural network, and so on, and used to determine the location of tags. readings used to determine tag location and/or preliminary tag locations may be filtered to produce reliable tag location indications. packages of user configurable parameters can be provided and used for the filtering of the preliminary tag locations. confidence levels may also be generated for determined tag locations and used, for example, to indicate how well an asset location system can distinguish between different tag locations. dated 2002-09-24"
6456261,head/helmet mounted passive and active infrared imaging system with/without parallax,"a passive/active infrared imaging system apparatus for mounting on a head/helmet includes a passive infrared camera head pack having a removable narrow band filter cover, an objective lens, a beam splitter, an uncooled focal plane array (ufpa) package, an interface board, and a display unit such a liquid crystal display (lcd), with forward/back, up/down, and tilt adjustment functions fitting any mask, mounted in the front of said head/helmet for converting infrared light images into electronic signals. an electronic unit coupled between the ufpa of the infrared camera and the display unit, includes a controller for processing video signals from the infrared camera and supplying them to the display unit. the electronic circuit includes a wireless video & audio transceiver, a piezoelectric microphone, a voice controller, and a neural network pattern recognition chip. the display unit (such as lcd)] is inside the head pack and mounted on the head/helmet for converting electronic signals into visible light images, so that it is in front of eyes of a user, so that the user can directly view an external scene without blocking his normal vision, if the optical axis of the display unit is aligned with the optical axis of the objective lens, the system parallax is eliminated. a battery pack having a video controller board and battery is mounted on the rear of the head/helmet so that it gives the video output and power to the infrared system. an eye-safe near infrared laser diode with corresponding optical and electronic attachments mounted on the head/helmet illuminates targets to get images through same passive infrared system.",2002-09-24,"The title of the patent is head/helmet mounted passive and active infrared imaging system with/without parallax and its abstract is a passive/active infrared imaging system apparatus for mounting on a head/helmet includes a passive infrared camera head pack having a removable narrow band filter cover, an objective lens, a beam splitter, an uncooled focal plane array (ufpa) package, an interface board, and a display unit such a liquid crystal display (lcd), with forward/back, up/down, and tilt adjustment functions fitting any mask, mounted in the front of said head/helmet for converting infrared light images into electronic signals. an electronic unit coupled between the ufpa of the infrared camera and the display unit, includes a controller for processing video signals from the infrared camera and supplying them to the display unit. the electronic circuit includes a wireless video & audio transceiver, a piezoelectric microphone, a voice controller, and a neural network pattern recognition chip. the display unit (such as lcd)] is inside the head pack and mounted on the head/helmet for converting electronic signals into visible light images, so that it is in front of eyes of a user, so that the user can directly view an external scene without blocking his normal vision, if the optical axis of the display unit is aligned with the optical axis of the objective lens, the system parallax is eliminated. a battery pack having a video controller board and battery is mounted on the rear of the head/helmet so that it gives the video output and power to the infrared system. an eye-safe near infrared laser diode with corresponding optical and electronic attachments mounted on the head/helmet illuminates targets to get images through same passive infrared system. dated 2002-09-24"
6456697,device and method of channel effect compensation for telephone speech recognition,device and method of channel effect compensation for a telephone speech recognition system is disclosed. the telephone speech recognition system comprises a compensatory neutral network and a recognize. the compensatory neural network receives an input signal and compensates the input signal with a bias to generate an output signal. the compensatory neural network provides a plurality of first parameters to determine the bias. the recognizer is coupled to the compensatory neural network for classifying the output signal according to a plurality of second parameters in acoustic models to generate a recognition result and determine a recognition loss. the first parameters and second parameters are adjusted according to the recognition loss and an adjustment means during a training process.,2002-09-24,The title of the patent is device and method of channel effect compensation for telephone speech recognition and its abstract is device and method of channel effect compensation for a telephone speech recognition system is disclosed. the telephone speech recognition system comprises a compensatory neutral network and a recognize. the compensatory neural network receives an input signal and compensates the input signal with a bias to generate an output signal. the compensatory neural network provides a plurality of first parameters to determine the bias. the recognizer is coupled to the compensatory neural network for classifying the output signal according to a plurality of second parameters in acoustic models to generate a recognition result and determine a recognition loss. the first parameters and second parameters are adjusted according to the recognition loss and an adjustment means during a training process. dated 2002-09-24
6456989,neuro-fuzzy-integrated data processing system,"the present invention relates to a data processing system in a hierarchical network configuration for executing applicable data processes in a comprehensible and executable form. an object of the present invention is to allow data processing capabilities to be established with high precision in a short time based on a fuzzy-neuro-integrated concept. a fuzzy model is generated by a data processing system in the form of membership functions and fuzzy rules as technical information relating to a control target. according to this fuzzy model, a weight value of the connection between neurons is set and a pre-wired neural network is established. then, the data of the control target are learned by the neural network. the connection state and a weight value of the neural network after the learning enable tuning of the fuzzy model.",2002-09-24,"The title of the patent is neuro-fuzzy-integrated data processing system and its abstract is the present invention relates to a data processing system in a hierarchical network configuration for executing applicable data processes in a comprehensible and executable form. an object of the present invention is to allow data processing capabilities to be established with high precision in a short time based on a fuzzy-neuro-integrated concept. a fuzzy model is generated by a data processing system in the form of membership functions and fuzzy rules as technical information relating to a control target. according to this fuzzy model, a weight value of the connection between neurons is set and a pre-wired neural network is established. then, the data of the control target are learned by the neural network. the connection state and a weight value of the neural network after the learning enable tuning of the fuzzy model. dated 2002-09-24"
6456990,method for transforming a fuzzy logic used to simulate a technical process into a neural network,"a method for transforming a fuzzy logic system into a neural network, where, in order to simulate membership functions, sigmoid functions are linked together in such a way that, even after the optimization of the neural network, back-transformation of the neural network into a, fuzzy logic system is possible. the advantage of the method described is that a fuzzy logic system can be transformed, in particular component by component, into a neural network and the latter can then be optimized as a whole, i.e. all the components together. the possibility of back-transforming the trained neural network ultimately means that an optimized fuzzy logic system can be obtained. this advantageously makes it possible to use, in particular, standardized fuzzy system software for describing the optimized fuzzy logic system.",2002-09-24,"The title of the patent is method for transforming a fuzzy logic used to simulate a technical process into a neural network and its abstract is a method for transforming a fuzzy logic system into a neural network, where, in order to simulate membership functions, sigmoid functions are linked together in such a way that, even after the optimization of the neural network, back-transformation of the neural network into a, fuzzy logic system is possible. the advantage of the method described is that a fuzzy logic system can be transformed, in particular component by component, into a neural network and the latter can then be optimized as a whole, i.e. all the components together. the possibility of back-transforming the trained neural network ultimately means that an optimized fuzzy logic system can be obtained. this advantageously makes it possible to use, in particular, standardized fuzzy system software for describing the optimized fuzzy logic system. dated 2002-09-24"
6456991,classification method and apparatus based on boosting and pruning of multiple classifiers,"a boosting and pruning system and method for utilizing a plurality of neural networks, preferably those based on adaptive resonance theory (art), in order to increase pattern classification accuracy is presented. the method utilizes a plurality of n randomly ordered copies of the input data, which is passed to a plurality of sets of booster networks. each of the plurality of n randomly ordered copies of the input data is divided into a plurality of portions, preferably with an equal allocation of the data corresponding to each class for which recognition is desired. the plurality of portions is used to train the set of booster networks. the rules generated by the set of booster networks are then pruned in an intra-booster pruning step, which uses a pair-wise fuzzy and operation to determine rule overlap and to eliminate rules which are sufficiently similar. this process results in a set of intra-booster pruned booster networks. a similar pruning process is applied in an inter-booster pruning process, which eliminates rules from the intra-booster pruned networks with sufficient overlap. the final, derivative booster network captures the essence of the plurality of sets of booster networks and provides for higher classification accuracy than available using a single network.",2002-09-24,"The title of the patent is classification method and apparatus based on boosting and pruning of multiple classifiers and its abstract is a boosting and pruning system and method for utilizing a plurality of neural networks, preferably those based on adaptive resonance theory (art), in order to increase pattern classification accuracy is presented. the method utilizes a plurality of n randomly ordered copies of the input data, which is passed to a plurality of sets of booster networks. each of the plurality of n randomly ordered copies of the input data is divided into a plurality of portions, preferably with an equal allocation of the data corresponding to each class for which recognition is desired. the plurality of portions is used to train the set of booster networks. the rules generated by the set of booster networks are then pruned in an intra-booster pruning step, which uses a pair-wise fuzzy and operation to determine rule overlap and to eliminate rules which are sufficiently similar. this process results in a set of intra-booster pruned booster networks. a similar pruning process is applied in an inter-booster pruning process, which eliminates rules from the intra-booster pruned networks with sufficient overlap. the final, derivative booster network captures the essence of the plurality of sets of booster networks and provides for higher classification accuracy than available using a single network. dated 2002-09-24"
6459975,method for recognizing the severity of a vehicle collision,"in a method for recognizing the severity of a vehicle collision, wherein the output signal of one or more acceleration sensors is processed and fed to a neural network that controls a release unit for an occupant protection device, and further wherein several occupant protection devices can be selected by the release unit in accordance with the severity and course of the vehicle collision, the future time progression of the output signal of the acceleration sensor is predicted with the help of the neural network based on the acceleration sensor signal values during at least one defined point in time.",2002-10-01,"The title of the patent is method for recognizing the severity of a vehicle collision and its abstract is in a method for recognizing the severity of a vehicle collision, wherein the output signal of one or more acceleration sensors is processed and fed to a neural network that controls a release unit for an occupant protection device, and further wherein several occupant protection devices can be selected by the release unit in accordance with the severity and course of the vehicle collision, the future time progression of the output signal of the acceleration sensor is predicted with the help of the neural network based on the acceleration sensor signal values during at least one defined point in time. dated 2002-10-01"
6460408,method for determining relevant variables representing the pressure in the cylinders of an internal combustion engine,a method for determining internal pressure in the cylinders of an internal combustion engine by measuring and analyzing structure borne noise signals using a neural network.,2002-10-08,The title of the patent is method for determining relevant variables representing the pressure in the cylinders of an internal combustion engine and its abstract is a method for determining internal pressure in the cylinders of an internal combustion engine by measuring and analyzing structure borne noise signals using a neural network. dated 2002-10-08
6462586,selectability of maximum magnitudes for k-winner take all circuit,"a k-wta circuit for selecting inputs which have a maximum-magnitude and outputting the results. k-wta is very useful in pattern classification such as k-nearest neighbor classifier, hamming neural classifier and some cascaded classification systems, since one classifier can not achieve very high performance, however, if a small set of candidates can be provided, for example k(k<<n), then a simple classifier with small complementary feature sets can be cascaded to realize multi-stage classification. a k-wta network is necessary to implement this function. the circuit as disclosed in the invention provides a design for a circuit in which the number of inputs with maximum magnitudes at any one time can be chosen. the circuit can have a plurality of inputs each of which has a corresponding output. in one embodiment of the invention external inputs can select the number of maximum-magnitudes inputs to be selected through the use of a current positive feedback loop. depending on the number of maximum-magnitude inputs selected the circuit will output voltage levels on the outputs that correspond to inputs whose current levels have higher magnitudes when compared to other inputs. the circuit has the further advantages of reconfigurability, self-adaptivity, low complexity, high accuracy, expandability, and a large dynamic range.",2002-10-08,"The title of the patent is selectability of maximum magnitudes for k-winner take all circuit and its abstract is a k-wta circuit for selecting inputs which have a maximum-magnitude and outputting the results. k-wta is very useful in pattern classification such as k-nearest neighbor classifier, hamming neural classifier and some cascaded classification systems, since one classifier can not achieve very high performance, however, if a small set of candidates can be provided, for example k(k<<n), then a simple classifier with small complementary feature sets can be cascaded to realize multi-stage classification. a k-wta network is necessary to implement this function. the circuit as disclosed in the invention provides a design for a circuit in which the number of inputs with maximum magnitudes at any one time can be chosen. the circuit can have a plurality of inputs each of which has a corresponding output. in one embodiment of the invention external inputs can select the number of maximum-magnitudes inputs to be selected through the use of a current positive feedback loop. depending on the number of maximum-magnitude inputs selected the circuit will output voltage levels on the outputs that correspond to inputs whose current levels have higher magnitudes when compared to other inputs. the circuit has the further advantages of reconfigurability, self-adaptivity, low complexity, high accuracy, expandability, and a large dynamic range. dated 2002-10-08"
6463341,orthogonal functional basis method for function approximation,"an orthogonal functional basis method for function approximation is disclosed. starting with the orthogonal least squares method, a new subset selection method for selecting a set of appropriate basis functions is explained where, instead of picking a subset from a given functional basis, the subset is selected from a combination of functional basis evolved from a set of heterogeneous basis functions. the method results in a more efficient neural network.",2002-10-08,"The title of the patent is orthogonal functional basis method for function approximation and its abstract is an orthogonal functional basis method for function approximation is disclosed. starting with the orthogonal least squares method, a new subset selection method for selecting a set of appropriate basis functions is explained where, instead of picking a subset from a given functional basis, the subset is selected from a combination of functional basis evolved from a set of heterogeneous basis functions. the method results in a more efficient neural network. dated 2002-10-08"
6463371,system for intelligent control of a vehicle suspension based on soft computing,a reduced control system suitable for control of an active suspension system as a controlled plant is described. the reduced control system is configured to use a reduced sensor set for controlling the suspension without significant loss of control quality (accuracy) as compared to an optimal control system with an optimum sensor set. the control system calculates the information content provided by the reduced sensor set as compared to the information content provided by the optimum set. the control system also calculates the difference between the entropy production rate of the plant and the entropy production rate of the controller. a genetic optimizer is used to tune a fuzzy neural network in the reduced controller. a fitness function for the genetic optimizer provides optimum control accuracy in the reduced control system by minimizing the difference in entropy production while maximizing the sensor information content.,2002-10-08,The title of the patent is system for intelligent control of a vehicle suspension based on soft computing and its abstract is a reduced control system suitable for control of an active suspension system as a controlled plant is described. the reduced control system is configured to use a reduced sensor set for controlling the suspension without significant loss of control quality (accuracy) as compared to an optimal control system with an optimum sensor set. the control system calculates the information content provided by the reduced sensor set as compared to the information content provided by the optimum set. the control system also calculates the difference between the entropy production rate of the plant and the entropy production rate of the controller. a genetic optimizer is used to tune a fuzzy neural network in the reduced controller. a fitness function for the genetic optimizer provides optimum control accuracy in the reduced control system by minimizing the difference in entropy production while maximizing the sensor information content. dated 2002-10-08
6463423,multi-winners feedforward neural network,"a multi-winners feedforward neural network which is constituted while using neuron-elements consisting of the modified neural network, enables learning to be carried out autonomously regarding any problems, and is capable of being constituted by small number of lsi circuit simply. in the neural network which has a hierarchical structure consisting of one layer and/or a plurality of layers constituted by plural pieces of neurons, there is provided a controller for controlling the number of firing of the neurons, in which the number of firing of the neurons is more than two, in such a way that the number of firing of the neurons is restrained depending on a specified value and/or range of the specified value in every layer, thus preventing dissipation of the number of the neurons. the controller is a detection controller for causing a value to be the specified value and/or the range of the specified value, depending on obtained value while detecting power supply current supplied to the whole neurons in every layer. otherwise, the controller is a number of firing detection neuron for detecting the number of firing of the neurons in every layer, thus causing value of output of the number of firing detection neuron to be the specified value and/or the range of the specified value.",2002-10-08,"The title of the patent is multi-winners feedforward neural network and its abstract is a multi-winners feedforward neural network which is constituted while using neuron-elements consisting of the modified neural network, enables learning to be carried out autonomously regarding any problems, and is capable of being constituted by small number of lsi circuit simply. in the neural network which has a hierarchical structure consisting of one layer and/or a plurality of layers constituted by plural pieces of neurons, there is provided a controller for controlling the number of firing of the neurons, in which the number of firing of the neurons is more than two, in such a way that the number of firing of the neurons is restrained depending on a specified value and/or range of the specified value in every layer, thus preventing dissipation of the number of the neurons. the controller is a detection controller for causing a value to be the specified value and/or the range of the specified value, depending on obtained value while detecting power supply current supplied to the whole neurons in every layer. otherwise, the controller is a number of firing detection neuron for detecting the number of firing of the neurons in every layer, thus causing value of output of the number of firing detection neuron to be the specified value and/or the range of the specified value. dated 2002-10-08"
6463425,neural network assisted multi-spectral segmentation system,"a neural network assisted multi-spectral segmentation method and system. according to the invention, three images having different optical bands are acquired for the same micrographic scene of a biological sample. the images are processed and a cellular material map is generated identifying cellular material. the cellular material map is then applied to a neural network. the neural network classifies the cellular material map into nuclear objects and cytoplasmic objects by determining a threshold surface in the 3-dimensional space separating the cytoplasmic and nuclear regions. in another aspect, the neural network comprises a hardware-encoded algorithm in the form of a look-up table.",2002-10-08,"The title of the patent is neural network assisted multi-spectral segmentation system and its abstract is a neural network assisted multi-spectral segmentation method and system. according to the invention, three images having different optical bands are acquired for the same micrographic scene of a biological sample. the images are processed and a cellular material map is generated identifying cellular material. the cellular material map is then applied to a neural network. the neural network classifies the cellular material map into nuclear objects and cytoplasmic objects by determining a threshold surface in the 3-dimensional space separating the cytoplasmic and nuclear regions. in another aspect, the neural network comprises a hardware-encoded algorithm in the form of a look-up table. dated 2002-10-08"
6463438,neural network for cell image analysis for identification of abnormal cells,"a neural network is used in a system to detect abnormalities in cells, including cancer in bladder tissue cells. the system has an image analysis system for generating data representative of imaging variables from an image of stained cells. the set of data is provided to a neural network which has been trained to detect abnormalities from known tissue cells with respect to the data from the same set of imaging variables. a conventional sigmoid-activated neural network, or alternatively, a hybrid neural network having a combination of sigmoid, gaussian and sinusoidal activation functions may be utilized. the trained neural network applies a set of weight factors obtained during training to the data to classify the unknown tissue cell as normal or abnormal.",2002-10-08,"The title of the patent is neural network for cell image analysis for identification of abnormal cells and its abstract is a neural network is used in a system to detect abnormalities in cells, including cancer in bladder tissue cells. the system has an image analysis system for generating data representative of imaging variables from an image of stained cells. the set of data is provided to a neural network which has been trained to detect abnormalities from known tissue cells with respect to the data from the same set of imaging variables. a conventional sigmoid-activated neural network, or alternatively, a hybrid neural network having a combination of sigmoid, gaussian and sinusoidal activation functions may be utilized. the trained neural network applies a set of weight factors obtained during training to the data to classify the unknown tissue cell as normal or abnormal. dated 2002-10-08"
6464392,tactical thermal luminescence sensor for ground path contamination detection,"chemical agent warfare materials and their simulant liquids are identified on terrestrial surfaces at a distance by recognizing the contaminant's infrared fingerprint spectrum brought out in thermal luminescence (tl). suspect surfaces are irradiated with microwave light that is absorbed into the surface and, subsequently, tl is released by the surface. an optics receiver collects the released tl radiant light, and a data acquisition system searches this tl radiant flux for the contaminant's fingerprint infrared spectrum. a decision on the presence or absence of any-of-n contaminants is done by a neural network system that acts as a filter through real-time pattern recognition of the contaminant's unique infrared absorption or emission spectra.",2002-10-15,"The title of the patent is tactical thermal luminescence sensor for ground path contamination detection and its abstract is chemical agent warfare materials and their simulant liquids are identified on terrestrial surfaces at a distance by recognizing the contaminant's infrared fingerprint spectrum brought out in thermal luminescence (tl). suspect surfaces are irradiated with microwave light that is absorbed into the surface and, subsequently, tl is released by the surface. an optics receiver collects the released tl radiant light, and a data acquisition system searches this tl radiant flux for the contaminant's fingerprint infrared spectrum. a decision on the presence or absence of any-of-n contaminants is done by a neural network system that acts as a filter through real-time pattern recognition of the contaminant's unique infrared absorption or emission spectra. dated 2002-10-15"
6466314,reticle design inspection system,"a method of reticle inspection, comprising generating a test reticle comprising a plurality of test pattern-features thereon; manufacturing a wafer using the reticle; and determining a transfer of at least one of said plurality of pattern features from said reticle to said wafer. preferably, a neural network is trained using the determination. preferably, a reticle is inspected by running detected defects through the neural network to determine if the detected defect has a consequence.",2002-10-15,"The title of the patent is reticle design inspection system and its abstract is a method of reticle inspection, comprising generating a test reticle comprising a plurality of test pattern-features thereon; manufacturing a wafer using the reticle; and determining a transfer of at least one of said plurality of pattern features from said reticle to said wafer. preferably, a neural network is trained using the determination. preferably, a reticle is inspected by running detected defects through the neural network to determine if the detected defect has a consequence. dated 2002-10-15"
6466888,neural network system for estimation of aircraft flight data,"input parameters which correspond to operational flight data of an aircraft within a predetermined flight domain, are defined through measured variable state parameters generated during aircraft flight, utilizing a neural network trained by exemplars corresponding to such variable state parameters and reference information on the aircraft for data processing of real time values of the variable state parameter measurement to calculate values of the input parameters and provide a corresponding output as a reliable estimate of aircraft flight data such as airspeed, sideslip and angle of attack.",2002-10-15,"The title of the patent is neural network system for estimation of aircraft flight data and its abstract is input parameters which correspond to operational flight data of an aircraft within a predetermined flight domain, are defined through measured variable state parameters generated during aircraft flight, utilizing a neural network trained by exemplars corresponding to such variable state parameters and reference information on the aircraft for data processing of real time values of the variable state parameter measurement to calculate values of the input parameters and provide a corresponding output as a reliable estimate of aircraft flight data such as airspeed, sideslip and angle of attack. dated 2002-10-15"
6466924,verification method of neural network and verification apparatus thereof,"a verification method and a verification apparatus of a neural network for guaranteeing the operation of the neural network to any input signals which might be inputted. the neural network verification apparatus comprises a data file storage section for storing a verification data file dfk concerning on the structure of the neural network to be verified, a network structuring section for structuring a neural network net to be verified by reading the neural network data file dfn from the data file storage section, an input data generating section for generating initial interval input vector database dbi by reading the verification data file dfk from the data file storage section, a verification executing section for executing the verification of the neural network net based on the neural network net structured by the network structuring section and the initial interval input vector database dbi created by the input data generating section and for generating a verification result database dbo based on the verification result, and others. the verification apparatus generates an interval input vector signal per input signal subspace which is obtained by dividing an input signal space formed by the input signal to the neural network net into a plurality of input. signal subspaces and accumulates them in a stack.",2002-10-15,"The title of the patent is verification method of neural network and verification apparatus thereof and its abstract is a verification method and a verification apparatus of a neural network for guaranteeing the operation of the neural network to any input signals which might be inputted. the neural network verification apparatus comprises a data file storage section for storing a verification data file dfk concerning on the structure of the neural network to be verified, a network structuring section for structuring a neural network net to be verified by reading the neural network data file dfn from the data file storage section, an input data generating section for generating initial interval input vector database dbi by reading the verification data file dfk from the data file storage section, a verification executing section for executing the verification of the neural network net based on the neural network net structured by the network structuring section and the initial interval input vector database dbi created by the input data generating section and for generating a verification result database dbo based on the verification result, and others. the verification apparatus generates an interval input vector signal per input signal subspace which is obtained by dividing an input signal space formed by the input signal to the neural network net into a plurality of input. signal subspaces and accumulates them in a stack. dated 2002-10-15"
6466925,method and means for simulation of communication systems,"a method and means for speeding up simulation in a communication system use importance sampling (is) involving fitted stochastic processes that are artificially altered and fed to the system as inputs in order to shorten the time between so called rare events. the simulation needs to be compensated for the alteration so that the simulation remains unbiased. the problem of finding optimal alternations can be reduced to find optimal bias parameters for the stochastic processes. here an artificial neural network (ann) together with a statistical bias optimizer is used in a first phase where the ann is trained with input parameters and their corresponding optimal output bias parameters. in an application phase, only the trained ann is used to very quickly provide optimal output samples for the final simulation.",2002-10-15,"The title of the patent is method and means for simulation of communication systems and its abstract is a method and means for speeding up simulation in a communication system use importance sampling (is) involving fitted stochastic processes that are artificially altered and fed to the system as inputs in order to shorten the time between so called rare events. the simulation needs to be compensated for the alteration so that the simulation remains unbiased. the problem of finding optimal alternations can be reduced to find optimal bias parameters for the stochastic processes. here an artificial neural network (ann) together with a statistical bias optimizer is used in a first phase where the ann is trained with input parameters and their corresponding optimal output bias parameters. in an application phase, only the trained ann is used to very quickly provide optimal output samples for the final simulation. dated 2002-10-15"
6468069,automatically optimized combustion control,"systems and methods are disclosed that optimize the combustion process in various reactors, furnaces, and internal combustion engines. video cameras are used to evaluate the combustion flame grade. depending on the desired form, standard or special video devices, or beam scanning devices, are used to image the combustion flame and by-products. the video device generates and outputs image signals during various phases of, and at various locations in, the combustion process. other forms of sensors monitor and generate data signals defining selected parameters of the combustion process, such as air flow, fuel flow, turbulence, exhaust and inlet valve openings, etc. in a preferred form, a neural networks initially processes the image data and characterizes the combustion flame. a fuzzy logic controller and associated fuzzy logic rule base analyzes the image data from the neural network, along with other sensor information. the fuzzy logic controller determines and generates control signals defining adjustments necessary to optimize the combustion process.",2002-10-22,"The title of the patent is automatically optimized combustion control and its abstract is systems and methods are disclosed that optimize the combustion process in various reactors, furnaces, and internal combustion engines. video cameras are used to evaluate the combustion flame grade. depending on the desired form, standard or special video devices, or beam scanning devices, are used to image the combustion flame and by-products. the video device generates and outputs image signals during various phases of, and at various locations in, the combustion process. other forms of sensors monitor and generate data signals defining selected parameters of the combustion process, such as air flow, fuel flow, turbulence, exhaust and inlet valve openings, etc. in a preferred form, a neural networks initially processes the image data and characterizes the combustion flame. a fuzzy logic controller and associated fuzzy logic rule base analyzes the image data from the neural network, along with other sensor information. the fuzzy logic controller determines and generates control signals defining adjustments necessary to optimize the combustion process. dated 2002-10-22"
6470261,automatic freeway incident detection system and method using artificial neural network and genetic algorithms,design of a neural network for automatic detection of incidents on a freeway is described. a neural network is trained using a combination of both back-propagation and genetic algorithm-based methods for optimizing the design of the neural network. the back-propagation and genetic algorithm work together in a collaborative manner in the neural network design. the training starts with incremental learning based on the instantaneous error and the global total error is accumulated for batch updating at the end of the training data being presented to the neural network. the genetic algorithm directly evaluates the performance of multiple sets of neural networks in parallel and then use the analyzed results to breed new neural networks that tend to be better suited to the problems at hand.,2002-10-22,The title of the patent is automatic freeway incident detection system and method using artificial neural network and genetic algorithms and its abstract is design of a neural network for automatic detection of incidents on a freeway is described. a neural network is trained using a combination of both back-propagation and genetic algorithm-based methods for optimizing the design of the neural network. the back-propagation and genetic algorithm work together in a collaborative manner in the neural network design. the training starts with incremental learning based on the instantaneous error and the global total error is accumulated for batch updating at the end of the training data being presented to the neural network. the genetic algorithm directly evaluates the performance of multiple sets of neural networks in parallel and then use the analyzed results to breed new neural networks that tend to be better suited to the problems at hand. dated 2002-10-22
6473746,method of verifying pretrained neural net mapping for use in safety-critical software,"a method of verifying pretrained, static, feedforward neural network mapping software using lipschitz constants for determining bounds on output values and estimation errors is disclosed. by way of example, two cases of interest from the point of view of safety-critical software, like aircraft fuel gauging systems, are discussed. the first case is the simpler case of when neural net mapping software is trained to replace look-up table mapping software. a detailed verification procedure is provided to establish functional equivalence of the neural net and look-up table mapping functions on the entire range of inputs accepted by the look-up table mapping function. the second case is when a neural net is trained to estimate the quantity of interest form the process (such as fuel mass, for example) from redundant and noisy sensor signals. given upper and lower bounds on sensor noises and on modeling inaccuracies, it is demonstrated how to verify the performance of such a neural network estimator (a &#8220;black box&#8221;) when compared to a true value of the estimated quantity.",2002-10-29,"The title of the patent is method of verifying pretrained neural net mapping for use in safety-critical software and its abstract is a method of verifying pretrained, static, feedforward neural network mapping software using lipschitz constants for determining bounds on output values and estimation errors is disclosed. by way of example, two cases of interest from the point of view of safety-critical software, like aircraft fuel gauging systems, are discussed. the first case is the simpler case of when neural net mapping software is trained to replace look-up table mapping software. a detailed verification procedure is provided to establish functional equivalence of the neural net and look-up table mapping functions on the entire range of inputs accepted by the look-up table mapping function. the second case is when a neural net is trained to estimate the quantity of interest form the process (such as fuel mass, for example) from redundant and noisy sensor signals. given upper and lower bounds on sensor noises and on modeling inaccuracies, it is demonstrated how to verify the performance of such a neural network estimator (a &#8220;black box&#8221;) when compared to a true value of the estimated quantity. dated 2002-10-29"
6473747,neural network trajectory command controller,an apparatus and method for controlling trajectory of an object (47) to a first predetermined position. the apparatus has an input layer (22) having nodes (22a-22f) for receiving input data indicative of the first predetermined position. first weighted connections (28) are connected to the nodes of the input layer (22). each of the first weighted connections (28) have a coefficient for weighting the input data. an output layer (26) having nodes (26a-26e) connected to the first weighted connections (28) determines trajectory data based upon the first weighted input data. the trajectory of the object is controlled based upon the determined trajectory data.,2002-10-29,The title of the patent is neural network trajectory command controller and its abstract is an apparatus and method for controlling trajectory of an object (47) to a first predetermined position. the apparatus has an input layer (22) having nodes (22a-22f) for receiving input data indicative of the first predetermined position. first weighted connections (28) are connected to the nodes of the input layer (22). each of the first weighted connections (28) have a coefficient for weighting the input data. an output layer (26) having nodes (26a-26e) connected to the first weighted connections (28) determines trajectory data based upon the first weighted input data. the trajectory of the object is controlled based upon the determined trajectory data. dated 2002-10-29
6474153,predicting system and predicting method configured to predict inflow volume of rainwater,"a predicting system configured to predict an inflow volume of rainwater includes a rainfall volume measuring unit for measuring a rainfall volume that has fallen by a present time, a rainfall volume predicting unit for predicting a rainfall volume in a future, and an inflow-volume measuring unit for measuring an inflow volume of rainwater that has flown into an objective facility by the present time. the system also includes a model identification unit, which has a non-linear model having a neural-network model for calculating an inflow volume of rainwater from a rainfall volume, and which has a parameter determining part for determining a degree and coefficient parameters of the non-linear model, based on the rainfall volume that has fallen by the present time measured by the rainfall volume measuring unit and the inflow volume of rainwater that has flown into the objective facility by the present time measured by the inflow-volume measuring unit. an inflow-volume predicting unit is adapted to predict an inflow volume of rainwater flowing into the objective facility in the future, based on the rainfall volume in the future predicted by the rainfall volume predicting unit, according to the non-linear model determined by the model identification unit. thus, the inflow volume of rainwater flowing into the objective facility can be predicted more accurately even if there is a non-linear relationship between the rainfall volume and the inflow volume of rainwater.",2002-11-05,"The title of the patent is predicting system and predicting method configured to predict inflow volume of rainwater and its abstract is a predicting system configured to predict an inflow volume of rainwater includes a rainfall volume measuring unit for measuring a rainfall volume that has fallen by a present time, a rainfall volume predicting unit for predicting a rainfall volume in a future, and an inflow-volume measuring unit for measuring an inflow volume of rainwater that has flown into an objective facility by the present time. the system also includes a model identification unit, which has a non-linear model having a neural-network model for calculating an inflow volume of rainwater from a rainfall volume, and which has a parameter determining part for determining a degree and coefficient parameters of the non-linear model, based on the rainfall volume that has fallen by the present time measured by the rainfall volume measuring unit and the inflow volume of rainwater that has flown into the objective facility by the present time measured by the inflow-volume measuring unit. an inflow-volume predicting unit is adapted to predict an inflow volume of rainwater flowing into the objective facility in the future, based on the rainfall volume in the future predicted by the rainfall volume predicting unit, according to the non-linear model determined by the model identification unit. thus, the inflow volume of rainwater flowing into the objective facility can be predicted more accurately even if there is a non-linear relationship between the rainfall volume and the inflow volume of rainwater. dated 2002-11-05"
6476391,infrared imaging system for advanced rescue vision system,"the system according to the invention uses cutting edge technologies such as uncooled staring focal plane detector array, hot pressed polycrystal objective lens, helmet mounted display using transparent image combiner, and neural network image colorization and recognition to dramatically enhance the system performance and reduce the weight and cost.the helmet mounted infrared imaging system can:detect and recognize flames, humans and other objects,reduce the weight of the helmet components (including camera head and combiner) to less than 0.5 pounds,view simultaneously visible and invisible surroundings without hindering operations.operate the imaging system hands-off; andtransmit the imaging data to and receive the map from a remote sight.alternative embodimentin an alternative embodiment, the infrared camera is mounted centered in front of the display.",2002-11-05,"The title of the patent is infrared imaging system for advanced rescue vision system and its abstract is the system according to the invention uses cutting edge technologies such as uncooled staring focal plane detector array, hot pressed polycrystal objective lens, helmet mounted display using transparent image combiner, and neural network image colorization and recognition to dramatically enhance the system performance and reduce the weight and cost.the helmet mounted infrared imaging system can:detect and recognize flames, humans and other objects,reduce the weight of the helmet components (including camera head and combiner) to less than 0.5 pounds,view simultaneously visible and invisible surroundings without hindering operations.operate the imaging system hands-off; andtransmit the imaging data to and receive the map from a remote sight.alternative embodimentin an alternative embodiment, the infrared camera is mounted centered in front of the display. dated 2002-11-05"
6477469,coarse-to-fine self-organizing map for automatic electrofacies ordering,"a method for ordering electrofacies to assist in identification of mineral deposits is disclosed. automated ordering of electrofacies allows geologists to draw inferences about the geological settings in which sediment deposit occurred without directly examining core samples or outcrops. the electrofacies order is determined by (a) training a one-dimensional linear self-organizing map to form an initial neural network that includes a plurality of neurons. the number of neurons is small in comparison to the number of electrofacies kernels (i.e., not greater than one-third the number of electrofacies kernels). (b1) a neuron is selected from the initial neural network. in the next step (b2), the processor determines if more than one electrofacies kernel is attached to the neuron. (b3) if more than one electrofacies kernel is attached to the neuron, then the neuron is split into the number of electrofacies kernels attached to the neuron. (c) steps (b1)-(b3) are repeated until all neurons in the initial neural network have been processed. in the next step, (d) a self-organizing map is trained to form a final neural network using the split neurons in the initial neural network as initial state. (e) steps (b1)-(d) are repeated if more than one electrofacies kernel is attached to a neuron with the initial neural network equal to the final neural network. in the last step (f), each electrofacies kernel corresponding to a neuron in the final neural network is correlated to an order number.",2002-11-05,"The title of the patent is coarse-to-fine self-organizing map for automatic electrofacies ordering and its abstract is a method for ordering electrofacies to assist in identification of mineral deposits is disclosed. automated ordering of electrofacies allows geologists to draw inferences about the geological settings in which sediment deposit occurred without directly examining core samples or outcrops. the electrofacies order is determined by (a) training a one-dimensional linear self-organizing map to form an initial neural network that includes a plurality of neurons. the number of neurons is small in comparison to the number of electrofacies kernels (i.e., not greater than one-third the number of electrofacies kernels). (b1) a neuron is selected from the initial neural network. in the next step (b2), the processor determines if more than one electrofacies kernel is attached to the neuron. (b3) if more than one electrofacies kernel is attached to the neuron, then the neuron is split into the number of electrofacies kernels attached to the neuron. (c) steps (b1)-(b3) are repeated until all neurons in the initial neural network have been processed. in the next step, (d) a self-organizing map is trained to form a final neural network using the split neurons in the initial neural network as initial state. (e) steps (b1)-(d) are repeated if more than one electrofacies kernel is attached to a neuron with the initial neural network equal to the final neural network. in the last step (f), each electrofacies kernel corresponding to a neuron in the final neural network is correlated to an order number. dated 2002-11-05"
6477516,system and method for predicting parameter of hydrocarbon with spectroscopy and neural networks,"a method for predicting parameters of hydrocarbons includes the steps of generating an nmr spectrum of a sample of a hydrocarbon having different hydrogen or carbon types related to structures or sample composition; dividing the nmr spectrum into regions corresponding to the different hydrogen or carbon types related to structures or sample composition; evaluating different spectral regions by either (i) determining average molecular parameters, and (ii) quantifying a signal intensity of said at least one region of said different regions, based upon a desired parameter to be predicted so as to provide spectrum extracted quantities; and applying the spectrum extracted quantities to a trained neural network trained to correlate spectrum extracted quantities with hydrocarbon parameters so as to predict the desired parameters from the spectrum extracted quantity. a system is also provided.",2002-11-05,"The title of the patent is system and method for predicting parameter of hydrocarbon with spectroscopy and neural networks and its abstract is a method for predicting parameters of hydrocarbons includes the steps of generating an nmr spectrum of a sample of a hydrocarbon having different hydrogen or carbon types related to structures or sample composition; dividing the nmr spectrum into regions corresponding to the different hydrogen or carbon types related to structures or sample composition; evaluating different spectral regions by either (i) determining average molecular parameters, and (ii) quantifying a signal intensity of said at least one region of said different regions, based upon a desired parameter to be predicted so as to provide spectrum extracted quantities; and applying the spectrum extracted quantities to a trained neural network trained to correlate spectrum extracted quantities with hydrocarbon parameters so as to predict the desired parameters from the spectrum extracted quantity. a system is also provided. dated 2002-11-05"
6480299,color printer characterization using optimization theory and neural networks,"a color management method/apparatus generates image color matching and international color consortium (icc) color printer profiles using a reduced number of color patch measurements. color printer characterization, and the generation of icc profiles usually require a large number of measured data points or color patches and complex interpolation techniques. this invention provides an optimization method/apparatus for performing lab to cmyk color space conversion, gamut mapping, and gray component replacement. a gamut trained network architecture performs lab to cmyk color space conversion to generate a color profile lookup table for a color printer, or alternatively, to directly control the color printer in accordance with the a plurality of color patches that accurately. represent the gamut of the color printer. more specifically, a feed forward neural network is trained using an ansi/it-8 basic data set consisting of 182 data points or color patches, or using a lesser number of data points such as 150 or 101 data points when redundant data points within linear regions of the 182 data point set are removed. a 5-to-7 neuron neural network architecture is preferred to perform the lab to cmyk color space conversion as the profile lookup table is built, or as the printer is directly controlled. for each cmyk signal, an ink optimization criteria is applied, to thereby control ink parameters such as the total quantity of ink in each cmyk ink printed pixel, and/or to control the total quantity of black ink in each cmyk ink printed pixel.",2002-11-12,"The title of the patent is color printer characterization using optimization theory and neural networks and its abstract is a color management method/apparatus generates image color matching and international color consortium (icc) color printer profiles using a reduced number of color patch measurements. color printer characterization, and the generation of icc profiles usually require a large number of measured data points or color patches and complex interpolation techniques. this invention provides an optimization method/apparatus for performing lab to cmyk color space conversion, gamut mapping, and gray component replacement. a gamut trained network architecture performs lab to cmyk color space conversion to generate a color profile lookup table for a color printer, or alternatively, to directly control the color printer in accordance with the a plurality of color patches that accurately. represent the gamut of the color printer. more specifically, a feed forward neural network is trained using an ansi/it-8 basic data set consisting of 182 data points or color patches, or using a lesser number of data points such as 150 or 101 data points when redundant data points within linear regions of the 182 data point set are removed. a 5-to-7 neuron neural network architecture is preferred to perform the lab to cmyk color space conversion as the profile lookup table is built, or as the printer is directly controlled. for each cmyk signal, an ink optimization criteria is applied, to thereby control ink parameters such as the total quantity of ink in each cmyk ink printed pixel, and/or to control the total quantity of black ink in each cmyk ink printed pixel. dated 2002-11-12"
6480621,statistical classifier with reduced weight memory requirements,"a neural network has reduced requirements for storing intermodal weight values, as a result of a dual-precision training process. in the forward propagation of training samples, low-resolution weight values are employed. during back-propagation of errors to train the network, higher-resolution values are used. after training, only the lower resolution values need to be stored for further run-time operation, thereby reducing memory requirements.",2002-11-12,"The title of the patent is statistical classifier with reduced weight memory requirements and its abstract is a neural network has reduced requirements for storing intermodal weight values, as a result of a dual-precision training process. in the forward propagation of training samples, low-resolution weight values are employed. during back-propagation of errors to train the network, higher-resolution values are used. after training, only the lower resolution values need to be stored for further run-time operation, thereby reducing memory requirements. dated 2002-11-12"
6480792,fatigue monitoring systems and methods incorporating neural networks,a fatigue monitoring system and method is disclosed in which a stream of data relating to the stresses experienced at a plurality of locations over the structure during operation is applied to a neural network trained to remove data stream values deemed to be in error. the data from the neural network is then processed to determine the fatigue life.,2002-11-12,The title of the patent is fatigue monitoring systems and methods incorporating neural networks and its abstract is a fatigue monitoring system and method is disclosed in which a stream of data relating to the stresses experienced at a plurality of locations over the structure during operation is applied to a neural network trained to remove data stream values deemed to be in error. the data from the neural network is then processed to determine the fatigue life. dated 2002-11-12
6484133,sensor response rate accelerator,"an apparatus and method for sensor signal prediction and for improving sensor signal response time, is disclosed. an adaptive filter or an artificial neural network is utilized to provide predictive sensor signal output and is further used to reduce sensor response time delay.",2002-11-19,"The title of the patent is sensor response rate accelerator and its abstract is an apparatus and method for sensor signal prediction and for improving sensor signal response time, is disclosed. an adaptive filter or an artificial neural network is utilized to provide predictive sensor signal output and is further used to reduce sensor response time delay. dated 2002-11-19"
6490527,method for characterization of rock strata in drilling operations,"a method and system for determining the relative strength and classification of rock strata in near real-time during drilling operations is provided for use in underground mines. neural network technology is used to classify mine roof strata in terms of, for example, relative strength or strength index as the roof bolt hole is being drilled (i.e., in near real-time). measurements taken while a layer of the rock strata is being drilled are used to compute the specific energy input and convert these data to suitably scaled features. a neural network is then used to classify the strength of the layer. the neural network can be trained using data of known rock strata classifications prior to using it to classify new measurements. the present system allows for detection of unsafe conditions within the rock strata being drilled, and allows appropriate warnings to be issued in near real-time so that appropriate actions can be taken.",2002-12-03,"The title of the patent is method for characterization of rock strata in drilling operations and its abstract is a method and system for determining the relative strength and classification of rock strata in near real-time during drilling operations is provided for use in underground mines. neural network technology is used to classify mine roof strata in terms of, for example, relative strength or strength index as the roof bolt hole is being drilled (i.e., in near real-time). measurements taken while a layer of the rock strata is being drilled are used to compute the specific energy input and convert these data to suitably scaled features. a neural network is then used to classify the strength of the layer. the neural network can be trained using data of known rock strata classifications prior to using it to classify new measurements. the present system allows for detection of unsafe conditions within the rock strata being drilled, and allows appropriate warnings to be issued in near real-time so that appropriate actions can be taken. dated 2002-12-03"
6490571,method and apparatus for neural networking using semantic attractor architecture,"a semantic attractor memory uses an evolving neural network architecture and learning rules derived from the study of human language acquisition and change to store, process and retrieve information. the architecture is based on multiple layer channels, with random connections from one layer to the next. one or more layers are devoted to processing input information. at least one processing layer is provided. one or more layers are devoted to processing outputs and feedback is provided from the outputs back to the processing layer or layers. inputs from parallel channels are also provided to the one or more processing layers. with the exception of the feedback loop and central processing layers, the network is feedforward unless it is employed in a hybrid back-propagation configuration. the learning rules are based on non-stationary statistical processes, such as the polya process or the processes leading to bose-einstein statistics, again derived from considerations of human language acquisition. the invention provides rapid, unsupervised processing of complex data sets, such as imagery or continuous human speech, and a means to capture successful processing or pattern classification constellations for implementation in other networks.",2002-12-03,"The title of the patent is method and apparatus for neural networking using semantic attractor architecture and its abstract is a semantic attractor memory uses an evolving neural network architecture and learning rules derived from the study of human language acquisition and change to store, process and retrieve information. the architecture is based on multiple layer channels, with random connections from one layer to the next. one or more layers are devoted to processing input information. at least one processing layer is provided. one or more layers are devoted to processing outputs and feedback is provided from the outputs back to the processing layer or layers. inputs from parallel channels are also provided to the one or more processing layers. with the exception of the feedback loop and central processing layers, the network is feedforward unless it is employed in a hybrid back-propagation configuration. the learning rules are based on non-stationary statistical processes, such as the polya process or the processes leading to bose-einstein statistics, again derived from considerations of human language acquisition. the invention provides rapid, unsupervised processing of complex data sets, such as imagery or continuous human speech, and a means to capture successful processing or pattern classification constellations for implementation in other networks. dated 2002-12-03"
6490573,neural network for modeling ecological and biological systems,"a method of operating a neural network for ecological and biological system modeling having a plurality of hidden layer neurons said method comprising: a plurality of network inputs and at least one network output, said plurality of neurons, each receiving a plurality of inputs applied to the network, reproduces the network using a regression model, and compares the output values with given target values, and using the comparison and goodness of fit to set the learning rules. the network does not require repetitive training and yields a global minimum for each given set of input variables.",2002-12-03,"The title of the patent is neural network for modeling ecological and biological systems and its abstract is a method of operating a neural network for ecological and biological system modeling having a plurality of hidden layer neurons said method comprising: a plurality of network inputs and at least one network output, said plurality of neurons, each receiving a plurality of inputs applied to the network, reproduces the network using a regression model, and compares the output values with given target values, and using the comparison and goodness of fit to set the learning rules. the network does not require repetitive training and yields a global minimum for each given set of input variables. dated 2002-12-03"
6493687,apparatus and method for detecting glass break,"a glass break detector is disclosed that uses a neural network to determine if an audio signal is breaking glass. a characteristic extraction unit is used to extract a set of signal characteristics from a time domain signal based on the audio signal. the set of signal characteristics is the set of the magnitudes of the discrete fourier transform coefficients of an acquired time domain signal, or the fourier transform coefficients themselves. a classifier is connected to the characteristic extraction unit. it is a two-layer neural network that uses the set of signal characteristics to accurately determine whether the acquired time domain signal represents breaking glass.",2002-12-10,"The title of the patent is apparatus and method for detecting glass break and its abstract is a glass break detector is disclosed that uses a neural network to determine if an audio signal is breaking glass. a characteristic extraction unit is used to extract a set of signal characteristics from a time domain signal based on the audio signal. the set of signal characteristics is the set of the magnitudes of the discrete fourier transform coefficients of an acquired time domain signal, or the fourier transform coefficients themselves. a classifier is connected to the characteristic extraction unit. it is a two-layer neural network that uses the set of signal characteristics to accurately determine whether the acquired time domain signal represents breaking glass. dated 2002-12-10"
6493689,neural net controller for noise and vibration reduction,"two neural networks are used to control adaptively a vibration and noise-producing plant. the first neural network, the emulator, models the complex, nonlinear output of the plant with respect to certain controls and stimuli applied to the plant. the second neural network, the controller, calculates a control signal which affects the vibration and noise producing characteristics of the plant. by using the emulator model to calculate the nonlinear plant gradient, the controller matrix coefficients can be adapted by backpropagation of the plant gradient to produce a control signal which results in the minimum vibration and noise possible, given the current operating characteristics of the plant.",2002-12-10,"The title of the patent is neural net controller for noise and vibration reduction and its abstract is two neural networks are used to control adaptively a vibration and noise-producing plant. the first neural network, the emulator, models the complex, nonlinear output of the plant with respect to certain controls and stimuli applied to the plant. the second neural network, the controller, calculates a control signal which affects the vibration and noise producing characteristics of the plant. by using the emulator model to calculate the nonlinear plant gradient, the controller matrix coefficients can be adapted by backpropagation of the plant gradient to produce a control signal which results in the minimum vibration and noise possible, given the current operating characteristics of the plant. dated 2002-12-10"
6496761,optimization control method for shock absorber,a control system for optimizing the performance of a vehicle suspension system by controlling the damping factor of one or more shock absorbers is described. the control system uses a fitness (performance) function that is based on the physical laws of minimum entropy. the control system uses a fuzzy neural network that is trained by a genetic analyzer. the genetic analyzer uses a fitness function that maximizes information while minimizing entropy production. the fitness function uses a difference between the time differential of entropy from a control signal produced in a learning control module and the time differential of the entropy calculated by a model of the suspension system that uses the control signal as an input the entropy calculation is based on a dynamic model of an equation of motion for the suspension system such that the suspension system is treated as an open dynamic system.,2002-12-17,The title of the patent is optimization control method for shock absorber and its abstract is a control system for optimizing the performance of a vehicle suspension system by controlling the damping factor of one or more shock absorbers is described. the control system uses a fitness (performance) function that is based on the physical laws of minimum entropy. the control system uses a fuzzy neural network that is trained by a genetic analyzer. the genetic analyzer uses a fitness function that maximizes information while minimizing entropy production. the fitness function uses a difference between the time differential of entropy from a control signal produced in a learning control module and the time differential of the entropy calculated by a model of the suspension system that uses the control signal as an input the entropy calculation is based on a dynamic model of an equation of motion for the suspension system such that the suspension system is treated as an open dynamic system. dated 2002-12-17
6496812,method and system for measuring and valuing contributions by group members to the achievement of a group goal,"a method and system for human or computer-based group-members to interact with peers to craft an action sequence to achieve a group goal. method includes means for guiding group members on how to integrate their activities in pursuit of a specific pre-defined group goal, when given only partial understanding of how they can achieve said goal. the method identifies, selects, values and integrates group-member actions that are causal to a group achievement. the system incorporates the method along with means for recording, assigning value and reporting contributions by group members. system also includes an apparatus consisting of head-mounted microphone, voice recognition software and miniature video screen in field of view to aid data collection in applications where events occur in rapid sequence. for computer-based group members, system includes unsupervised neural network embodied in a computer mechanism and means to evaluate the instant activity and immediately relate processed information to guide the integration of group members actions.",2002-12-17,"The title of the patent is method and system for measuring and valuing contributions by group members to the achievement of a group goal and its abstract is a method and system for human or computer-based group-members to interact with peers to craft an action sequence to achieve a group goal. method includes means for guiding group members on how to integrate their activities in pursuit of a specific pre-defined group goal, when given only partial understanding of how they can achieve said goal. the method identifies, selects, values and integrates group-member actions that are causal to a group achievement. the system incorporates the method along with means for recording, assigning value and reporting contributions by group members. system also includes an apparatus consisting of head-mounted microphone, voice recognition software and miniature video screen in field of view to aid data collection in applications where events occur in rapid sequence. for computer-based group members, system includes unsupervised neural network embodied in a computer mechanism and means to evaluate the instant activity and immediately relate processed information to guide the integration of group members actions. dated 2002-12-17"
6496815,"neuron, hierarchical neural network using the neuron, and multiplying circuit used for multiplying process in the neuron","there is provided a neuron which is capable of expressing an excitative coupling and a suppressive coupling by one signal by devising signals processed in the neuron to reduce a circuit area of a neural network in constructing the neural network by a digital electronic circuit. a multiplying block calculates a numerical value following a normal distribution n(wx, 1) by using a corresponding link weight w under the supposition that delay time of each pulse of an input signal follows a normal distribution of n(x, 1). next, an adding block adds the numerical values calculated by the respective multiplying blocks one after another and a non-linear operating block counts a number of positive values within the added value obtained by the adding block. a pulse delaying block delays output pulse following a normal distribution in which delay time is 0 in average generated by a basic pulse generating block based on the result of operation of the non-linear operating block to output as an output signal.",2002-12-17,"The title of the patent is neuron, hierarchical neural network using the neuron, and multiplying circuit used for multiplying process in the neuron and its abstract is there is provided a neuron which is capable of expressing an excitative coupling and a suppressive coupling by one signal by devising signals processed in the neuron to reduce a circuit area of a neural network in constructing the neural network by a digital electronic circuit. a multiplying block calculates a numerical value following a normal distribution n(wx, 1) by using a corresponding link weight w under the supposition that delay time of each pulse of an input signal follows a normal distribution of n(x, 1). next, an adding block adds the numerical values calculated by the respective multiplying blocks one after another and a non-linear operating block counts a number of positive values within the added value obtained by the adding block. a pulse delaying block delays output pulse following a normal distribution in which delay time is 0 in average generated by a basic pulse generating block based on the result of operation of the non-linear operating block to output as an output signal. dated 2002-12-17"
6501294,neuron circuit,"a neuron circuit that can be served as a building block for a neural network implemented in an integrated circuit is disclosed. the neuron circuit includes a synapse circuit block and a neuron body circuit block. the synapse circuit block has three transistors, and the body of one of the three transistors is controlled by a weighted input. the neuron body circuit block includes a current mirror circuit, a summing circuit, and an invertor circuit. the neuron body circuit is coupled to the synapse circuit block to generate an output pulse.",2002-12-31,"The title of the patent is neuron circuit and its abstract is a neuron circuit that can be served as a building block for a neural network implemented in an integrated circuit is disclosed. the neuron circuit includes a synapse circuit block and a neuron body circuit block. the synapse circuit block has three transistors, and the body of one of the three transistors is controlled by a weighted input. the neuron body circuit block includes a current mirror circuit, a summing circuit, and an invertor circuit. the neuron body circuit is coupled to the synapse circuit block to generate an output pulse. dated 2002-12-31"
6502083,neuron architecture having a dual structure and neural networks incorporating the same,"the improved neuron is connected to input buses which transport input data and control signals. it basically consists of a computation block, a register block, an evaluation block and a daisy chain block. all these blocks, except the computation block substantially have a symmetric construction. registers are used to store data: the local norm and context, the distance, the aif value and the category. the improved neuron further needs some r/w memory capacity which may be placed either in the neuron or outside. the evaluation circuit is connected to an output bus to generate global signals thereon. the daisy chain block allows to chain the improved neuron with others to form an artificial neural network (ann). the improved neuron may work either as a single neuron (single mode) or as two independent neurons (dual mode). in the latter case, the computation block, which is common to the two dual neurons, must operate sequentially to service one neuron after the other. the selection between the two modes (single/dual) is made by the user which stores a specific logic value in a dedicated register of the control logic circuitry in each improved neuron.",2002-12-31,"The title of the patent is neuron architecture having a dual structure and neural networks incorporating the same and its abstract is the improved neuron is connected to input buses which transport input data and control signals. it basically consists of a computation block, a register block, an evaluation block and a daisy chain block. all these blocks, except the computation block substantially have a symmetric construction. registers are used to store data: the local norm and context, the distance, the aif value and the category. the improved neuron further needs some r/w memory capacity which may be placed either in the neuron or outside. the evaluation circuit is connected to an output bus to generate global signals thereon. the daisy chain block allows to chain the improved neuron with others to form an artificial neural network (ann). the improved neuron may work either as a single neuron (single mode) or as two independent neurons (dual mode). in the latter case, the computation block, which is common to the two dual neurons, must operate sequentially to service one neuron after the other. the selection between the two modes (single/dual) is made by the user which stores a specific logic value in a dedicated register of the control logic circuitry in each improved neuron. dated 2002-12-31"
6505130,laser doppler vibrometer for remote assessment of structural components,"a method and system for remotely inspecting the integrity of a structure. this can be performed by a method creating a vibratory response in the structure from a remote location and then measuring the vibratory response of the structure remotely. alternatively, this can be performed by a system for remotely measuring the integrity of a structure using a vehicle and an artificial neural network, where the vehicle is equipped with a vibratory response device. the vibratory response can be produced by infrasonic and audio frequencies that can be produced by at least a vehicle, motor, or sound recording. the vibratory response can be measured with a laser vibrometer or an audio recording device.",2003-01-07,"The title of the patent is laser doppler vibrometer for remote assessment of structural components and its abstract is a method and system for remotely inspecting the integrity of a structure. this can be performed by a method creating a vibratory response in the structure from a remote location and then measuring the vibratory response of the structure remotely. alternatively, this can be performed by a system for remotely measuring the integrity of a structure using a vehicle and an artificial neural network, where the vehicle is equipped with a vibratory response device. the vibratory response can be produced by infrasonic and audio frequencies that can be produced by at least a vehicle, motor, or sound recording. the vibratory response can be measured with a laser vibrometer or an audio recording device. dated 2003-01-07"
6505181,recognition system,"a recognition system of the self-organizing artificial neural network type is arranged to classify input data according to stored categories which have been determined by a training process. in the training process the initial category representations are selectively iteratively updated in response to a series of training patterns and in accordance with a competitive learning routine. this routine uses measures of category utilization based on the proportion of all inputs received over a representative period, particularly long term utilisation and short term utilization, to ensure that all available categories will be used and that the system is stable. the training rate which determines the amount of modification to a category representation at an up-date is local to each category and is based upon the maturity of the category and on the similarity measure between the internal representative pattern and the training input so that the training duration can be minimized. a user-operated selectively-operable suggestion learning input is provided to each category to modify the training process or to enable secondary training to proceed during classification of input data using that input data as the training patterns. the categories are represented by multiple reference patterns with respective importance values from which the degree of compatibility between an input and a category is computed taking into account the importance values.",2003-01-07,"The title of the patent is recognition system and its abstract is a recognition system of the self-organizing artificial neural network type is arranged to classify input data according to stored categories which have been determined by a training process. in the training process the initial category representations are selectively iteratively updated in response to a series of training patterns and in accordance with a competitive learning routine. this routine uses measures of category utilization based on the proportion of all inputs received over a representative period, particularly long term utilisation and short term utilization, to ensure that all available categories will be used and that the system is stable. the training rate which determines the amount of modification to a category representation at an up-date is local to each category and is based upon the maturity of the category and on the similarity measure between the internal representative pattern and the training input so that the training duration can be minimized. a user-operated selectively-operable suggestion learning input is provided to each category to modify the training process or to enable secondary training to proceed during classification of input data using that input data as the training patterns. the categories are represented by multiple reference patterns with respective importance values from which the degree of compatibility between an input and a category is computed taking into account the importance values. dated 2003-01-07"
6507803,method for determining spraying parameters for a paint spraying unit,"a method for determining spraying parameters that are suitable as input values for a paint spraying unit that can electrostatically charge a liquid paint. in this case, at least one artificial neural network is used to determine the spraying parameters, an output of such a neural network being available for each spraying parameter. a suitable number of real measured values are fed to the one neural network or a plurality of neural networks as input values, initially in a learning phase. the measured values further contain associated real spraying parameters in addition to a paint thickness distribution in the form of discrete values. input values are fed to the one neural network or a plurality of neural networks in the application phase. the input values being the result of an analysis of the paint thickness distribution of a targeted, that is to say prescribed, spraying result.",2003-01-14,"The title of the patent is method for determining spraying parameters for a paint spraying unit and its abstract is a method for determining spraying parameters that are suitable as input values for a paint spraying unit that can electrostatically charge a liquid paint. in this case, at least one artificial neural network is used to determine the spraying parameters, an output of such a neural network being available for each spraying parameter. a suitable number of real measured values are fed to the one neural network or a plurality of neural networks as input values, initially in a learning phase. the measured values further contain associated real spraying parameters in addition to a paint thickness distribution in the form of discrete values. input values are fed to the one neural network or a plurality of neural networks in the application phase. the input values being the result of an analysis of the paint thickness distribution of a targeted, that is to say prescribed, spraying result. dated 2003-01-14"
6511424,method of and apparatus for evaluation and mitigation of microsleep events,"a method and apparatus for determining, monitoring and predicting levels of alertness by detecting microsleep episodes includes a plurality of channel processing units and a channel combining unit. each of the channel processing units receives an information channel which conveys information associated with the mental state of the subject, such as an eeg channel, and classifies the information into a distinct category. such categories may include microsleep, non-microsleep, one or more of a plurality of stages of sleep, one or more of a plurality of stages of wakefulness, or a transition state characterized by a transition from one of the aforementioned states to another. each of the channel processing units includes a neural network which has been trained with a set of example input/result vector pairs. the example input/result vector pairs are generated by correlating actual information channel outputs with observed fatigue related events such as nodding off, head snapping, multiple blinks, blank stares, wide eyes, yawning, prolonged eyelid closures, and slow rolling eye movements.",2003-01-28,"The title of the patent is method of and apparatus for evaluation and mitigation of microsleep events and its abstract is a method and apparatus for determining, monitoring and predicting levels of alertness by detecting microsleep episodes includes a plurality of channel processing units and a channel combining unit. each of the channel processing units receives an information channel which conveys information associated with the mental state of the subject, such as an eeg channel, and classifies the information into a distinct category. such categories may include microsleep, non-microsleep, one or more of a plurality of stages of sleep, one or more of a plurality of stages of wakefulness, or a transition state characterized by a transition from one of the aforementioned states to another. each of the channel processing units includes a neural network which has been trained with a set of example input/result vector pairs. the example input/result vector pairs are generated by correlating actual information channel outputs with observed fatigue related events such as nodding off, head snapping, multiple blinks, blank stares, wide eyes, yawning, prolonged eyelid closures, and slow rolling eye movements. dated 2003-01-28"
6513023,artificial neural network with hardware training and hardware refresh,"a neural network circuit is provided having a plurality of circuits capable of charge storage. also provided is a plurality of circuits each coupled to at least one of the plurality of charge storage circuits and constructed to generate an output in accordance with a neuron transfer function. each of a plurality of circuits is coupled to one of the plurality of neuron transfer function circuits and constructed to generate a derivative of the output. a weight update circuit updates the charge storage circuits based upon output from the plurality of transfer function circuits and output from the plurality of derivative circuits. in preferred embodiments, separate training and validation networks share the same set of charge storage circuits and may operate concurrently. the validation network has a separate transfer function circuits each being coupled to the charge storage circuits so as to replicate the training network's coupling of the plurality of charge storage to the plurality of transfer function circuits. the plurality of transfer function circuits may be constructed each having a transconductance amplifier providing differential currents combined to provide an output in accordance with a transfer function. the derivative circuits may have a circuit constructed to generate a biased differential currents combined so as to provide the derivative of the transfer function.",2003-01-28,"The title of the patent is artificial neural network with hardware training and hardware refresh and its abstract is a neural network circuit is provided having a plurality of circuits capable of charge storage. also provided is a plurality of circuits each coupled to at least one of the plurality of charge storage circuits and constructed to generate an output in accordance with a neuron transfer function. each of a plurality of circuits is coupled to one of the plurality of neuron transfer function circuits and constructed to generate a derivative of the output. a weight update circuit updates the charge storage circuits based upon output from the plurality of transfer function circuits and output from the plurality of derivative circuits. in preferred embodiments, separate training and validation networks share the same set of charge storage circuits and may operate concurrently. the validation network has a separate transfer function circuits each being coupled to the charge storage circuits so as to replicate the training network's coupling of the plurality of charge storage to the plurality of transfer function circuits. the plurality of transfer function circuits may be constructed each having a transconductance amplifier providing differential currents combined to provide an output in accordance with a transfer function. the derivative circuits may have a circuit constructed to generate a biased differential currents combined so as to provide the derivative of the transfer function. dated 2003-01-28"
6516309,method and apparatus for evolving a neural network,"a method of evolving a neural network that includes a plurality of processing elements interconnected by a plurality of weighted connections includes the step of obtaining a definition for the neural network by evolving a plurality of weights for the plurality of weighted connections, and evolving a plurality of activation function parameters associated with the plurality of processing elements. another step of the method includes determining whether the definition for the neural network may be simplified based upon at least one activation function parameter of the plurality of activation function parameters. yet another step of the method includes updating the definition for the neural network in response to determining that the definition for the neural network may be simplified. the method utilizes particle swarm optimization techniques to evolve the plurality of weights and the plurality of activation parameters. moreover, the method simplifies activation functions of processing elements in response to corresponding activation parameters meeting certain criteria, and removes processing elements from the definition of the neural network in response to corresponding activation parameters satisfying certain criteria. various apparatus are also disclosed for implementing network evolution and simplification.",2003-02-04,"The title of the patent is method and apparatus for evolving a neural network and its abstract is a method of evolving a neural network that includes a plurality of processing elements interconnected by a plurality of weighted connections includes the step of obtaining a definition for the neural network by evolving a plurality of weights for the plurality of weighted connections, and evolving a plurality of activation function parameters associated with the plurality of processing elements. another step of the method includes determining whether the definition for the neural network may be simplified based upon at least one activation function parameter of the plurality of activation function parameters. yet another step of the method includes updating the definition for the neural network in response to determining that the definition for the neural network may be simplified. the method utilizes particle swarm optimization techniques to evolve the plurality of weights and the plurality of activation parameters. moreover, the method simplifies activation functions of processing elements in response to corresponding activation parameters meeting certain criteria, and removes processing elements from the definition of the neural network in response to corresponding activation parameters satisfying certain criteria. various apparatus are also disclosed for implementing network evolution and simplification. dated 2003-02-04"
6518536,joining equipment,"a joining equipment in which a neural network is employed for controlling a joining process. a dynamic analog model is used for neuron elements configuring the network. the equipment includes a detector, a controller, and a neural network. the detector detects a joining state of a joining portion when work pieces are joined with each other. the controller controls the output of the joining equipment. in response to output signals from the detector, the neural network transmits signals to the controller. such a structure allows the joining equipment to flexibly respond to complicated changes in joining states. besides, using similarity with a thermal conduction equation enables to minimize the number of input items fed into the neural network. furthermore, using an approximate solution to the thermal conduction equation realizes to accelerate the time for numerical calculation without loss of accuracy.",2003-02-11,"The title of the patent is joining equipment and its abstract is a joining equipment in which a neural network is employed for controlling a joining process. a dynamic analog model is used for neuron elements configuring the network. the equipment includes a detector, a controller, and a neural network. the detector detects a joining state of a joining portion when work pieces are joined with each other. the controller controls the output of the joining equipment. in response to output signals from the detector, the neural network transmits signals to the controller. such a structure allows the joining equipment to flexibly respond to complicated changes in joining states. besides, using similarity with a thermal conduction equation enables to minimize the number of input items fed into the neural network. furthermore, using an approximate solution to the thermal conduction equation realizes to accelerate the time for numerical calculation without loss of accuracy. dated 2003-02-11"
6519535,eddy current technique for predicting burst pressure,"a signal processing technique which correlates eddy current inspection data from a tube having a critical tubing defect with a range of predicted burst pressures for the tube is provided. the method can directly correlate the raw eddy current inspection data representing the critical tubing defect with the range of burst pressures using a regression technique, preferably an artificial neural network. alternatively, the technique deconvolves the raw eddy current inspection data into a set of undistorted signals, each of which represents a separate defect of the tube. the undistorted defect signal which represents the critical tubing defect is related to a range of burst pressures utilizing a regression technique.",2003-02-11,"The title of the patent is eddy current technique for predicting burst pressure and its abstract is a signal processing technique which correlates eddy current inspection data from a tube having a critical tubing defect with a range of predicted burst pressures for the tube is provided. the method can directly correlate the raw eddy current inspection data representing the critical tubing defect with the range of burst pressures using a regression technique, preferably an artificial neural network. alternatively, the technique deconvolves the raw eddy current inspection data into a set of undistorted signals, each of which represents a separate defect of the tube. the undistorted defect signal which represents the critical tubing defect is related to a range of burst pressures utilizing a regression technique. dated 2003-02-11"
6523018,neural chip architecture and neural networks incorporated therein,"the neural semiconductor chip first includes: a global register and control logic circuit block, a r/w memory block and a plurality of neurons fed by buses transporting data such as the input vector data, set-up parameters, etc., and signals such as the feed back and control signals. the r/w memory block, typically a ram, is common to all neurons to avoid circuit duplication, increasing thereby the number of neurons integrated in the chip. the r/w memory stores the prototype components. each neuron comprises a computation block, a register block, an evaluation block and a daisy chain block to chain the neurons. all these blocks (except the computation block) have a symmetric structure and are designed so that each neuron may operate in a dual manner, i.e. either as a single neuron (single mode) or as two independent neurons (dual mode). each neuron generates local signals. the neural chip further includes an or circuit which performs an or function for all corresponding local signals to generate global signals that are merged in an on-chip common communication bus shared by all neurons of the chip. the r/w memory block, the neurons and the or circuit form an artificial neural network having high flexibility due to this dual mode feature which allows to mix single and dual neurons in the ann.",2003-02-18,"The title of the patent is neural chip architecture and neural networks incorporated therein and its abstract is the neural semiconductor chip first includes: a global register and control logic circuit block, a r/w memory block and a plurality of neurons fed by buses transporting data such as the input vector data, set-up parameters, etc., and signals such as the feed back and control signals. the r/w memory block, typically a ram, is common to all neurons to avoid circuit duplication, increasing thereby the number of neurons integrated in the chip. the r/w memory stores the prototype components. each neuron comprises a computation block, a register block, an evaluation block and a daisy chain block to chain the neurons. all these blocks (except the computation block) have a symmetric structure and are designed so that each neuron may operate in a dual manner, i.e. either as a single neuron (single mode) or as two independent neurons (dual mode). each neuron generates local signals. the neural chip further includes an or circuit which performs an or function for all corresponding local signals to generate global signals that are merged in an on-chip common communication bus shared by all neurons of the chip. the r/w memory block, the neurons and the or circuit form an artificial neural network having high flexibility due to this dual mode feature which allows to mix single and dual neurons in the ann. dated 2003-02-18"
6526167,image processing apparatus and method and provision medium,"an image processing apparatus and method and a provision medium arranged to perform image learning and image recognition in a short time. an image difference detector computes a differential image between an image stored in a frame buffer and an image stored in another frame buffer, and also computes the centroid of the differential image. a information collector forms rgb histogram data and binary data of a peripheral area about the centroid obtained by the image difference detector. a category former formed by a kohonen network forms a category based on the rgb histogram data and binary data. a category statistic processor performs statistical processing of the categories output from the category former, and outputs a processing result to a learner formed by a recurrent neural network.",2003-02-25,"The title of the patent is image processing apparatus and method and provision medium and its abstract is an image processing apparatus and method and a provision medium arranged to perform image learning and image recognition in a short time. an image difference detector computes a differential image between an image stored in a frame buffer and an image stored in another frame buffer, and also computes the centroid of the differential image. a information collector forms rgb histogram data and binary data of a peripheral area about the centroid obtained by the image difference detector. a category former formed by a kohonen network forms a category based on the rgb histogram data and binary data. a category statistic processor performs statistical processing of the categories output from the category former, and outputs a processing result to a learner formed by a recurrent neural network. dated 2003-02-25"
6526168,visual neural classifier,"a neural classifier that allows visualization of the query, the training data and the decision regions in a single two-dimensional display, providing benefits for both the designer and the user. the visual neural classifier is formed from a set of experts and a visualization network. visualization is accomplished by a funnel-shaped multilayer dimensionality reduction network configured to learn one or more classification tasks. if a single dimensionality reduction network does not provide sufficiently accurate classification results, a group of these dimensionality reduction networks may be arranged in a modular architecture. among these dimensionality reduction networks, the experts receive the input data and the visualization network combines the decisions of the experts to form the final classification decision.",2003-02-25,"The title of the patent is visual neural classifier and its abstract is a neural classifier that allows visualization of the query, the training data and the decision regions in a single two-dimensional display, providing benefits for both the designer and the user. the visual neural classifier is formed from a set of experts and a visualization network. visualization is accomplished by a funnel-shaped multilayer dimensionality reduction network configured to learn one or more classification tasks. if a single dimensionality reduction network does not provide sufficiently accurate classification results, a group of these dimensionality reduction networks may be arranged in a modular architecture. among these dimensionality reduction networks, the experts receive the input data and the visualization network combines the decisions of the experts to form the final classification decision. dated 2003-02-25"
6526361,battery testing and classification,method and apparatus for battery evaluation and classification applies transient microcharge and/or microload pulses to an automotive battery. classification is made on the basis of analysis of the resultant voltage profile or portions or dimensions thereof. in one embodiment the analysis utilizes a neural network or algorithm to assess a microcycle sequence of microload/microcharge tests utilizing one of a series of battery parameters including impedance as well as voltage characteristics to effect classification. another embodiment adopts an optimized (not maximum) level of prior test-based data-training for a self-organizing neural network. a third embodiment utilizes prior test data correlation to enable algorithm-based classification without use of a neural network.,2003-02-25,The title of the patent is battery testing and classification and its abstract is method and apparatus for battery evaluation and classification applies transient microcharge and/or microload pulses to an automotive battery. classification is made on the basis of analysis of the resultant voltage profile or portions or dimensions thereof. in one embodiment the analysis utilizes a neural network or algorithm to assess a microcycle sequence of microload/microcharge tests utilizing one of a series of battery parameters including impedance as well as voltage characteristics to effect classification. another embodiment adopts an optimized (not maximum) level of prior test-based data-training for a self-organizing neural network. a third embodiment utilizes prior test data correlation to enable algorithm-based classification without use of a neural network. dated 2003-02-25
6526547,method for efficient manufacturing of integrated circuits,"this invention pertains to a method for the systematic development of integrated chip technology. the method may include obtaining empirical data of parameters for an existing integrated circuit manufacturing process and extrapolating the known data to a new technology to assess potential yields of the new technology from the known process. further, process variables of the new process may be adjusted based upon the empirical data in order to optimize the yields of the new technology. a logic based computing system such as a fuzzy logic or neural-network system may be utilized. the computing system may also be utilized to improve the yields of an existing manufacturing process by adjust process variables within downstream process tools based upon data collected in upstream process for a particular semiconductor substrate or lot.",2003-02-25,"The title of the patent is method for efficient manufacturing of integrated circuits and its abstract is this invention pertains to a method for the systematic development of integrated chip technology. the method may include obtaining empirical data of parameters for an existing integrated circuit manufacturing process and extrapolating the known data to a new technology to assess potential yields of the new technology from the known process. further, process variables of the new process may be adjusted based upon the empirical data in order to optimize the yields of the new technology. a logic based computing system such as a fuzzy logic or neural-network system may be utilized. the computing system may also be utilized to improve the yields of an existing manufacturing process by adjust process variables within downstream process tools based upon data collected in upstream process for a particular semiconductor substrate or lot. dated 2003-02-25"
6529150,photonic analog to digital conversion based on temporal and spatial oversampling techniques,"a method of converting an analog signal to a digital signal includes (a) filtering the analog signal to the range 0&le;fx&le;fb; (b) sampling the filtered signal at a rate fs>>fn, where fs is the sampling frequency, fn=2fx is the nyquist frequency of the sampled signal, and fb&le;fs/2 is the constrained signal bandwidth; (c) converting the sampled signal to an optical sampled signal: (d) converting the optical sampled signal from a temporal signal to a spatial signal; (e) illuminating a smart pixel array with the spatial signal; (f) processing the spatial signal with an error diffusion neural network to produce a 2-d binary image; and (g) averaging rows and columns of the 2-d binary image using a digital low pass filter and a decimation circuit.",2003-03-04,"The title of the patent is photonic analog to digital conversion based on temporal and spatial oversampling techniques and its abstract is a method of converting an analog signal to a digital signal includes (a) filtering the analog signal to the range 0&le;fx&le;fb; (b) sampling the filtered signal at a rate fs>>fn, where fs is the sampling frequency, fn=2fx is the nyquist frequency of the sampled signal, and fb&le;fs/2 is the constrained signal bandwidth; (c) converting the sampled signal to an optical sampled signal: (d) converting the optical sampled signal from a temporal signal to a spatial signal; (e) illuminating a smart pixel array with the spatial signal; (f) processing the spatial signal with an error diffusion neural network to produce a 2-d binary image; and (g) averaging rows and columns of the 2-d binary image using a digital low pass filter and a decimation circuit. dated 2003-03-04"
6529780,method for automatic operation of industrial plants,"the invention relates to an automation system for the erection and operation of industrial plants, in particular for the design, project engineering, implementation, commissioning, maintenance and optimization of individual plant components or complete plants in the basic materials industry, which plants have a computer-based control system which, for a description of the process in control engineering terms, has recourse to process models, for example in he form of mathematical/physical models, neural network models or knowledge-based system, in order to develop the system to the extent that straightforward and cost-effective decentral process management and optimization may be achieved remote from the plant, decentralized process management and optimization by means of one or more interlinked control points is proposed, process changes are continuously monitored online or offline or at least checked by modelling, using modern, public communication means, and the process models, parameters and software are adaptable specifically to the plant.",2003-03-04,"The title of the patent is method for automatic operation of industrial plants and its abstract is the invention relates to an automation system for the erection and operation of industrial plants, in particular for the design, project engineering, implementation, commissioning, maintenance and optimization of individual plant components or complete plants in the basic materials industry, which plants have a computer-based control system which, for a description of the process in control engineering terms, has recourse to process models, for example in he form of mathematical/physical models, neural network models or knowledge-based system, in order to develop the system to the extent that straightforward and cost-effective decentral process management and optimization may be achieved remote from the plant, decentralized process management and optimization by means of one or more interlinked control points is proposed, process changes are continuously monitored online or offline or at least checked by modelling, using modern, public communication means, and the process models, parameters and software are adaptable specifically to the plant. dated 2003-03-04"
6535795,method for chemical addition utilizing adaptive optimization,"the present invention provides a method for chemical addition utilizing adaptive process control optimizations having a combination of expert system(s), neural network(s) and genetic algorithm(s).",2003-03-18,"The title of the patent is method for chemical addition utilizing adaptive optimization and its abstract is the present invention provides a method for chemical addition utilizing adaptive process control optimizations having a combination of expert system(s), neural network(s) and genetic algorithm(s). dated 2003-03-18"
6535862,method and circuit for performing the integrity diagnostic of an artificial neural network,"a diagnostic method engages all the neurons of an artificial neural network (ann) based on mapping an input space defined by vector components based on category, context, and actual field of influence (aif). the method includes the steps loading the components of a first and second input vectors into the ann; engaging all the neurons of a same prototype; having all the neurons compute their own distance between the respective prototypes and the second input vector (which should be the same if the neurons were good); determining the minimum distance dmin and comparing dmin with a distance d measured between the first and the second input vectors. if dmin<d, it is indicative that at least one failing neuron exists (i.e., either the distance, the category or the aif value differs from a predetermined expected value), in which case the failing neuron is isolated. if dmin&ge;d, the good neurons of the ann are deselected, preventing the good neurons from further learning/recognition processing; and repeating the previous steps until the ann is empty. the present method provides a fast and inexpensive integrity diagnostic of the neurons forming an ann at the cost of only a few logic gates.",2003-03-18,"The title of the patent is method and circuit for performing the integrity diagnostic of an artificial neural network and its abstract is a diagnostic method engages all the neurons of an artificial neural network (ann) based on mapping an input space defined by vector components based on category, context, and actual field of influence (aif). the method includes the steps loading the components of a first and second input vectors into the ann; engaging all the neurons of a same prototype; having all the neurons compute their own distance between the respective prototypes and the second input vector (which should be the same if the neurons were good); determining the minimum distance dmin and comparing dmin with a distance d measured between the first and the second input vectors. if dmin<d, it is indicative that at least one failing neuron exists (i.e., either the distance, the category or the aif value differs from a predetermined expected value), in which case the failing neuron is isolated. if dmin&ge;d, the good neurons of the ann are deselected, preventing the good neurons from further learning/recognition processing; and repeating the previous steps until the ann is empty. the present method provides a fast and inexpensive integrity diagnostic of the neurons forming an ann at the cost of only a few logic gates. dated 2003-03-18"
6539304,gps navigation system using neural networks,a gps receiver includes a satellite receiver/processor having an input that receives input signals from at least one gps satellite. the output of the receiver/processor provides satellite-related navigation information. a neural network receives the satellite-related information to obtain an output signal representative of receiver-related navigation information. the neural network includes a first node layer connected to a second node layer through a first connection layer and a third node layer connected to the second node layer through a second connection layer. each of the node layers comprises a plurality of neurons.,2003-03-25,The title of the patent is gps navigation system using neural networks and its abstract is a gps receiver includes a satellite receiver/processor having an input that receives input signals from at least one gps satellite. the output of the receiver/processor provides satellite-related navigation information. a neural network receives the satellite-related information to obtain an output signal representative of receiver-related navigation information. the neural network includes a first node layer connected to a second node layer through a first connection layer and a third node layer connected to the second node layer through a second connection layer. each of the node layers comprises a plurality of neurons. dated 2003-03-25
6539368,"neural processor, saturation unit, calculation unit and adder circuit","the present invention relates to the field of computer science and can be used for neural network emulation and digital signal processing. increasing of the neural processor performance is achieved using the ability to change word lengths of results in program mode. the neural processor includes six registers, a shift register, a and gate, two fifos, a switch, a multiplexer, two saturation units, a calculation unit and a adder circuit to execute operations over vectors of programmable word length data. increasing of the saturation unit performance is achieved using the ability to process vector of input operands with programmable word length at a time. increasing of the adder circuit performance is achieved using ability to sum two vectors of input operands of programmable word lengths.",2003-03-25,"The title of the patent is neural processor, saturation unit, calculation unit and adder circuit and its abstract is the present invention relates to the field of computer science and can be used for neural network emulation and digital signal processing. increasing of the neural processor performance is achieved using the ability to change word lengths of results in program mode. the neural processor includes six registers, a shift register, a and gate, two fifos, a switch, a multiplexer, two saturation units, a calculation unit and a adder circuit to execute operations over vectors of programmable word length data. increasing of the saturation unit performance is achieved using the ability to process vector of input operands with programmable word length at a time. increasing of the adder circuit performance is achieved using ability to sum two vectors of input operands of programmable word lengths. dated 2003-03-25"
6542879,neural network trajectory command controller,an apparatus and method for controlling trajectory of an object (47) to a first predetermined position. the apparatus has an input layer (22) having nodes (22a-22f) for receiving input data indicative of the first predetermined position. first weighted connections (28) are connected to the nodes of the input layer (22). each of the first weighted connections (28) have a coefficient for weighting the input data. an output layer (26) having nodes (26a-26e) connected to the first weighted connections (28) determines trajectory data based upon the first weighted input data. the trajectory of the object is controlled based upon the determined trajectory data.,2003-04-01,The title of the patent is neural network trajectory command controller and its abstract is an apparatus and method for controlling trajectory of an object (47) to a first predetermined position. the apparatus has an input layer (22) having nodes (22a-22f) for receiving input data indicative of the first predetermined position. first weighted connections (28) are connected to the nodes of the input layer (22). each of the first weighted connections (28) have a coefficient for weighting the input data. an output layer (26) having nodes (26a-26e) connected to the first weighted connections (28) determines trajectory data based upon the first weighted input data. the trajectory of the object is controlled based upon the determined trajectory data. dated 2003-04-01
6548316,monolithic semiconductor device and method of manufacturing the same,"a monolithic semiconductor device comprising a substrate, a layer of photoconductive material formed on the substrate, a transparent insulator formed on the photoconductive material and a layer of material which emits light when electrically stimulated, said layer of light emitting material being formed on the transparent insulator. the light emitting material is preferably an organic electro-luminescent material such as a polymer. particular application of the device is in implementing an analog based neural network and by selection and arrangement of various components the device may also act as a display. a method of manufacturing the device is also disclosed.",2003-04-15,"The title of the patent is monolithic semiconductor device and method of manufacturing the same and its abstract is a monolithic semiconductor device comprising a substrate, a layer of photoconductive material formed on the substrate, a transparent insulator formed on the photoconductive material and a layer of material which emits light when electrically stimulated, said layer of light emitting material being formed on the transparent insulator. the light emitting material is preferably an organic electro-luminescent material such as a polymer. particular application of the device is in implementing an analog based neural network and by selection and arrangement of various components the device may also act as a display. a method of manufacturing the device is also disclosed. dated 2003-04-15"
6549661,pattern recognition apparatus and pattern recognition method,"recognizable data storage apparatus stores an image feature parameter of an object and its classification result for those objects for which said classification results are evaluated as having high reliability and recognizable and for which a classification result is outputted by first pattern recognition apparatus, and recognition suspension data storage apparatus stores an image feature parameter for those objects for which said classification results are evaluated as having low reliability and as being suspended from recognition. pattern recognition method constitution apparatus constructs second pattern recognition apparatus on the basis of the classification result stored in the recognizable data storage apparatus, and re-classification is conducted for those objects recognition of which is to be suspended. a neural network, for example, is used for the second pattern recognition apparatus. wrong classification of objects due to an individual difference can be reduced and automated pattern recognition having high accuracy without depending on an individual difference can be accomplished.",2003-04-15,"The title of the patent is pattern recognition apparatus and pattern recognition method and its abstract is recognizable data storage apparatus stores an image feature parameter of an object and its classification result for those objects for which said classification results are evaluated as having high reliability and recognizable and for which a classification result is outputted by first pattern recognition apparatus, and recognition suspension data storage apparatus stores an image feature parameter for those objects for which said classification results are evaluated as having low reliability and as being suspended from recognition. pattern recognition method constitution apparatus constructs second pattern recognition apparatus on the basis of the classification result stored in the recognizable data storage apparatus, and re-classification is conducted for those objects recognition of which is to be suspended. a neural network, for example, is used for the second pattern recognition apparatus. wrong classification of objects due to an individual difference can be reduced and automated pattern recognition having high accuracy without depending on an individual difference can be accomplished. dated 2003-04-15"
6550320,system and method for predicting tire forces using tire deformation sensors,"a system and method for predicting the forces generated in the tire contact patch from measurements of tire deformations, including separating the lateral force, the vertical force, and the circumferential torque using measurements of tire deformations. a system and method for using a trained neural network or bilinear equations to determine any combination or permutation of one or more of any of the following from tire sidewall deformation sensors, e.g., magnetic tire sidewall torsion measuring (swt) sensors: the lateral force acting on the tire, the circumferential torque acting on the tire, the longitudinal force acting on the tire, the vertical force acting on the tire, and forces and/or torques having any one or more of the foregoing as components thereof.",2003-04-22,"The title of the patent is system and method for predicting tire forces using tire deformation sensors and its abstract is a system and method for predicting the forces generated in the tire contact patch from measurements of tire deformations, including separating the lateral force, the vertical force, and the circumferential torque using measurements of tire deformations. a system and method for using a trained neural network or bilinear equations to determine any combination or permutation of one or more of any of the following from tire sidewall deformation sensors, e.g., magnetic tire sidewall torsion measuring (swt) sensors: the lateral force acting on the tire, the circumferential torque acting on the tire, the longitudinal force acting on the tire, the vertical force acting on the tire, and forces and/or torques having any one or more of the foregoing as components thereof. dated 2003-04-22"
6553357,method for improving neural network architectures using evolutionary algorithms,"the noise associated with conventional techniques for evolutionary improvement of neural network architectures is reduced so that of an optimum architecture can be determined more efficiently and more effectively. parameters that affect the initialization of a neural network architecture are included within the encoding that is used by an evolutionary algorithm to optimize the neural network architecture. the example initialization parameters include an encoding that determines the initial nodal weights used in each architecture at the commencement of the training cycle. by including the initialization parameters within the encoding used by the evolutionary algorithm, the initialization parameters that have a positive effect on the performance of the resultant evolved network architecture are propagated and potentially improved from generation to generation. conversely, initialization parameters that, for example, cause the resultant evolved network to be poorly trained, will not be propagated. in accordance with a second aspect of this invention, the encoding also includes parameters that affect the training process, such as the duration of the training cycle, the training inputs applied, and so on. in accordance with a third aspect of this invention, the same set of training or evaluation inputs are applied to all members whose performances are directly compared.",2003-04-22,"The title of the patent is method for improving neural network architectures using evolutionary algorithms and its abstract is the noise associated with conventional techniques for evolutionary improvement of neural network architectures is reduced so that of an optimum architecture can be determined more efficiently and more effectively. parameters that affect the initialization of a neural network architecture are included within the encoding that is used by an evolutionary algorithm to optimize the neural network architecture. the example initialization parameters include an encoding that determines the initial nodal weights used in each architecture at the commencement of the training cycle. by including the initialization parameters within the encoding used by the evolutionary algorithm, the initialization parameters that have a positive effect on the performance of the resultant evolved network architecture are propagated and potentially improved from generation to generation. conversely, initialization parameters that, for example, cause the resultant evolved network to be poorly trained, will not be propagated. in accordance with a second aspect of this invention, the encoding also includes parameters that affect the training process, such as the duration of the training cycle, the training inputs applied, and so on. in accordance with a third aspect of this invention, the same set of training or evaluation inputs are applied to all members whose performances are directly compared. dated 2003-04-22"
6556291,defect inspection method and defect inspection apparatus,"a defect inspection apparatus for inspecting a presence of a defect on an object includes: a first input unit which inputs wavelength characteristics of each of a plurality of samples with wavelength variation of an illumination light for inspection; a second input unit which inputs inspection conditions which an inspector sets for each of the samples as a teaching signal; a third input unit which inputs a wavelength characteristic of the object with the wavelength variation of the illumination light; a neural network which learns and stores a relationship between the inputted wavelength characteristic of each sample and the inputted inspection condition for each sample, and determines an inspection condition for the object based on the inputted wavelength characteristic of the object and the learned relationship; and a defect detector which detects a defect of the object based on the determined inspection condition of the object.",2003-04-29,"The title of the patent is defect inspection method and defect inspection apparatus and its abstract is a defect inspection apparatus for inspecting a presence of a defect on an object includes: a first input unit which inputs wavelength characteristics of each of a plurality of samples with wavelength variation of an illumination light for inspection; a second input unit which inputs inspection conditions which an inspector sets for each of the samples as a teaching signal; a third input unit which inputs a wavelength characteristic of the object with the wavelength variation of the illumination light; a neural network which learns and stores a relationship between the inputted wavelength characteristic of each sample and the inputted inspection condition for each sample, and determines an inspection condition for the object based on the inputted wavelength characteristic of the object and the learned relationship; and a defect detector which detects a defect of the object based on the determined inspection condition of the object. dated 2003-04-29"
6556699,method for combining automated detections from medical images with observed detections of a human interpreter,"a method and system for detecting and displaying clustered microcalcifications in a digital mammogram, wherein a single digital mammogram is first automatically cropped to a breast area sub-image which is then processed by means of an optimized difference of gaussians filter to enhance the appearance of potential microcalcifications in the sub-image. the potential microcalcifications are thresholded, clusters are detected, features are computed for the detected clusters, and the clusters are classified as either suspicious or not suspicious by means of a neural network. thresholding is preferably by sloping local thresholding, but may also be performed by global and dual-local thresholding. the locations in the original digital mammogram of the suspicious detected clustered microcalcifications are indicated. parameters for use in the detection and thresholding portions of the system are computer-optimized by means of a genetic algorithm. the results of the system are optimally combined with a radiologist's observation of the original mammogram by combining the observations with the results, after the radiologist has first accepted or rejected individual detections reported by the system.",2003-04-29,"The title of the patent is method for combining automated detections from medical images with observed detections of a human interpreter and its abstract is a method and system for detecting and displaying clustered microcalcifications in a digital mammogram, wherein a single digital mammogram is first automatically cropped to a breast area sub-image which is then processed by means of an optimized difference of gaussians filter to enhance the appearance of potential microcalcifications in the sub-image. the potential microcalcifications are thresholded, clusters are detected, features are computed for the detected clusters, and the clusters are classified as either suspicious or not suspicious by means of a neural network. thresholding is preferably by sloping local thresholding, but may also be performed by global and dual-local thresholding. the locations in the original digital mammogram of the suspicious detected clustered microcalcifications are indicated. parameters for use in the detection and thresholding portions of the system are computer-optimized by means of a genetic algorithm. the results of the system are optimally combined with a radiologist's observation of the original mammogram by combining the observations with the results, after the radiologist has first accepted or rejected individual detections reported by the system. dated 2003-04-29"
6556951,system and method for intelligent quality control of a process,"a method and system for detecting errors in a process such as laboratory analysis of patient specimens and generation of test results is described. the steps of the method include collecting data elements having a range of values from the process. the number of data elements having values within predetermined intervals of the range are then counted. the counts of the data elements are applied as inputs to nodes of a neural network, each count being applied to a node representing the predetermined interval corresponding to the count. output is then generated from the neural network based on the inputs, the output indicative of whether an error in the process (such as bias error or a precision error) has occurred. if the technology is used with a laboratory instrument, the output is generated in real time and available immediately for automatic or manual correction of the instrument.",2003-04-29,"The title of the patent is system and method for intelligent quality control of a process and its abstract is a method and system for detecting errors in a process such as laboratory analysis of patient specimens and generation of test results is described. the steps of the method include collecting data elements having a range of values from the process. the number of data elements having values within predetermined intervals of the range are then counted. the counts of the data elements are applied as inputs to nodes of a neural network, each count being applied to a node representing the predetermined interval corresponding to the count. output is then generated from the neural network based on the inputs, the output indicative of whether an error in the process (such as bias error or a precision error) has occurred. if the technology is used with a laboratory instrument, the output is generated in real time and available immediately for automatic or manual correction of the instrument. dated 2003-04-29"
6556979,method and system for identifying consumer credit revolvers with neural network time series segmentation,"initially, customer time series credit files are acquired. the credit files are organized in a data mart environment for supporting a query system. time series utilization attributes are created and a neural network time series segmentation process is applied and n&times;n dimension segments are generated for analysis. the chart may be modified to more accurately depict profitable credit revolvers. credit data from each potential new customer is processed in a similar fashion by the neural network segmentation process. profitable credit revolvers are identified by having credit utilization patterns belonging to profitable segments previously identified.",2003-04-29,"The title of the patent is method and system for identifying consumer credit revolvers with neural network time series segmentation and its abstract is initially, customer time series credit files are acquired. the credit files are organized in a data mart environment for supporting a query system. time series utilization attributes are created and a neural network time series segmentation process is applied and n&times;n dimension segments are generated for analysis. the chart may be modified to more accurately depict profitable credit revolvers. credit data from each potential new customer is processed in a similar fashion by the neural network segmentation process. profitable credit revolvers are identified by having credit utilization patterns belonging to profitable segments previously identified. dated 2003-04-29"
6556980,model-free adaptive control for industrial processes,"an enhanced model-free adaptive controller is disclosed, which consists of a linear dynamic neural network that can be easily configured and put in automatic mode to control simple to complex processes. two multivariable model-free adaptive controller designs are disclosed. an enhanced anti-delay model-free adaptive controller is introduced to control processes with large time delays. a feedforward/feedback model-free adaptive control system with two designs is introduced to compensate for measurable disturbances.",2003-04-29,"The title of the patent is model-free adaptive control for industrial processes and its abstract is an enhanced model-free adaptive controller is disclosed, which consists of a linear dynamic neural network that can be easily configured and put in automatic mode to control simple to complex processes. two multivariable model-free adaptive controller designs are disclosed. an enhanced anti-delay model-free adaptive controller is introduced to control processes with large time delays. a feedforward/feedback model-free adaptive control system with two designs is introduced to compensate for measurable disturbances. dated 2003-04-29"
6559450,gamma camera with two sequential correction maps,"a method of correcting errors in imaging data in a gamma camera including determining a first correction map based on one or both of (1) calculated corrections and (2) a first data acquisition, determining a second correction map based on a second data acquisition and correcting the imaging data based on the first and second correction maps. in a preferred embodiment of the invention, error correction is implemented using a neural network. alternatively, a neural network can be used to perform the entire calculation of event position and/or energy.",2003-05-06,"The title of the patent is gamma camera with two sequential correction maps and its abstract is a method of correcting errors in imaging data in a gamma camera including determining a first correction map based on one or both of (1) calculated corrections and (2) a first data acquisition, determining a second correction map based on a second data acquisition and correcting the imaging data based on the first and second correction maps. in a preferred embodiment of the invention, error correction is implemented using a neural network. alternatively, a neural network can be used to perform the entire calculation of event position and/or energy. dated 2003-05-06"
6560360,feed forward feed back multiple neural network with context driven recognition,"a recognition system is disclosed, including a representation of an object in terms of its constituent parts that is translationally invariant, and which provides scale invariant recognition. the system further provides effective recognition of patterns that are partially present in the input signal, or that are partially occluded, and also provides an effective representation for sequences within the input signal. the system utilizes dynamically determined, context based expectations, for identifying individual features/parts of an object to be recognized. the system is computationally efficient, and capable of highly parallel implementation, and further includes a mechanism for improving the preprocessing of individual sections of an input pattern, either by applying one or more preprocessors selected from a set of several preprocessors, or by changing the parameters within a single preprocessor.",2003-05-06,"The title of the patent is feed forward feed back multiple neural network with context driven recognition and its abstract is a recognition system is disclosed, including a representation of an object in terms of its constituent parts that is translationally invariant, and which provides scale invariant recognition. the system further provides effective recognition of patterns that are partially present in the input signal, or that are partially occluded, and also provides an effective representation for sequences within the input signal. the system utilizes dynamically determined, context based expectations, for identifying individual features/parts of an object to be recognized. the system is computationally efficient, and capable of highly parallel implementation, and further includes a mechanism for improving the preprocessing of individual sections of an input pattern, either by applying one or more preprocessors selected from a set of several preprocessors, or by changing the parameters within a single preprocessor. dated 2003-05-06"
6560498,formation method and device for curved plates,"this invention includes the generation of forming information and its manipulation scheme as a method to form curved plates in ship hull-pieces. this invention consists of three components as follows: one is to construct and utilize a database which includes data about flat plates, objective curved plates, plates which are being formed, and their forming information, another is to infer new forming information with an artificial neural network system, and the third is to obtain forming information through calculating in-plane and bending strains. in the third, initial forming information is obtained by calculating strains from relationship between flat plates and objective curved plates. and new forming information is yielded through calculating the strains from relationship between partially formed curved plates and objective curved plates. final objective plate are reached by repeatedly performing the measurement of the difference between plates in the proceeding steps and final objective plates and the calculation of the new strains in each process. therefore, through this invention standardization and automation can be realized in the formation of curved plates.",2003-05-06,"The title of the patent is formation method and device for curved plates and its abstract is this invention includes the generation of forming information and its manipulation scheme as a method to form curved plates in ship hull-pieces. this invention consists of three components as follows: one is to construct and utilize a database which includes data about flat plates, objective curved plates, plates which are being formed, and their forming information, another is to infer new forming information with an artificial neural network system, and the third is to obtain forming information through calculating in-plane and bending strains. in the third, initial forming information is obtained by calculating strains from relationship between flat plates and objective curved plates. and new forming information is yielded through calculating the strains from relationship between partially formed curved plates and objective curved plates. final objective plate are reached by repeatedly performing the measurement of the difference between plates in the proceeding steps and final objective plates and the calculation of the new strains in each process. therefore, through this invention standardization and automation can be realized in the formation of curved plates. dated 2003-05-06"
6560540,method for mapping seismic attributes using neural networks,a method for training a probabilistic neural network to map seismic attributes or similar quantities.,2003-05-06,The title of the patent is method for mapping seismic attributes using neural networks and its abstract is a method for training a probabilistic neural network to map seismic attributes or similar quantities. dated 2003-05-06
6560582,dynamic memory processor,"a dynamic memory processor for time variant pattern recognition and an input data dimensionality reduction is provided having a multi-layer harmonic neural network and a classifier network. the multi-layer harmonic neural network receives a fused feature vector of the pattern to be recognized from a neural sensor and generates output vectors which aid in discrimination between similar patterns. the fused feature vector and each output vector are separately provided to corresponding positional king of the mountain (pkom) circuits within the classifier network. each pkom circuit generates a positional output vector with only one element having a value corresponding to one, the element corresponding to the element of its input vector having the highest contribution. the positional output vectors are mapped into a multidimensional memory space and read by a recognition vector array which generates a plurality of recognition vectors.",2003-05-06,"The title of the patent is dynamic memory processor and its abstract is a dynamic memory processor for time variant pattern recognition and an input data dimensionality reduction is provided having a multi-layer harmonic neural network and a classifier network. the multi-layer harmonic neural network receives a fused feature vector of the pattern to be recognized from a neural sensor and generates output vectors which aid in discrimination between similar patterns. the fused feature vector and each output vector are separately provided to corresponding positional king of the mountain (pkom) circuits within the classifier network. each pkom circuit generates a positional output vector with only one element having a value corresponding to one, the element corresponding to the element of its input vector having the highest contribution. the positional output vectors are mapped into a multidimensional memory space and read by a recognition vector array which generates a plurality of recognition vectors. dated 2003-05-06"
6560583,method and apparatus for constructing a self-adapting smart transmission device to control information transmission between elements of a network,"a process and apparatus that enable continual self-adaptation of dynamic transmission devices (sastds), allow &#8220;smart&#8221; filtering of information for transmission from one element to another within a network of interacting elements, such as a &#8220;neural network&#8221;. an adaptation algorithm (mapsa) incorporated in stds allows multiple parameters of stds (which determine its filtering properties) to continuously adapt simultaneously and interdependently. in this manner, complex correlations can be established between the parameters in all stds within a network of interacting elements. the process according to the invention therefore establishes unique patterns of connection parameters within the network which in turn dictates a novel sequence of information processing steps by the network.",2003-05-06,"The title of the patent is method and apparatus for constructing a self-adapting smart transmission device to control information transmission between elements of a network and its abstract is a process and apparatus that enable continual self-adaptation of dynamic transmission devices (sastds), allow &#8220;smart&#8221; filtering of information for transmission from one element to another within a network of interacting elements, such as a &#8220;neural network&#8221;. an adaptation algorithm (mapsa) incorporated in stds allows multiple parameters of stds (which determine its filtering properties) to continuously adapt simultaneously and interdependently. in this manner, complex correlations can be established between the parameters in all stds within a network of interacting elements. the process according to the invention therefore establishes unique patterns of connection parameters within the network which in turn dictates a novel sequence of information processing steps by the network. dated 2003-05-06"
6560586,multiresolution learning paradigm and signal prediction,"a neural network learning process provides a trained network that has good generalization ability for even highly nonlinear dynamic systems, and is trained with approximations of a signal obtained, each at a different respective resolution, using wavelet transformation. approximations are used in order from low to high. the trained neural network is used to predict values. in a preferred embodiment of the invention, the trained neural network is used in predicting network traffic patterns.",2003-05-06,"The title of the patent is multiresolution learning paradigm and signal prediction and its abstract is a neural network learning process provides a trained network that has good generalization ability for even highly nonlinear dynamic systems, and is trained with approximations of a signal obtained, each at a different respective resolution, using wavelet transformation. approximations are used in order from low to high. the trained neural network is used to predict values. in a preferred embodiment of the invention, the trained neural network is used in predicting network traffic patterns. dated 2003-05-06"
6564198,fuzzy expert system for interpretable rule extraction from neural networks,"an method and apparatus for extracting an interpretable, meaningful, and concise rule set from neural networks is presented. the method involves adjustment of gain parameter, &lgr; and the threshold, tj for the sigmoid activation function of the interactive-or operator used in the extraction/development of a rule set from an artificial neural network. a multi-stage procedure involving coarse and fine adjustment is used in order to constrain the range of the antecedents of the extracted rules to the range of values of the inputs to the artificial neural network. furthermore, the consequents of the extracted rules are provided based on degree of membership such that they are easily understandable by human beings. the method disclosed may be applied to any pattern recognition task, and is particularly useful in applications such as vehicle occupant sensing and recognition, object recognition, gesture recognition, and facial pattern recognition, among others.",2003-05-13,"The title of the patent is fuzzy expert system for interpretable rule extraction from neural networks and its abstract is an method and apparatus for extracting an interpretable, meaningful, and concise rule set from neural networks is presented. the method involves adjustment of gain parameter, &lgr; and the threshold, tj for the sigmoid activation function of the interactive-or operator used in the extraction/development of a rule set from an artificial neural network. a multi-stage procedure involving coarse and fine adjustment is used in order to constrain the range of the antecedents of the extracted rules to the range of values of the inputs to the artificial neural network. furthermore, the consequents of the extracted rules are provided based on degree of membership such that they are easily understandable by human beings. the method disclosed may be applied to any pattern recognition task, and is particularly useful in applications such as vehicle occupant sensing and recognition, object recognition, gesture recognition, and facial pattern recognition, among others. dated 2003-05-13"
6565039,wing-drive mechanism and vehicle employing same,"a wing-drive mechanism is described that permits, with proper control, movement of a wing about multiple wing trajectories. the wing-drive is capable of independent movement about three rotational degrees of movement; movement about a flap axis is independent of movement about a yaw axis, and both are independent of changes in the pitch of the wing. methods of controlling the wing-drive mechanism to affect a desired wing trajectory include the use of a non-linear automated controller that generates input signals to the wing-drive mechanism by comparing actual and desired wing trajectories in real time. specification of wing trajectories is preferably also accomplished in real time using an automated trajectory specification system, which can include a fuzzy logic processor or a neural network.",2003-05-20,"The title of the patent is wing-drive mechanism and vehicle employing same and its abstract is a wing-drive mechanism is described that permits, with proper control, movement of a wing about multiple wing trajectories. the wing-drive is capable of independent movement about three rotational degrees of movement; movement about a flap axis is independent of movement about a yaw axis, and both are independent of changes in the pitch of the wing. methods of controlling the wing-drive mechanism to affect a desired wing trajectory include the use of a non-linear automated controller that generates input signals to the wing-drive mechanism by comparing actual and desired wing trajectories in real time. specification of wing trajectories is preferably also accomplished in real time using an automated trajectory specification system, which can include a fuzzy logic processor or a neural network. dated 2003-05-20"
6567485,apparatus for communicating between a neural network and a user system via a bus,"apparatus for communication between a neural network and a user system via a bus includes an activity/frequency converter for each neurone of the network. the activity/frequency converter produces activity pulses which are encoded by encoders and then placed on the communication bus. arbitration arrangements for each converter include an inhibition control circuit and a blocking circuit connected in common to all the converters to transmit a temporary blocking command to them. each control circuit detects the presence of a pulse at the output of its associated converter and, while any such pulse is present, activates the blocking circuit so that it transmits the command for temporarily blocking their operation to the other converters.",2003-05-20,"The title of the patent is apparatus for communicating between a neural network and a user system via a bus and its abstract is apparatus for communication between a neural network and a user system via a bus includes an activity/frequency converter for each neurone of the network. the activity/frequency converter produces activity pulses which are encoded by encoders and then placed on the communication bus. arbitration arrangements for each converter include an inhibition control circuit and a blocking circuit connected in common to all the converters to transmit a temporary blocking command to them. each control circuit detects the presence of a pulse at the output of its associated converter and, while any such pulse is present, activates the blocking circuit so that it transmits the command for temporarily blocking their operation to the other converters. dated 2003-05-20"
6567795,artificial neural network and fuzzy logic based boiler tube leak detection systems,"power industry boiler tube failures are a major cause of utility forced outages in the united states, with approximately 41,000 tube failures occurring every year at a cost of $5 billion a year. accordingly, early tube leak detection and isolation is highly desirable. early detection allows scheduling of a repair rather than suffering a forced outage, and significantly increases the chance of preventing damage to adjacent tubes. the instant detection scheme starts with identification of boiler tube leak process variables which are divided into universal sensitive variables, local leak sensitive variables, group leak sensitive variables, and subgroup leak sensitive variables, and which may be automatically be obtained using a data driven approach and a leak sensitivity function. one embodiment uses artificial neural networks (ann) to learn the map between appropriate leak sensitive variables and the leak behavior. the second design philosophy integrates anns with approximate reasoning using fuzzy logic and fuzzy sets. in the second design, anns are used for learning, while approximate reasoning and inference engines are used for decision making. advantages include use of already monitored process variables, no additional hardware and/or maintenance requirements, systematic processing does not require an expert system and/or a skilled operator, and the systems are portable and can be easily tailored for use on a variety of different boilers.",2003-05-20,"The title of the patent is artificial neural network and fuzzy logic based boiler tube leak detection systems and its abstract is power industry boiler tube failures are a major cause of utility forced outages in the united states, with approximately 41,000 tube failures occurring every year at a cost of $5 billion a year. accordingly, early tube leak detection and isolation is highly desirable. early detection allows scheduling of a repair rather than suffering a forced outage, and significantly increases the chance of preventing damage to adjacent tubes. the instant detection scheme starts with identification of boiler tube leak process variables which are divided into universal sensitive variables, local leak sensitive variables, group leak sensitive variables, and subgroup leak sensitive variables, and which may be automatically be obtained using a data driven approach and a leak sensitivity function. one embodiment uses artificial neural networks (ann) to learn the map between appropriate leak sensitive variables and the leak behavior. the second design philosophy integrates anns with approximate reasoning using fuzzy logic and fuzzy sets. in the second design, anns are used for learning, while approximate reasoning and inference engines are used for decision making. advantages include use of already monitored process variables, no additional hardware and/or maintenance requirements, systematic processing does not require an expert system and/or a skilled operator, and the systems are portable and can be easily tailored for use on a variety of different boilers. dated 2003-05-20"
6568634,wing-drive mechanism and vehicle employing same,"a wing-drive mechanism is described that permits, with proper control, movement of a wing about multiple wing trajectories. the wing-drive is capable of independent movement about three rotational degrees of movement; movement about a flap axis is independent of movement about a yaw axis, and both are independent of changes in the pitch of the wing. methods of controlling the wing-drive mechanism to affect a desired wing trajectory include the use of a non-linear automated controller that generates input signals to the wing-drive mechanism by comparing actual and desired wing trajectories in real time. specification of wing trajectories is preferably also accomplished in real time using an automated trajectory specification system, which can include a fuzzy logic processor or a neural network. a vehicle that derives controlled motion as a whole from the wing-drive mechanism is also disclosed.",2003-05-27,"The title of the patent is wing-drive mechanism and vehicle employing same and its abstract is a wing-drive mechanism is described that permits, with proper control, movement of a wing about multiple wing trajectories. the wing-drive is capable of independent movement about three rotational degrees of movement; movement about a flap axis is independent of movement about a yaw axis, and both are independent of changes in the pitch of the wing. methods of controlling the wing-drive mechanism to affect a desired wing trajectory include the use of a non-linear automated controller that generates input signals to the wing-drive mechanism by comparing actual and desired wing trajectories in real time. specification of wing trajectories is preferably also accomplished in real time using an automated trajectory specification system, which can include a fuzzy logic processor or a neural network. a vehicle that derives controlled motion as a whole from the wing-drive mechanism is also disclosed. dated 2003-05-27"
6571228,hybrid neural networks for color identification,"a method uses a hybrid neural network including a self organizing mapping neural network (som nn) and a, back-propagation neural network (bp nn) for color identification. in the method the red, green and blue (rgb) of color samples are input as features of training samples and are automatically classified by way of som nn. afterwards, the outcomes of som nn are respectively delivered to various bp nn for further learning; and the map relationship of the input and the output defines the x,y, z corresponding the x, y and z values of a coordinate system of the standard color samples of rgb and it8. by way of the above learning structure, a non-linear model of color identification can be set up. after color samples are self organized and classified by som nn network, data can be categorized in clusters as a result of characteristic difference thereof. then the data are respectively sent to bp nn for learning whereby-the learning system not only can be quickly converged but also lower error discrepancy in operation effectively.",2003-05-27,"The title of the patent is hybrid neural networks for color identification and its abstract is a method uses a hybrid neural network including a self organizing mapping neural network (som nn) and a, back-propagation neural network (bp nn) for color identification. in the method the red, green and blue (rgb) of color samples are input as features of training samples and are automatically classified by way of som nn. afterwards, the outcomes of som nn are respectively delivered to various bp nn for further learning; and the map relationship of the input and the output defines the x,y, z corresponding the x, y and z values of a coordinate system of the standard color samples of rgb and it8. by way of the above learning structure, a non-linear model of color identification can be set up. after color samples are self organized and classified by som nn network, data can be categorized in clusters as a result of characteristic difference thereof. then the data are respectively sent to bp nn for learning whereby-the learning system not only can be quickly converged but also lower error discrepancy in operation effectively. dated 2003-05-27"
6572560,multi-modal cardiac diagnostic decision support system and method,"a method for extracting features from cardiac acoustic signals includes the steps of obtaining a cardiac acoustic signal, and extracting physiologically significant features from the cardiac acoustic signal using a neural network. a method for evaluating cardiac acoustic signals includes the steps of obtaining a cardiac acoustic signal, analyzing the cardiac acoustic signal with a wavelet decomposition to extract time-frequency information, and identifying basic heart sounds using neural networks applied to the extracted time-frequency information. a method for determining cardiac event sequences from cardiac acoustic signals includes the steps of obtaining a cardiac acoustic signal, and processing a sequence of features extracted from the cardiac acoustic signal by a probabilistic finite-state automaton to determine a most probable sequence of cardiac events given the cardiac acoustic signal. a method for extracting findings from cardiac acoustic signals includes the steps of obtaining a cardiac acoustic signal, processing the cardiac acoustic signal to determine a most probable sequence of cardiac events given the cardiac acoustic signal, and extracting the clinical findings from the sequence of cardiac events. a method for determining a status of heart murmurs includes the steps of obtaining a cardiac acoustic signal, detecting a murmur, if any, from the cardiac acoustic signal, and determining whether the murmur is one of functional and pathological based upon expert rules.",2003-06-03,"The title of the patent is multi-modal cardiac diagnostic decision support system and method and its abstract is a method for extracting features from cardiac acoustic signals includes the steps of obtaining a cardiac acoustic signal, and extracting physiologically significant features from the cardiac acoustic signal using a neural network. a method for evaluating cardiac acoustic signals includes the steps of obtaining a cardiac acoustic signal, analyzing the cardiac acoustic signal with a wavelet decomposition to extract time-frequency information, and identifying basic heart sounds using neural networks applied to the extracted time-frequency information. a method for determining cardiac event sequences from cardiac acoustic signals includes the steps of obtaining a cardiac acoustic signal, and processing a sequence of features extracted from the cardiac acoustic signal by a probabilistic finite-state automaton to determine a most probable sequence of cardiac events given the cardiac acoustic signal. a method for extracting findings from cardiac acoustic signals includes the steps of obtaining a cardiac acoustic signal, processing the cardiac acoustic signal to determine a most probable sequence of cardiac events given the cardiac acoustic signal, and extracting the clinical findings from the sequence of cardiac events. a method for determining a status of heart murmurs includes the steps of obtaining a cardiac acoustic signal, detecting a murmur, if any, from the cardiac acoustic signal, and determining whether the murmur is one of functional and pathological based upon expert rules. dated 2003-06-03"
6574565,system and method for enhanced hydrocarbon recovery,"a neural network based system, method, and process for the automated delineation of spatially dependent objects is disclosed. the method is applicable to objects such as hydrocarbon accumulations, aeromagnetic profiles, astronomical clusters, weather clusters, objects from radar, sonar, seismic and infrared returns, etc. one of the novelties in the present invention is that the method can be utilized whether or not known data is available to provide traditional training sets. the output consists of a classification of the input data into clearly delineated accumulations, clusters, objects, etc. that have various types and properties. a preferred but non-exclusive application of the present invention is the automated delineation of hydrocarbon accumulations and sub-regions within the accumulations with various properties, in an oil and gas field, prior to the commencement of drilling operations. the invention may also be used to increase the effectiveness of enhanced oil recovery techniques.",2003-06-03,"The title of the patent is system and method for enhanced hydrocarbon recovery and its abstract is a neural network based system, method, and process for the automated delineation of spatially dependent objects is disclosed. the method is applicable to objects such as hydrocarbon accumulations, aeromagnetic profiles, astronomical clusters, weather clusters, objects from radar, sonar, seismic and infrared returns, etc. one of the novelties in the present invention is that the method can be utilized whether or not known data is available to provide traditional training sets. the output consists of a classification of the input data into clearly delineated accumulations, clusters, objects, etc. that have various types and properties. a preferred but non-exclusive application of the present invention is the automated delineation of hydrocarbon accumulations and sub-regions within the accumulations with various properties, in an oil and gas field, prior to the commencement of drilling operations. the invention may also be used to increase the effectiveness of enhanced oil recovery techniques. dated 2003-06-03"
6574632,multiple engine information retrieval and visualization system,"an information retrieval and visualization system utilizes multiple search engines for retrieving documents from a document database based upon user input queries. search engines include an n-gram search engine and a vector space model search engine using a neural network training algorithm. each search engine produces a common mathematical representation of each retrieved document. the retrieved documents are then combined and ranked. mathematical representations for each respective document is mapped onto a display. information displayed includes a three-dimensional display of keywords from the user input query. the three-dimensional visualization capability based upon the mathematical representation of information within the information retrieval and visualization system provides users with an intuitive understanding, with relevance feedback/query refinement techniques that can be better utilized, resulting in higher retrieval accuracy (precision).",2003-06-03,"The title of the patent is multiple engine information retrieval and visualization system and its abstract is an information retrieval and visualization system utilizes multiple search engines for retrieving documents from a document database based upon user input queries. search engines include an n-gram search engine and a vector space model search engine using a neural network training algorithm. each search engine produces a common mathematical representation of each retrieved document. the retrieved documents are then combined and ranked. mathematical representations for each respective document is mapped onto a display. information displayed includes a three-dimensional display of keywords from the user input query. the three-dimensional visualization capability based upon the mathematical representation of information within the information retrieval and visualization system provides users with an intuitive understanding, with relevance feedback/query refinement techniques that can be better utilized, resulting in higher retrieval accuracy (precision). dated 2003-06-03"
6574754,self-monitoring storage device using neural networks,"a digital data storage device such as a rotating magnetic disk drive contains an on-board condition monitoring system, comprising a neural network coupled to multiple inputs derived from measured parameters of disk drive operation. the neural network uses a configurable set of weights to compute one or more quantities representing disk drive condition as a function of the various inputs. the weights are stored in a configuration table, which can be overwritten by a host computer. the drive is sold and installed with one set of weights, based on the then existing knowledge of the disk drive designers, and may be updated in the field as the designers acquire experience data by simply writing the weights to the configuration table of the disk drive, without altering disk drive control code or other disk drive features. preferably, the disk drive designers include as input to the neural network any parameter which might conceivably be useful, even if the designers initially believe that the parameter has no significance. in this case, the designers can assign the parameter a weight of zero during initial release. if subsequent experience then shows that the parameter has some unexpected significance, the neural network can be corrected simply by changing weighting factors, without altering the control programming code.",2003-06-03,"The title of the patent is self-monitoring storage device using neural networks and its abstract is a digital data storage device such as a rotating magnetic disk drive contains an on-board condition monitoring system, comprising a neural network coupled to multiple inputs derived from measured parameters of disk drive operation. the neural network uses a configurable set of weights to compute one or more quantities representing disk drive condition as a function of the various inputs. the weights are stored in a configuration table, which can be overwritten by a host computer. the drive is sold and installed with one set of weights, based on the then existing knowledge of the disk drive designers, and may be updated in the field as the designers acquire experience data by simply writing the weights to the configuration table of the disk drive, without altering disk drive control code or other disk drive features. preferably, the disk drive designers include as input to the neural network any parameter which might conceivably be useful, even if the designers initially believe that the parameter has no significance. in this case, the designers can assign the parameter a weight of zero during initial release. if subsequent experience then shows that the parameter has some unexpected significance, the neural network can be corrected simply by changing weighting factors, without altering the control programming code. dated 2003-06-03"
6577700,neural network based multi-criteria optimization image reconstruction technique for imaging two- and three-phase flow systems using electrical capacitance tomography,"a new image reconstruction technique for imaging two- and three-phase flows using electrical capacitance tomography (ect) has been developed based on multi-criteria optimization using an analog neural network, hereafter referred to as neural network multi-criteria optimization image reconstruction (nn-moirt)). the reconstruction technique is a combination between multi-criteria optimization image reconstruction technique for linear tomography, and the so-called linear back projection (lbp) technique commonly used for capacitance tomography. the multi-criteria optimization image reconstruction problem is solved using hopfield model dynamic neural-network computing. for three-component imaging, the single-step sigmoid function in the hopfield networks is replaced by a double-step sigmoid function, allowing the neural computation to converge to three-distinct stable regions in the output space corresponding to the three components, enabling the differentiation among the single phases.",2003-06-10,"The title of the patent is neural network based multi-criteria optimization image reconstruction technique for imaging two- and three-phase flow systems using electrical capacitance tomography and its abstract is a new image reconstruction technique for imaging two- and three-phase flows using electrical capacitance tomography (ect) has been developed based on multi-criteria optimization using an analog neural network, hereafter referred to as neural network multi-criteria optimization image reconstruction (nn-moirt)). the reconstruction technique is a combination between multi-criteria optimization image reconstruction technique for linear tomography, and the so-called linear back projection (lbp) technique commonly used for capacitance tomography. the multi-criteria optimization image reconstruction problem is solved using hopfield model dynamic neural-network computing. for three-component imaging, the single-step sigmoid function in the hopfield networks is replaced by a double-step sigmoid function, allowing the neural computation to converge to three-distinct stable regions in the output space corresponding to the three components, enabling the differentiation among the single phases. dated 2003-06-10"
6577960,liquid gauging apparatus using a time delay neural network,"liquid gauging apparatus using a time delay neural network for determining a quantity of liquid in a container that is not directly measurable by sensors is disclosed. the apparatus comprises a plurality of sensors and a processor. each of the sensors are capable of measuring a respective parameter of the liquid and for producing a time varying sensor output signal representative of the respective parameter measured thereby. the processor is programmed to process the sensor output signals by a time delay neural network algorithm to determine a current quantity of the liquid in the container based on current and past parameter measurements of the sensor output signals. also disclosed is a method of training a time delay neural network algorithm for computing a quantity of liquid in a container from current and past liquid parameter sensor measurements. the method comprises the steps of: establishing a dynamic model of liquid behavior in the container and parameter measurements of the liquid behavior sensed by a plurality of sensors; deriving from the dynamic model training data sets for a plurality of liquid quantity values, each data set comprising current and past liquid parameter sensor measurement values corresponding to a liquid quantity value of the plurality, and the corresponding liquid quantity value; and training the time delay neural network algorithm with the derived training data sets.",2003-06-10,"The title of the patent is liquid gauging apparatus using a time delay neural network and its abstract is liquid gauging apparatus using a time delay neural network for determining a quantity of liquid in a container that is not directly measurable by sensors is disclosed. the apparatus comprises a plurality of sensors and a processor. each of the sensors are capable of measuring a respective parameter of the liquid and for producing a time varying sensor output signal representative of the respective parameter measured thereby. the processor is programmed to process the sensor output signals by a time delay neural network algorithm to determine a current quantity of the liquid in the container based on current and past parameter measurements of the sensor output signals. also disclosed is a method of training a time delay neural network algorithm for computing a quantity of liquid in a container from current and past liquid parameter sensor measurements. the method comprises the steps of: establishing a dynamic model of liquid behavior in the container and parameter measurements of the liquid behavior sensed by a plurality of sensors; deriving from the dynamic model training data sets for a plurality of liquid quantity values, each data set comprising current and past liquid parameter sensor measurement values corresponding to a liquid quantity value of the plurality, and the corresponding liquid quantity value; and training the time delay neural network algorithm with the derived training data sets. dated 2003-06-10"
6578020,method and system for converting code to executable code using neural networks implemented in a very large scale integration (vlsi) integrated circuit,disclosed is a an integrated circuit method and system for generating a compiler to map a code set to object code capable of being executed on an operating system platform. the integrated circuit is encoded with logic including at least one neural network. the at least one neural network in the integrated circuit is trained to convert the code set to object code. the at least one trained neural network is then used to convert the code set to object code.,2003-06-10,The title of the patent is method and system for converting code to executable code using neural networks implemented in a very large scale integration (vlsi) integrated circuit and its abstract is disclosed is a an integrated circuit method and system for generating a compiler to map a code set to object code capable of being executed on an operating system platform. the integrated circuit is encoded with logic including at least one neural network. the at least one neural network in the integrated circuit is trained to convert the code set to object code. the at least one trained neural network is then used to convert the code set to object code. dated 2003-06-10
6578021,method and system for classifying network devices in virtual lans,"network management information stored by network devices in a switched network is obtained at a network management workstation. this is information that relates to the activity of the network devices on the network, such as the logical address of the network devices in communication with other devices. for tcp/ip networks utilizing the nmp protocol, this information is stored in the mib or the rmon matrix group variables. this information feeds a neural network. the output of the neural network is a list of network devices grouped in virtual lans (vlans) such that network devices communicating, or having recently communicated, are grouped in the same vlan. the network management information is periodically updated so the vlan grouping can also be periodically refreshed to reflect current network device activity and thus optimize the network bandwidth.",2003-06-10,"The title of the patent is method and system for classifying network devices in virtual lans and its abstract is network management information stored by network devices in a switched network is obtained at a network management workstation. this is information that relates to the activity of the network devices on the network, such as the logical address of the network devices in communication with other devices. for tcp/ip networks utilizing the nmp protocol, this information is stored in the mib or the rmon matrix group variables. this information feeds a neural network. the output of the neural network is a list of network devices grouped in virtual lans (vlans) such that network devices communicating, or having recently communicated, are grouped in the same vlan. the network management information is periodically updated so the vlan grouping can also be periodically refreshed to reflect current network device activity and thus optimize the network bandwidth. dated 2003-06-10"
6578172,method and arrangement for implementing convolutional decoding,"the invention relates to a method and arrangement for advantageously decoding and channel correcting a convolutionally encoded signal received over a transmission path. the signal comprises code words and the arrangement comprises a neural network comprising a set of neurons which comprise a set of inputs and an output. the received code words (400) and at least some of the output signals (402) of the neural network neurons are connected to the inputs of the neurons, and the neurons comprise means (404) for multiplying at least some of the neuron inputs prior to combining means (406). the arrangement also comprises means (123) for estimating the transmission channel. further, estimated channel data (400) is connected to the inputs of the neurons and a predetermined neuron is arranged to give an estimate of the channel-corrected and decoded symbol in its output signal.",2003-06-10,"The title of the patent is method and arrangement for implementing convolutional decoding and its abstract is the invention relates to a method and arrangement for advantageously decoding and channel correcting a convolutionally encoded signal received over a transmission path. the signal comprises code words and the arrangement comprises a neural network comprising a set of neurons which comprise a set of inputs and an output. the received code words (400) and at least some of the output signals (402) of the neural network neurons are connected to the inputs of the neurons, and the neurons comprise means (404) for multiplying at least some of the neuron inputs prior to combining means (406). the arrangement also comprises means (123) for estimating the transmission channel. further, estimated channel data (400) is connected to the inputs of the neurons and a predetermined neuron is arranged to give an estimate of the channel-corrected and decoded symbol in its output signal. dated 2003-06-10"
6581048,3-brain architecture for an intelligent decision and control system,"a method and system for intelligent control of external devices using a mammalian brain-like structure having three parts. the method and system include a computer-implemented neural network system which is an extension of the model-based adaptive critic design and is applicable to real-time control (e.g., robotic control) and real-time distributed control. additional uses include data visualization, data mining, and other tasks requiring complex analysis of inter-relationships between data.",2003-06-17,"The title of the patent is 3-brain architecture for an intelligent decision and control system and its abstract is a method and system for intelligent control of external devices using a mammalian brain-like structure having three parts. the method and system include a computer-implemented neural network system which is an extension of the model-based adaptive critic design and is applicable to real-time control (e.g., robotic control) and real-time distributed control. additional uses include data visualization, data mining, and other tasks requiring complex analysis of inter-relationships between data. dated 2003-06-17"
6583651,neural network output sensing and decision circuit and method,"a device and method for selecting within a group of analog signals the one with the lowest or with the highest value. in one embodiment the device has a differential amplifier configuration having an input to receive a comparison signal, a plurality of inputs to receive analog signals and a corresponding plurality of outputs to provide digital voltage signals. this device also has at least one logic circuit having a plurality of input terminals, each connected to a corresponding output of the differential amplifier configuration, and having at least one output terminal.",2003-06-24,"The title of the patent is neural network output sensing and decision circuit and method and its abstract is a device and method for selecting within a group of analog signals the one with the lowest or with the highest value. in one embodiment the device has a differential amplifier configuration having an input to receive a comparison signal, a plurality of inputs to receive analog signals and a corresponding plurality of outputs to provide digital voltage signals. this device also has at least one logic circuit having a plurality of input terminals, each connected to a corresponding output of the differential amplifier configuration, and having at least one output terminal. dated 2003-06-24"
6584366,method and arrangement for neural modeling of a paper winding device,"a method for modeling a paper winding device, particularly for modeling a tambour drum cutter. influencing and control quantities are determined at a real paper winding device and are stored dependent on time. said quantities are used to determine the web strength depending on the wound number of layers or, respectively, or a correlative quantity is used as a target quantity along with the other relevant control and influencing quantities to enable a neural network to be trained as a model for a nip (ni) for this winding device. new data for training the network can be continually obtained during the operation of the arrangement, thereby improving the model. optimal control parameters can be determined for various production requirements by means of an iterative process dependent on the winding quality, which can also be determined from the web strength.",2003-06-24,"The title of the patent is method and arrangement for neural modeling of a paper winding device and its abstract is a method for modeling a paper winding device, particularly for modeling a tambour drum cutter. influencing and control quantities are determined at a real paper winding device and are stored dependent on time. said quantities are used to determine the web strength depending on the wound number of layers or, respectively, or a correlative quantity is used as a target quantity along with the other relevant control and influencing quantities to enable a neural network to be trained as a model for a nip (ni) for this winding device. new data for training the network can be continually obtained during the operation of the arrangement, thereby improving the model. optimal control parameters can be determined for various production requirements by means of an iterative process dependent on the winding quality, which can also be determined from the web strength. dated 2003-06-24"
6585647,method and means for synthetic structural imaging and volume estimation of biological tissue organs,"a system (10) having a processor, a memory, generates low frequency ultrasound signals to be applied to tissue (14) to generate a weighted sum of tissue ramp, step, and impulse signatures. the system (10) analyzes the tissue signatures to determine the low frequency target profile which is used to generate a graphic representation of the tissue as well as to estimate the volume of the tissue; to classify the tissue (14) as to type, and condition of the tissue using a set of stored tissue data. the classifier may include a neural network (34), and/or a nearest neighbor rule processor (36). the system (10) performs as a non-invasive acoustic measurement, an imaging system, and method which uses synthetic structural imaging (ssi) techniques to provide unique information concerning the size, shape of biological structures for classification, visualization of normal, abnormal tissues, organs, biological structures, etc.",2003-07-01,"The title of the patent is method and means for synthetic structural imaging and volume estimation of biological tissue organs and its abstract is a system (10) having a processor, a memory, generates low frequency ultrasound signals to be applied to tissue (14) to generate a weighted sum of tissue ramp, step, and impulse signatures. the system (10) analyzes the tissue signatures to determine the low frequency target profile which is used to generate a graphic representation of the tissue as well as to estimate the volume of the tissue; to classify the tissue (14) as to type, and condition of the tissue using a set of stored tissue data. the classifier may include a neural network (34), and/or a nearest neighbor rule processor (36). the system (10) performs as a non-invasive acoustic measurement, an imaging system, and method which uses synthetic structural imaging (ssi) techniques to provide unique information concerning the size, shape of biological structures for classification, visualization of normal, abnormal tissues, organs, biological structures, etc. dated 2003-07-01"
6587580,stencil printing process optimization for circuit pack assembly using neural network modeling,a system for determining the optimal settings for parameter of a stencil printing machine. the system generates a model mapping parameter inputs to output results. the model is then used to determine the optimal settings of parameter inputs in order to produce the desired results. one form of mapping is to generate a neural network model of a system. the neural network is generated from data that includes multiple sets of input parameter settings and the resulting output associated with the inputs. back propagation is then performed on the neural network to determine the optimal settings.,2003-07-01,The title of the patent is stencil printing process optimization for circuit pack assembly using neural network modeling and its abstract is a system for determining the optimal settings for parameter of a stencil printing machine. the system generates a model mapping parameter inputs to output results. the model is then used to determine the optimal settings of parameter inputs in order to produce the desired results. one form of mapping is to generate a neural network model of a system. the neural network is generated from data that includes multiple sets of input parameter settings and the resulting output associated with the inputs. back propagation is then performed on the neural network to determine the optimal settings. dated 2003-07-01
6587845,method and apparatus for identification and optimization of bioactive compounds using a neural network,a computational method for the discovery and design of therapeutically valuable bioactive compounds is presented. the method employed has successfully analyzed enzymatic inhibitors for their chemical properties through the use of a neural network and associated algorithms. this method is an improvement over the current methods of drug discovery which often employs a random search through a large library of synthesized chemical compounds or biological samples for bioactivity related to a specific therapeutic use. this time-consuming process is the most expensive portion of current drug discovery methods. the development of computational methods for the prediction of specific molecular activity will facilitate the design of novel chemotherapeutics or other chemically useful compounds. the novel neural network provided in the current invention is &#8220;trained&#8221; with the bioactivity of known compounds and then used to predict the bioactivity of unknown compounds.,2003-07-01,The title of the patent is method and apparatus for identification and optimization of bioactive compounds using a neural network and its abstract is a computational method for the discovery and design of therapeutically valuable bioactive compounds is presented. the method employed has successfully analyzed enzymatic inhibitors for their chemical properties through the use of a neural network and associated algorithms. this method is an improvement over the current methods of drug discovery which often employs a random search through a large library of synthesized chemical compounds or biological samples for bioactivity related to a specific therapeutic use. this time-consuming process is the most expensive portion of current drug discovery methods. the development of computational methods for the prediction of specific molecular activity will facilitate the design of novel chemotherapeutics or other chemically useful compounds. the novel neural network provided in the current invention is &#8220;trained&#8221; with the bioactivity of known compounds and then used to predict the bioactivity of unknown compounds. dated 2003-07-01
6590362,method and system for early detection of incipient faults in electric motors,"a method and system for early detection of incipient faults in an electric motor are disclosed. first, current and voltage values for one or more phases of the electric motor are measured during motor operations. a set of current predictions is then determined via a neural network-based current predictor based on the measured voltage values and an estimate of motor speed values of the electric motor. next, a set of residuals is generated by combining the set of current predictions with the measured current values. a set of fault indicators is subsequently computed from the set of residuals and the measured current values. finally, a determination is made as to whether or not there is an incipient electrical, mechanical, and/or electromechanical fault occurring based on the comparison result of the set of fault indicators and a set of predetermined baseline values.",2003-07-08,"The title of the patent is method and system for early detection of incipient faults in electric motors and its abstract is a method and system for early detection of incipient faults in an electric motor are disclosed. first, current and voltage values for one or more phases of the electric motor are measured during motor operations. a set of current predictions is then determined via a neural network-based current predictor based on the measured voltage values and an estimate of motor speed values of the electric motor. next, a set of residuals is generated by combining the set of current predictions with the measured current values. a set of fault indicators is subsequently computed from the set of residuals and the measured current values. finally, a determination is made as to whether or not there is an incipient electrical, mechanical, and/or electromechanical fault occurring based on the comparison result of the set of fault indicators and a set of predetermined baseline values. dated 2003-07-08"
6591254,method and apparatus for operating a neural network with missing and/or incomplete data,"a neural network system is provided that models the system in a system model (12) with the output thereof providing a predicted output. this predicted output is modified or controlled by an output control (14). input data is processed in a data preprocess step (10) to reconcile the data for input to the system model (12). additionally, the error resulted from the reconciliation is input to an uncertainty model to predict the uncertainty in the predicted output. this is input to a decision processor (20) which is utilized to control the output control (14). the output control (14) is controlled to either vary the predicted output or to inhibit the predicted output whenever the output of the uncertainty model (18) exceeds a predetermined decision threshold, input by a decision threshold block (22). additionally, a validity model (16) is also provided which represents the reliability or validity of the output as a function of the number of data points in a given data region during training of the system model (12). this predicts the confidence in the predicted output which is also input to the decision processor (20). the decision processor (20) therefore bases its decision on the predicted confidence and the predicted uncertainty. additionally, the uncertainty output by the data preprocess block (10) can be utilized to train the system model (12).",2003-07-08,"The title of the patent is method and apparatus for operating a neural network with missing and/or incomplete data and its abstract is a neural network system is provided that models the system in a system model (12) with the output thereof providing a predicted output. this predicted output is modified or controlled by an output control (14). input data is processed in a data preprocess step (10) to reconcile the data for input to the system model (12). additionally, the error resulted from the reconciliation is input to an uncertainty model to predict the uncertainty in the predicted output. this is input to a decision processor (20) which is utilized to control the output control (14). the output control (14) is controlled to either vary the predicted output or to inhibit the predicted output whenever the output of the uncertainty model (18) exceeds a predetermined decision threshold, input by a decision threshold block (22). additionally, a validity model (16) is also provided which represents the reliability or validity of the output as a function of the number of data points in a given data region during training of the system model (12). this predicts the confidence in the predicted output which is also input to the decision processor (20). the decision processor (20) therefore bases its decision on the predicted confidence and the predicted uncertainty. additionally, the uncertainty output by the data preprocess block (10) can be utilized to train the system model (12). dated 2003-07-08"
6591255,"automatic data extraction, error correction and forecasting system","a &#8220;rapid learner client service&#8221; (rlcs) system that allows a large number of end-users to obtain the benefits of a sophisticated neural-network forecasting system. rather than purchasing or developing a forecasting system of their own, rlcs clients subscribe to a forecasting service performed by forecasting equipment located at a remote site. this allows a single highly sophisticated forecasting system to meet the forecasting needs of a large number of subscribers. this forecasting service is performed by an rlcs server that periodically and automatically accesses the subscriber's computer to obtain a fresh set of input data. alternatively, the subscriber's computer may contact the rlcs server to initiate the process. this input data is then downloaded to the rlcs server, where it is checked and corrected for errors by imputing values for missing or deviant input values. the error-corrected input data is then used to compute a forecast of output values, which are downloaded to the client's computer. the rlcs server also computes and downloads a set accuracy statistics for the client's review.",2003-07-08,"The title of the patent is automatic data extraction, error correction and forecasting system and its abstract is a &#8220;rapid learner client service&#8221; (rlcs) system that allows a large number of end-users to obtain the benefits of a sophisticated neural-network forecasting system. rather than purchasing or developing a forecasting system of their own, rlcs clients subscribe to a forecasting service performed by forecasting equipment located at a remote site. this allows a single highly sophisticated forecasting system to meet the forecasting needs of a large number of subscribers. this forecasting service is performed by an rlcs server that periodically and automatically accesses the subscriber's computer to obtain a fresh set of input data. alternatively, the subscriber's computer may contact the rlcs server to initiate the process. this input data is then downloaded to the rlcs server, where it is checked and corrected for errors by imputing values for missing or deviant input values. the error-corrected input data is then used to compute a forecast of output values, which are downloaded to the client's computer. the rlcs server also computes and downloads a set accuracy statistics for the client's review. dated 2003-07-08"
6592222,flicker and frequency doubling in virtual reality,"a system for testing and quantifying visual field and other visual function information in a head-mounted virtual reality environment, utilizing a directed image formation device for scanning of a flickering image for display to the test subject. a method and an apparatus are also provided for utilizing a central neural network and a central data bank to perform automatic interpretation of the visual function test parameters obtained in a plurality of visual field testing systems, for a plurality of patients, with control and response signals being transmitted via the internet. the data produced by the testing systems are automatically analyzed and compared with patterns on which the neural network was previously trained, and clinical diagnoses for pathological conditions are thereby suggested to the respective clinician for each patient.",2003-07-15,"The title of the patent is flicker and frequency doubling in virtual reality and its abstract is a system for testing and quantifying visual field and other visual function information in a head-mounted virtual reality environment, utilizing a directed image formation device for scanning of a flickering image for display to the test subject. a method and an apparatus are also provided for utilizing a central neural network and a central data bank to perform automatic interpretation of the visual function test parameters obtained in a plurality of visual field testing systems, for a plurality of patients, with control and response signals being transmitted via the internet. the data produced by the testing systems are automatically analyzed and compared with patterns on which the neural network was previously trained, and clinical diagnoses for pathological conditions are thereby suggested to the respective clinician for each patient. dated 2003-07-15"
6594382,neural sensors,"a neural sensor is provided which receives raw input data defining a pattern, such as image or sound data, and generates a classification identifier for the pattern. the neural sensor has a pattern array former which organizes the raw input data into the proper array format. a first order processing section receives the pattern array and generates a first order feature vector illustrative of first order features of the input data. a second order processing section also receives the pattern array and generates at least one second order feature vector illustrative of gradients in the input data. a vector fusion section receives the feature vectors from the first and second order processing sections and generates a single fused feature vector which is provided to a pattern classifier network. the pattern classifier network, in turn, generates a pattern classification for the input data.",2003-07-15,"The title of the patent is neural sensors and its abstract is a neural sensor is provided which receives raw input data defining a pattern, such as image or sound data, and generates a classification identifier for the pattern. the neural sensor has a pattern array former which organizes the raw input data into the proper array format. a first order processing section receives the pattern array and generates a first order feature vector illustrative of first order features of the input data. a second order processing section also receives the pattern array and generates at least one second order feature vector illustrative of gradients in the input data. a vector fusion section receives the feature vectors from the first and second order processing sections and generates a single fused feature vector which is provided to a pattern classifier network. the pattern classifier network, in turn, generates a pattern classification for the input data. dated 2003-07-15"
6594602,methods of calibrating pressure and temperature transducers and associated apparatus,"a calibration method provides enhanced accuracy in calibrating outputs of sensors. in embodiments described herein, the outputs of one or more sensors are input to a neural network and the neural network is trained to generate calibrated outputs in response thereto. in one method, the neural network is trained to simulate the output of a known accurate reference sensor in response to input to the neural network of the output of a subject sensor. in another method, the neural network is trained to simulate the output of a known accurate reference sensor in response to input to the neural network of the output of a subject sensor and the output of a second sensor. additional methods are provided which compensate for changes in a stimulus applied to a sensor, the output which is indicative of another stimulus.",2003-07-15,"The title of the patent is methods of calibrating pressure and temperature transducers and associated apparatus and its abstract is a calibration method provides enhanced accuracy in calibrating outputs of sensors. in embodiments described herein, the outputs of one or more sensors are input to a neural network and the neural network is trained to generate calibrated outputs in response thereto. in one method, the neural network is trained to simulate the output of a known accurate reference sensor in response to input to the neural network of the output of a subject sensor. in another method, the neural network is trained to simulate the output of a known accurate reference sensor in response to input to the neural network of the output of a subject sensor and the output of a second sensor. additional methods are provided which compensate for changes in a stimulus applied to a sensor, the output which is indicative of another stimulus. dated 2003-07-15"
6594622,system and method for extracting symbols from numeric time series for forecasting extreme events,"a method for predicting extreme changes in numeric time series data includes converting a numeric time series into a sequence of symbols. a prediction method, such as a neural network or nearest neighbor algorithm is used to make the forecast. a numeric time series data is identified with extreme changes in them, and a window of length w that precedes the extreme change is extracted. those extracts of a time series are built into a matrix (characteristic matrix) for singular value decomposition. the built matrix undergoes singular value decomposition, which reveals the characteristic vectors (symbols) that are indicative of time series that have characteristics that precede an extreme event. to perform forecasting, a window of length w in a new time series is generated and the dot product of the windows is taken against a predetermined number of columns of characteristic matrix, and, forecasting is performed on the new series.",2003-07-15,"The title of the patent is system and method for extracting symbols from numeric time series for forecasting extreme events and its abstract is a method for predicting extreme changes in numeric time series data includes converting a numeric time series into a sequence of symbols. a prediction method, such as a neural network or nearest neighbor algorithm is used to make the forecast. a numeric time series data is identified with extreme changes in them, and a window of length w that precedes the extreme change is extracted. those extracts of a time series are built into a matrix (characteristic matrix) for singular value decomposition. the built matrix undergoes singular value decomposition, which reveals the characteristic vectors (symbols) that are indicative of time series that have characteristics that precede an extreme event. to perform forecasting, a window of length w in a new time series is generated and the dot product of the windows is taken against a predetermined number of columns of characteristic matrix, and, forecasting is performed on the new series. dated 2003-07-15"
6596973,pyrometer calibrated wafer temperature estimator,"a wafer temperature estimator calibrates contact-type temperature sensor measurements that are used by a temperature controller to control substrate temperature in a high temperature processing chamber. wafer temperature estimator parameters provide an estimated wafer temperature from contact-type temperature sensor measurements. the estimator parameters are refined using non-contact-type temperature sensor measurements during periods when the substrate temperature is decreasing or the heaters are off. a corresponding temperature control system includes a heater, a contact-type temperature sensor in close proximity to the substrate, and an optical pyrometer placed to read temperature directly from the substrate. a wafer temperature estimator uses the estimator parameters and measurements from the contact-type sensor to determine an estimated wafer temperature. a temperature controller reads the estimated wafer temperature and makes changes to the heater power accordingly. the wafer temperature estimator has a nonlinear neural network system that is trained using inputs from the various sensors.",2003-07-22,"The title of the patent is pyrometer calibrated wafer temperature estimator and its abstract is a wafer temperature estimator calibrates contact-type temperature sensor measurements that are used by a temperature controller to control substrate temperature in a high temperature processing chamber. wafer temperature estimator parameters provide an estimated wafer temperature from contact-type temperature sensor measurements. the estimator parameters are refined using non-contact-type temperature sensor measurements during periods when the substrate temperature is decreasing or the heaters are off. a corresponding temperature control system includes a heater, a contact-type temperature sensor in close proximity to the substrate, and an optical pyrometer placed to read temperature directly from the substrate. a wafer temperature estimator uses the estimator parameters and measurements from the contact-type sensor to determine an estimated wafer temperature. a temperature controller reads the estimated wafer temperature and makes changes to the heater power accordingly. the wafer temperature estimator has a nonlinear neural network system that is trained using inputs from the various sensors. dated 2003-07-22"
6597660,method for real-time traffic analysis on packet networks,"an architecture for capture and generation, and a set of methods for characterization, prediction, and classification of traffic in packet networks are disclosed. the architecture consists of a device that stores packet timing information and processes the data so that characterization, prediction, and classification algorithms can perform operations in real-time. a methodology is disclosed for real-time traffic analysis, characterization, prediction, and classification in packet networks. the methodology is based on the simultaneous aggregation of packet arrival times at different times scales. the traffic is represented at the synchronous carrier level by the arrival or non-arrival of a packet. the invention does not require knowledge about the information source, nor needs to decode the information contents of the packets. only the arrival timing information is required. the invention provides a characterization of the traffic on packet networks suitable for a real-time implementation. the methodology can be applied in real-time traffic classification by training a neural network from calculated second order statistics of the traffic of several known sources. performance descriptors for the network can also be obtained by calculating the deviation of the traffic distribution from calculated models. traffic prediction can also be done by training a neural network from a vector of the results of a given processing against a vector of results of the subsequent processing unit; noticing that the latter vector contains information at a larger time scale than the previous. the invention also provides a method of estimating an effective bandwidth measure in real time which can be used for connection admission control and dynamic routing in packet networks. the invention provides appropriate traffic descriptors that can be applied in more efficient traffic control on packet networks.",2003-07-22,"The title of the patent is method for real-time traffic analysis on packet networks and its abstract is an architecture for capture and generation, and a set of methods for characterization, prediction, and classification of traffic in packet networks are disclosed. the architecture consists of a device that stores packet timing information and processes the data so that characterization, prediction, and classification algorithms can perform operations in real-time. a methodology is disclosed for real-time traffic analysis, characterization, prediction, and classification in packet networks. the methodology is based on the simultaneous aggregation of packet arrival times at different times scales. the traffic is represented at the synchronous carrier level by the arrival or non-arrival of a packet. the invention does not require knowledge about the information source, nor needs to decode the information contents of the packets. only the arrival timing information is required. the invention provides a characterization of the traffic on packet networks suitable for a real-time implementation. the methodology can be applied in real-time traffic classification by training a neural network from calculated second order statistics of the traffic of several known sources. performance descriptors for the network can also be obtained by calculating the deviation of the traffic distribution from calculated models. traffic prediction can also be done by training a neural network from a vector of the results of a given processing against a vector of results of the subsequent processing unit; noticing that the latter vector contains information at a larger time scale than the previous. the invention also provides a method of estimating an effective bandwidth measure in real time which can be used for connection admission control and dynamic routing in packet networks. the invention provides appropriate traffic descriptors that can be applied in more efficient traffic control on packet networks. dated 2003-07-22"
6598047,method and system for searching text,"an associative search methodology is presented whereby words, phrases, entire documents or technical jargon are input to the system for the purposes of searching a database for like items. a self-organizing associative data structure utilizing latent semantic analysis (lsa) combined with a hierarchical mixture of experts (hme) neural network learning system tracks relationships between words and word groupings as well as user significance feedback to automatically expand the search so that data elements can be found which cross technological/taxonomic boundaries. this results in an increase in both recall and precision parameters for each individual user's search.",2003-07-22,"The title of the patent is method and system for searching text and its abstract is an associative search methodology is presented whereby words, phrases, entire documents or technical jargon are input to the system for the purposes of searching a database for like items. a self-organizing associative data structure utilizing latent semantic analysis (lsa) combined with a hierarchical mixture of experts (hme) neural network learning system tracks relationships between words and word groupings as well as user significance feedback to automatically expand the search so that data elements can be found which cross technological/taxonomic boundaries. this results in an increase in both recall and precision parameters for each individual user's search. dated 2003-07-22"
6600961,intelligent control method for injection machine,"an intelligent control method for injection machine is to transplant the intelligent control and prediction techniques of a neural network to an injection machine, which has been exemplified capable of deciding the quasi best machine parameters rapidly in couple processing cycles for increasing yield with least loss, and for detecting and adjusting conditions until a desired operation environment is obtained.",2003-07-29,"The title of the patent is intelligent control method for injection machine and its abstract is an intelligent control method for injection machine is to transplant the intelligent control and prediction techniques of a neural network to an injection machine, which has been exemplified capable of deciding the quasi best machine parameters rapidly in couple processing cycles for increasing yield with least loss, and for detecting and adjusting conditions until a desired operation environment is obtained. dated 2003-07-29"
6600984,method for judging the seriousness of a motor vehicle crash,"like in a conventional crash detection algorithm, a crash is detected if a threshold value of an acceleration or an integrated acceleration is exceeded. various characteristic values for characterizing the sensor signal paths are collected and supplied to the neural network. said neural network returns the characteristic values for the seriousness of the crash, thereby facilitating the various retention systems to reaction individually. the invention provides a method for reliably, simply and quickly making statements regarding the seriousness and the course of a motor vehicle crash and for controlling occupant protection systems according to demand.",2003-07-29,"The title of the patent is method for judging the seriousness of a motor vehicle crash and its abstract is like in a conventional crash detection algorithm, a crash is detected if a threshold value of an acceleration or an integrated acceleration is exceeded. various characteristic values for characterizing the sensor signal paths are collected and supplied to the neural network. said neural network returns the characteristic values for the seriousness of the crash, thereby facilitating the various retention systems to reaction individually. the invention provides a method for reliably, simply and quickly making statements regarding the seriousness and the course of a motor vehicle crash and for controlling occupant protection systems according to demand. dated 2003-07-29"
6601049,self-adjusting multi-layer neural network architectures and methods therefor,"a method and apparatus for using a neural network to process information includes multiple nodes arrayed in multiple layers for transforming input arrays from prior layers or the environment into output arrays for subsequent layers or output devices. learning rules based on reinforcement are applied. interconnections between nodes are provided in a manner whereby the number and structure of the interconnections are self-adjusted by the learning rules during learning. at least one of the layers is used as a processing layer, and multiple lateral inputs to each node of each processing layer are used to retrieve information. the invention provides rapid, unsupervised processing of complex data sets, such as imagery or continuous human speech, and captures successful processing or pattern classification constellations for implementation in other networks. the invention includes application-specific self-adjusting multi-layer architectures that employ reinforcement learning rules to create updated data arrays for computation.",2003-07-29,"The title of the patent is self-adjusting multi-layer neural network architectures and methods therefor and its abstract is a method and apparatus for using a neural network to process information includes multiple nodes arrayed in multiple layers for transforming input arrays from prior layers or the environment into output arrays for subsequent layers or output devices. learning rules based on reinforcement are applied. interconnections between nodes are provided in a manner whereby the number and structure of the interconnections are self-adjusted by the learning rules during learning. at least one of the layers is used as a processing layer, and multiple lateral inputs to each node of each processing layer are used to retrieve information. the invention provides rapid, unsupervised processing of complex data sets, such as imagery or continuous human speech, and captures successful processing or pattern classification constellations for implementation in other networks. the invention includes application-specific self-adjusting multi-layer architectures that employ reinforcement learning rules to create updated data arrays for computation. dated 2003-07-29"
6601051,neural systems with range reducers and/or extenders,"a neural system is disclosed for processing an exogenous input process to produce a good outward output process with respect to a performance criterion, even if the range of one or both of these processes is necessarily large and/or keeps necessarily expanding during the operation of the neural system. the disclosed neural system comprises a recurrent neural network (rnn) and at least one range extender or reducer, each of which is a dynamic transformer. a range reducer transforms dynamically at least one component of the exogenous input process into inputs to at least one input neuron of said rnn. a range extender transforms dynamically outputs of at least one output neuron of said rnn into at least one component of the outward output process. there are many types of range extender and reducer, which have different degrees of effectiveness and computational costs. for a neural system under design, the types of range extenders and/or reducers are selected jointly with the architecture of the rnn in consideration of the neural system's processing performance with respect to the mentioned performance criterion, the rnn's size and the computational cost of selected range extenders and reducers so as to maximize the cost effectiveness of the neural system.",2003-07-29,"The title of the patent is neural systems with range reducers and/or extenders and its abstract is a neural system is disclosed for processing an exogenous input process to produce a good outward output process with respect to a performance criterion, even if the range of one or both of these processes is necessarily large and/or keeps necessarily expanding during the operation of the neural system. the disclosed neural system comprises a recurrent neural network (rnn) and at least one range extender or reducer, each of which is a dynamic transformer. a range reducer transforms dynamically at least one component of the exogenous input process into inputs to at least one input neuron of said rnn. a range extender transforms dynamically outputs of at least one output neuron of said rnn into at least one component of the outward output process. there are many types of range extender and reducer, which have different degrees of effectiveness and computational costs. for a neural system under design, the types of range extenders and/or reducers are selected jointly with the architecture of the rnn in consideration of the neural system's processing performance with respect to the mentioned performance criterion, the rnn's size and the computational cost of selected range extenders and reducers so as to maximize the cost effectiveness of the neural system. dated 2003-07-29"
6601052,selective attention method using neural network,"the present invention discloses an implementation of the selective attention mechanism occurring in the human brain using a conventional neural network, multi-layer perceptron and the error back-propagation method as a conventional learning method, and an application of the selective attention mechanism to perception of patterns such as voices or characters. in contrast to the conventional multi-layer perceptron and error back-propagation method in which the weighted value of the network is changed based on a given input signal, the selective attention algorithm of the present invention involves learning a present input pattern to minimize the error of the output layer with the weighted value set to a fixed value, so that the network can receive only a desired input signal to simulate the selective attention mechanism in the aspect of the biology. the present invention also used the selective attention algorithm to define the degree of attention to a plurality of candidate classes as a new criterion for perception, thus providing high perception performance relative to the conventional recognition system for a single candidate class.",2003-07-29,"The title of the patent is selective attention method using neural network and its abstract is the present invention discloses an implementation of the selective attention mechanism occurring in the human brain using a conventional neural network, multi-layer perceptron and the error back-propagation method as a conventional learning method, and an application of the selective attention mechanism to perception of patterns such as voices or characters. in contrast to the conventional multi-layer perceptron and error back-propagation method in which the weighted value of the network is changed based on a given input signal, the selective attention algorithm of the present invention involves learning a present input pattern to minimize the error of the output layer with the weighted value set to a fixed value, so that the network can receive only a desired input signal to simulate the selective attention mechanism in the aspect of the biology. the present invention also used the selective attention algorithm to define the degree of attention to a plurality of candidate classes as a new criterion for perception, thus providing high perception performance relative to the conventional recognition system for a single candidate class. dated 2003-07-29"
6601053,optimized artificial neural networks,neural network architectures are represented by symbol strings. an initial population of networks is trained and evaluated. the strings representing the fittest networks are modified according to a genetic algorithm and the process is repeated until an optimized network is produced.,2003-07-29,The title of the patent is optimized artificial neural networks and its abstract is neural network architectures are represented by symbol strings. an initial population of networks is trained and evaluated. the strings representing the fittest networks are modified according to a genetic algorithm and the process is repeated until an optimized network is produced. dated 2003-07-29
6601054,active acoustic and structural vibration control without online controller adjustment and path modeling,"active vibration control (avc) systems without online path modeling and controller adjustment are provided that are able to adapt to an uncertain operating environment. the controller (250, 280, 315, 252, 282, 317, 254, 319) of such an avc system is an adaptive recursive neural network whose weights are determined in an offline training and are held fixed online during the operation of the system. avc feedforward, feedback, and feedforward-feedback systems in accordance with the present invention are described. an avc feedforward system has no error sensor and an avc feedback system has no reference sensor. all sensor outputs of an avc system are processed by the controller for generating control signals to drive at least one secondary source (240). while an error sensor (480, 481) must be a vibrational sensor, a reference sensor (230, 270, 295, 305, 330) may be either a vibrational or nonvibrational sensor. the provided avc systems reduce or eliminate most of such shortcomings of the prior-art avc systems as use of an error sensor, relatively slow convergence of a weight/waveform adjustment algorithm, frequent adjustment of a path model, use of a high-order adaptive linear transversal filter, instability of an adaptive linear recursive filter, failure to use a useful nonvibrational reference sensor, failure to deal with the nonlinear behavior of a primary or secondary path, weight adjustment using control predicted values, use of an identification neural network, and online adjustment of the weights of a neural network.",2003-07-29,"The title of the patent is active acoustic and structural vibration control without online controller adjustment and path modeling and its abstract is active vibration control (avc) systems without online path modeling and controller adjustment are provided that are able to adapt to an uncertain operating environment. the controller (250, 280, 315, 252, 282, 317, 254, 319) of such an avc system is an adaptive recursive neural network whose weights are determined in an offline training and are held fixed online during the operation of the system. avc feedforward, feedback, and feedforward-feedback systems in accordance with the present invention are described. an avc feedforward system has no error sensor and an avc feedback system has no reference sensor. all sensor outputs of an avc system are processed by the controller for generating control signals to drive at least one secondary source (240). while an error sensor (480, 481) must be a vibrational sensor, a reference sensor (230, 270, 295, 305, 330) may be either a vibrational or nonvibrational sensor. the provided avc systems reduce or eliminate most of such shortcomings of the prior-art avc systems as use of an error sensor, relatively slow convergence of a weight/waveform adjustment algorithm, frequent adjustment of a path model, use of a high-order adaptive linear transversal filter, instability of an adaptive linear recursive filter, failure to use a useful nonvibrational reference sensor, failure to deal with the nonlinear behavior of a primary or secondary path, weight adjustment using control predicted values, use of an identification neural network, and online adjustment of the weights of a neural network. dated 2003-07-29"
6604029,multi-function air data probes using neural network for sideslip compensation,"an air data sensing probe such as a multi-function probe includes a barrel having multiple pressure sensing ports for sensing multiple pressures. instrumentation coupled to the pressure sensing ports provides electrical signals indicative of the pressures. an inertial navigation system input of the probe receives electrical signals indicative of inertial navigation data for the aircraft. a neural network of the probe receives as inputs the electrical signals indicative of the multiple pressures and the electrical signals indicative of the inertial navigation data. the neural network is trained or configured to provide as an output, electrical signals indicative of an air data parameter.",2003-08-05,"The title of the patent is multi-function air data probes using neural network for sideslip compensation and its abstract is an air data sensing probe such as a multi-function probe includes a barrel having multiple pressure sensing ports for sensing multiple pressures. instrumentation coupled to the pressure sensing ports provides electrical signals indicative of the pressures. an inertial navigation system input of the probe receives electrical signals indicative of inertial navigation data for the aircraft. a neural network of the probe receives as inputs the electrical signals indicative of the multiple pressures and the electrical signals indicative of the inertial navigation data. the neural network is trained or configured to provide as an output, electrical signals indicative of an air data parameter. dated 2003-08-05"
6604178,hard disk drive employing neural network for performing expected access time calculations,"a method and apparatus for calculating an expected access time associated with one of a plurality of disk drive commands employs one or more neural networks. a plurality of disk drive commands received from an external source are stored in a memory, typically in a queue. using a neural network, an expected access time associated with each of the queued commands is determined. determining the expected access time associated with each of the queued commands involves determining a time for performing a seek and settle operation for each of the queued commands and a latency time associated with each of the queued commands. the command indicated by the neural network as having a minimum expected access time relative to access times associated with other ones of the queued commands is identified for execution. a first neural network may be used to determine an expected access time associated with each read command stored in a read command queue and a second neural network may be used to determine an expected access time associated with each write command stored in a write command queue. a pair of read and write neural networks may be associated with each of a number of read/write transducers employed in a disk drive system. the neural network may be trained at the time of manufacture and on a periodic basis during the service life of the disk drive.",2003-08-05,"The title of the patent is hard disk drive employing neural network for performing expected access time calculations and its abstract is a method and apparatus for calculating an expected access time associated with one of a plurality of disk drive commands employs one or more neural networks. a plurality of disk drive commands received from an external source are stored in a memory, typically in a queue. using a neural network, an expected access time associated with each of the queued commands is determined. determining the expected access time associated with each of the queued commands involves determining a time for performing a seek and settle operation for each of the queued commands and a latency time associated with each of the queued commands. the command indicated by the neural network as having a minimum expected access time relative to access times associated with other ones of the queued commands is identified for execution. a first neural network may be used to determine an expected access time associated with each read command stored in a read command queue and a second neural network may be used to determine an expected access time associated with each write command stored in a write command queue. a pair of read and write neural networks may be associated with each of a number of read/write transducers employed in a disk drive system. the neural network may be trained at the time of manufacture and on a periodic basis during the service life of the disk drive. dated 2003-08-05"
6606612,method for constructing composite response surfaces by combining neural networks with other interpolation or estimation techniques,"a method and system for design optimization that incorporates the advantages of both traditional response surface methodology (rsm) and neural networks is disclosed. the present invention employs a unique strategy called parameter-based partitioning of the given design space. in the design procedure, a sequence of composite response surfaces based on both neural networks and polynomial fits is used to traverse the design space to identify an optimal solution. the composite response surface has both the power of neural networks and the economy of low-order polynomials (in terms of the number of simulations needed and the network training requirements). the present invention handles design problems with many more parameters than would be possible using neural networks alone and permits a designer to rapidly perform a variety of trade-off studies before arriving at the final design.",2003-08-12,"The title of the patent is method for constructing composite response surfaces by combining neural networks with other interpolation or estimation techniques and its abstract is a method and system for design optimization that incorporates the advantages of both traditional response surface methodology (rsm) and neural networks is disclosed. the present invention employs a unique strategy called parameter-based partitioning of the given design space. in the design procedure, a sequence of composite response surfaces based on both neural networks and polynomial fits is used to traverse the design space to identify an optimal solution. the composite response surface has both the power of neural networks and the economy of low-order polynomials (in terms of the number of simulations needed and the network training requirements). the present invention handles design problems with many more parameters than would be possible using neural networks alone and permits a designer to rapidly perform a variety of trade-off studies before arriving at the final design. dated 2003-08-12"
6606614,neural network integrated circuit with fewer pins,"a neural network integrated circuit comprises many neuron circuits each with a distance resister that is compared in a competition for the closest-hit with all the other neurons. such closest-hit comparison is conducted bit-by-bit over the many bit positions of a distance measure in binary format each time after the neurons fire. a single-wire and-bus interconnects every neuron in a whole system. each neuron drives the single-wire and-bus with an open-collector buffer. all neurons press the single-wire and-bus with their respective distance measures in successive cycles, starting with the most significant bit. for example, a fourteen-bit binary distance word requires fourteen comparison cycles. any neuron that sees a &#8220;0&#8221; on the single-wire and-bus when its own corresponding bit in its distance measure is a &#8220;1&#8221;, automatically drops from the competition. by the time the least significant bit cycle is run, a single closest distance will have been determined. such neuron that wins announces itself with an identifying code.",2003-08-12,"The title of the patent is neural network integrated circuit with fewer pins and its abstract is a neural network integrated circuit comprises many neuron circuits each with a distance resister that is compared in a competition for the closest-hit with all the other neurons. such closest-hit comparison is conducted bit-by-bit over the many bit positions of a distance measure in binary format each time after the neurons fire. a single-wire and-bus interconnects every neuron in a whole system. each neuron drives the single-wire and-bus with an open-collector buffer. all neurons press the single-wire and-bus with their respective distance measures in successive cycles, starting with the most significant bit. for example, a fourteen-bit binary distance word requires fourteen comparison cycles. any neuron that sees a &#8220;0&#8221; on the single-wire and-bus when its own corresponding bit in its distance measure is a &#8220;1&#8221;, automatically drops from the competition. by the time the least significant bit cycle is run, a single closest distance will have been determined. such neuron that wins announces itself with an identifying code. dated 2003-08-12"
6608924,neural network model for compressing/decompressing image/acoustic data files,"a new neural model for direct classification, dc, is introduced for acoustic/pictorial data compression. it is based on the adaptive resonance theorem and kohonen self organizing feature map neural models. in the adaptive training of the dc model, an input data file is vectorized into a domain of same size vector subunits. the result of the training (step 10 to 34) is to cluster the input vector domain into classes of similar subunits, and develop a center of mass called a centroid for each class to be stored in a codebook (cb) table. in the compression process, which is parallel to the training (step 33), for each input subunit, we obtain the index of the closest centroid in the cb. all indices and the cb will form the compressed file, cf. in the decompression phase (steps 42 to 52), for each index in the cf, a lookup process is performed into the cb to obtain the centroid representative of the original subunit. the obtained centroid is placed in the decompressed file. the compression is realized because the size of the input subunit ((8 or 24)*n2 bits) is an order of magnitude larger than its encoding index log2 [size of cb] bits. in order to achieve a better compression ratio, lzw is performed on cf (step 38) before storing (or transmitting) it.",2003-08-19,"The title of the patent is neural network model for compressing/decompressing image/acoustic data files and its abstract is a new neural model for direct classification, dc, is introduced for acoustic/pictorial data compression. it is based on the adaptive resonance theorem and kohonen self organizing feature map neural models. in the adaptive training of the dc model, an input data file is vectorized into a domain of same size vector subunits. the result of the training (step 10 to 34) is to cluster the input vector domain into classes of similar subunits, and develop a center of mass called a centroid for each class to be stored in a codebook (cb) table. in the compression process, which is parallel to the training (step 33), for each input subunit, we obtain the index of the closest centroid in the cb. all indices and the cb will form the compressed file, cf. in the decompression phase (steps 42 to 52), for each index in the cf, a lookup process is performed into the cb to obtain the centroid representative of the original subunit. the obtained centroid is placed in the decompressed file. the compression is realized because the size of the input subunit ((8 or 24)*n2 bits) is an order of magnitude larger than its encoding index log2 [size of cb] bits. in order to achieve a better compression ratio, lzw is performed on cf (step 38) before storing (or transmitting) it. dated 2003-08-19"
6609024,method of making a judgment on emotional positivity or negativity by detecting asymmetry of brain waves of left and right cerebral hemispheres,"a method of making a judgment on emotional positivist or negativity of a person, comprises the steps of obtaining asymmetry ratio between the brain waves of the left and right cerebral hemisphere measured in a given unit time by means of electrodes attached to the left and right side of the scalp, calculating asymmetry ratio versus time at each frequency of the measured brain waves if the time taken for measuring the brain waves exceeds a given time interval, calculating an increase and a decrease of the asymmetry ratio of the previous step with time, and entering the increase and decrease into an artificial neural network to make a judgment on the emotional positivity or negativity.",2003-08-19,"The title of the patent is method of making a judgment on emotional positivity or negativity by detecting asymmetry of brain waves of left and right cerebral hemispheres and its abstract is a method of making a judgment on emotional positivist or negativity of a person, comprises the steps of obtaining asymmetry ratio between the brain waves of the left and right cerebral hemisphere measured in a given unit time by means of electrodes attached to the left and right side of the scalp, calculating asymmetry ratio versus time at each frequency of the measured brain waves if the time taken for measuring the brain waves exceeds a given time interval, calculating an increase and a decrease of the asymmetry ratio of the previous step with time, and entering the increase and decrease into an artificial neural network to make a judgment on the emotional positivity or negativity. dated 2003-08-19"
6609060,system for intelligent control of an engine based on soft computing,a reduced control system suitable for control of an engine as a nonlinear plant is described. the reduced control system is configured to use a reduced sensor set for controlling the plant without significant loss of control quality (accuracy) as compared to an optimal control system with an optimum sensor set. the control system calculates the information content provided by the reduced sensor set as compared to the information content provided by the optimum set. the control system also calculates the difference between the entropy production rate of the plant and the entropy production rate of the controller. a genetic optimizer is used to tune a fuzzy neural network in the reduced controller. a fitness function for the genetic optimizer provides optimum control accuracy in the reduced control system by minimizing the difference in entropy production while maximizing the sensor information content.,2003-08-19,The title of the patent is system for intelligent control of an engine based on soft computing and its abstract is a reduced control system suitable for control of an engine as a nonlinear plant is described. the reduced control system is configured to use a reduced sensor set for controlling the plant without significant loss of control quality (accuracy) as compared to an optimal control system with an optimum sensor set. the control system calculates the information content provided by the reduced sensor set as compared to the information content provided by the optimum set. the control system also calculates the difference between the entropy production rate of the plant and the entropy production rate of the controller. a genetic optimizer is used to tune a fuzzy neural network in the reduced controller. a fitness function for the genetic optimizer provides optimum control accuracy in the reduced control system by minimizing the difference in entropy production while maximizing the sensor information content. dated 2003-08-19
6609118,methods and systems for automated property valuation,"the present invention is a method and system for automating the process for valuing a property that produces an estimated value of a subject property, and a quality assessment of the estimated value, that is based on the fusion of multiple processes for valuing a property. in one embodiment, three processes for valuing a subject property are fused. the first process, called locval, uses the location and living area to provide an estimate of the subject property's value. the second process, called aigen, is a generative artificial intelligence method that trains a fuzzy-neural network using a subset of cases from a case-base, and produces a run-time system to provide an estimate of the subject property's value. the third process, called aicomp, uses a case based reasoning process similar to the sales comparison approach to determine an estimate of the subject property's value.",2003-08-19,"The title of the patent is methods and systems for automated property valuation and its abstract is the present invention is a method and system for automating the process for valuing a property that produces an estimated value of a subject property, and a quality assessment of the estimated value, that is based on the fusion of multiple processes for valuing a property. in one embodiment, three processes for valuing a subject property are fused. the first process, called locval, uses the location and living area to provide an estimate of the subject property's value. the second process, called aigen, is a generative artificial intelligence method that trains a fuzzy-neural network using a subset of cases from a case-base, and produces a run-time system to provide an estimate of the subject property's value. the third process, called aicomp, uses a case based reasoning process similar to the sales comparison approach to determine an estimate of the subject property's value. dated 2003-08-19"
6611771,method and apparatus to detect a stator turn fault in an ac motor,"a stator turn fault detection system and method capable of real-time detection of a stator turn fault in an electric motor is provided. the stator turn fault detector includes a feed forward neural network that when trained, using fundamental frequency sequence components of the voltage and current supplying the electric motor, will estimate a fundamental frequency sequence component of current indicative of a stator turn fault. a method for detecting a stator turn fault in an electric motor as well as a method for training a feed forward neural network for use with the stator turn fault detector is disclosed.",2003-08-26,"The title of the patent is method and apparatus to detect a stator turn fault in an ac motor and its abstract is a stator turn fault detection system and method capable of real-time detection of a stator turn fault in an electric motor is provided. the stator turn fault detector includes a feed forward neural network that when trained, using fundamental frequency sequence components of the voltage and current supplying the electric motor, will estimate a fundamental frequency sequence component of current indicative of a stator turn fault. a method for detecting a stator turn fault in an electric motor as well as a method for training a feed forward neural network for use with the stator turn fault detector is disclosed. dated 2003-08-26"
6611823,backlash compensation using neural network,"methods and systems for backlash compensation. restrictive assumptions on the backlash nonlinearity (e.g. the same slopes of the lines, etc.) are not required. the compensator scheme has dynamic inversion structure, with a neural network in the feedforward path that approximates the backlash inversion error plus filter dynamics needed for backstepping design. the neural network controller does not require preliminary off-line training. neural network tuning is based on a modified hebbian tuning law, which requires less computation than backpropagation. the backstepping controller uses a practical filtered derivative, unlike the usual differentiation required by earlier backstepping routines. rigorous stability proofs are given using lyapunov theory. simulation results show that the proposed compensation scheme is an efficient way of improving the tracking performance of a vast array of nonlinear systems with backlash.",2003-08-26,"The title of the patent is backlash compensation using neural network and its abstract is methods and systems for backlash compensation. restrictive assumptions on the backlash nonlinearity (e.g. the same slopes of the lines, etc.) are not required. the compensator scheme has dynamic inversion structure, with a neural network in the feedforward path that approximates the backlash inversion error plus filter dynamics needed for backstepping design. the neural network controller does not require preliminary off-line training. neural network tuning is based on a modified hebbian tuning law, which requires less computation than backpropagation. the backstepping controller uses a practical filtered derivative, unlike the usual differentiation required by earlier backstepping routines. rigorous stability proofs are given using lyapunov theory. simulation results show that the proposed compensation scheme is an efficient way of improving the tracking performance of a vast array of nonlinear systems with backlash. dated 2003-08-26"
6618712,particle analysis using laser ablation mass spectroscopy,"the present invention provides a method of quickly identifying bioaerosols by class, even if the subject bioaerosol has not been previously encountered. the method begins by collecting laser ablation mass spectra from known particles. the spectra are correlated with the known particles, including the species of particle and the classification (e.g., bacteria). the spectra can then be used to train a neural network, for example using genetic algorithm-based training, to recognize each spectra and to recognize characteristics of the classifications. the spectra can also be used in a multivariate patch algorithm. laser ablation mass specta from unknown particles can be presented as inputs to the trained neural net for identification as to classification. the description below first describes suitable intelligent algorithms and multivariate patch algorithms, then presents an example of the present invention including results.",2003-09-09,"The title of the patent is particle analysis using laser ablation mass spectroscopy and its abstract is the present invention provides a method of quickly identifying bioaerosols by class, even if the subject bioaerosol has not been previously encountered. the method begins by collecting laser ablation mass spectra from known particles. the spectra are correlated with the known particles, including the species of particle and the classification (e.g., bacteria). the spectra can then be used to train a neural network, for example using genetic algorithm-based training, to recognize each spectra and to recognize characteristics of the classifications. the spectra can also be used in a multivariate patch algorithm. laser ablation mass specta from unknown particles can be presented as inputs to the trained neural net for identification as to classification. the description below first describes suitable intelligent algorithms and multivariate patch algorithms, then presents an example of the present invention including results. dated 2003-09-09"
6618713,neural directors,"a neural director is provided which is a neural network constructed with weights that are determined a priori. the neural director receives an input vector x comprising &#8220;i&#8221; input components &#8220;xi&#8221; and generates in response an output vector y comprising &#8220;j&#8221; output components. the neural director has an input processing node layer which receives the input vector x and an output processing node layer which generates the output vector y. the connections between the input and output processing node layers are a unique weighting set w(i,j) that contains an internal representation of a uniform spatial distribution of &#8220;j&#8221; unit vectors throughout a unit sphere of &#8220;i&#8221; dimensions. thus the cosine value between any two adjacent unit vectors is a constant everywhere in the unit sphere.",2003-09-09,"The title of the patent is neural directors and its abstract is a neural director is provided which is a neural network constructed with weights that are determined a priori. the neural director receives an input vector x comprising &#8220;i&#8221; input components &#8220;xi&#8221; and generates in response an output vector y comprising &#8220;j&#8221; output components. the neural director has an input processing node layer which receives the input vector x and an output processing node layer which generates the output vector y. the connections between the input and output processing node layers are a unique weighting set w(i,j) that contains an internal representation of a uniform spatial distribution of &#8220;j&#8221; unit vectors throughout a unit sphere of &#8220;i&#8221; dimensions. thus the cosine value between any two adjacent unit vectors is a constant everywhere in the unit sphere. dated 2003-09-09"
6622125,automatic sales promotion selection system and method,"an automated sales promotion selection system uses neural networks to identify promising sales promotions based on recent customer purchases. the system includes a customer information device that receives customer data relating to customer purchases of items from an inventory of items, a central processing unit having a sales promotion neural network and a storage unit containing a plurality of item identifiers comprising potential customer purchases of additional items from the inventory, wherein the sales opportunity neural network responds to customer data received from the customer information device by determining if one or more of the item identifiers in the storage unit corresponds to an item likely to be purchased by one of the customers, and an output device that receives the item identifiers of the likely purchases determined by the sales promotion neural network and produces a sales promotion relating to at least one of the item identifiers.",2003-09-16,"The title of the patent is automatic sales promotion selection system and method and its abstract is an automated sales promotion selection system uses neural networks to identify promising sales promotions based on recent customer purchases. the system includes a customer information device that receives customer data relating to customer purchases of items from an inventory of items, a central processing unit having a sales promotion neural network and a storage unit containing a plurality of item identifiers comprising potential customer purchases of additional items from the inventory, wherein the sales opportunity neural network responds to customer data received from the customer information device by determining if one or more of the item identifiers in the storage unit corresponds to an item likely to be purchased by one of the customers, and an output device that receives the item identifiers of the likely purchases determined by the sales promotion neural network and produces a sales promotion relating to at least one of the item identifiers. dated 2003-09-16"
6625588,associative neuron in an artificial neural network,"an associative artificial neuron and method of forming output signals of an associative artificial neuron includes receiving a number of auxiliary input signals; forming from the auxiliary input signals a sum weighted by coefficients and applying a non-linear function to the weighted sum to generate a non-linear signal. the neuron and method further include receiving a main input signal and forming, based on the main signal and the non-linear signal, the function s or v, which is used to generate a main output signal, and at lest one of three logical functions s and v, not s and v, and s and not v. the at least one logical function is used to generate an additional output signal for the associative artificial neuron.",2003-09-23,"The title of the patent is associative neuron in an artificial neural network and its abstract is an associative artificial neuron and method of forming output signals of an associative artificial neuron includes receiving a number of auxiliary input signals; forming from the auxiliary input signals a sum weighted by coefficients and applying a non-linear function to the weighted sum to generate a non-linear signal. the neuron and method further include receiving a main input signal and forming, based on the main signal and the non-linear signal, the function s or v, which is used to generate a main output signal, and at lest one of three logical functions s and v, not s and v, and s and not v. the at least one logical function is used to generate an additional output signal for the associative artificial neuron. dated 2003-09-23"
6627154,electronic techniques for analyte detection,"techniques are used to detect and identify analytes. techniques are used to fabricate and manufacture sensors to detect analytes. an analyte (1810) is sensed by sensors (1820) that output electrical signals in response to the analyte. the electrical signals are preprocessed (1830) by filtering and amplification. in an embodiment, this preprocessing includes adapting the sensor and electronics to the environment in which the analyte exists. the electrical signals are further processed (1840) to classify and identify the analyte, which may be by a neural network.",2003-09-30,"The title of the patent is electronic techniques for analyte detection and its abstract is techniques are used to detect and identify analytes. techniques are used to fabricate and manufacture sensors to detect analytes. an analyte (1810) is sensed by sensors (1820) that output electrical signals in response to the analyte. the electrical signals are preprocessed (1830) by filtering and amplification. in an embodiment, this preprocessing includes adapting the sensor and electronics to the environment in which the analyte exists. the electrical signals are further processed (1840) to classify and identify the analyte, which may be by a neural network. dated 2003-09-30"
6627464,adaptive plasma characterization system,"an adaptive plasma characterization system and method characterize a semiconductor plasma process using fuzzy logic and neural networks. the method includes the step of collecting input and output training data, where the input training data is based on variables associated with electrical power used to control a plasma chamber and results from execution of the plasma process. the method further includes the step of generating fuzzy logic-based input and output membership functions based on the training data. the membership functions enable estimation of an output parameter value of the plasma process, such that the membership functions characterize the plasma process with regard to the output parameter. modifying the membership functions based on a neural network learning algorithm and output data provides ability to learn. thus, etching process parameters such as etch rate, end point detection, and chamber maintenance can all be characterized in a manner that allows the system to operate autonomously.",2003-09-30,"The title of the patent is adaptive plasma characterization system and its abstract is an adaptive plasma characterization system and method characterize a semiconductor plasma process using fuzzy logic and neural networks. the method includes the step of collecting input and output training data, where the input training data is based on variables associated with electrical power used to control a plasma chamber and results from execution of the plasma process. the method further includes the step of generating fuzzy logic-based input and output membership functions based on the training data. the membership functions enable estimation of an output parameter value of the plasma process, such that the membership functions characterize the plasma process with regard to the output parameter. modifying the membership functions based on a neural network learning algorithm and output data provides ability to learn. thus, etching process parameters such as etch rate, end point detection, and chamber maintenance can all be characterized in a manner that allows the system to operate autonomously. dated 2003-09-30"
6629037,optimal paths for marine data collection,a method for finding an optimal path in seismic in fill shooting using a neural network to estimate cable feathering and uses multiple off set cost maps and potentials in cellular automata. the process determines cable coordinates at each cell location and store the coordinates of a predecessor cell to eliminate multiple paths.,2003-09-30,The title of the patent is optimal paths for marine data collection and its abstract is a method for finding an optimal path in seismic in fill shooting using a neural network to estimate cable feathering and uses multiple off set cost maps and potentials in cellular automata. the process determines cable coordinates at each cell location and store the coordinates of a predecessor cell to eliminate multiple paths. dated 2003-09-30
6629085,method of training a neural network for the guidance of a missile to a target,"a method of training a neural network for the guidance of a missile to a target includes the steps of: computing a solution of the non-linear guidance problem in analytical form, generating numerical solutions for a number of flights of a virtual missile to a virtual target, determining a human pilot's behavior and of the missile by simulating a number of flights of the virtual missile to a virtual target, and &#8220;cloning&#8221; a neural or fuzzy-neural network with the knowledge about the guidance of the missile to the target as obtained by the preceding steps. for determining the behavior of the human pilot, a scenario of missile and target is represented. this scenario is transformed into slow-motion. the flight of the missile to the target is simulated in slow-motion, the human pilot guiding the missile to the target. the pilot's behavior and the behavior of the missile resulting therefrom is stored for a number of such simulated flights. the thus stored data are re-transformed into real time. a guidance unit having a neural or fuzzy-neural network is trained with the re-transformed behavior of pilot and missile.",2003-09-30,"The title of the patent is method of training a neural network for the guidance of a missile to a target and its abstract is a method of training a neural network for the guidance of a missile to a target includes the steps of: computing a solution of the non-linear guidance problem in analytical form, generating numerical solutions for a number of flights of a virtual missile to a virtual target, determining a human pilot's behavior and of the missile by simulating a number of flights of the virtual missile to a virtual target, and &#8220;cloning&#8221; a neural or fuzzy-neural network with the knowledge about the guidance of the missile to the target as obtained by the preceding steps. for determining the behavior of the human pilot, a scenario of missile and target is represented. this scenario is transformed into slow-motion. the flight of the missile to the target is simulated in slow-motion, the human pilot guiding the missile to the target. the pilot's behavior and the behavior of the missile resulting therefrom is stored for a number of such simulated flights. the thus stored data are re-transformed into real time. a guidance unit having a neural or fuzzy-neural network is trained with the re-transformed behavior of pilot and missile. dated 2003-09-30"
6629086,method for interpreting petroleum characteristics of geological sediments,"the invention relates to a method for automatic interpretation of geochemical measurements obtained by pyrolysis of a rock sample in order to obtain information pertaining to the organic matter contained in the sample. in this method, rock samples having known petroleum characteristics are used to carry out a phase of training of an artificial neural network, the neural network is used to obtain parameters pertaining to the organic matter of a rock sample, interpretation of the parameters is refined at the network output by using fuzzy sets for refining interpretation of the parameters at an output of the neural network.",2003-09-30,"The title of the patent is method for interpreting petroleum characteristics of geological sediments and its abstract is the invention relates to a method for automatic interpretation of geochemical measurements obtained by pyrolysis of a rock sample in order to obtain information pertaining to the organic matter contained in the sample. in this method, rock samples having known petroleum characteristics are used to carry out a phase of training of an artificial neural network, the neural network is used to obtain parameters pertaining to the organic matter of a rock sample, interpretation of the parameters is refined at the network output by using fuzzy sets for refining interpretation of the parameters at an output of the neural network. dated 2003-09-30"
6629089,artificial neural network voice coil motor controller,"a robust artificial neural network controller is proposed for the motion control of a magnetic disk drive voice coil motor (voice coil motor). the neural controller is used to approximate the nonlinear functions (actuator electromechanical dynamics) of the voice coil motor while having on line training. one main advantage of this approach, when compared with standard adaptive control, is that complex dynamical analysis is not needed. using this design, not only strong robustness with respect to uncertain dynamics and non-linearities can be obtained, but also the output tracking error between the plant output and the desired reference can asymptotically converge to zero. additionally, standard offline training, utilizing training vectors to stimulate the voice coil motor, is not required.",2003-09-30,"The title of the patent is artificial neural network voice coil motor controller and its abstract is a robust artificial neural network controller is proposed for the motion control of a magnetic disk drive voice coil motor (voice coil motor). the neural controller is used to approximate the nonlinear functions (actuator electromechanical dynamics) of the voice coil motor while having on line training. one main advantage of this approach, when compared with standard adaptive control, is that complex dynamical analysis is not needed. using this design, not only strong robustness with respect to uncertain dynamics and non-linearities can be obtained, but also the output tracking error between the plant output and the desired reference can asymptotically converge to zero. additionally, standard offline training, utilizing training vectors to stimulate the voice coil motor, is not required. dated 2003-09-30"
6633855,"method, system, and program for filtering content using neural networks","disclosed is a method, system, and program for filtering a data object for content deemed unacceptable by a user. a data object requested by a viewer program is received. the data object is processed to determine predefined language statements. information on the determined language statements is inputted into a neural network to produce an output value. a determination is then made as to whether the output value indicates that the data object is unacceptable. viewer program access to the data object is inhibited upon determining that the data object is unacceptable.",2003-10-14,"The title of the patent is method, system, and program for filtering content using neural networks and its abstract is disclosed is a method, system, and program for filtering a data object for content deemed unacceptable by a user. a data object requested by a viewer program is received. the data object is processed to determine predefined language statements. information on the determined language statements is inputted into a neural network to produce an output value. a determination is then made as to whether the output value indicates that the data object is unacceptable. viewer program access to the data object is inhibited upon determining that the data object is unacceptable. dated 2003-10-14"
6636814,light rail vehicle having predictive diagnostic system for motor driven automated doors,"disclosed is a light rail vehicle having a predictive diagnostic system for a motor driven automated door (100) to enable condition-based maintenance. the light rail vehicle (110) has an automated door system (112), at least one data acquisition board (114), a data collection program (116), an exponential smoothing algorithm (118), and a neural network (120). the need for maintenance is identified through the collection of various door system (112) parameters, calculating current energy and time consumption from these parameters, and determining the rate of degradation based on current energy and time consumption of the door system (112) as compared with historical energy and time consumption. from the rate of degradation, maintenance can be scheduled as needed.",2003-10-21,"The title of the patent is light rail vehicle having predictive diagnostic system for motor driven automated doors and its abstract is disclosed is a light rail vehicle having a predictive diagnostic system for a motor driven automated door (100) to enable condition-based maintenance. the light rail vehicle (110) has an automated door system (112), at least one data acquisition board (114), a data collection program (116), an exponential smoothing algorithm (118), and a neural network (120). the need for maintenance is identified through the collection of various door system (112) parameters, calculating current energy and time consumption from these parameters, and determining the rate of degradation based on current energy and time consumption of the door system (112) as compared with historical energy and time consumption. from the rate of degradation, maintenance can be scheduled as needed. dated 2003-10-21"
6636841,system and method for telecommunications system fault diagnostics,"a telecommunications fault location and diagnostic system employs a remote test unit (rtu) to collect system parameter data. the rtu is operatively coupled to a trained neural network, which receives the system parameter data from the rtu. the neural network is trained using pre-screened historical fault data, which is stored in a database. once trained, the neural network classifies the rtu data into one of a predetermined number of fault probabilities.",2003-10-21,"The title of the patent is system and method for telecommunications system fault diagnostics and its abstract is a telecommunications fault location and diagnostic system employs a remote test unit (rtu) to collect system parameter data. the rtu is operatively coupled to a trained neural network, which receives the system parameter data from the rtu. the neural network is trained using pre-screened historical fault data, which is stored in a database. once trained, the neural network classifies the rtu data into one of a predetermined number of fault probabilities. dated 2003-10-21"
6639900,use of generic classifiers to determine physical topology in heterogeneous networking environments,"round trip time, bottleneck link speed, and hop count information from one node to the remaining nodes within a network is collected and processed by an adaptive resonance theory (art) neural network to classify the nodes by physical location or site group. for each site group, round trip time from one node to the remaining nodes is then collected and processed utilizing an art neural network to classify the nodes into one or more physical groups. the resulting breakdown of site groups within the network and physical groups within the site groups forms a model which may be employed by networking and system management applications. no private or proprietary vendor specific information from communications devices within the network need be employed to develop the model, only publicly available information regarding communications parameters.",2003-10-28,"The title of the patent is use of generic classifiers to determine physical topology in heterogeneous networking environments and its abstract is round trip time, bottleneck link speed, and hop count information from one node to the remaining nodes within a network is collected and processed by an adaptive resonance theory (art) neural network to classify the nodes by physical location or site group. for each site group, round trip time from one node to the remaining nodes is then collected and processed utilizing an art neural network to classify the nodes into one or more physical groups. the resulting breakdown of site groups within the network and physical groups within the site groups forms a model which may be employed by networking and system management applications. no private or proprietary vendor specific information from communications devices within the network need be employed to develop the model, only publicly available information regarding communications parameters. dated 2003-10-28"
6641746,control of semiconductor processing,"an integrated metrology and lithography/etch system and method (10) for micro-electronics device manufacturing. a process control neural network (30) is used to develop an estimated process control parameter (32) for controlling an etching process (28). the process control neural network is responsive to a multi-parameter characterization of a patterned resist feature mpc(pr) (16) developed on a substrate. the process control parameter is used as a feed-forward control for the etching process to develop an actual final mask feature. a multi-parameter characterization of the actual final mask feature mpc(hm) (36) is used as an input to a training neural network (40) for mapping to an ideal process control parameter. the ideal process control parameter is compared to the estimated control parameter to develop an error parameter (46), which is then used to train the process control neural network.",2003-11-04,"The title of the patent is control of semiconductor processing and its abstract is an integrated metrology and lithography/etch system and method (10) for micro-electronics device manufacturing. a process control neural network (30) is used to develop an estimated process control parameter (32) for controlling an etching process (28). the process control neural network is responsive to a multi-parameter characterization of a patterned resist feature mpc(pr) (16) developed on a substrate. the process control parameter is used as a feed-forward control for the etching process to develop an actual final mask feature. a multi-parameter characterization of the actual final mask feature mpc(hm) (36) is used as an input to a training neural network (40) for mapping to an ideal process control parameter. the ideal process control parameter is compared to the estimated control parameter to develop an error parameter (46), which is then used to train the process control neural network. dated 2003-11-04"
6643628,cellular automata neural network method for process modeling of film-substrate interactions and other dynamic processes,a cellular automata neural network method for process modeling of film-substrate interactions utilizes a cellular automaton system having variable rules for each cell. the variable rules describe a state change algorithm for atoms or other objects near a substrate. the state change algorithm is used to create a training set of solutions for training a neural network. the cellular automaton system is run to model the film-substrate interactions with the neural network providing the state change solutions in place of the more computationally complex state change algorithm to achieve real-time or near real-time simulations.,2003-11-04,The title of the patent is cellular automata neural network method for process modeling of film-substrate interactions and other dynamic processes and its abstract is a cellular automata neural network method for process modeling of film-substrate interactions utilizes a cellular automaton system having variable rules for each cell. the variable rules describe a state change algorithm for atoms or other objects near a substrate. the state change algorithm is used to create a training set of solutions for training a neural network. the cellular automaton system is run to model the film-substrate interactions with the neural network providing the state change solutions in place of the more computationally complex state change algorithm to achieve real-time or near real-time simulations. dated 2003-11-04
6646996,use of adaptive resonance theory to differentiate network device types (routers vs switches),"to determine a network communications device type, (switch or router) without reference to internal information within the network communications device, two packets having preselected, differing sizes (e.g., 64 bytes and 1500 bytes) are sequentially transmitted from one network node to another through the network communications device. the difference between the transmission start times for the two packets, determined by time references set up based on internal data processing system high resolution counters and placed in the ip packet payload, and the difference between the receipt stop times&#8212;that is, when the last portions of the two packets are received&#8212;are compared. if the two differences are substantially the same, the network communications device is classified as a switch. if the two differences are unequal by an appreciable amount, the network communications device is classified as a router. classification may optionally be performed by a neural network trained with the expect relative relationship between the transmission start times difference and the receipt stop times difference for the preselected packet sizes when transmitted through switches and routers.",2003-11-11,"The title of the patent is use of adaptive resonance theory to differentiate network device types (routers vs switches) and its abstract is to determine a network communications device type, (switch or router) without reference to internal information within the network communications device, two packets having preselected, differing sizes (e.g., 64 bytes and 1500 bytes) are sequentially transmitted from one network node to another through the network communications device. the difference between the transmission start times for the two packets, determined by time references set up based on internal data processing system high resolution counters and placed in the ip packet payload, and the difference between the receipt stop times&#8212;that is, when the last portions of the two packets are received&#8212;are compared. if the two differences are substantially the same, the network communications device is classified as a switch. if the two differences are unequal by an appreciable amount, the network communications device is classified as a router. classification may optionally be performed by a neural network trained with the expect relative relationship between the transmission start times difference and the receipt stop times difference for the preselected packet sizes when transmitted through switches and routers. dated 2003-11-11"
6647368,"sensor pair for detecting changes within a human ear and producing a signal corresponding to thought, movement, biological function and/or speech","a pair of sensors are used for detecting an air pressure change signal within an ear of a person caused by the person's initiating action (thought, movement, biological function and/or speech). one of the microphones is placed at least partially within an ear of the person and the other is placed adjacent to and external to the ear, to produce two electrical signals, respectively corresponding to internally detected and to externally detected changes in air pressure. comparison of the unmodified signal strength difference between these two signals is used to distinguish an initiating action component of each signal from an external source component of each signal. the electrical signals are processed to produce an output signal corresponding to the initiating action, which signal is then recognized by a neural network or speech recognizer, and used for control or communication.",2003-11-11,"The title of the patent is sensor pair for detecting changes within a human ear and producing a signal corresponding to thought, movement, biological function and/or speech and its abstract is a pair of sensors are used for detecting an air pressure change signal within an ear of a person caused by the person's initiating action (thought, movement, biological function and/or speech). one of the microphones is placed at least partially within an ear of the person and the other is placed adjacent to and external to the ear, to produce two electrical signals, respectively corresponding to internally detected and to externally detected changes in air pressure. comparison of the unmodified signal strength difference between these two signals is used to distinguish an initiating action component of each signal from an external source component of each signal. the electrical signals are processed to produce an output signal corresponding to the initiating action, which signal is then recognized by a neural network or speech recognizer, and used for control or communication. dated 2003-11-11"
6647377,"multi-kernel neural network concurrent learning, monitoring, and forecasting system","a multi-kernel neural network computing architecture configured to learn correlations among feature values 34, 38 as the network monitors and imputes measured input values 30 and also predicts future output values 46. this computing architecture includes a multi-kernel neural network array 14 with the capability to learn and predict in real time. the cip 10 also includes a manager 16 and an input-output transducer 12 that may be used for input-output refinement. these components allow the computing capacity of the multi-kernel array 14 to be reassigned in response to measured performance or other factors. the output feature values 46 computed by the multi-kernel array 14 and processed by an output processor 44 of the transducer 12 are supplied to a response unit 18 that may be configured to perform a variety of monitoring, forecasting, and control operations in response to the computed output values.",2003-11-11,"The title of the patent is multi-kernel neural network concurrent learning, monitoring, and forecasting system and its abstract is a multi-kernel neural network computing architecture configured to learn correlations among feature values 34, 38 as the network monitors and imputes measured input values 30 and also predicts future output values 46. this computing architecture includes a multi-kernel neural network array 14 with the capability to learn and predict in real time. the cip 10 also includes a manager 16 and an input-output transducer 12 that may be used for input-output refinement. these components allow the computing capacity of the multi-kernel array 14 to be reassigned in response to measured performance or other factors. the output feature values 46 computed by the multi-kernel array 14 and processed by an output processor 44 of the transducer 12 are supplied to a response unit 18 that may be configured to perform a variety of monitoring, forecasting, and control operations in response to the computed output values. dated 2003-11-11"
6647379,method and apparatus for interpreting information,"a computer product for analyzing data representative of account usage, the computer product disposed on a computer readable medium and including instructions for causing a processor to provide a data classifier the data representative of account usage, wherein the data classifier produces in response to the data representative of account usage, classification outputs, provide to a rule inducer input data based on the data representative of account usage and the classification outputs, wherein the rule inducer produces rules descriptive of relationships between the data representative of account usage and the classification outputs. the data classifier can include a neural network and a bayesian classifier. the input data based on the data representative of account usage can include a combination of the data representative of account usage and the classification outputs.",2003-11-11,"The title of the patent is method and apparatus for interpreting information and its abstract is a computer product for analyzing data representative of account usage, the computer product disposed on a computer readable medium and including instructions for causing a processor to provide a data classifier the data representative of account usage, wherein the data classifier produces in response to the data representative of account usage, classification outputs, provide to a rule inducer input data based on the data representative of account usage and the classification outputs, wherein the rule inducer produces rules descriptive of relationships between the data representative of account usage and the classification outputs. the data classifier can include a neural network and a bayesian classifier. the input data based on the data representative of account usage can include a combination of the data representative of account usage and the classification outputs. dated 2003-11-11"
6650766,method for combining automated detections from medical images with observed detections of a human interpreter,"a method and system for detecting and displaying clustered microcalcifications in a digital mammogram, wherein a single digital mammogram is first automatically cropped to a breast area sub-image which is then processed by means of an optimized difference of gaussians filter to enhance the appearance of potential microcalcifications in the sub-image. the potential microcalcifications are thresholded, clusters are detected, features are computed for the detected clusters, and the clusters are classified as either suspicious or not suspicious by means of a neural network. thresholding is preferably by sloping local thresholding, but may also be performed by global and dual-local thresholding. the locations in the original digital mammogram of the suspicious detected clustered microcalcifications are indicated. parameters for use in the detection and thresholding portions of the system are computer-optimized by means of a genetic algorithm. the results of the system are optimally combined with a radiologist's observation of the original mammogram by combining the observations with the results, after the radiologist has first accepted or rejected individual detections reported by the system.",2003-11-18,"The title of the patent is method for combining automated detections from medical images with observed detections of a human interpreter and its abstract is a method and system for detecting and displaying clustered microcalcifications in a digital mammogram, wherein a single digital mammogram is first automatically cropped to a breast area sub-image which is then processed by means of an optimized difference of gaussians filter to enhance the appearance of potential microcalcifications in the sub-image. the potential microcalcifications are thresholded, clusters are detected, features are computed for the detected clusters, and the clusters are classified as either suspicious or not suspicious by means of a neural network. thresholding is preferably by sloping local thresholding, but may also be performed by global and dual-local thresholding. the locations in the original digital mammogram of the suspicious detected clustered microcalcifications are indicated. parameters for use in the detection and thresholding portions of the system are computer-optimized by means of a genetic algorithm. the results of the system are optimally combined with a radiologist's observation of the original mammogram by combining the observations with the results, after the radiologist has first accepted or rejected individual detections reported by the system. dated 2003-11-18"
6650779,method and apparatus for analyzing an image to detect and identify patterns,"a method and apparatus is provided which analyzes an image of an object to detect and identify defects in the object utilizing multi-dimensional wavelet neural networks. &#8220;the present invention generates a signal representing part of the object, then extracts certain features of the signal. these features are then provided to a multidimensional neural network for classification, which indicates if the features correlate with a predetermined pattern. this process of analyzing the features to detect and identify predetermined patterns results in a robust fault detection and identification system which is computationally efficient and economical because of the learning element contained therein which lessens the need for human assistance.&#8221;",2003-11-18,"The title of the patent is method and apparatus for analyzing an image to detect and identify patterns and its abstract is a method and apparatus is provided which analyzes an image of an object to detect and identify defects in the object utilizing multi-dimensional wavelet neural networks. &#8220;the present invention generates a signal representing part of the object, then extracts certain features of the signal. these features are then provided to a multidimensional neural network for classification, which indicates if the features correlate with a predetermined pattern. this process of analyzing the features to detect and identify predetermined patterns results in a robust fault detection and identification system which is computationally efficient and economical because of the learning element contained therein which lessens the need for human assistance.&#8221; dated 2003-11-18"
6654730,neural network arithmetic apparatus and neutral network operation method,"when neuron operations are computed in parallel using a large number of arithmetic units, arithmetic units for neuron operations and arithmetic units for error signal operations need not be provided separately, and a neural network arithmetic apparatus that consumes the bus band less is provided for updating of synapse connection weights. operation results of arithmetic units and setting information of a master node are exchanged between them through a local bus. during neuron operations, partial sums of neuron output values from the arithmetic units are accumulated by the master node to generate and output a neuron output value, and an arithmetic unit to which neuron operations of the specific neuron are assigned receives and stores the neuron output value outputted from the master node.",2003-11-25,"The title of the patent is neural network arithmetic apparatus and neutral network operation method and its abstract is when neuron operations are computed in parallel using a large number of arithmetic units, arithmetic units for neuron operations and arithmetic units for error signal operations need not be provided separately, and a neural network arithmetic apparatus that consumes the bus band less is provided for updating of synapse connection weights. operation results of arithmetic units and setting information of a master node are exchanged between them through a local bus. during neuron operations, partial sums of neuron output values from the arithmetic units are accumulated by the master node to generate and output a neuron output value, and an arithmetic unit to which neuron operations of the specific neuron are assigned receives and stores the neuron output value outputted from the master node. dated 2003-11-25"
6658396,neural network drug dosage estimation,"neural networks are constructed (programmed), trained on historical data, and used to predict any of (1) optimal patient dosage of a single drug, (2) optimal patient dosage of one drug in respect of the patient's concurrent usage of another drug, (3a) optimal patient drug dosage in respect of diverse patient characteristics, (3b) sensitivity of recommended patient drug dosage to the patient characteristics, (4a) expected outcome versus patient drug dosage, (4b) sensitivity of the expected outcome to variant drug dosage(s), (5) expected outcome(s) from drug dosage(s) other than the projected optimal dosage. both human and economic costs of both optimal and sub-optimal drug therapies may be extrapolated from the exercise of various optimized and trained neural networks. heretofore little recognized sensitivities&#8212;such as, for example, patient race in the administration of psychotropic drugs&#8212;are made manifest. individual prescribing physicians employing deviant patterns of drug therapy may be recognized. although not intended to prescribe drugs, nor even to set prescription drug dosage, the neural networks are very sophisticated and authoritative &#8220;helps&#8221; to physicians, and to physician reviewers, in answering &#8220;what if&#8221; questions.",2003-12-02,"The title of the patent is neural network drug dosage estimation and its abstract is neural networks are constructed (programmed), trained on historical data, and used to predict any of (1) optimal patient dosage of a single drug, (2) optimal patient dosage of one drug in respect of the patient's concurrent usage of another drug, (3a) optimal patient drug dosage in respect of diverse patient characteristics, (3b) sensitivity of recommended patient drug dosage to the patient characteristics, (4a) expected outcome versus patient drug dosage, (4b) sensitivity of the expected outcome to variant drug dosage(s), (5) expected outcome(s) from drug dosage(s) other than the projected optimal dosage. both human and economic costs of both optimal and sub-optimal drug therapies may be extrapolated from the exercise of various optimized and trained neural networks. heretofore little recognized sensitivities&#8212;such as, for example, patient race in the administration of psychotropic drugs&#8212;are made manifest. individual prescribing physicians employing deviant patterns of drug therapy may be recognized. although not intended to prescribe drugs, nor even to set prescription drug dosage, the neural networks are very sophisticated and authoritative &#8220;helps&#8221; to physicians, and to physician reviewers, in answering &#8220;what if&#8221; questions. dated 2003-12-02"
6658980,combat pilot aid system,"a combat pilot system for an aircraft comprises corresponding combat pilot devices 12, 12&#8242; provided in a runtime module 14 on board the aircraft and a ground-based system training module 16 respectively. the run-time module 12&#8242; is trained using model data generated simulation model 18 as well as feedback data from actual missile firings. the matrix of weights derived from the training routines are then programmed into the neural network in the corresponding combat pilot aid system 12 on board the aircraft so that the runtime module is capable of processing the input data producing the four parameters required for launch of a missile and also a fire/no fire indication to the pilot.",2003-12-09,"The title of the patent is combat pilot aid system and its abstract is a combat pilot system for an aircraft comprises corresponding combat pilot devices 12, 12&#8242; provided in a runtime module 14 on board the aircraft and a ground-based system training module 16 respectively. the run-time module 12&#8242; is trained using model data generated simulation model 18 as well as feedback data from actual missile firings. the matrix of weights derived from the training routines are then programmed into the neural network in the corresponding combat pilot aid system 12 on board the aircraft so that the runtime module is capable of processing the input data producing the four parameters required for launch of a missile and also a fire/no fire indication to the pilot. dated 2003-12-09"
6661908,signature recognition system and method,"a method of authenticating a signature including the steps of sampling a signature and storing data representative of the signature, converting the data to high dimensions vectors, feeding the high dimension vectors to an unsupervised neural network, performing a high order principal component extraction process on the high dimensions vectors to thereby identifying clusters of high dimension points, and analyzing the clusters of high dimension points to determine, based on previously stored information, the authenticity of the signature. also an apparatus for such authentication including a sampling device for sampling a signature and storing data representative of the signature, a converting device connected downsteam of the sampling device for converting the data to high dimension vectors, an unsupervised neural network for receiving the high dimension and performing a high order principal component extraction process on the high dimensions vectors to thereby identify clusters of high dimension points, and an analyzing device connected to the unsupervised neural network for analyzing the clusters of high dimension points to determine the authenticity of the signature.",2003-12-09,"The title of the patent is signature recognition system and method and its abstract is a method of authenticating a signature including the steps of sampling a signature and storing data representative of the signature, converting the data to high dimensions vectors, feeding the high dimension vectors to an unsupervised neural network, performing a high order principal component extraction process on the high dimensions vectors to thereby identifying clusters of high dimension points, and analyzing the clusters of high dimension points to determine, based on previously stored information, the authenticity of the signature. also an apparatus for such authentication including a sampling device for sampling a signature and storing data representative of the signature, a converting device connected downsteam of the sampling device for converting the data to high dimension vectors, an unsupervised neural network for receiving the high dimension and performing a high order principal component extraction process on the high dimensions vectors to thereby identify clusters of high dimension points, and an analyzing device connected to the unsupervised neural network for analyzing the clusters of high dimension points to determine the authenticity of the signature. dated 2003-12-09"
6662059,characteristic adjusting method in process of manufacturing products,"in a process control data base, intermediate characteristics, processing condition for controlling a characteristic and a final characteristic in a process of manufacturing products are stored as a set of data for each product lot. in process 1, a set of data for each product lot are prepared. next, in process 2, cluster processing is conducted on each set of data obtained in process 1. in process 3, using the set of data obtained in process 2, the intermediate characteristics and the processing condition for controlling a characteristic are inputted and the final characteristic is outputted, and a causal relation between the input and output is quantified by a neural network so as to make a learning model. in a model applying stage, the most appropriate processing condition for controlling a characteristic is retrieved by using the learning model and the intermediate characteristics in the previous steps.",2003-12-09,"The title of the patent is characteristic adjusting method in process of manufacturing products and its abstract is in a process control data base, intermediate characteristics, processing condition for controlling a characteristic and a final characteristic in a process of manufacturing products are stored as a set of data for each product lot. in process 1, a set of data for each product lot are prepared. next, in process 2, cluster processing is conducted on each set of data obtained in process 1. in process 3, using the set of data obtained in process 2, the intermediate characteristics and the processing condition for controlling a characteristic are inputted and the final characteristic is outputted, and a causal relation between the input and output is quantified by a neural network so as to make a learning model. in a model applying stage, the most appropriate processing condition for controlling a characteristic is retrieved by using the learning model and the intermediate characteristics in the previous steps. dated 2003-12-09"
6662112,method for classifying avo data using an interpreter-trained neural network,"avo anomalies are classified in near-offset and far-offset seismic data volumes, by first calculating a plurality of initial avo seismic attributes representative of the offset seismic data volumes. a probabilistic neural network is constructed from the calculated initial avo seismic attributes. avo anomaly classifications are calculated in a portion of the offset seismic data volumes. the preceding steps are repeated until the calculated avo anomaly classifications in the portion of the offset seismic data volumes are satisfactory. avo anomaly classifications are calculated throughout the offset seismic data volumes using the constructed probabilistic neural network.",2003-12-09,"The title of the patent is method for classifying avo data using an interpreter-trained neural network and its abstract is avo anomalies are classified in near-offset and far-offset seismic data volumes, by first calculating a plurality of initial avo seismic attributes representative of the offset seismic data volumes. a probabilistic neural network is constructed from the calculated initial avo seismic attributes. avo anomaly classifications are calculated in a portion of the offset seismic data volumes. the preceding steps are repeated until the calculated avo anomaly classifications in the portion of the offset seismic data volumes are satisfactory. avo anomaly classifications are calculated throughout the offset seismic data volumes using the constructed probabilistic neural network. dated 2003-12-09"
6665651,control system and technique employing reinforcement learning having stability and learning phases,"a feedback control system for automatic on-line training of a controller for a plant, the system having a reinforcement learning agent connected in parallel with the controller. the learning agent comprises an actor network and a critic network operatively arranged to carry out at least one sequence of a stability phase followed by a learning phase. during the stability phase, a multi-dimensional boundary of values is determined. during the learning phase, a plurality of updated weight values is generated in connection with the on-line training, if and until one of the updated weight values reaches the boundary, at which time a next sequence is carried out to determine a next multi-dimensional boundary of values followed by a next learning phase. also, a method for automatic on-line training of a feedback controller within a system comprising the controller and a plant by employing a reinforcement learning agent comprising a neural network to carry out at least one sequence comprising a stability phase followed by a learning phase. further included, a computer executable program code on a computer readable storage medium, for on-line training of a feedback controller within a system comprising the controller and a plant.",2003-12-16,"The title of the patent is control system and technique employing reinforcement learning having stability and learning phases and its abstract is a feedback control system for automatic on-line training of a controller for a plant, the system having a reinforcement learning agent connected in parallel with the controller. the learning agent comprises an actor network and a critic network operatively arranged to carry out at least one sequence of a stability phase followed by a learning phase. during the stability phase, a multi-dimensional boundary of values is determined. during the learning phase, a plurality of updated weight values is generated in connection with the on-line training, if and until one of the updated weight values reaches the boundary, at which time a next sequence is carried out to determine a next multi-dimensional boundary of values followed by a next learning phase. also, a method for automatic on-line training of a feedback controller within a system comprising the controller and a plant by employing a reinforcement learning agent comprising a neural network to carry out at least one sequence comprising a stability phase followed by a learning phase. further included, a computer executable program code on a computer readable storage medium, for on-line training of a feedback controller within a system comprising the controller and a plant. dated 2003-12-16"
6671391,pose-adaptive face detection system and process,"a face detection system and process capable of detecting a person depicted in an input image and identifying their face pose. prepared training images are used to train a 2-stage classifier which includes a bank of support vector machines (svms) as an initial pre-classifier layer, and a neural network forming a subsequent decision classifier layer. once the svms and the neural network are trained, input image regions are prepared and input into the system. an output is produced from the neural network which indicates first whether the region under consideration depicts a face, and secondly, if a face is present, into what pose range the pose of the face falls.",2003-12-30,"The title of the patent is pose-adaptive face detection system and process and its abstract is a face detection system and process capable of detecting a person depicted in an input image and identifying their face pose. prepared training images are used to train a 2-stage classifier which includes a bank of support vector machines (svms) as an initial pre-classifier layer, and a neural network forming a subsequent decision classifier layer. once the svms and the neural network are trained, input image regions are prepared and input into the system. an output is produced from the neural network which indicates first whether the region under consideration depicts a face, and secondly, if a face is present, into what pose range the pose of the face falls. dated 2003-12-30"
6671400,panoramic image navigation system using neural network for correction of image distortion,"a system permitting one or more users to navigate a wide-angle image, and in particular a panoramic image, using a neural network to correct distortion in that image. to train the neural network, a calibration pattern which defines an array of calibration points is disposed to occupy a field of view of a wide-angle imaging apparatus, and a calibration image is thereby captured by the apparatus. respective view directions of the calibration points and the positions of these points in the calibration image are used as data for training the neural network to correctly match view directions to corresponding intra-image positions. subsequently, to generate an undistorted sub-image of an arbitrary distorted wide-angle image, an array of display pixel positions are expressed as a corresponding set of view directions, and the neural network used to convert these to a corresponding set of positions within the wide-angle image, with the video attributes at these positions being then assigned to the corresponding pixels of the sub-image.",2003-12-30,"The title of the patent is panoramic image navigation system using neural network for correction of image distortion and its abstract is a system permitting one or more users to navigate a wide-angle image, and in particular a panoramic image, using a neural network to correct distortion in that image. to train the neural network, a calibration pattern which defines an array of calibration points is disposed to occupy a field of view of a wide-angle imaging apparatus, and a calibration image is thereby captured by the apparatus. respective view directions of the calibration points and the positions of these points in the calibration image are used as data for training the neural network to correctly match view directions to corresponding intra-image positions. subsequently, to generate an undistorted sub-image of an arbitrary distorted wide-angle image, an array of display pixel positions are expressed as a corresponding set of view directions, and the neural network used to convert these to a corresponding set of positions within the wide-angle image, with the video attributes at these positions being then assigned to the corresponding pixels of the sub-image. dated 2003-12-30"
6674855,high performance multifrequency signal detection,"a multifrequency detector, such as a dtmf detector, includes a goertzel module detecting energies of a current frame of an incoming signal at nominal tone frequencies. a number of ratios are computed from the detected energies and supplied as an input to a multidimensional thresholding device such as a neural network. the neural network generates a decision level to be used in a subsequent frame decision. a frequency estimation module also receives the incoming signal and generates a number of frequency estimation indications to be used in the final determination of whether the incoming signal contains a valid multifrequency signal. a low-level decision module returns a frame decision result (d/n) for the current frame based on inputs from the goertzel module and the decision level. a high-level decision module receives the frame decision result and information based on the frequency estimation indications. a state machine generates a digit start signal when the frame decision result and the frequency estimation information indicate a transition from an out-of-digit state to an in-digit state. the detector outputs a suggested digit provided by the goertzel module and an affirmative digit detection result when the state machine transitions from out-of-digit to in-digit.",2004-01-06,"The title of the patent is high performance multifrequency signal detection and its abstract is a multifrequency detector, such as a dtmf detector, includes a goertzel module detecting energies of a current frame of an incoming signal at nominal tone frequencies. a number of ratios are computed from the detected energies and supplied as an input to a multidimensional thresholding device such as a neural network. the neural network generates a decision level to be used in a subsequent frame decision. a frequency estimation module also receives the incoming signal and generates a number of frequency estimation indications to be used in the final determination of whether the incoming signal contains a valid multifrequency signal. a low-level decision module returns a frame decision result (d/n) for the current frame based on inputs from the goertzel module and the decision level. a high-level decision module receives the frame decision result and information based on the frequency estimation indications. a state machine generates a digit start signal when the frame decision result and the frequency estimation information indicate a transition from an out-of-digit state to an in-digit state. the detector outputs a suggested digit provided by the goertzel module and an affirmative digit detection result when the state machine transitions from out-of-digit to in-digit. dated 2004-01-06"
6674867,neurofuzzy based device for programmable hearing aids,"a neurofuzzy device is described that provides a fuzzy logic based user-machine interface for optimal fitting of programmable hearing prosthesis using a neural network that generates targets to be matched by the hearing prosthesis based on individual audiometric and other relevant data to the specific impairment and on the neural network accumulated learning from previous successful fittings. the incorporated learning process can occur on or off line and implements fitting rationales that can satisfy the needs of a general or specific clientele. the parameters of the programmable prosthetic device are set as a group in order to achieve optimal matching to the targets. the user-machine interface realized by a fuzzy logic system deciphers the commends/responses of the user while listening to various stimuli and modifies the targets accordingly thus, providing a closed loop system for in-situ interactive fitting.",2004-01-06,"The title of the patent is neurofuzzy based device for programmable hearing aids and its abstract is a neurofuzzy device is described that provides a fuzzy logic based user-machine interface for optimal fitting of programmable hearing prosthesis using a neural network that generates targets to be matched by the hearing prosthesis based on individual audiometric and other relevant data to the specific impairment and on the neural network accumulated learning from previous successful fittings. the incorporated learning process can occur on or off line and implements fitting rationales that can satisfy the needs of a general or specific clientele. the parameters of the programmable prosthetic device are set as a group in order to achieve optimal matching to the targets. the user-machine interface realized by a fuzzy logic system deciphers the commends/responses of the user while listening to various stimuli and modifies the targets accordingly thus, providing a closed loop system for in-situ interactive fitting. dated 2004-01-06"
6675134,performance assessment of data classifiers,"a method for assessing the performance of a data classifier operable to generate an element of output data in response to an element of input data, such as a neural network, is disclosed. the method includes using the data classifier to generate elements of result output data in response to elements of test input data, determining a measure of difference between each element of test output data and each corresponding element of result output data, forming a distribution function of the measures of differences, and forming a measure of performance from the distribution function.",2004-01-06,"The title of the patent is performance assessment of data classifiers and its abstract is a method for assessing the performance of a data classifier operable to generate an element of output data in response to an element of input data, such as a neural network, is disclosed. the method includes using the data classifier to generate elements of result output data in response to elements of test input data, determining a measure of difference between each element of test output data and each corresponding element of result output data, forming a distribution function of the measures of differences, and forming a measure of performance from the distribution function. dated 2004-01-06"
6675162,"method for scanning, analyzing and handling various kinds of digital information content","computer-implemented methods are described for, first, characterizing a specific category of information content&#8212;pornography, for example&#8212;and then accurately identifying instances of that category of content within a real-time media stream, such as a web page, e-mail or other digital dataset. this content-recognition technology enables a new class of highly scalable applications to manage such content, including filtering, classifying, prioritizing, tracking etc. an illustrative application of the invention is a software product for use in conjunction with web-browser client software for screening access to web pages that contain pornography or other potentially harmful or offensive content. a target attribute set of regular expression, such as natural language words and/or phrases, is formed by statistical analysis of a number of samples of datasets characterized as &#8220;containing,&#8221; and another set of samples characterized as &#8220;not containing,&#8221; the selected category of information content. this list of expressions is refined by applying correlation analysis to the samples or &#8220;training data.&#8221; neural-network feed-forward techniques are then applied, again using a substantial training dataset, for adaptively assigning relative weights to each of the expressions in the target attribute set, thereby forming an awaited list that is highly predictive of the information content category of interest.",2004-01-06,"The title of the patent is method for scanning, analyzing and handling various kinds of digital information content and its abstract is computer-implemented methods are described for, first, characterizing a specific category of information content&#8212;pornography, for example&#8212;and then accurately identifying instances of that category of content within a real-time media stream, such as a web page, e-mail or other digital dataset. this content-recognition technology enables a new class of highly scalable applications to manage such content, including filtering, classifying, prioritizing, tracking etc. an illustrative application of the invention is a software product for use in conjunction with web-browser client software for screening access to web pages that contain pornography or other potentially harmful or offensive content. a target attribute set of regular expression, such as natural language words and/or phrases, is formed by statistical analysis of a number of samples of datasets characterized as &#8220;containing,&#8221; and another set of samples characterized as &#8220;not containing,&#8221; the selected category of information content. this list of expressions is refined by applying correlation analysis to the samples or &#8220;training data.&#8221; neural-network feed-forward techniques are then applied, again using a substantial training dataset, for adaptively assigning relative weights to each of the expressions in the target attribute set, thereby forming an awaited list that is highly predictive of the information content category of interest. dated 2004-01-06"
6678421,subband coefficient prediction with pattern recognition techniques,"a method and an apparatus for estimating the upper frequency band coefficients solely from low frequency information in a subband multiresolution decomposition. in operation, a bayesian classifier predicts the significance or insignificance of a high frequency signal, then a neural network estimates the sign and magnitude of the visually significant information and effectively eliminates insignificant information. finally an algorithm estimates upper frequencies based on lower frequencies within the same band. this estimation is performed recursively for each level of a multiresolution decomposition pyramid until a reconstructed version of the data product is returned.",2004-01-13,"The title of the patent is subband coefficient prediction with pattern recognition techniques and its abstract is a method and an apparatus for estimating the upper frequency band coefficients solely from low frequency information in a subband multiresolution decomposition. in operation, a bayesian classifier predicts the significance or insignificance of a high frequency signal, then a neural network estimates the sign and magnitude of the visually significant information and effectively eliminates insignificant information. finally an algorithm estimates upper frequencies based on lower frequencies within the same band. this estimation is performed recursively for each level of a multiresolution decomposition pyramid until a reconstructed version of the data product is returned. dated 2004-01-13"
6678618,neural network methods to predict enzyme inhibitor or receptor ligand potency,"a new method to analyze and predict the binding energy for enzyme-transition state inhibitor interactions is presented. computational neural networks are employed to discovery quantum mechanical features of transition states and putative inhibitors necessary for binding. the method is able to generate its own relationship between the quantum mechanical structure of the inhibitor and the strength of binding. feed-forward neural networks with back propagation of error can be trained to recognize the quantum mechanical electrostatic potential at the entire van der waals surface, rather than a collapsed representation, of a group of training inhibitors and to predict the strength of interactions between the enzyme and a group of novel inhibitors. the experimental results show that the neural networks can predict with quantitative accuracy the binding strength of new inhibitors. the method is in fact able to predict the large binding free energy of the transition state, when trained with less tightly bound inhibitors. the present method is also applicable to prediction of the binding free energy of a ligand to a receptor. the application of this approach to the study of transition state inhibitors and ligands would permit evaluation of chemical libraries of potential inhibitory, agonistic, or antagonistic agents. the method is amenable to incorporation in a computer-readable medium accessible by general-purpose computers.",2004-01-13,"The title of the patent is neural network methods to predict enzyme inhibitor or receptor ligand potency and its abstract is a new method to analyze and predict the binding energy for enzyme-transition state inhibitor interactions is presented. computational neural networks are employed to discovery quantum mechanical features of transition states and putative inhibitors necessary for binding. the method is able to generate its own relationship between the quantum mechanical structure of the inhibitor and the strength of binding. feed-forward neural networks with back propagation of error can be trained to recognize the quantum mechanical electrostatic potential at the entire van der waals surface, rather than a collapsed representation, of a group of training inhibitors and to predict the strength of interactions between the enzyme and a group of novel inhibitors. the experimental results show that the neural networks can predict with quantitative accuracy the binding strength of new inhibitors. the method is in fact able to predict the large binding free energy of the transition state, when trained with less tightly bound inhibitors. the present method is also applicable to prediction of the binding free energy of a ligand to a receptor. the application of this approach to the study of transition state inhibitors and ligands would permit evaluation of chemical libraries of potential inhibitory, agonistic, or antagonistic agents. the method is amenable to incorporation in a computer-readable medium accessible by general-purpose computers. dated 2004-01-13"
6678640,"method and apparatus for parameter estimation, parameter estimation control and learning control","a parameter estimation device for estimating parameters relating to input and output of a control object by using a neural network, including control domain division means for selecting at least one parameter having strong correlation with non-linearity amongst input-output characteristics of the control object, as a parameter of a premise part of fuzzy operation, and dividing a control domain of the control object into a plurality of small domains by fuzzy estimation based on the parameter of the premise part; and estimation means for estimating dynamic behavior of the control object by using, as a consequent part of the fuzzy operation, a neural network which receives a parameter indicating the operation state of each of the small domain into which the control domain is divided, so that the output from the neural network can be checked in advance.",2004-01-13,"The title of the patent is method and apparatus for parameter estimation, parameter estimation control and learning control and its abstract is a parameter estimation device for estimating parameters relating to input and output of a control object by using a neural network, including control domain division means for selecting at least one parameter having strong correlation with non-linearity amongst input-output characteristics of the control object, as a parameter of a premise part of fuzzy operation, and dividing a control domain of the control object into a plurality of small domains by fuzzy estimation based on the parameter of the premise part; and estimation means for estimating dynamic behavior of the control object by using, as a consequent part of the fuzzy operation, a neural network which receives a parameter indicating the operation state of each of the small domain into which the control domain is divided, so that the output from the neural network can be checked in advance. dated 2004-01-13"
6678669,method for selecting medical and biochemical diagnostic tests using neural network-related applications,"methods are provided for developing medical diagnostic tests using decision-support systems, such as neural networks. patient data or information, typically patient history or clinical data, are analyzed by the decision-support systems to identify important or relevant variables and decision-support systems are trained on the patient data. patient data are augmented by biochemical test data, or results, where available, to refine performance. the resulting decision-support systems are employed to evaluate specific observation values and test results, to guide the development of biochemical or other diagnostic tests, too assess a course of treatment, to identify new diagnostic tests and disease markers, to identify useful therapies, and to provide the decision-support functionality for the test. methods for identification of important input variables for a medical diagnostic tests for use in training the decision-support systems to guide the development of the tests, for improving the sensitivity and specificity of such tests, and for selecting diagnostic tests that improve overall diagnosis of, or potential for, a disease state and that permit the effectiveness of a selected therapeutic protocol to be assessed are provided. the methods for identification can be applied in any field in which statistics are used to determine outcomes. a method for evaluating the effectiveness of any given diagnostic test is also provided.",2004-01-13,"The title of the patent is method for selecting medical and biochemical diagnostic tests using neural network-related applications and its abstract is methods are provided for developing medical diagnostic tests using decision-support systems, such as neural networks. patient data or information, typically patient history or clinical data, are analyzed by the decision-support systems to identify important or relevant variables and decision-support systems are trained on the patient data. patient data are augmented by biochemical test data, or results, where available, to refine performance. the resulting decision-support systems are employed to evaluate specific observation values and test results, to guide the development of biochemical or other diagnostic tests, too assess a course of treatment, to identify new diagnostic tests and disease markers, to identify useful therapies, and to provide the decision-support functionality for the test. methods for identification of important input variables for a medical diagnostic tests for use in training the decision-support systems to guide the development of the tests, for improving the sensitivity and specificity of such tests, and for selecting diagnostic tests that improve overall diagnosis of, or potential for, a disease state and that permit the effectiveness of a selected therapeutic protocol to be assessed are provided. the methods for identification can be applied in any field in which statistics are used to determine outcomes. a method for evaluating the effectiveness of any given diagnostic test is also provided. dated 2004-01-13"
6678670,non-integer order dynamic systems,"a circuit implementing a non-integer order dynamic system includes a neural network that receives at least one input signal and generates therefrom at least one output signal. the input and output signals are related to each by a non-integer order integro-differential relationship through the coefficients of the neural network. a plurality of such circuits, implementing respective non-integer order controllers can be interconnected in an arrangement wherein any of the integral or differential blocks included in one of these circuits generates a signal which is fed to any of the integral or differential blocks of another circuit in the system.",2004-01-13,"The title of the patent is non-integer order dynamic systems and its abstract is a circuit implementing a non-integer order dynamic system includes a neural network that receives at least one input signal and generates therefrom at least one output signal. the input and output signals are related to each by a non-integer order integro-differential relationship through the coefficients of the neural network. a plurality of such circuits, implementing respective non-integer order controllers can be interconnected in an arrangement wherein any of the integral or differential blocks included in one of these circuits generates a signal which is fed to any of the integral or differential blocks of another circuit in the system. dated 2004-01-13"
6681161,voltage control on a train system,"the present invention provides methods for preventing low train voltages and managing interference, thereby improving the efficiency, reliability, and passenger comfort associated with commuter trains. an algorithm implementing neural network technology is used to predict low voltages before they occur. once voltages are predicted, then multiple trains can be controlled to prevent low voltage events. further, algorithms for managing inference are presented in the present invention. different types of interference problems are addressed in the present invention such as &#8220;interference during acceleration&#8221;, &#8220;interference near station stops&#8221;, and &#8220;interference during delay recovery.&#8221; managing such interference avoids unnecessary brake/acceleration cycles during acceleration, immediately before station stops, and after substantial delays. algorithms are demonstrated to avoid oscillatory brake/acceleration cycles due to interference and to smooth the trajectories of closely following trains. this is achieved by maintaining sufficient following distances to avoid unnecessary braking/accelerating. these methods generate smooth train trajectories, making for a more comfortable ride, and improve train motor reliability by avoiding unnecessary mode-changes between propulsion and braking. these algorithms can also have a favorable impact on traction power system requirements and energy consumption.",2004-01-20,"The title of the patent is voltage control on a train system and its abstract is the present invention provides methods for preventing low train voltages and managing interference, thereby improving the efficiency, reliability, and passenger comfort associated with commuter trains. an algorithm implementing neural network technology is used to predict low voltages before they occur. once voltages are predicted, then multiple trains can be controlled to prevent low voltage events. further, algorithms for managing inference are presented in the present invention. different types of interference problems are addressed in the present invention such as &#8220;interference during acceleration&#8221;, &#8220;interference near station stops&#8221;, and &#8220;interference during delay recovery.&#8221; managing such interference avoids unnecessary brake/acceleration cycles during acceleration, immediately before station stops, and after substantial delays. algorithms are demonstrated to avoid oscillatory brake/acceleration cycles due to interference and to smooth the trajectories of closely following trains. this is achieved by maintaining sufficient following distances to avoid unnecessary braking/accelerating. these methods generate smooth train trajectories, making for a more comfortable ride, and improve train motor reliability by avoiding unnecessary mode-changes between propulsion and braking. these algorithms can also have a favorable impact on traction power system requirements and energy consumption. dated 2004-01-20"
6682491,method for artifact reduction in intracranial pressure measurements,"a method for artifact reduction in sonomicrometer obtained intracranial pressure measurements generally comprises isolating a component of a sonomicrometer waveform attributable solely to changes in intracranial volume by using a neural network or other nonlinear engine to extract a heartbeat component from the sonomicrometer output. because the heartbeat is so characteristic, no actual measurement of the heartbeat as the forcing function is required to isolate the resulting changes in distance from the artifact induced changes in distance. the neural network then be utilized to directly map measured changes in skull distance over time to changes in intracranial pressure over a volume change, the inverse of compliance. the method is generally extendable to use with other volumetric based measurement techniques.",2004-01-27,"The title of the patent is method for artifact reduction in intracranial pressure measurements and its abstract is a method for artifact reduction in sonomicrometer obtained intracranial pressure measurements generally comprises isolating a component of a sonomicrometer waveform attributable solely to changes in intracranial volume by using a neural network or other nonlinear engine to extract a heartbeat component from the sonomicrometer output. because the heartbeat is so characteristic, no actual measurement of the heartbeat as the forcing function is required to isolate the resulting changes in distance from the artifact induced changes in distance. the neural network then be utilized to directly map measured changes in skull distance over time to changes in intracranial pressure over a volume change, the inverse of compliance. the method is generally extendable to use with other volumetric based measurement techniques. dated 2004-01-27"
6684194,subscriber identification system,"a subscriber identification system is presented in which subscriber selection data including channel changes, volume changes, and time-of-day viewing information is used to identify a subscriber from a group of subscribers. in one instance, the subscriber viewing data is recorded and a signal processing algorithm such as a fourier transform is used to produce a processed version of the subscriber selection data. the processed version of the subscriber selection data can be correlated with stored common identifiers of subscriber profiles to determine which subscriber from the group is presently viewing the programming. a neural network or fuzzy logic can be used as the mechanism for identifying the subscriber from clusters of information which are associated with individual subscribers.",2004-01-27,"The title of the patent is subscriber identification system and its abstract is a subscriber identification system is presented in which subscriber selection data including channel changes, volume changes, and time-of-day viewing information is used to identify a subscriber from a group of subscribers. in one instance, the subscriber viewing data is recorded and a signal processing algorithm such as a fourier transform is used to produce a processed version of the subscriber selection data. the processed version of the subscriber selection data can be correlated with stored common identifiers of subscriber profiles to determine which subscriber from the group is presently viewing the programming. a neural network or fuzzy logic can be used as the mechanism for identifying the subscriber from clusters of information which are associated with individual subscribers. dated 2004-01-27"
6687336,line qualification with neural networks,a method tests a subscriber line. the method includes determining values of electrical line features from electrical measurements on the subscriber line and processing a portion of the values of the electrical features with a neural network. the neural network predicts whether the line qualifies to support one or more preselected data services from the portion of the values.,2004-02-03,The title of the patent is line qualification with neural networks and its abstract is a method tests a subscriber line. the method includes determining values of electrical line features from electrical measurements on the subscriber line and processing a portion of the values of the electrical features with a neural network. the neural network predicts whether the line qualifies to support one or more preselected data services from the portion of the values. dated 2004-02-03
6687597,neural control system and method for alternatively fueled engines,"a powertrain controller of a vehicle provides fuel injection pulses based on gasoline operation. the pulse widths of the fuel injection pulses are modified with reference to air temperature, engine speed, and exhaust gas oxygen (ego) content to control fuel injectors for an alternative fuel such as natural gas. the ego content, based on alternative fuel operation, is detected and compared to a desired air-fuel ratio or desired fuel trims to provide error information that is used to adjust the modification of the pulse widths. in response to the error information, a neural network (as an example) dynamically adjust the pulse widths of the alternative fuel injection based on the weights of measured, detected engine speed, ego, universal exhaust gas oxygen, or air temperatures. the engine operating on alternative fuel is provided with the proper mixture of alternative fuel and air to respond to various engine loads and meet emission standards.",2004-02-03,"The title of the patent is neural control system and method for alternatively fueled engines and its abstract is a powertrain controller of a vehicle provides fuel injection pulses based on gasoline operation. the pulse widths of the fuel injection pulses are modified with reference to air temperature, engine speed, and exhaust gas oxygen (ego) content to control fuel injectors for an alternative fuel such as natural gas. the ego content, based on alternative fuel operation, is detected and compared to a desired air-fuel ratio or desired fuel trims to provide error information that is used to adjust the modification of the pulse widths. in response to the error information, a neural network (as an example) dynamically adjust the pulse widths of the alternative fuel injection based on the weights of measured, detected engine speed, ego, universal exhaust gas oxygen, or air temperatures. the engine operating on alternative fuel is provided with the proper mixture of alternative fuel and air to respond to various engine loads and meet emission standards. dated 2004-02-03"
6691073,"adaptive state space signal separation, discrimination and recovery","this invention unifies a set of statistical signal processing, neuromorphic systems, and microelectronic implementation techniques for blind separation and recovery of mixed signals. a set of architectures, frameworks, algorithms, and devices for separating, discriminating, and recovering original signal sources by processing a set of received mixtures and functions of said signals are described. the adaptation inherent in the referenced architectures, frameworks, algorithms, and devices is based on processing of the received, measured, recorded or otherwise stored signals or functions thereof. there are multiple criteria that can be used alone or in conjunction with other criteria for achieving the separation and recovery of the original signal content from the signal mixtures. the composition adopts both discrete-time and continuous-time formulations with a view towards implementations in the digital as well as the analog domains of microelectronic circuits. this invention focuses on the development and formulation of dynamic architectures with adaptive update laws for multi-source blind signal separation/recovery. the system of the invention seeks to permit the adaptive blind separation and recovery of several unknown signals mixed together in changing interference environments with very minimal assumption on the original signals. the system of this invention has practical applications to non-multiplexed media sharing, adaptive interferer rejection, acoustic sensors, acoustic diagnostics, medical diagnostics and instrumentation, speech, voice, language recognition and processing, wired and wireless modulated communication signal receivers, and cellular communications.this invention also introduces a set of update laws and links minimization of mutual information and the information maximization of the output entropy function of a nonlinear neural network, specifically in relation to techniques for blind separation, discrimination and recovery of mixed signals. the system of the invention seeks to permit the adaptive blind separation and recovery of several unknown signals mixed together in changing interference environments with very minimal assumption on the original signals.",2004-02-10,"The title of the patent is adaptive state space signal separation, discrimination and recovery and its abstract is this invention unifies a set of statistical signal processing, neuromorphic systems, and microelectronic implementation techniques for blind separation and recovery of mixed signals. a set of architectures, frameworks, algorithms, and devices for separating, discriminating, and recovering original signal sources by processing a set of received mixtures and functions of said signals are described. the adaptation inherent in the referenced architectures, frameworks, algorithms, and devices is based on processing of the received, measured, recorded or otherwise stored signals or functions thereof. there are multiple criteria that can be used alone or in conjunction with other criteria for achieving the separation and recovery of the original signal content from the signal mixtures. the composition adopts both discrete-time and continuous-time formulations with a view towards implementations in the digital as well as the analog domains of microelectronic circuits. this invention focuses on the development and formulation of dynamic architectures with adaptive update laws for multi-source blind signal separation/recovery. the system of the invention seeks to permit the adaptive blind separation and recovery of several unknown signals mixed together in changing interference environments with very minimal assumption on the original signals. the system of this invention has practical applications to non-multiplexed media sharing, adaptive interferer rejection, acoustic sensors, acoustic diagnostics, medical diagnostics and instrumentation, speech, voice, language recognition and processing, wired and wireless modulated communication signal receivers, and cellular communications.this invention also introduces a set of update laws and links minimization of mutual information and the information maximization of the output entropy function of a nonlinear neural network, specifically in relation to techniques for blind separation, discrimination and recovery of mixed signals. the system of the invention seeks to permit the adaptive blind separation and recovery of several unknown signals mixed together in changing interference environments with very minimal assumption on the original signals. dated 2004-02-10"
6691082,method and system for sub-band hybrid coding,"a system and method are provided for processing audio and speech signals using a pitch and voicing dependent spectral estimation algorithm (voicing algorithm) to accurately represent voiced speech, unvoiced speech, and mixed speech in the presence of background noise, and background noise with a single model. the present invention also modifies the synthesis model based on an estimate of the current input signal to improve the perceptual quality of the speech and background noise under a variety of input conditions. the present invention also improves the voicing dependent spectral estimation algorithm robustness by introducing the use of a multi-layer neural network in the estimation process. the voicing dependent spectral estimation algorithm provides an accurate and robust estimate of the voicing probability under a variety of background noise conditions. this is essential to providing high quality intelligible speech in the presence of background noise. in one embodiment, the waveform coding is implemented by separating the input signal into at least two sub-band signals and encoding one of the at least two sub-band signals using a first encoding algorithm to produce at least one encoded output signal; and encoding another of said at least two sub-band signals using a second encoding algorithm to produce at least one other encoded output signal, where the first encoding algorithm is different from the second encoding algorithm. in accordance with the described embodiment, the present invention provides an encoder that codes n user defined sub-band signal in the baseband with one of a plurality of waveform coding algorithms, and encodes n user defined sub-band signals with one of a plurality of parametric coding algorithms. that is, the selected waveform/parametric encoding algorithm may be different in each sub-band.",2004-02-10,"The title of the patent is method and system for sub-band hybrid coding and its abstract is a system and method are provided for processing audio and speech signals using a pitch and voicing dependent spectral estimation algorithm (voicing algorithm) to accurately represent voiced speech, unvoiced speech, and mixed speech in the presence of background noise, and background noise with a single model. the present invention also modifies the synthesis model based on an estimate of the current input signal to improve the perceptual quality of the speech and background noise under a variety of input conditions. the present invention also improves the voicing dependent spectral estimation algorithm robustness by introducing the use of a multi-layer neural network in the estimation process. the voicing dependent spectral estimation algorithm provides an accurate and robust estimate of the voicing probability under a variety of background noise conditions. this is essential to providing high quality intelligible speech in the presence of background noise. in one embodiment, the waveform coding is implemented by separating the input signal into at least two sub-band signals and encoding one of the at least two sub-band signals using a first encoding algorithm to produce at least one encoded output signal; and encoding another of said at least two sub-band signals using a second encoding algorithm to produce at least one other encoded output signal, where the first encoding algorithm is different from the second encoding algorithm. in accordance with the described embodiment, the present invention provides an encoder that codes n user defined sub-band signal in the baseband with one of a plurality of waveform coding algorithms, and encodes n user defined sub-band signals with one of a plurality of parametric coding algorithms. that is, the selected waveform/parametric encoding algorithm may be different in each sub-band. dated 2004-02-10"
6694046,automated computerized scheme for distinction between benign and malignant solitary pulmonary nodules on chest images,"an automated method for analyzing a nodule and a computer storage medium storing computer instructions by which the method can be implemented when the instructions are loaded into a computer to program the computer. the method includes obtaining a digital image including the nodule; segmenting the nodule to obtain an outline of the nodule, including generating a difference image from chest image, identifying image intensity contour lines representative of respective image intensities in a region of interest including the nodule, and obtaining an outline of the nodule based on the image intensity contours; extracting features of the nodule based on the outline; applying features including the extracted features to at least one image classifier; and determining a likelihood of malignancy of the nodule based on the output of the at least one classifier. in one embodiment, extracted features are applied to a linear discriminant analyzer and/or an artificial neural network analyzer, the outputs of which are thresholded and the nodule determined to be non-malignant if each classifier output is below the threshold. in another embodiment, a common nodule appearing in an x-ray chest image and a ct image is segmented in each image, features extracted based on the outlines of each segmented nodule in the respective x-ray chest and ct images, and the extracted features from the x-ray chest image and ct images merged as inputs to a common classifier, with the output of the common classifier indicating the likelihood of malignancy.",2004-02-17,"The title of the patent is automated computerized scheme for distinction between benign and malignant solitary pulmonary nodules on chest images and its abstract is an automated method for analyzing a nodule and a computer storage medium storing computer instructions by which the method can be implemented when the instructions are loaded into a computer to program the computer. the method includes obtaining a digital image including the nodule; segmenting the nodule to obtain an outline of the nodule, including generating a difference image from chest image, identifying image intensity contour lines representative of respective image intensities in a region of interest including the nodule, and obtaining an outline of the nodule based on the image intensity contours; extracting features of the nodule based on the outline; applying features including the extracted features to at least one image classifier; and determining a likelihood of malignancy of the nodule based on the output of the at least one classifier. in one embodiment, extracted features are applied to a linear discriminant analyzer and/or an artificial neural network analyzer, the outputs of which are thresholded and the nodule determined to be non-malignant if each classifier output is below the threshold. in another embodiment, a common nodule appearing in an x-ray chest image and a ct image is segmented in each image, features extracted based on the outlines of each segmented nodule in the respective x-ray chest and ct images, and the extracted features from the x-ray chest image and ct images merged as inputs to a common classifier, with the output of the common classifier indicating the likelihood of malignancy. dated 2004-02-17"
6694049,multimode invariant processor,"a multimode invariant processor is provided to simultaneously classify one or more patterns in multidimensional or in two dimensional &#8220;real world&#8221; images. the classification is invariant to a translation, a change in scale size and a rotation of a whole or partially hidden photonic image. the multimode invariant image processor comprises a retina portion, a nonlinear processing portion, a convergence processing portion and a classifier portion. the retina portion processes the photonic image to obtain an image data array of pixels and further process the array of pixels through a window difference network to obtain gradients of the image data. the neural directors of the nonlinear processing portion receive the gradients and generate respective feature vectors, which may have a greater dimensionality than the gradient information, to aid in discrimination between similar patterns in the image data. the convergence portion processes the feature information to generate a convergence of common feature information representing at least one image feature in the image data. the classifier portion receives the common feature information and generates in response feature classification information indicating the likelihood that selected features are present in the image.",2004-02-17,"The title of the patent is multimode invariant processor and its abstract is a multimode invariant processor is provided to simultaneously classify one or more patterns in multidimensional or in two dimensional &#8220;real world&#8221; images. the classification is invariant to a translation, a change in scale size and a rotation of a whole or partially hidden photonic image. the multimode invariant image processor comprises a retina portion, a nonlinear processing portion, a convergence processing portion and a classifier portion. the retina portion processes the photonic image to obtain an image data array of pixels and further process the array of pixels through a window difference network to obtain gradients of the image data. the neural directors of the nonlinear processing portion receive the gradients and generate respective feature vectors, which may have a greater dimensionality than the gradient information, to aid in discrimination between similar patterns in the image data. the convergence portion processes the feature information to generate a convergence of common feature information representing at least one image feature in the image data. the classifier portion receives the common feature information and generates in response feature classification information indicating the likelihood that selected features are present in the image. dated 2004-02-17"
6694301,goal-oriented clustering,"clustering for purposes of data visualization and making predictions is disclosed. embodiments of the invention are operable on a number of variables that have a predetermined representation. the variables include input-only variables, output-only variables, and both input-and-output variables. embodiments of the invention generate a model that has a bottleneck architecture. the model includes a top layer of nodes of at least the input-only variables, one or more middle layer of hidden nodes, and a bottom layer of nodes of the output-only and the input-and-output variables. at least one cluster is determined from this model. the model can be a probabilistic neural network and/or a bayesian network.",2004-02-17,"The title of the patent is goal-oriented clustering and its abstract is clustering for purposes of data visualization and making predictions is disclosed. embodiments of the invention are operable on a number of variables that have a predetermined representation. the variables include input-only variables, output-only variables, and both input-and-output variables. embodiments of the invention generate a model that has a bottleneck architecture. the model includes a top layer of nodes of at least the input-only variables, one or more middle layer of hidden nodes, and a bottom layer of nodes of the output-only and the input-and-output variables. at least one cluster is determined from this model. the model can be a probabilistic neural network and/or a bayesian network. dated 2004-02-17"
6700648,method for controlling a processing apparatus,"in a method for controlling a processing apparatus, an error value between an input value of the processing apparatus for processing a subject to be processed, and a measurement value obtained by measuring the subject being processed is obtained. a correction value is computed for correcting the input value of the processing apparatus in the direction of decreasing the error value, and the values are managed as processing data to be utilized in computing a next correction value. previous processing data having a history identical to that of the subject loaded to the processing apparatus is searched, and a current bias correction value is predicted from a plurality of most recent correction values having the identical history. also, a current random correction value is predicted by means of a neural network on the basis of a plurality of most recent random correction values. the predicted bias correction value is summed with the random correction value as a current correction value of the processing apparatus.",2004-03-02,"The title of the patent is method for controlling a processing apparatus and its abstract is in a method for controlling a processing apparatus, an error value between an input value of the processing apparatus for processing a subject to be processed, and a measurement value obtained by measuring the subject being processed is obtained. a correction value is computed for correcting the input value of the processing apparatus in the direction of decreasing the error value, and the values are managed as processing data to be utilized in computing a next correction value. previous processing data having a history identical to that of the subject loaded to the processing apparatus is searched, and a current bias correction value is predicted from a plurality of most recent correction values having the identical history. also, a current random correction value is predicted by means of a neural network on the basis of a plurality of most recent random correction values. the predicted bias correction value is summed with the random correction value as a current correction value of the processing apparatus. dated 2004-03-02"
6701029,ring-wedge data analysis of digital images,"computer software for and a method of calculating ring-wedge data from a digital image by performing a discrete fourier transform of the digital image. a discrete autocorrelation, discrete cosine transform, and/or hadamard transform is also preferably performed, together with providing the results to a neural network (most preferably a fully connected, three-layer, feed-forward neural network with sigmoidal activation functions) to perform pattern recognition on the data.",2004-03-02,"The title of the patent is ring-wedge data analysis of digital images and its abstract is computer software for and a method of calculating ring-wedge data from a digital image by performing a discrete fourier transform of the digital image. a discrete autocorrelation, discrete cosine transform, and/or hadamard transform is also preferably performed, together with providing the results to a neural network (most preferably a fully connected, three-layer, feed-forward neural network with sigmoidal activation functions) to perform pattern recognition on the data. dated 2004-03-02"
6701236,intelligent mechatronic control suspension system based on soft computing,"a control system for optimizing a shock absorber having a non-linear kinetic characteristic is described. the control system uses a fitness (performance) function that is based on the physical laws of minimum entropy and biologically inspired constraints relating to mechanical constraints and/or rider comfort, driveability, etc. in one embodiment, a genetic analyzer is used in an off-line mode to develop a teaching signal. an information filter is used to filter the teaching signal to produce a compressed teaching signal. the compressed teaching signal can be approximated online by a fuzzy controller that operates using knowledge from a knowledge base. in one embodiment, the control system includes a learning system, such as a neural network that is trained by the compressed training signal. the learning system is used to create a knowledge base for use by an online fuzzy controller. the online fuzzy controller is used to program a linear controller.",2004-03-02,"The title of the patent is intelligent mechatronic control suspension system based on soft computing and its abstract is a control system for optimizing a shock absorber having a non-linear kinetic characteristic is described. the control system uses a fitness (performance) function that is based on the physical laws of minimum entropy and biologically inspired constraints relating to mechanical constraints and/or rider comfort, driveability, etc. in one embodiment, a genetic analyzer is used in an off-line mode to develop a teaching signal. an information filter is used to filter the teaching signal to produce a compressed teaching signal. the compressed teaching signal can be approximated online by a fuzzy controller that operates using knowledge from a knowledge base. in one embodiment, the control system includes a learning system, such as a neural network that is trained by the compressed training signal. the learning system is used to create a knowledge base for use by an online fuzzy controller. the online fuzzy controller is used to program a linear controller. dated 2004-03-02"
6701318,multiple engine information retrieval and visualization system,"an information retrieval and visualization system utilizes multiple search engines for retrieving documents from a document database based upon user input queries. search engines include an n-gram search engine and a vector space model search engine using a neural network training algorithm. each search engine produces a common mathematical representation of each retrieved document. the retrieved documents are then combined and ranked. mathematical representations for each respective document is mapped onto a display. information displayed includes a three-dimensional display of keywords from the user input query. the three-dimensional visualization capability based upon the mathematical representation of information within the information retrieval and visualization system provides users with an intuitive understanding, with relevance feedback/query refinement techniques that can be better utilized, resulting in higher retrieval accuracy (precision).",2004-03-02,"The title of the patent is multiple engine information retrieval and visualization system and its abstract is an information retrieval and visualization system utilizes multiple search engines for retrieving documents from a document database based upon user input queries. search engines include an n-gram search engine and a vector space model search engine using a neural network training algorithm. each search engine produces a common mathematical representation of each retrieved document. the retrieved documents are then combined and ranked. mathematical representations for each respective document is mapped onto a display. information displayed includes a three-dimensional display of keywords from the user input query. the three-dimensional visualization capability based upon the mathematical representation of information within the information retrieval and visualization system provides users with an intuitive understanding, with relevance feedback/query refinement techniques that can be better utilized, resulting in higher retrieval accuracy (precision). dated 2004-03-02"
6704717,analytic algorithm for enhanced back-propagation neural network processing,"a method, apparatus, and article of manufacture for performing data mining applications in a relational database management system. at least one analytic algorithm for enhanced back-propagation neural network processing is performed by a computer, wherein the analytic algorithm for enhanced back-propagation neural network processing includes sql statements performed by the relational database management system directly against the relational database and programmatic iteration. the analytic algorithm for enhanced back-propagation neural network processing operates on data in the relational database that has been partitioned into training, testing and validation data sets. the analytic algorithm for enhanced back-propagation neural network processing maps the data in the training data sets to nodes in the neural network wherein the data is processed as it moves from an input node of the neural network through a hidden node of the neural network to an output node of the neural network. in addition, the analytic algorithm for enhanced back-propagation neural network processing determines an error difference between the output node's value and a target value as the data is mapped to the output node in the neural network, and changes a weight value for one or more of the nodes based on an accumulation of the error difference for the node, in order to get the neural network to converge on a solution. finally, the analytic algorithm for enhanced back-propagation neural network processing cross-validates the changed weight value to prevent overfitting the node.",2004-03-09,"The title of the patent is analytic algorithm for enhanced back-propagation neural network processing and its abstract is a method, apparatus, and article of manufacture for performing data mining applications in a relational database management system. at least one analytic algorithm for enhanced back-propagation neural network processing is performed by a computer, wherein the analytic algorithm for enhanced back-propagation neural network processing includes sql statements performed by the relational database management system directly against the relational database and programmatic iteration. the analytic algorithm for enhanced back-propagation neural network processing operates on data in the relational database that has been partitioned into training, testing and validation data sets. the analytic algorithm for enhanced back-propagation neural network processing maps the data in the training data sets to nodes in the neural network wherein the data is processed as it moves from an input node of the neural network through a hidden node of the neural network to an output node of the neural network. in addition, the analytic algorithm for enhanced back-propagation neural network processing determines an error difference between the output node's value and a target value as the data is mapped to the output node in the neural network, and changes a weight value for one or more of the nodes based on an accumulation of the error difference for the node, in order to get the neural network to converge on a solution. finally, the analytic algorithm for enhanced back-propagation neural network processing cross-validates the changed weight value to prevent overfitting the node. dated 2004-03-09"
6708160,object nets,"a method, system and computer program product for implementing at least one of a learning-based diagnostics system and a control system (e.g., using a neural network). by using objectnets to model general object types, it is possible to design a control system that represents system components as relational structures rather than fixed vectors. such an advance is possible by exploiting non-euclidean principles of symmetry.",2004-03-16,"The title of the patent is object nets and its abstract is a method, system and computer program product for implementing at least one of a learning-based diagnostics system and a control system (e.g., using a neural network). by using objectnets to model general object types, it is possible to design a control system that represents system components as relational structures rather than fixed vectors. such an advance is possible by exploiting non-euclidean principles of symmetry. dated 2004-03-16"
6713978,method and system for determining induction motor speed,"a non-linear, semi-parametric neural network-based adaptive filter is utilized to determine the dynamic speed of a rotating rotor within an induction motor, without the explicit use of a speed sensor, such as a tachometer, is disclosed. the neural network-based filter is developed using actual motor current measurements, voltage measurements, and nameplate information. the neural network-based adaptive filter is trained using an estimated speed calculator derived from the actual current and voltage measurements. the neural network-based adaptive filter uses voltage and current measurements to determine the instantaneous speed of a rotating rotor. the neural network-based adaptive filter also includes an on-line adaptation scheme that permits the filter to be readily adapted for new operating conditions during operations.",2004-03-30,"The title of the patent is method and system for determining induction motor speed and its abstract is a non-linear, semi-parametric neural network-based adaptive filter is utilized to determine the dynamic speed of a rotating rotor within an induction motor, without the explicit use of a speed sensor, such as a tachometer, is disclosed. the neural network-based filter is developed using actual motor current measurements, voltage measurements, and nameplate information. the neural network-based adaptive filter is trained using an estimated speed calculator derived from the actual current and voltage measurements. the neural network-based adaptive filter uses voltage and current measurements to determine the instantaneous speed of a rotating rotor. the neural network-based adaptive filter also includes an on-line adaptation scheme that permits the filter to be readily adapted for new operating conditions during operations. dated 2004-03-30"
6714917,subscriber identification based on electronic program guide data,"a subscriber identification system is presented in which epg related data including scrolling rates, paging rates, information screen viewing times, and manner and frequency of epg activation are used to identify a subscriber from a group of subscribers. in one instance, the subscriber viewing data is recorded and a signal processing algorithm, such as a fourier transform, is used to produce a processed version of the epg related data. the processed version of the epg related data can be correlated with stored common identifiers of subscriber profiles to determine which subscriber from the group is presently viewing the programming. a neural network or fuzzy logic can be used as the mechanism for identifying the subscriber from clusters of information, which are associated with individual subscribers.",2004-03-30,"The title of the patent is subscriber identification based on electronic program guide data and its abstract is a subscriber identification system is presented in which epg related data including scrolling rates, paging rates, information screen viewing times, and manner and frequency of epg activation are used to identify a subscriber from a group of subscribers. in one instance, the subscriber viewing data is recorded and a signal processing algorithm, such as a fourier transform, is used to produce a processed version of the epg related data. the processed version of the epg related data can be correlated with stored common identifiers of subscriber profiles to determine which subscriber from the group is presently viewing the programming. a neural network or fuzzy logic can be used as the mechanism for identifying the subscriber from clusters of information, which are associated with individual subscribers. dated 2004-03-30"
6714924,computer-implemented neural network color matching formulation system,"a method and apparatus for color matching are provided, in which paint recipe neural networks are utilized. the color of a standard is expressed as color values. the neural network includes an input layer having nodes for receiving input data related to paint bases. weighted connections connect to the nodes of the input layer and have coefficients for weighting the input data. an output layer having nodes are either directly or indirectly connected to the weighted connections and generates output data related to color values. the data to the input layer and the data from the output layer are interrelated through the neural network's nonlinear relationship. the paint color matching neural network can be used for, but not limited to, color formula correction, matching from scratch, effect pigment identification, selection of targets for color tools, searching existing formulas for the closest match, identification of formula mistakes, development of color tolerances and enhancing conversion routines.",2004-03-30,"The title of the patent is computer-implemented neural network color matching formulation system and its abstract is a method and apparatus for color matching are provided, in which paint recipe neural networks are utilized. the color of a standard is expressed as color values. the neural network includes an input layer having nodes for receiving input data related to paint bases. weighted connections connect to the nodes of the input layer and have coefficients for weighting the input data. an output layer having nodes are either directly or indirectly connected to the weighted connections and generates output data related to color values. the data to the input layer and the data from the output layer are interrelated through the neural network's nonlinear relationship. the paint color matching neural network can be used for, but not limited to, color formula correction, matching from scratch, effect pigment identification, selection of targets for color tools, searching existing formulas for the closest match, identification of formula mistakes, development of color tolerances and enhancing conversion routines. dated 2004-03-30"
6717728,system and method for visualization of stereo and multi aspect images,"the method and system of the present invention displays autostereographic images without parallax barriers or loss of resolution. a stereopair is processed and sent to a distant display and one or more transmissive displays placed in front of it. each display has a calculated images containing at least some of the image information destined for both eyes of a viewer. each display acts as a mask for the other displays. the processing of the stereopairs to produce calculated images comprises summing the predicted image data, comparing the predicted image data to the desired stereopair, and minimizing the error. in a preferred embodiment, this processing is performed by an artificial neural network. a spatial mask, such as a diffuser, can be placed between displays to suppress moir&eacute; patterns.",2004-04-06,"The title of the patent is system and method for visualization of stereo and multi aspect images and its abstract is the method and system of the present invention displays autostereographic images without parallax barriers or loss of resolution. a stereopair is processed and sent to a distant display and one or more transmissive displays placed in front of it. each display has a calculated images containing at least some of the image information destined for both eyes of a viewer. each display acts as a mask for the other displays. the processing of the stereopairs to produce calculated images comprises summing the predicted image data, comparing the predicted image data to the desired stereopair, and minimizing the error. in a preferred embodiment, this processing is performed by an artificial neural network. a spatial mask, such as a diffuser, can be placed between displays to suppress moir&eacute; patterns. dated 2004-04-06"
6718196,multimodality instrument for tissue characterization,"a system with multimodality instrument for tissue identification includes a computer-controlled motor driven heuristic probe with a multisensory tip. for neurosurgical applications, the instrument is mounted on a stereotactic frame for the probe to penetrate the brain in a precisely controlled fashion. the resistance of the brain tissue being penetrated is continually monitored by a miniaturized strain gauge attached to the probe tip. other modality sensors may be mounted near the probe tip to provide real-time tissue characterizations and the ability to detect the proximity of blood vessels, thus eliminating errors normally associated with registration of pre-operative scans, tissue swelling, elastic tissue deformation, human judgement, etc., and rendering surgical procedures safer, more accurate, and efficient. a neural network program adaptively learns the information on resistance and other characteristic features of normal brain tissue during the surgery and provides near real-time modeling. a fuzzy logic interface to the neural network program incorporates expert medical knowledge in the learning process. identification of abnormal brain tissue is determined by the detection of change and comparison with previously learned models of abnormal brain tissues. the operation of the instrument is controlled through a user friendly graphical interface. patient data is presented in a 3d stereographics display. acoustic feedback of selected information may optionally be provided. upon detection of the close proximity to blood vessels or abnormal brain tissue, the computer-controlled motor immediately stops probe penetration. the use of this system will make surgical procedures safer, more accurate, and more efficient. other applications of this system include the detection, prognosis and treatment of breast cancer, prostate cancer, spinal diseases, and use in general exploratory surgery.",2004-04-06,"The title of the patent is multimodality instrument for tissue characterization and its abstract is a system with multimodality instrument for tissue identification includes a computer-controlled motor driven heuristic probe with a multisensory tip. for neurosurgical applications, the instrument is mounted on a stereotactic frame for the probe to penetrate the brain in a precisely controlled fashion. the resistance of the brain tissue being penetrated is continually monitored by a miniaturized strain gauge attached to the probe tip. other modality sensors may be mounted near the probe tip to provide real-time tissue characterizations and the ability to detect the proximity of blood vessels, thus eliminating errors normally associated with registration of pre-operative scans, tissue swelling, elastic tissue deformation, human judgement, etc., and rendering surgical procedures safer, more accurate, and efficient. a neural network program adaptively learns the information on resistance and other characteristic features of normal brain tissue during the surgery and provides near real-time modeling. a fuzzy logic interface to the neural network program incorporates expert medical knowledge in the learning process. identification of abnormal brain tissue is determined by the detection of change and comparison with previously learned models of abnormal brain tissues. the operation of the instrument is controlled through a user friendly graphical interface. patient data is presented in a 3d stereographics display. acoustic feedback of selected information may optionally be provided. upon detection of the close proximity to blood vessels or abnormal brain tissue, the computer-controlled motor immediately stops probe penetration. the use of this system will make surgical procedures safer, more accurate, and more efficient. other applications of this system include the detection, prognosis and treatment of breast cancer, prostate cancer, spinal diseases, and use in general exploratory surgery. dated 2004-04-06"
6718316,neural network noise anomaly recognition system and method,"a system and method for a neural network is disclosed that is trained to recognize noise characteristics or other types of interference and to determine when an input waveform deviates from learned noise characteristics. a plurality of neural networks is preferably provided, which each receives a plurality of samples of intervals or windows of the input waveform. each of the neural networks produces an output based on whether an anomaly is detected with respect to the noise, which the neural network is trained to detect. the plurality of outputs of the neural networks is preferably applied to a decision aid for deciding whether the input waveform contains a non-noise component. the decision aid may include a database, a computational section and a decision module. the system and method may provide a preliminary processing of the input waveform and is used to recognize the particular noise rather than a non-noise signal.",2004-04-06,"The title of the patent is neural network noise anomaly recognition system and method and its abstract is a system and method for a neural network is disclosed that is trained to recognize noise characteristics or other types of interference and to determine when an input waveform deviates from learned noise characteristics. a plurality of neural networks is preferably provided, which each receives a plurality of samples of intervals or windows of the input waveform. each of the neural networks produces an output based on whether an anomaly is detected with respect to the noise, which the neural network is trained to detect. the plurality of outputs of the neural networks is preferably applied to a decision aid for deciding whether the input waveform contains a non-noise component. the decision aid may include a database, a computational section and a decision module. the system and method may provide a preliminary processing of the input waveform and is used to recognize the particular noise rather than a non-noise signal. dated 2004-04-06"
6721718,system for intelligent control based on soft computing,a reduced control system suitable for control of a nonlinear or unstable plant is described. the reduced control system is configured to use a reduced sensor set for controlling the plant without significant loss of control quality (accuracy) as compared to an optimal control system with an optimum sensor set. the control system calculates the information content provided by the reduced sensor set as compared to the information content provided by the optimum set. the control system also calculates the difference between the entropy production rate of the plant and the entropy production rate of the controller. a genetic optimizer is used to tune a fuzzy neural network in the reduced controller. a fitness function for the genetic optimizer provides optimum control accuracy in the reduced control system by minimizing the difference in entropy production while maximizing the sensor information content.,2004-04-13,The title of the patent is system for intelligent control based on soft computing and its abstract is a reduced control system suitable for control of a nonlinear or unstable plant is described. the reduced control system is configured to use a reduced sensor set for controlling the plant without significant loss of control quality (accuracy) as compared to an optimal control system with an optimum sensor set. the control system calculates the information content provided by the reduced sensor set as compared to the information content provided by the optimum set. the control system also calculates the difference between the entropy production rate of the plant and the entropy production rate of the controller. a genetic optimizer is used to tune a fuzzy neural network in the reduced controller. a fitness function for the genetic optimizer provides optimum control accuracy in the reduced control system by minimizing the difference in entropy production while maximizing the sensor information content. dated 2004-04-13
6722173,method and device for determining the rolling force in a roll stand,"in a method for determining the rolling force in a roll stand for rolling metallic material which is to be rolled, the rolling force is determined by means of at least one neural network. according to the invention, the neural network is trained using, in particular measured, values for the rolling force under different operating conditions with a view to improving the determination of the rolling force. the neural network is advantageously trained using values for the rolling force and values for the different operating conditions for roll stands of different rolling trains.",2004-04-20,"The title of the patent is method and device for determining the rolling force in a roll stand and its abstract is in a method for determining the rolling force in a roll stand for rolling metallic material which is to be rolled, the rolling force is determined by means of at least one neural network. according to the invention, the neural network is trained using, in particular measured, values for the rolling force under different operating conditions with a view to improving the determination of the rolling force. the neural network is advantageously trained using values for the rolling force and values for the different operating conditions for roll stands of different rolling trains. dated 2004-04-20"
6724687,"characterizing oil, gasor geothermal wells, including fractures thereof","an excitation event in an oil, gas or geothermal well creates a responsive signal having lower and higher frequency components, which higher frequency component provides information about one or more characteristics of the well. examples of such characteristics pertaining to a subterranean fracture include: breakdown pressure at fracture initiation, time it takes proppant to reach and to screenout the tip of the fracture, fracture geometry and fracture growth, fracture closure pressure, relative fluid flow through respective perforations, and horsepower requirements to perform a fracture treatment. one excitation event includes creating an excitation signal having a maximum amplitude change occurring within a time t1, which is less than a period t2 of the higher frequency component. wavelet processing may be used to separate or distinguish the higher frequency waveform from the lower frequency waveform. the information can be used to control a process (for example, a fracturing process) applied to the respective well or one or more other wells. in another aspect, an unidentified signature waveform is compared to identified signature waveforms in a neural network computer database to create an identity for the unidentified signature waveform relative to an identified signature waveform in the database. a system to determine a characteristic of an oil, gas or geothermal well is also disclosed.",2004-04-20,"The title of the patent is characterizing oil, gasor geothermal wells, including fractures thereof and its abstract is an excitation event in an oil, gas or geothermal well creates a responsive signal having lower and higher frequency components, which higher frequency component provides information about one or more characteristics of the well. examples of such characteristics pertaining to a subterranean fracture include: breakdown pressure at fracture initiation, time it takes proppant to reach and to screenout the tip of the fracture, fracture geometry and fracture growth, fracture closure pressure, relative fluid flow through respective perforations, and horsepower requirements to perform a fracture treatment. one excitation event includes creating an excitation signal having a maximum amplitude change occurring within a time t1, which is less than a period t2 of the higher frequency component. wavelet processing may be used to separate or distinguish the higher frequency waveform from the lower frequency waveform. the information can be used to control a process (for example, a fracturing process) applied to the respective well or one or more other wells. in another aspect, an unidentified signature waveform is compared to identified signature waveforms in a neural network computer database to create an identity for the unidentified signature waveform relative to an identified signature waveform in the database. a system to determine a characteristic of an oil, gas or geothermal well is also disclosed. dated 2004-04-20"
6725163,method for processing seismic measured data with a neuronal network,"the invention relates to a method for processing a 3-d measurement data set provided with seismic attributes, by means of a neural network, whereby a self-organizing map is trained with selected training data, and the data to be examined are classified according to the self-organizing map. in this connection, the measurement data are processed by way of sub-volumes comprising a spatial environment of each sample. the classification is stored in a result data set, and displayed, for example in a two-dimensional map (fig. 3).",2004-04-20,"The title of the patent is method for processing seismic measured data with a neuronal network and its abstract is the invention relates to a method for processing a 3-d measurement data set provided with seismic attributes, by means of a neural network, whereby a self-organizing map is trained with selected training data, and the data to be examined are classified according to the self-organizing map. in this connection, the measurement data are processed by way of sub-volumes comprising a spatial environment of each sample. the classification is stored in a result data set, and displayed, for example in a two-dimensional map (fig. 3). dated 2004-04-20"
6725207,media selection using a neural network,"a method and system for automatically classifying a print medium entering a printing device as being a print medium type having known properties relevant to print operations. a detection system captures data indicative of optical characteristics of the incoming medium. the data is spectrally examined to derive frequency-related information. at least one neural network utilizes the frequency-related information to determine a medium type. in one embodiment, a major category network determines the medium type as one of five major medium types. subsequently, the medium is subjected to analysis with a specific neural network for differentiating the identified major media type into narrower categories. each neural network comprises a layer of adaptive decision-making nodes. each node includes an activation function for processing the sum of multiple weighted inputs for generating an output. the output is directed to the output level that is at least partially utilized for a medium type determination.",2004-04-20,"The title of the patent is media selection using a neural network and its abstract is a method and system for automatically classifying a print medium entering a printing device as being a print medium type having known properties relevant to print operations. a detection system captures data indicative of optical characteristics of the incoming medium. the data is spectrally examined to derive frequency-related information. at least one neural network utilizes the frequency-related information to determine a medium type. in one embodiment, a major category network determines the medium type as one of five major medium types. subsequently, the medium is subjected to analysis with a specific neural network for differentiating the identified major media type into narrower categories. each neural network comprises a layer of adaptive decision-making nodes. each node includes an activation function for processing the sum of multiple weighted inputs for generating an output. the output is directed to the output level that is at least partially utilized for a medium type determination. dated 2004-04-20"
6725208,bayesian neural networks for optimization and control,an optimization system is provided utilizing a bayesian neural network calculation of a derivative wherein an output is optimized with respect to an input utilizing a stochastical method that averages over many regression models. this is done such that constraints from first principal models are incorporated in terms of prior art distributions.,2004-04-20,The title of the patent is bayesian neural networks for optimization and control and its abstract is an optimization system is provided utilizing a bayesian neural network calculation of a derivative wherein an output is optimized with respect to an input utilizing a stochastical method that averages over many regression models. this is done such that constraints from first principal models are incorporated in terms of prior art distributions. dated 2004-04-20
6726113,temperature control strategy utilizing neural network processing of occupancy and activity level sensing,a temperature control system utilizes detected occupancy and activity levels to automatically condition a response of an hvac system to a difference between a setpoint temperature and an actual temperature within a zone. the illustrated example includes an infrared sensor that provides at least one signal indicating the activity and occupancy levels in the zone. a neural network processes the sensor signal to provide an indication of the occupancy and activity levels to a controller. the controller automatically adjusts at least one control parameter of the hvac system to compensate for changes in the occupancy or activity levels that would affect the temperature comfort setting in the zone.,2004-04-27,The title of the patent is temperature control strategy utilizing neural network processing of occupancy and activity level sensing and its abstract is a temperature control system utilizes detected occupancy and activity levels to automatically condition a response of an hvac system to a difference between a setpoint temperature and an actual temperature within a zone. the illustrated example includes an infrared sensor that provides at least one signal indicating the activity and occupancy levels in the zone. a neural network processes the sensor signal to provide an indication of the occupancy and activity levels to a controller. the controller automatically adjusts at least one control parameter of the hvac system to compensate for changes in the occupancy or activity levels that would affect the temperature comfort setting in the zone. dated 2004-04-27
6728404,method for recognizing object images and learning method for neural networks,"a method for recognizing an object image comprises the steps of extracting a candidate for a predetermined object image from an overall image, and making a judgment as to whether the extracted candidate for the predetermined object image is or is not the predetermined object image. the candidate for the predetermined object image is extracted by causing the center point of a view window, which has a predetermined size, to travel to the position of the candidate for the predetermined object image, and determining an extraction area in accordance with the size and/or the shape of the candidate for the predetermined object image, the center point of the view window being taken as a reference during the determination of the extraction area. a learning method for a neural network comprises the steps of extracting a target object image, for which learning operations are to be carried out, from an image, feeding a signal, which represents the extracted target object image, into a neural network, and carrying out the learning operations of the neural network in accordance with the input target object image.",2004-04-27,"The title of the patent is method for recognizing object images and learning method for neural networks and its abstract is a method for recognizing an object image comprises the steps of extracting a candidate for a predetermined object image from an overall image, and making a judgment as to whether the extracted candidate for the predetermined object image is or is not the predetermined object image. the candidate for the predetermined object image is extracted by causing the center point of a view window, which has a predetermined size, to travel to the position of the candidate for the predetermined object image, and determining an extraction area in accordance with the size and/or the shape of the candidate for the predetermined object image, the center point of the view window being taken as a reference during the determination of the extraction area. a learning method for a neural network comprises the steps of extracting a target object image, for which learning operations are to be carried out, from an image, feeding a signal, which represents the extracted target object image, into a neural network, and carrying out the learning operations of the neural network in accordance with the input target object image. dated 2004-04-27"
6728687,electronic circuit for controlling a movement by a fuzzy cellular architecture,"a method of controlling the movements of a multi-actuator electro-mechanical system having a matrix of locally interconnected analog cells associated therewith is provided. each cell represents a hardware implementation of a model of fuzzy inference rules. the model includes a fuzzy circuit architecture which may be implemented in an integrated circuit with vlsi cmos technology that generates and controls a reaction diffusion mechanism typical of auto-waves using a fuzzy neural network. the fuzzy neural network defines the functional relationships that may duplicate simultaneous reaction diffusion equations. the duplication of the simultaneous reaction diffusion equations is provided using two sets of fuzzy rules processing, in a linguistic manner, the state variables of the cells. an oscillatory type dynamic is imposed on each cell where two dynamic processes having different kinetic characteristics coexist.",2004-04-27,"The title of the patent is electronic circuit for controlling a movement by a fuzzy cellular architecture and its abstract is a method of controlling the movements of a multi-actuator electro-mechanical system having a matrix of locally interconnected analog cells associated therewith is provided. each cell represents a hardware implementation of a model of fuzzy inference rules. the model includes a fuzzy circuit architecture which may be implemented in an integrated circuit with vlsi cmos technology that generates and controls a reaction diffusion mechanism typical of auto-waves using a fuzzy neural network. the fuzzy neural network defines the functional relationships that may duplicate simultaneous reaction diffusion equations. the duplication of the simultaneous reaction diffusion equations is provided using two sets of fuzzy rules processing, in a linguistic manner, the state variables of the cells. an oscillatory type dynamic is imposed on each cell where two dynamic processes having different kinetic characteristics coexist. dated 2004-04-27"
6731788,symbol classification with shape features applied to neural network,"an image processing device and method for classifying symbols, such as text, in a video stream employs a back propagation neural network (bpnn) whose feature space is derived from size, translation, and rotation invariant shape-dependent features. various example feature spaces are discussed such as regular and invariant moments and an angle histogram derived from a delaunay triangulation of a thinned, thresholded, symbol. such feature spaces provide a good match to bpnn as a classifier because of the poor resolution of characters in video streams.",2004-05-04,"The title of the patent is symbol classification with shape features applied to neural network and its abstract is an image processing device and method for classifying symbols, such as text, in a video stream employs a back propagation neural network (bpnn) whose feature space is derived from size, translation, and rotation invariant shape-dependent features. various example feature spaces are discussed such as regular and invariant moments and an angle histogram derived from a delaunay triangulation of a thinned, thresholded, symbol. such feature spaces provide a good match to bpnn as a classifier because of the poor resolution of characters in video streams. dated 2004-05-04"
6731804,thermal luminescence liquid monitoring system and method,"a thermal luminescence water monitor system and method for real-time remote sensing and identification of chemical and biological materials (cbms) in a liquid source, comprising an irradiation component having a microwave radiation source tuned to water's vibration-rotation exciting energy, a glass cell for holding a liquid sample contained within a sealed chamber for its irradiation and concomitant liberation of thermal luminescence, a spectrometer analysis component for collecting and processing thermal luminescence emissions, a neural network component for filtering thermal luminescence difference-spectra components and pattern recognition of predetermined cbms to determine their presence in the liquid source.",2004-05-04,"The title of the patent is thermal luminescence liquid monitoring system and method and its abstract is a thermal luminescence water monitor system and method for real-time remote sensing and identification of chemical and biological materials (cbms) in a liquid source, comprising an irradiation component having a microwave radiation source tuned to water's vibration-rotation exciting energy, a glass cell for holding a liquid sample contained within a sealed chamber for its irradiation and concomitant liberation of thermal luminescence, a spectrometer analysis component for collecting and processing thermal luminescence emissions, a neural network component for filtering thermal luminescence difference-spectra components and pattern recognition of predetermined cbms to determine their presence in the liquid source. dated 2004-05-04"
6732052,method and apparatus for prediction control in drilling dynamics using neural networks,the present invention provides a drilling system that utilizes a neural network for predictive control of drilling operations. a downhole processor controls the operation of the various devices in a bottom hole assembly to effect changes to drilling parameters and drilling direction to autonomously optimize the drilling effectiveness. the neural network iteratively updates a prediction model of the drilling operations and provides recommendations for drilling corrections to a drilling operator.,2004-05-04,The title of the patent is method and apparatus for prediction control in drilling dynamics using neural networks and its abstract is the present invention provides a drilling system that utilizes a neural network for predictive control of drilling operations. a downhole processor controls the operation of the various devices in a bottom hole assembly to effect changes to drilling parameters and drilling direction to autonomously optimize the drilling effectiveness. the neural network iteratively updates a prediction model of the drilling operations and provides recommendations for drilling corrections to a drilling operator. dated 2004-05-04
6735467,method and system for detecting seizures using electroencephalograms,"a method and a system are provided for detection of seizures by applying advanced numerical analysis techniques to digitized waveforms of an electroencephalogram (eeg) recording. in an embodiment, the advanced numerical analysis techniques implemented in the method and system for detecting seizures include a combination of matching pursuit, neural network rules, and connected-object clustering algorithms.",2004-05-11,"The title of the patent is method and system for detecting seizures using electroencephalograms and its abstract is a method and a system are provided for detection of seizures by applying advanced numerical analysis techniques to digitized waveforms of an electroencephalogram (eeg) recording. in an embodiment, the advanced numerical analysis techniques implemented in the method and system for detecting seizures include a combination of matching pursuit, neural network rules, and connected-object clustering algorithms. dated 2004-05-11"
6735579,static memory processor,a static memory processor for pattern recognition and an input data dimensionality reduction is provided having a multi-layer harmonic neural network and a classifier network. the multi-layer harmonic neural network receives a fused feature vector of the pattern to be recognized from a neural sensor and generates output vectors which aid in discrimination between similar patterns. the fused feature vector and each output vector are separately provided to corresponding positional king of the mountain (pkom) circuits within the classifier network. each pkom circuit generates a positional output vector with only the element corresponding to the element of the fused feature vector or output vector having the highest contribution in its respective vector having a value corresponding to one. the positional output vectors are active in a multidimensional memory space and are read by a recognition vector array which generates class likelihood outputs determined by the occupied memory space. the class likelihood outputs are provided to a class pkom circuit which outputs classification identifiers to provide the desired pattern recognition.,2004-05-11,The title of the patent is static memory processor and its abstract is a static memory processor for pattern recognition and an input data dimensionality reduction is provided having a multi-layer harmonic neural network and a classifier network. the multi-layer harmonic neural network receives a fused feature vector of the pattern to be recognized from a neural sensor and generates output vectors which aid in discrimination between similar patterns. the fused feature vector and each output vector are separately provided to corresponding positional king of the mountain (pkom) circuits within the classifier network. each pkom circuit generates a positional output vector with only the element corresponding to the element of the fused feature vector or output vector having the highest contribution in its respective vector having a value corresponding to one. the positional output vectors are active in a multidimensional memory space and are read by a recognition vector array which generates class likelihood outputs determined by the occupied memory space. the class likelihood outputs are provided to a class pkom circuit which outputs classification identifiers to provide the desired pattern recognition. dated 2004-05-11
6735580,artificial neural network based universal time series,"a neural network based universal time series prediction system for financial securities includes a pipelined recurrent ann architecutre having a plurality of identical modules to first adjust internal weights and biases in response to a first training set representing a nonlinear financial time series of samples of a financial quantity and a target value, and then determine and store an estimated prediction error of the ann in order to adjust short time stock price predictions in accordance with the stored prediction error. the prediction system is also designed to output upper and lower prediction bounds within a confidence region.",2004-05-11,"The title of the patent is artificial neural network based universal time series and its abstract is a neural network based universal time series prediction system for financial securities includes a pipelined recurrent ann architecutre having a plurality of identical modules to first adjust internal weights and biases in response to a first training set representing a nonlinear financial time series of samples of a financial quantity and a target value, and then determine and store an estimated prediction error of the ann in order to adjust short time stock price predictions in accordance with the stored prediction error. the prediction system is also designed to output upper and lower prediction bounds within a confidence region. dated 2004-05-11"
6735748,method and apparatus for performing extraction using a model trained with bayesian inference,"a machine-learning model may be created to perform integrated circuit layout extraction. using such a machine-learning system has two main phases: model creation and model application. the model creation phase comprises creating one or more extraction models using machine-learning techniques. the machine learning is performed by first creating training data sets composed of the identified parameters from typical examples of the smaller extraction problem and the answers to those example extraction problems as solved using a highly accurate physics-based field solver. next, the system performs machine learning using bayesian inference in order to train the neural network models. the bayesian inference may be implemented with normal monte carlo techniques, hybrid monte carlo techniques, or other bayesian learning techniques. after the creation of a set of models for each of the smaller simpler extraction problems, the machine-learning based models may be used for extraction.",2004-05-11,"The title of the patent is method and apparatus for performing extraction using a model trained with bayesian inference and its abstract is a machine-learning model may be created to perform integrated circuit layout extraction. using such a machine-learning system has two main phases: model creation and model application. the model creation phase comprises creating one or more extraction models using machine-learning techniques. the machine learning is performed by first creating training data sets composed of the identified parameters from typical examples of the smaller extraction problem and the answers to those example extraction problems as solved using a highly accurate physics-based field solver. next, the system performs machine learning using bayesian inference in order to train the neural network models. the bayesian inference may be implemented with normal monte carlo techniques, hybrid monte carlo techniques, or other bayesian learning techniques. after the creation of a set of models for each of the smaller simpler extraction problems, the machine-learning based models may be used for extraction. dated 2004-05-11"
6736089,method and system for sootblowing optimization,"a controller determines and adjusts system parameters, including cleanliness levels or sootblower operating settings, that are useful for maintaining the cleanliness of a fossil fuel boiler at an efficient level. some embodiments use a direct controller to determine cleanliness levels and/or sootblower operating settings. some embodiments use an indirect controller, with a system model, to determine cleanliness levels and/or sootblower settings. the controller may use a model that is, for example, a neural network, or a mass energy balance, or a genetically programmed model. the controller uses input about the actual performance or slate of the boiler for adaptation. the controller may operate in conjunction with a sootblower optimization system that controls the actual settings of the sootblowers. the controller may coordinate cleanliness settings for multiple sootblowers and/or across a plurality of heat zones in the boiler.",2004-05-18,"The title of the patent is method and system for sootblowing optimization and its abstract is a controller determines and adjusts system parameters, including cleanliness levels or sootblower operating settings, that are useful for maintaining the cleanliness of a fossil fuel boiler at an efficient level. some embodiments use a direct controller to determine cleanliness levels and/or sootblower operating settings. some embodiments use an indirect controller, with a system model, to determine cleanliness levels and/or sootblower settings. the controller may use a model that is, for example, a neural network, or a mass energy balance, or a genetically programmed model. the controller uses input about the actual performance or slate of the boiler for adaptation. the controller may operate in conjunction with a sootblower optimization system that controls the actual settings of the sootblowers. the controller may coordinate cleanliness settings for multiple sootblowers and/or across a plurality of heat zones in the boiler. dated 2004-05-18"
6738500,method and system for detecting small structures in images,"the invention is a method and apparatus for automated detection of small structures in images. one specific use is to detect malignant microcalcification clusters in mammograms. a digitized and filtered mammogram image is stored in a computer. seed pixels, which are pixels that are brighter than their immediate neighbors, are identified to indicate candidate structures and used to construct two regions. various features are then measured using the two regions around each seed point. the features characterize each candidate structure and are input to a classifier, such as a neural network. the classifier then distinguishes between structures of interest and background. the structures detected by the classifier are then presented to a clustering algorithm. a detected structure that is less than a threshold distance away from the nearest structure and a cluster is included in that cluster. finally, the results are displayed, either on a monitor or on hard copy, with a frame around the detected cluster.",2004-05-18,"The title of the patent is method and system for detecting small structures in images and its abstract is the invention is a method and apparatus for automated detection of small structures in images. one specific use is to detect malignant microcalcification clusters in mammograms. a digitized and filtered mammogram image is stored in a computer. seed pixels, which are pixels that are brighter than their immediate neighbors, are identified to indicate candidate structures and used to construct two regions. various features are then measured using the two regions around each seed point. the features characterize each candidate structure and are input to a classifier, such as a neural network. the classifier then distinguishes between structures of interest and background. the structures detected by the classifier are then presented to a clustering algorithm. a detected structure that is less than a threshold distance away from the nearest structure and a cluster is included in that cluster. finally, the results are displayed, either on a monitor or on hard copy, with a frame around the detected cluster. dated 2004-05-18"
6738760,method and system for providing electronic discovery on computer databases and archives using artificial intelligence to recover legally relevant data,"a method for providing electronic discovery on computer systems and archives is provided by using artificial intelligence to produce smart search agents to retrieve relevant data, particularly legally relevant documents. information relevant to desired data related to an issue is input into a neural network to train said neural network to produce search algorithms in the form of smart search agent. the smart search agents are released onto target computer systems and/or archives to search for responsive data and documents. notification, reports, and indexing of responsive data and documents can be provided to produce relevant results or prevent the production of relevant results.",2004-05-18,"The title of the patent is method and system for providing electronic discovery on computer databases and archives using artificial intelligence to recover legally relevant data and its abstract is a method for providing electronic discovery on computer systems and archives is provided by using artificial intelligence to produce smart search agents to retrieve relevant data, particularly legally relevant documents. information relevant to desired data related to an issue is input into a neural network to train said neural network to produce search algorithms in the form of smart search agent. the smart search agents are released onto target computer systems and/or archives to search for responsive data and documents. notification, reports, and indexing of responsive data and documents can be provided to produce relevant results or prevent the production of relevant results. dated 2004-05-18"
6741568,use of adaptive resonance theory (art) neural networks to compute bottleneck link speed in heterogeneous networking environments,"bottleneck link speed, or the transmission speed of the slowest link within a path between two nodes, is determining by transmitting a sequence of icmp echo data packets from the source node to the target node at a selected interval and measuring the return data packet intervals. rather than using statistical analysis methods, the return data packet interval measurements are input into an adaptive resonance theory neural network trained with the expected interval for every known, existing network transmission speed. the neural network will then classify the return data packet interval measurements, indicating the bottleneck link speed. since most of the computation&#8212;that required to train the neural network&#8212;may be performed before the data packet interval measurements are made rather than after, the bottleneck link speed may be determined from the return data packet interval measurements significantly faster and using less computational resources than with statistical analysis techniques. moreover, fewer measurements are required to determine bottleneck link speed to the same degree of accuracy.",2004-05-25,"The title of the patent is use of adaptive resonance theory (art) neural networks to compute bottleneck link speed in heterogeneous networking environments and its abstract is bottleneck link speed, or the transmission speed of the slowest link within a path between two nodes, is determining by transmitting a sequence of icmp echo data packets from the source node to the target node at a selected interval and measuring the return data packet intervals. rather than using statistical analysis methods, the return data packet interval measurements are input into an adaptive resonance theory neural network trained with the expected interval for every known, existing network transmission speed. the neural network will then classify the return data packet interval measurements, indicating the bottleneck link speed. since most of the computation&#8212;that required to train the neural network&#8212;may be performed before the data packet interval measurements are made rather than after, the bottleneck link speed may be determined from the return data packet interval measurements significantly faster and using less computational resources than with statistical analysis techniques. moreover, fewer measurements are required to determine bottleneck link speed to the same degree of accuracy. dated 2004-05-25"
6741956,analog computation using hybridization-capable oligomers,"the present invention is directed to an analog, oligomer-based method for determining a mathematical result of carrying out an operation of matrix algebra on input data. the method comprises representing at least one m-component vector v=&sgr;iviei by a set of single-stranded oligomers ei and ei which are in 1:1 correspondence with the basis vectors ei, i=1, 2, . . . , m in an abstract m-dimensional vector space. a composition comprising at least one set of oligomers ei and ei representing the components of a vector is obtained as input date and is subjected to at least one physical or chemical treatment having an effect on the oligomers that is an analog representation of an operation of matrix algebra. the method can be used to represent the operations of a neural network; for example, to produce a content-addressable memory, or a multilayer perceptron.",2004-05-25,"The title of the patent is analog computation using hybridization-capable oligomers and its abstract is the present invention is directed to an analog, oligomer-based method for determining a mathematical result of carrying out an operation of matrix algebra on input data. the method comprises representing at least one m-component vector v=&sgr;iviei by a set of single-stranded oligomers ei and ei which are in 1:1 correspondence with the basis vectors ei, i=1, 2, . . . , m in an abstract m-dimensional vector space. a composition comprising at least one set of oligomers ei and ei representing the components of a vector is obtained as input date and is subjected to at least one physical or chemical treatment having an effect on the oligomers that is an analog representation of an operation of matrix algebra. the method can be used to represent the operations of a neural network; for example, to produce a content-addressable memory, or a multilayer perceptron. dated 2004-05-25"
6745158,method and device for determining the layer thickness distribution in a paint layer,"a method for determining a layer thickness distribution in a paint layer produced during paint spraying after inputting specific spraying parameters into an electrostatically based paint spraying device. a data processing device sets up and uses a phenomenological mathematical model of a quasi-stationary three-dimensional spray pattern. specific parameters, such as an angle of rotation of electrodes and a rate of movement of the spraying device are input into the phenomenological model as fixed input parameters. in addition, real physical input parameters such as paint volume, directing air data and a voltage value, whose influence on the spraying result is not accurately known, are fed to an artificial neural network. the neural network having been previously trained using real input data such as a configuration of the spraying device, a paint type, operating parameters, and measured values of the layer thickness distribution. the neural network carries out a conversion of the input parameters into model input parameters which are fed to the phenomenological model. spray patterns formed by the phenomenological model are integrated in a further functional unit as a function of movement data of the spraying device which are contained in the input parameters to form the overall paint layer which is output.",2004-06-01,"The title of the patent is method and device for determining the layer thickness distribution in a paint layer and its abstract is a method for determining a layer thickness distribution in a paint layer produced during paint spraying after inputting specific spraying parameters into an electrostatically based paint spraying device. a data processing device sets up and uses a phenomenological mathematical model of a quasi-stationary three-dimensional spray pattern. specific parameters, such as an angle of rotation of electrodes and a rate of movement of the spraying device are input into the phenomenological model as fixed input parameters. in addition, real physical input parameters such as paint volume, directing air data and a voltage value, whose influence on the spraying result is not accurately known, are fed to an artificial neural network. the neural network having been previously trained using real input data such as a configuration of the spraying device, a paint type, operating parameters, and measured values of the layer thickness distribution. the neural network carries out a conversion of the input parameters into model input parameters which are fed to the phenomenological model. spray patterns formed by the phenomenological model are integrated in a further functional unit as a function of movement data of the spraying device which are contained in the input parameters to form the overall paint layer which is output. dated 2004-06-01"
6745169,learning process for a neural network,"a learning process for a neural network for open-loop or closed-loop control of an industrial process with time-variable parameters. the neural network is configured either as an open-loop or closed-loop-control network with which the process is controlled. the neural network is trained with the current process data so that it builds a model of the current process. the neural network can also be configured as a background network which is trained during operation with representative process data so that it builds an averaged model of the process over a longer period of time. after a certain learning time or upon the occurrence of an external event, the open-loop or closed-control network is replaced by the background network.",2004-06-01,"The title of the patent is learning process for a neural network and its abstract is a learning process for a neural network for open-loop or closed-loop control of an industrial process with time-variable parameters. the neural network is configured either as an open-loop or closed-loop-control network with which the process is controlled. the neural network is trained with the current process data so that it builds a model of the current process. the neural network can also be configured as a background network which is trained during operation with representative process data so that it builds an averaged model of the process over a longer period of time. after a certain learning time or upon the occurrence of an external event, the open-loop or closed-control network is replaced by the background network. dated 2004-06-01"
6746960,electronic techniques for analyte detection,"techniques are used to detect and identify analytes. techniques are used to fabricate and manufacture sensors to detect analytes. an analyte (1810) is sensed by sensors (1820) that output electrical signals in response to the analyte. the electrical signals are preprocessed (1830) by filtering and amplification. in an embodiment, this preprocessing includes adapting the sensor and electronics to the environment in which the analyte exists. the electrical signals are further processed (1840) to classify and identify the analyte, which may be by a neural network.",2004-06-08,"The title of the patent is electronic techniques for analyte detection and its abstract is techniques are used to detect and identify analytes. techniques are used to fabricate and manufacture sensors to detect analytes. an analyte (1810) is sensed by sensors (1820) that output electrical signals in response to the analyte. the electrical signals are preprocessed (1830) by filtering and amplification. in an embodiment, this preprocessing includes adapting the sensor and electronics to the environment in which the analyte exists. the electrical signals are further processed (1840) to classify and identify the analyte, which may be by a neural network. dated 2004-06-08"
6748369,method and system for automated property valuation,"a method and system for automating a process for valuing a property that produces an estimated value of a subject property, and a reliability assessment of the estimated value. the process is a generative artificial intelligence method that trains a fuzzy-neural network using a subset of cases from a case-base, and produces a run-time system to provide an estimate of the subject property's value. a network-based implementation of fuzzy inference is based on a system that implements a fuzzy system as a five-layer neural network so that the structure of the network can be interpreted in terms of high-level rules. the neural network is trained automatically from data. if/then rules are used to map inputs to outputs by a fuzzy logic inference system. different models for the same problem can be obtained by changing the inputs to the neuro-fuzzy network, or by varying its architecture.",2004-06-08,"The title of the patent is method and system for automated property valuation and its abstract is a method and system for automating a process for valuing a property that produces an estimated value of a subject property, and a reliability assessment of the estimated value. the process is a generative artificial intelligence method that trains a fuzzy-neural network using a subset of cases from a case-base, and produces a run-time system to provide an estimate of the subject property's value. a network-based implementation of fuzzy inference is based on a system that implements a fuzzy system as a five-layer neural network so that the structure of the network can be interpreted in terms of high-level rules. the neural network is trained automatically from data. if/then rules are used to map inputs to outputs by a fuzzy logic inference system. different models for the same problem can be obtained by changing the inputs to the neuro-fuzzy network, or by varying its architecture. dated 2004-06-08"
6751529,system and method for controlling model aircraft,"in one embodiment, a method for controlling an aircraft comprises providing an attitude error as a first input into a neural controller and an attitude rate as a second input into the neural controller. the attitude error is calculated from a commanded attitude and a current measured attitude, and the attitude rate is derived from the current measured attitude. the method also comprises processing the first input and the second input to generate a commanded servo actuator rate as an output of the neural controller. the method further comprises generating a commanded actuator position from the commanded servo actuator rate and a current servo position, and inputting the commanded actuator position to a servo motor configured to drive an attitude actuator to the commanded actuator position. the neural controller is developed from a neural network, wherein the neural network is designed without using conventional control laws, and the neural network is trained to eliminate the attitude error.",2004-06-15,"The title of the patent is system and method for controlling model aircraft and its abstract is in one embodiment, a method for controlling an aircraft comprises providing an attitude error as a first input into a neural controller and an attitude rate as a second input into the neural controller. the attitude error is calculated from a commanded attitude and a current measured attitude, and the attitude rate is derived from the current measured attitude. the method also comprises processing the first input and the second input to generate a commanded servo actuator rate as an output of the neural controller. the method further comprises generating a commanded actuator position from the commanded servo actuator rate and a current servo position, and inputting the commanded actuator position to a servo motor configured to drive an attitude actuator to the commanded actuator position. the neural controller is developed from a neural network, wherein the neural network is designed without using conventional control laws, and the neural network is trained to eliminate the attitude error. dated 2004-06-15"
6751602,neural net controller for noise and vibration reduction,"two neural networks are used to control adaptively a vibration and noise-producing plant. the first neural network, the emulator, models the complex, nonlinear output of the plant with respect to certain controls and stimuli applied to the plant. the second neural network, the controller, calculates a control signal which affects the vibration and noise producing characteristics of the plant. by using the emulator model to calculate the nonlinear plant gradient, the controller matrix coefficients can be adapted by backpropagation of the plant gradient to produce a control signal which results in the minimum vibration and noise possible, given the current operating characteristics of the plant.",2004-06-15,"The title of the patent is neural net controller for noise and vibration reduction and its abstract is two neural networks are used to control adaptively a vibration and noise-producing plant. the first neural network, the emulator, models the complex, nonlinear output of the plant with respect to certain controls and stimuli applied to the plant. the second neural network, the controller, calculates a control signal which affects the vibration and noise producing characteristics of the plant. by using the emulator model to calculate the nonlinear plant gradient, the controller matrix coefficients can be adapted by backpropagation of the plant gradient to produce a control signal which results in the minimum vibration and noise possible, given the current operating characteristics of the plant. dated 2004-06-15"
6754380,method of training massive training artificial neural networks (mtann) for the detection of abnormalities in medical images,"a method, system, and computer program product of selecting a set of training images for a massive training artificial neural network (mtann). the method comprises selecting the set of training images from a set of domain images; training the mtann with the set of training images; applying a plurality of images from the set of domain images to the trained mtann to obtain a corresponding plurality of scores; and determining the set of training images based on the plurality of images, the corresponding plurality of scores, and the set of training images. the method is useful for the reduction of false positives in computerized detection of abnormalities in medical images. in particular, the mtaan can be used for the detection of lung nodules in low-dose ct (ldct). the mtann consists of a modified multilayer artificial neural network capable of operating on image data directly.",2004-06-22,"The title of the patent is method of training massive training artificial neural networks (mtann) for the detection of abnormalities in medical images and its abstract is a method, system, and computer program product of selecting a set of training images for a massive training artificial neural network (mtann). the method comprises selecting the set of training images from a set of domain images; training the mtann with the set of training images; applying a plurality of images from the set of domain images to the trained mtann to obtain a corresponding plurality of scores; and determining the set of training images based on the plurality of images, the corresponding plurality of scores, and the set of training images. the method is useful for the reduction of false positives in computerized detection of abnormalities in medical images. in particular, the mtaan can be used for the detection of lung nodules in low-dose ct (ldct). the mtann consists of a modified multilayer artificial neural network capable of operating on image data directly. dated 2004-06-22"
6754589,system and method for enhanced hydrocarbon recovery,"a neural network based system, method, and process for the automated delineation of spatially dependent objects is disclosed. the method is applicable to objects such as hydrocarbon accumulations, aeromagnetic profiles, astronomical clusters, weather clusters, objects from radar, sonar, seismic and infrared returns, etc. one of the novelties in the present invention is that the method can be utilized whether or not known data is available to provide traditional training sets. the output consists of a classification of the input data into clearly delineated accumulations, clusters, objects, etc. that have various types and properties. a preferred but non-exclusive application of the present invention is the automated delineation of hydrocarbon accumulations and sub-regions within the accumulations with various properties, in an oil and gas field, prior to the commencement of drilling operations. the invention may also be used to increase the effectiveness of enhanced oil recovery techniques.",2004-06-22,"The title of the patent is system and method for enhanced hydrocarbon recovery and its abstract is a neural network based system, method, and process for the automated delineation of spatially dependent objects is disclosed. the method is applicable to objects such as hydrocarbon accumulations, aeromagnetic profiles, astronomical clusters, weather clusters, objects from radar, sonar, seismic and infrared returns, etc. one of the novelties in the present invention is that the method can be utilized whether or not known data is available to provide traditional training sets. the output consists of a classification of the input data into clearly delineated accumulations, clusters, objects, etc. that have various types and properties. a preferred but non-exclusive application of the present invention is the automated delineation of hydrocarbon accumulations and sub-regions within the accumulations with various properties, in an oil and gas field, prior to the commencement of drilling operations. the invention may also be used to increase the effectiveness of enhanced oil recovery techniques. dated 2004-06-22"
6757570,system and method for adaptive control of uncertain nonlinear processes,"a process and neural network architecture for on-line adjustment of the weights of the neural network in a manner that corrects errors made by a nonlinear controller designed based on a model for the dynamics of a process under control. a computer system is provided for controlling the dynamic output response signal of a nonlinear physical process, where the physical process is represented by a fixed model of the process. the computer system includes a controlled device for responding to the output response signal of the system. the computer system also includes a linear controller for providing a pseudo control signal that is based on the fixed model for the process and provides a second controller, connected to the linear controller, for receiving the pseudo control signal and for providing a modified pseudo control signal to correct for the errors made in modeling the nonlinearities in the process. a response network is also included as part of the computer system. the response network receives the modified pseudo control signal and provides the output response signal to the controlled device. the second controller preferably is a neural network. the computer system may include a plurality of neural networks with each neural network designated to control a selected variable or degree of freedom within the system.",2004-06-29,"The title of the patent is system and method for adaptive control of uncertain nonlinear processes and its abstract is a process and neural network architecture for on-line adjustment of the weights of the neural network in a manner that corrects errors made by a nonlinear controller designed based on a model for the dynamics of a process under control. a computer system is provided for controlling the dynamic output response signal of a nonlinear physical process, where the physical process is represented by a fixed model of the process. the computer system includes a controlled device for responding to the output response signal of the system. the computer system also includes a linear controller for providing a pseudo control signal that is based on the fixed model for the process and provides a second controller, connected to the linear controller, for receiving the pseudo control signal and for providing a modified pseudo control signal to correct for the errors made in modeling the nonlinearities in the process. a response network is also included as part of the computer system. the response network receives the modified pseudo control signal and provides the output response signal to the controlled device. the second controller preferably is a neural network. the computer system may include a plurality of neural networks with each neural network designated to control a selected variable or degree of freedom within the system. dated 2004-06-29"
6757665,detection of pump cavitation/blockage and seal failure via current signature analysis,"a system and method is provided for monitoring the operating condition of a pump by evaluating fault data encoded in the instantaneous current of the motor driving the pump. the data is converted to a frequency spectrum which is analyzed to create a fault signature having fault attributes relating to various fault conditions associated with the pump. the fault signature is then input to a neural network that operates in conjunction with a preprocessing and post processing module to perform decisions and output those decisions to a user interface. a stand alone module is also provided that includes an adaptive preprocessing module, a one-shot unsupervised neural network and a fuzzy based expert system to provide a decision making module that operates with limited human supervision.",2004-06-29,"The title of the patent is detection of pump cavitation/blockage and seal failure via current signature analysis and its abstract is a system and method is provided for monitoring the operating condition of a pump by evaluating fault data encoded in the instantaneous current of the motor driving the pump. the data is converted to a frequency spectrum which is analyzed to create a fault signature having fault attributes relating to various fault conditions associated with the pump. the fault signature is then input to a neural network that operates in conjunction with a preprocessing and post processing module to perform decisions and output those decisions to a user interface. a stand alone module is also provided that includes an adaptive preprocessing module, a one-shot unsupervised neural network and a fuzzy based expert system to provide a decision making module that operates with limited human supervision. dated 2004-06-29"
6757666,locally connected neural network with improved feature vector,a pattern recognizer which uses neuromorphs with a fixed amount of energy that is distributed among the elements. the distribution of the energy is used to form a histogram which is used as a feature vector.,2004-06-29,The title of the patent is locally connected neural network with improved feature vector and its abstract is a pattern recognizer which uses neuromorphs with a fixed amount of energy that is distributed among the elements. the distribution of the energy is used to form a histogram which is used as a feature vector. dated 2004-06-29
6758277,system and method for fluid flow optimization,"a controllable gas-lift well having controllable gas-lift valves and sensors for detecting flow regime is provided. the well uses production tubing and casing to communicate with and power the controllable valve from the surface. a signal impedance apparatus in the form of induction chokes at the surface and downhole electrically isolate the tubing from the casing. a high band-width, adaptable spread spectrum communication system is used to communicate between the controllable valve and the surface. sensors, such as pressure, temperature, and acoustic sensors, may be provided downhole to more accurately assess downhole conditions and in particular, the flow regime of the fluid within the tubing. operating conditions, such as gas injection rate, back pressure on the tubing, and position of downhole controllable valves are varied depending on flow regime, downhole conditions, oil production, gas usage and availability, to optimize production. an artificial neural network (ann) is trained to detect a taylor flow regime using downhole acoustic sensors, plus other sensors as desired. the detection and control system and method thereof is useful in many applications involving multi-phase flow in a conduit.",2004-07-06,"The title of the patent is system and method for fluid flow optimization and its abstract is a controllable gas-lift well having controllable gas-lift valves and sensors for detecting flow regime is provided. the well uses production tubing and casing to communicate with and power the controllable valve from the surface. a signal impedance apparatus in the form of induction chokes at the surface and downhole electrically isolate the tubing from the casing. a high band-width, adaptable spread spectrum communication system is used to communicate between the controllable valve and the surface. sensors, such as pressure, temperature, and acoustic sensors, may be provided downhole to more accurately assess downhole conditions and in particular, the flow regime of the fluid within the tubing. operating conditions, such as gas injection rate, back pressure on the tubing, and position of downhole controllable valves are varied depending on flow regime, downhole conditions, oil production, gas usage and availability, to optimize production. an artificial neural network (ann) is trained to detect a taylor flow regime using downhole acoustic sensors, plus other sensors as desired. the detection and control system and method thereof is useful in many applications involving multi-phase flow in a conduit. dated 2004-07-06"
6760061,traffic sensor,a traffic sensor system for detecting and tracking vehicles is describede. a video camera is employed to obtain a video image of a section of a roadway. motion is detected through changes in luminance and edges in frames of the video image. predetermined sets of pixels (&#8220;tiles&#8221;) in the frames are designated to be in either an &#8220;active&#8221; state or an &#8220;inactive&#8221; state. a tile becomes active when the luminance or edge values of the pixels of the tile differ from the respective luminance or edge values of a corresponding tile in a reference frame in accordance with predetermined criteria. the tile becomes inactive when the luminance or edge values of the pixels of the tile do not differ from the corresponding reference frame tile in accordance with the is predetermined criteria. shape and motion of groups of active tiles (&#8220;quanta&#8221;) are analyzed with software and a neural network to detect and track vehicles.,2004-07-06,The title of the patent is traffic sensor and its abstract is a traffic sensor system for detecting and tracking vehicles is describede. a video camera is employed to obtain a video image of a section of a roadway. motion is detected through changes in luminance and edges in frames of the video image. predetermined sets of pixels (&#8220;tiles&#8221;) in the frames are designated to be in either an &#8220;active&#8221; state or an &#8220;inactive&#8221; state. a tile becomes active when the luminance or edge values of the pixels of the tile differ from the respective luminance or edge values of a corresponding tile in a reference frame in accordance with predetermined criteria. the tile becomes inactive when the luminance or edge values of the pixels of the tile do not differ from the corresponding reference frame tile in accordance with the is predetermined criteria. shape and motion of groups of active tiles (&#8220;quanta&#8221;) are analyzed with software and a neural network to detect and track vehicles. dated 2004-07-06
6760468,method and system for the detection of lung nodule in radiological images using digital image processing and artificial neural network,"a method and system improve the detection of abnormalities, such as lung nodules, in radiological images using digital image processing and artificial neural network techniques. the detection method and system use a nodule phantom for matching in order to enhance the efficiency in detection. the detection method and system use spherical parameters to characterize true nodules, thus enabling detection of the nodules in the mediastinum. the detection method and system use a multi-layer back-propagation neural network architecture not only for the classification of lung nodules but also for the integration of detection results from different classifiers. in addition, this method and system improve the detection efficiency by recommending the ranking of true nodules and several false positive nodules prior to the training of the neural network classifier. the method and system use image segmentation to remove regions outside the chest in order to reduce the false positives outside the chest region.",2004-07-06,"The title of the patent is method and system for the detection of lung nodule in radiological images using digital image processing and artificial neural network and its abstract is a method and system improve the detection of abnormalities, such as lung nodules, in radiological images using digital image processing and artificial neural network techniques. the detection method and system use a nodule phantom for matching in order to enhance the efficiency in detection. the detection method and system use spherical parameters to characterize true nodules, thus enabling detection of the nodules in the mediastinum. the detection method and system use a multi-layer back-propagation neural network architecture not only for the classification of lung nodules but also for the integration of detection results from different classifiers. in addition, this method and system improve the detection efficiency by recommending the ranking of true nodules and several false positive nodules prior to the training of the neural network classifier. the method and system use image segmentation to remove regions outside the chest in order to reduce the false positives outside the chest region. dated 2004-07-06"
6760714,representation and retrieval of images using content vectors derived from image information elements,"image features are generated by performing wavelet transformations at sample points on images stored in electronic form. multiple wavelet transformations at a point are combined to form an image feature vector. a prototypical set of feature vectors, or atoms, is derived from the set of feature vectors to form an &#8220;atomic vocabulary.&#8221; the prototypical feature vectors are derived using a vector quantization method (e.g., using neural network self-organization techniques) in which a vector quantization network is also generated. the atomic vocabulary is used to define new images. meaning is established between atoms in the atomic vocabulary. high-dimensional context vectors are assigned to each atom. the context vectors are then trained as a function of the proximity and co-occurrence of each atom to other atoms in the image. after training, the context vectors associated with the atoms that comprise an image are combined to form a summary vector for the image. images are retrieved using a number of query methods (e.g., images, image portions, vocabulary atoms, index terms). the user's query is converted into a query context vector. a dot product is calculated between the query vector and the summary vectors to locate images having the closest meaning. the invention is also applicable to video or temporally related images, and can also be used in conjunction with other context vector data domains such as text or audio, thereby linking images to such data domains.",2004-07-06,"The title of the patent is representation and retrieval of images using content vectors derived from image information elements and its abstract is image features are generated by performing wavelet transformations at sample points on images stored in electronic form. multiple wavelet transformations at a point are combined to form an image feature vector. a prototypical set of feature vectors, or atoms, is derived from the set of feature vectors to form an &#8220;atomic vocabulary.&#8221; the prototypical feature vectors are derived using a vector quantization method (e.g., using neural network self-organization techniques) in which a vector quantization network is also generated. the atomic vocabulary is used to define new images. meaning is established between atoms in the atomic vocabulary. high-dimensional context vectors are assigned to each atom. the context vectors are then trained as a function of the proximity and co-occurrence of each atom to other atoms in the image. after training, the context vectors associated with the atoms that comprise an image are combined to form a summary vector for the image. images are retrieved using a number of query methods (e.g., images, image portions, vocabulary atoms, index terms). the user's query is converted into a query context vector. a dot product is calculated between the query vector and the summary vectors to locate images having the closest meaning. the invention is also applicable to video or temporally related images, and can also be used in conjunction with other context vector data domains such as text or audio, thereby linking images to such data domains. dated 2004-07-06"
6760716,adaptive predictive model in a process control system,"an adaptive predictive model includes a standard predictive model, such as a neural network or a natural model, constructed to produce an output that predicts a process parameter and a combiner network that combines the output of the predictive model with one or more measured values of the process parameter to produce an adjusted predicted process parameter during operation of a process. the adaptive predictive model reduces or corrects for non-linear as well as linear errors in the prediction of a process variable without having to reform the predictive model itself, while requiring only minor increases in processing power and time.",2004-07-06,"The title of the patent is adaptive predictive model in a process control system and its abstract is an adaptive predictive model includes a standard predictive model, such as a neural network or a natural model, constructed to produce an output that predicts a process parameter and a combiner network that combines the output of the predictive model with one or more measured values of the process parameter to produce an adjusted predicted process parameter during operation of a process. the adaptive predictive model reduces or corrects for non-linear as well as linear errors in the prediction of a process variable without having to reform the predictive model itself, while requiring only minor increases in processing power and time. dated 2004-07-06"
6763303,system for classifying seafloor roughness,"the present invention relates to a novel system for seafloor classification using artificial neural network (ann) hybrid layout with the use of unprocessed multi-beam backscatter data and more importantly, this invention relates to an real-time seafloor roughness classifier using backscatter data after training the self-organized mapping (som) network and learning vector quantization (lvq) network wherein, the system of the present invention has the unique capability for the combined use of unsupervised som followed by supervised lvq to achieve a highly improved performance in the said roughness classification, which was hitherto non-existent and has the additional capability for the use of a combination of the two variants of the lvq layout to work together to achieve the best results in seafloor roughness classification.",2004-07-13,"The title of the patent is system for classifying seafloor roughness and its abstract is the present invention relates to a novel system for seafloor classification using artificial neural network (ann) hybrid layout with the use of unprocessed multi-beam backscatter data and more importantly, this invention relates to an real-time seafloor roughness classifier using backscatter data after training the self-organized mapping (som) network and learning vector quantization (lvq) network wherein, the system of the present invention has the unique capability for the combined use of unsupervised som followed by supervised lvq to achieve a highly improved performance in the said roughness classification, which was hitherto non-existent and has the additional capability for the use of a combination of the two variants of the lvq layout to work together to achieve the best results in seafloor roughness classification. dated 2004-07-13"
6763338,machine decisions based on preferential voting techniques,"a method and apparatus for computing an overall or aggregate decision based on intermediate decisions as to which of a set of alternatives best characterize an object. the alternatives are partitioned into at least two series of preferences corresponding to at least two intermediate rankings. various embodiments may base the intermediate rankings on: a machine learning technique; a decision tree; a belief network; a neural network; a static model; a program; or an evolutionary training method. based on the preferences, a weak alternative is selected and removed from the series. the selection of the weak alternative may include identifying which alternatives lose pairwise to the other alternatives, are excluded from the first preferences, are included in the last preferences, or have a lowest average preference ranking. the selecting and removing continue until the remaining alternatives are the aggregate decision. various embodiments may be applied to classification problems, prediction problems or selection problems.",2004-07-13,"The title of the patent is machine decisions based on preferential voting techniques and its abstract is a method and apparatus for computing an overall or aggregate decision based on intermediate decisions as to which of a set of alternatives best characterize an object. the alternatives are partitioned into at least two series of preferences corresponding to at least two intermediate rankings. various embodiments may base the intermediate rankings on: a machine learning technique; a decision tree; a belief network; a neural network; a static model; a program; or an evolutionary training method. based on the preferences, a weak alternative is selected and removed from the series. the selection of the weak alternative may include identifying which alternatives lose pairwise to the other alternatives, are excluded from the first preferences, are included in the last preferences, or have a lowest average preference ranking. the selecting and removing continue until the remaining alternatives are the aggregate decision. various embodiments may be applied to classification problems, prediction problems or selection problems. dated 2004-07-13"
6763339,biologically-based signal processing system applied to noise removal for signal extraction,the method and system described herein use a biologically-based signal processing system for noise removal for signal extraction. a wavelet transform may be used in conjunction with a neural network to imitate a biological system. the neural network may be trained using ideal data derived from physical principles or noiseless signals to determine to remove noise from the signal.,2004-07-13,The title of the patent is biologically-based signal processing system applied to noise removal for signal extraction and its abstract is the method and system described herein use a biologically-based signal processing system for noise removal for signal extraction. a wavelet transform may be used in conjunction with a neural network to imitate a biological system. the neural network may be trained using ideal data derived from physical principles or noiseless signals to determine to remove noise from the signal. dated 2004-07-13
6763340,microelectromechanical system artificial neural network device,"a novel microelectromechanical system artificial neural network (mems ann) device performs the function of a conventional artificial neural network node element. micro-machined polysilicon or high aspect ratio composite beam micro-resonators replace as computational elements the silicon transistors and software simulations of prior-art anns. the basic memsann device forms a non-linear (e.g., sigmoid) function of a sum of products. products of the magnitudes of sine waves, applied to the input drive comb and shuttle magnitudes, are formed in the frequency domain and summed by coupling a plurality of resonators with a mechanical coupling frame, or by integrating them into one resonator. a sigmoid function is applied to the sum of products by shaping the overlap capacitance of the output comb fingers of the resonator. methods of building and using various single mems ann devices and multi-layered arrays of mems ann circuits are also described. these novel mems anns exhibit an attractive combination of performance characteristics, compared to conventional hardware anns that use silicon transistors or simulations of anns running in software on digital computers, including lower cost, simpler design, wider temperature range, greater radiation tolerance, and lower operating and standby power. these advantages favorably impact system weight and size because of reduced shielding, cooling, and power requirements.",2004-07-13,"The title of the patent is microelectromechanical system artificial neural network device and its abstract is a novel microelectromechanical system artificial neural network (mems ann) device performs the function of a conventional artificial neural network node element. micro-machined polysilicon or high aspect ratio composite beam micro-resonators replace as computational elements the silicon transistors and software simulations of prior-art anns. the basic memsann device forms a non-linear (e.g., sigmoid) function of a sum of products. products of the magnitudes of sine waves, applied to the input drive comb and shuttle magnitudes, are formed in the frequency domain and summed by coupling a plurality of resonators with a mechanical coupling frame, or by integrating them into one resonator. a sigmoid function is applied to the sum of products by shaping the overlap capacitance of the output comb fingers of the resonator. methods of building and using various single mems ann devices and multi-layered arrays of mems ann circuits are also described. these novel mems anns exhibit an attractive combination of performance characteristics, compared to conventional hardware anns that use silicon transistors or simulations of anns running in software on digital computers, including lower cost, simpler design, wider temperature range, greater radiation tolerance, and lower operating and standby power. these advantages favorably impact system weight and size because of reduced shielding, cooling, and power requirements. dated 2004-07-13"
6766316,method and system of ranking and clustering for document indexing and retrieval,"a relevancy ranking and clustering method and system that determines the relevance of a document relative to a user's query using a similarity comparison process. input queries are parsed into one or more query predicate structures using an ontological parser. the ontological parser parses a set of known documents to generate one or more document predicate structures. a comparison of each query predicate structure with each document predicate structure is performed to determine a matching degree, represented by a real number. a multilevel modifier strategy is implemented to assign different relevance values to the different parts of each predicate structure match to calculate the predicate structure's matching degree. the relevance of a document to a user's query is determined by calculating a similarity coefficient, based on the structures of each pair of query predicates and document predicates. documents are autonomously clustered using a self-organizing neural network that provides a coordinate system that makes judgments in a non-subjective fashion.",2004-07-20,"The title of the patent is method and system of ranking and clustering for document indexing and retrieval and its abstract is a relevancy ranking and clustering method and system that determines the relevance of a document relative to a user's query using a similarity comparison process. input queries are parsed into one or more query predicate structures using an ontological parser. the ontological parser parses a set of known documents to generate one or more document predicate structures. a comparison of each query predicate structure with each document predicate structure is performed to determine a matching degree, represented by a real number. a multilevel modifier strategy is implemented to assign different relevance values to the different parts of each predicate structure match to calculate the predicate structure's matching degree. the relevance of a document to a user's query is determined by calculating a similarity coefficient, based on the structures of each pair of query predicates and document predicates. documents are autonomously clustered using a self-organizing neural network that provides a coordinate system that makes judgments in a non-subjective fashion. dated 2004-07-20"
6769016,intelligent spam detection system using an updateable neural analysis engine,"a system, method and computer program product are provided for detecting an unwanted message. first, an electronic mail message is received. text in the electronic mail message is decomposed. statistics associated with the text are gathered using a statistical analyzer. a neural network engine coupled to the statistical analyzer is taught to recognize unwanted messages based on statistical indicators. the statistical indicators are analyzed utilizing the neural network engine for determining whether the electronic mail message is an unwanted message.",2004-07-27,"The title of the patent is intelligent spam detection system using an updateable neural analysis engine and its abstract is a system, method and computer program product are provided for detecting an unwanted message. first, an electronic mail message is received. text in the electronic mail message is decomposed. statistics associated with the text are gathered using a statistical analyzer. a neural network engine coupled to the statistical analyzer is taught to recognize unwanted messages based on statistical indicators. the statistical indicators are analyzed utilizing the neural network engine for determining whether the electronic mail message is an unwanted message. dated 2004-07-27"
6769066,method and apparatus for training a neural network model for use in computer network intrusion detection,detecting harmful or illegal intrusions into a computer network or into restricted portions of a computer network uses a process of synthesizing anomalous data to be used in training a neural network-based model for use in a computer network intrusion detection system. anomalous data for artificially creating a set of features reflecting anomalous behavior for a particular activity is performed. this is done in conjunction with the creation of normal-behavior feature values. a distribution of users of normal feature values and an expected distribution of users of anomalous feature values are then defined in the form of histograms. the anomalous-feature histogram is then sampled to produce anomalous-behavior feature values. these values are then used to train a model having a neural network training algorithm where the model is used in the computer network intrusion detection system. the model is trained such that it can efficiently recognize anomalous behavior by users in a dynamic computing environment where user behavior can change frequently.,2004-07-27,The title of the patent is method and apparatus for training a neural network model for use in computer network intrusion detection and its abstract is detecting harmful or illegal intrusions into a computer network or into restricted portions of a computer network uses a process of synthesizing anomalous data to be used in training a neural network-based model for use in a computer network intrusion detection system. anomalous data for artificially creating a set of features reflecting anomalous behavior for a particular activity is performed. this is done in conjunction with the creation of normal-behavior feature values. a distribution of users of normal feature values and an expected distribution of users of anomalous feature values are then defined in the form of histograms. the anomalous-feature histogram is then sampled to produce anomalous-behavior feature values. these values are then used to train a model having a neural network training algorithm where the model is used in the computer network intrusion detection system. the model is trained such that it can efficiently recognize anomalous behavior by users in a dynamic computing environment where user behavior can change frequently. dated 2004-07-27
6775564,non-invasive glucose measuring device and method for measuring blood glucose,"a glucose measuring device for determining the concentration of glucose in intravascular blood within a body part of a subject. the device includes at least one light source having a wavelength of 650, 880, 940 or 1300 nm to illuminate the fluid. at least one receptor (14) associated with the light source (12) for receiving light and generating a transmission signal representing the light transmitted is also provided. a support piece is including for supporting the light source associated with the respective receptor. the support piece is adapted to engage a body part of a subject. finally, a signal analyzer determines the glucose concentration in the blood of the subject. a method for determining the glucose concentration is also provided which calibrates a measuring device and sets the operating current for illuminating the light sources during operation of the device. once a transmission signal is generated by receptors (14) receiving light via the light source and illuminated blood, and the high and low values from each of the signals are selected and stored in the device (20), the values are subtracted to obtain a single transmission value for each of the light sources. these calculated values are then compared to a database of target transmission values, either using a neural network, or directly compared to determine the glucose concentration, which value is then displayed (28) on the device.",2004-08-10,"The title of the patent is non-invasive glucose measuring device and method for measuring blood glucose and its abstract is a glucose measuring device for determining the concentration of glucose in intravascular blood within a body part of a subject. the device includes at least one light source having a wavelength of 650, 880, 940 or 1300 nm to illuminate the fluid. at least one receptor (14) associated with the light source (12) for receiving light and generating a transmission signal representing the light transmitted is also provided. a support piece is including for supporting the light source associated with the respective receptor. the support piece is adapted to engage a body part of a subject. finally, a signal analyzer determines the glucose concentration in the blood of the subject. a method for determining the glucose concentration is also provided which calibrates a measuring device and sets the operating current for illuminating the light sources during operation of the device. once a transmission signal is generated by receptors (14) receiving light via the light source and illuminated blood, and the high and low values from each of the signals are selected and stored in the device (20), the values are subtracted to obtain a single transmission value for each of the light sources. these calculated values are then compared to a database of target transmission values, either using a neural network, or directly compared to determine the glucose concentration, which value is then displayed (28) on the device. dated 2004-08-10"
6775619,neural net prediction of seismic streamer shape,"a neural network to predict seismic streamer shape during seismic operations having an input layer, an optional hidden layer, and an output layer, each layer having one or more nodes. the first layer comprises input nodes attached to seismic data acquisition operational parameters as follows: vessel coordinates, receiver coordinates, time, vessel velocity, current velocity, wind velocity, water temperature, salinity, tidal information, water depth, streamer density, and streamer dimensions. each node in the input layer is connected to each node in the hidden layer and each node in the hidden layer is connected to each node in the output layer, the output layer outputting a predicted cable shape. the layer maybe omitted. when the hidden lay is omitted, each node in the input layer is attached to each node in the output layer.",2004-08-10,"The title of the patent is neural net prediction of seismic streamer shape and its abstract is a neural network to predict seismic streamer shape during seismic operations having an input layer, an optional hidden layer, and an output layer, each layer having one or more nodes. the first layer comprises input nodes attached to seismic data acquisition operational parameters as follows: vessel coordinates, receiver coordinates, time, vessel velocity, current velocity, wind velocity, water temperature, salinity, tidal information, water depth, streamer density, and streamer dimensions. each node in the input layer is connected to each node in the hidden layer and each node in the hidden layer is connected to each node in the output layer, the output layer outputting a predicted cable shape. the layer maybe omitted. when the hidden lay is omitted, each node in the input layer is attached to each node in the output layer. dated 2004-08-10"
6782373,method and circuits for associating a norm to each component of an input pattern presented to a neural network,"the method and circuits of the present invention aim to associate a norm to each component of an input pattern presented to an input space mapping algorithm based artificial neural network (ann) during the distance evaluation process. the set of norms, referred to as the &#8220;component&#8221; norms is memorized in specific memorization means in the ann. in a first embodiment, the ann is provided with a global memory, common for all the neurons of the ann, that memorizes all the component norms. for each component of the input pattern, all the neurons perform the elementary (or partial) distance calculation with the corresponding prototype components stored therein during the distance evaluation process using the associated component norm. the distance elementary calculations are then combined using a &#8220;distance&#8221; norm to determine the final distance between the input pattern and the prototypes stored in the neurons. in another embodiment, the set of component norms is memorized in the neurons themselves in the prototype memorization means, so that the global memory is no longer physically necessary. this implementation allows to significantly optimize the consumed silicon area when the ann is integrated in a silicon chip.",2004-08-24,"The title of the patent is method and circuits for associating a norm to each component of an input pattern presented to a neural network and its abstract is the method and circuits of the present invention aim to associate a norm to each component of an input pattern presented to an input space mapping algorithm based artificial neural network (ann) during the distance evaluation process. the set of norms, referred to as the &#8220;component&#8221; norms is memorized in specific memorization means in the ann. in a first embodiment, the ann is provided with a global memory, common for all the neurons of the ann, that memorizes all the component norms. for each component of the input pattern, all the neurons perform the elementary (or partial) distance calculation with the corresponding prototype components stored therein during the distance evaluation process using the associated component norm. the distance elementary calculations are then combined using a &#8220;distance&#8221; norm to determine the final distance between the input pattern and the prototypes stored in the neurons. in another embodiment, the set of component norms is memorized in the neurons themselves in the prototype memorization means, so that the global memory is no longer physically necessary. this implementation allows to significantly optimize the consumed silicon area when the ann is integrated in a silicon chip. dated 2004-08-24"
6782375,neural network based decision processor and method,a computer network-based customer acquisition server and method of selecting preferred products includes using a neural network-based decision engine that automatically generate queries and select preferred products as a function of responses to the queries.,2004-08-24,The title of the patent is neural network based decision processor and method and its abstract is a computer network-based customer acquisition server and method of selecting preferred products includes using a neural network-based decision engine that automatically generate queries and select preferred products as a function of responses to the queries. dated 2004-08-24
6783097,wing-drive mechanism and vehicle employing same,"a wing-drive mechanism is described that permits, with proper control, movement of a wing about multiple wing trajectories. the wing-drive is capable of independent movement about three rotational degrees of movement; movement about a flap axis is independent of movement about a yaw axis, and both are independent of changes in the pitch of the wing. methods of controlling the wing-drive mechanism to affect a desired wing trajectory include the use of a non-linear automated controller that generates input signals to the wing-drive mechanism by comparing actual and desired wing trajectories in real time. specification of wing trajectories is preferably also accomplished in real time using an automated trajectory specification system, which can include a fuzzy logic processor or a neural network.",2004-08-31,"The title of the patent is wing-drive mechanism and vehicle employing same and its abstract is a wing-drive mechanism is described that permits, with proper control, movement of a wing about multiple wing trajectories. the wing-drive is capable of independent movement about three rotational degrees of movement; movement about a flap axis is independent of movement about a yaw axis, and both are independent of changes in the pitch of the wing. methods of controlling the wing-drive mechanism to affect a desired wing trajectory include the use of a non-linear automated controller that generates input signals to the wing-drive mechanism by comparing actual and desired wing trajectories in real time. specification of wing trajectories is preferably also accomplished in real time using an automated trajectory specification system, which can include a fuzzy logic processor or a neural network. dated 2004-08-31"
6785736,method and system for optimizing the network path of mobile programs,"a method and system for computing the shortest path for traveling inside a network and visiting a predefined list of network addresses. the method can be used by a system management workstation communicating and a mobile program visiting the list of networks addresses; the system management workstation communicates with said mobile program to get the list of network addresses to be visited; the system management workstation communicates also with all the networks addresses of the list to get the parameter values for determining the shortest path. the system management platform computes the shortest path by running a kohonen neural network reading in input the references to nodes and their,parameter values which form bi dimensional coordinates; the output is the ordered list of network addresses to be visited by the mobile program.",2004-08-31,"The title of the patent is method and system for optimizing the network path of mobile programs and its abstract is a method and system for computing the shortest path for traveling inside a network and visiting a predefined list of network addresses. the method can be used by a system management workstation communicating and a mobile program visiting the list of networks addresses; the system management workstation communicates with said mobile program to get the list of network addresses to be visited; the system management workstation communicates also with all the networks addresses of the list to get the parameter values for determining the shortest path. the system management platform computes the shortest path by running a kohonen neural network reading in input the references to nodes and their,parameter values which form bi dimensional coordinates; the output is the ordered list of network addresses to be visited by the mobile program. dated 2004-08-31"
6786635,turbine blade (bucket) health monitoring and prognosis using neural network based diagnostic techniques in conjunction with pyrometer signals,oxidation of turbine buckets may cause unexpected and expensive turbine failures. turbine bucket oxidation condition may be estimated to predict remaining useful bucket life during operation of a turbine by processing time-varying temperature distributions measured with a pyrometer of at least one rotating turbine bucket.,2004-09-07,The title of the patent is turbine blade (bucket) health monitoring and prognosis using neural network based diagnostic techniques in conjunction with pyrometer signals and its abstract is oxidation of turbine buckets may cause unexpected and expensive turbine failures. turbine bucket oxidation condition may be estimated to predict remaining useful bucket life during operation of a turbine by processing time-varying temperature distributions measured with a pyrometer of at least one rotating turbine bucket. dated 2004-09-07
6787747,fast phase diversity wavefront correction using a neural network,"a phase diversity wavefront correction system for use in a multiple aperture optical imaging system forms an in-focus image as a composite, focused image from the multiple apertures of the system and also forms an additional image which is deliberately made out of focus to a known extent. taken together, the two images are processed to create one or more metrics, such as the power metric and sharpness metric. neural networks are provided, each having an output corresponding to a parameter of an aperture of the imaging system, such as a piston position (axial displacement) or tip/tilt (angular displacement) of one telescope with respect to the others in the system. the neural networks each correspond to one parameter of a telescope or a combinations of parameters and are trained to identify a subset of elements within the metrics that, when input into the network, produce the best estimate of the piston or tip/tilt position relative to a reference telescope or an estimate of a combination of parameters, such as the average of a subset of telescopes. during active use of the system, metrics generated from the in-focus and out-of-focus images of the object scene and the trained neural networks are used to provide estimates of piston and/or tip/tilt positions which are in turn used to drive the pistons and/or tip/tilt controllers to correct for aberrant movement and keep the telescopes phased.",2004-09-07,"The title of the patent is fast phase diversity wavefront correction using a neural network and its abstract is a phase diversity wavefront correction system for use in a multiple aperture optical imaging system forms an in-focus image as a composite, focused image from the multiple apertures of the system and also forms an additional image which is deliberately made out of focus to a known extent. taken together, the two images are processed to create one or more metrics, such as the power metric and sharpness metric. neural networks are provided, each having an output corresponding to a parameter of an aperture of the imaging system, such as a piston position (axial displacement) or tip/tilt (angular displacement) of one telescope with respect to the others in the system. the neural networks each correspond to one parameter of a telescope or a combinations of parameters and are trained to identify a subset of elements within the metrics that, when input into the network, produce the best estimate of the piston or tip/tilt position relative to a reference telescope or an estimate of a combination of parameters, such as the average of a subset of telescopes. during active use of the system, metrics generated from the in-focus and out-of-focus images of the object scene and the trained neural networks are used to provide estimates of piston and/or tip/tilt positions which are in turn used to drive the pistons and/or tip/tilt controllers to correct for aberrant movement and keep the telescopes phased. dated 2004-09-07"
6789620,downhole sensing and flow control utilizing neural networks,"methods are provided for downhole sensing and flow control utilizing neural networks. in a described embodiment, a temporary sensor is positioned downhole with a permanent sensor. outputs of the temporary and permanent sensors are recorded as training data sets. a neural network is trained using the training data sets. when the temporary sensor is no longer present or no longer operational in the well, the neural network is capable of determining the temporary sensor's output in response to the input to the neural network of the permanent sensor's output.",2004-09-14,"The title of the patent is downhole sensing and flow control utilizing neural networks and its abstract is methods are provided for downhole sensing and flow control utilizing neural networks. in a described embodiment, a temporary sensor is positioned downhole with a permanent sensor. outputs of the temporary and permanent sensors are recorded as training data sets. a neural network is trained using the training data sets. when the temporary sensor is no longer present or no longer operational in the well, the neural network is capable of determining the temporary sensor's output in response to the input to the neural network of the permanent sensor's output. dated 2004-09-14"
6792388,method and system for monitoring and analyzing a paper manufacturing process,"a method for monitoring and analyzing a paper manufacturing process, is disclosed in which a large number of quantities (xi, t) are measured from the process, the measured quantities (xi, t) are entered as an input vector ({overscore (x)}, t) into a neural network, which, in response to the input vector, produces an output vector ({overscore (y)}, t) as a continuous quantity, at least one fingerprint consistent with a good process situation in regard of runnability, i.e. an optimal output vector ({overscore (y)}o, t) is determined and stored in memory, the stored fingerprints and fingerprints or output vectors obtained in a normal process situation are compared substantially in real time, and based on the comparison, a difference to be presented in a graphic form to the user is determined. accordingly, a continuous time-dependent scalar quantity k=k(t) to be presented as a result to the user is determined as a geometric distance between the instantaneous measured fingerprint ({overscore (y)}, t) and the taught fingerprint ({overscore (y)}o), and the scalar quantity k=k(t) is displayed via a display device to the user.",2004-09-14,"The title of the patent is method and system for monitoring and analyzing a paper manufacturing process and its abstract is a method for monitoring and analyzing a paper manufacturing process, is disclosed in which a large number of quantities (xi, t) are measured from the process, the measured quantities (xi, t) are entered as an input vector ({overscore (x)}, t) into a neural network, which, in response to the input vector, produces an output vector ({overscore (y)}, t) as a continuous quantity, at least one fingerprint consistent with a good process situation in regard of runnability, i.e. an optimal output vector ({overscore (y)}o, t) is determined and stored in memory, the stored fingerprints and fingerprints or output vectors obtained in a normal process situation are compared substantially in real time, and based on the comparison, a difference to be presented in a graphic form to the user is determined. accordingly, a continuous time-dependent scalar quantity k=k(t) to be presented as a result to the user is determined as a geometric distance between the instantaneous measured fingerprint ({overscore (y)}, t) and the taught fingerprint ({overscore (y)}o), and the scalar quantity k=k(t) is displayed via a display device to the user. dated 2004-09-14"
6792412,neural network system and method for controlling information output based on user feedback,"a system and method for controlling information output based on user feedback about the information that includes a plurality of information sources. at least one neural network module selects one or more of a plurality of objects to receive information from the plurality of information sources based on a plurality of inputs and a plurality of weight values during that epoch. at least one server, associated with the neural network module, provides one or more of the objects to a plurality of recipients. the recipients provide feedback during an epoch. at the conclusion of an epoch, the neural network takes the feedback that has been provided from the recipients and generates a rating value for each of the objects. based on the rating value and selections made, the neural network redetermines the weight values. the neural network then selects the objects to receive information during a subsequent epoch.",2004-09-14,"The title of the patent is neural network system and method for controlling information output based on user feedback and its abstract is a system and method for controlling information output based on user feedback about the information that includes a plurality of information sources. at least one neural network module selects one or more of a plurality of objects to receive information from the plurality of information sources based on a plurality of inputs and a plurality of weight values during that epoch. at least one server, associated with the neural network module, provides one or more of the objects to a plurality of recipients. the recipients provide feedback during an epoch. at the conclusion of an epoch, the neural network takes the feedback that has been provided from the recipients and generates a rating value for each of the objects. based on the rating value and selections made, the neural network redetermines the weight values. the neural network then selects the objects to receive information during a subsequent epoch. dated 2004-09-14"
6792413,"data processing apparatus and method, recording medium, and program","this invention provides a data processing apparatus which can store and recall more complicated time-series data than those processed in related art technologies. in the data processing apparatus, a recurrent neural network (rnn) of higher layer generates long-period parameter and supplies it to an input layer of rnn of lower layer via a computing block. the rnn uses this input as a parameter and computes short-period input.",2004-09-14,"The title of the patent is data processing apparatus and method, recording medium, and program and its abstract is this invention provides a data processing apparatus which can store and recall more complicated time-series data than those processed in related art technologies. in the data processing apparatus, a recurrent neural network (rnn) of higher layer generates long-period parameter and supplies it to an input layer of rnn of lower layer via a computing block. the rnn uses this input as a parameter and computes short-period input. dated 2004-09-14"
6795778,system and method for facilitating welding system diagnostics,"a system and method for facilitating welding system diagnostics is provided. the invention includes a welder, a local system, a remote system, and/or an alarm component. the invention further provides for receiving sensor input(s), performing test sequence(s) based, at least in part, upon the sensor input(s) and/or performing internal diagnostics. the invention further provides for determining a health status of the welder and communicating the health status of the welder to the local system, the remote system and/or the alarm component. the health status of the welder can include welder alarm(s) and/or fault(s). information regarding the health status of the welder can be sent by telephone, voicemail, e-mail and/or beeper. the welder can communicate with the local system and/or remote system to schedule maintenance. the invention further provides for a expert component to facilitate welding diagnostics based, at least in part, upon the health status of the welder, welder data, an expert data store, a local service support data store, a remote expert data store and/or a remote service support data store. the expert component can employ various artificial intelligence technique(s) (e.g., bayesian model, probability tree network, fuzzy logic and/or neural network) to facilitate welding diagnostics based, at least in part, upon the health status received from the welder. the expert component can adaptively modify its modeling technique(s) based upon historical success (e.g., learn from success of previous welding diagnostics). the invention further provides for the welder, local system and/or remote system to initiate corrective action, at least temporarily, based, at least in part, upon the health status of the welder.",2004-09-21,"The title of the patent is system and method for facilitating welding system diagnostics and its abstract is a system and method for facilitating welding system diagnostics is provided. the invention includes a welder, a local system, a remote system, and/or an alarm component. the invention further provides for receiving sensor input(s), performing test sequence(s) based, at least in part, upon the sensor input(s) and/or performing internal diagnostics. the invention further provides for determining a health status of the welder and communicating the health status of the welder to the local system, the remote system and/or the alarm component. the health status of the welder can include welder alarm(s) and/or fault(s). information regarding the health status of the welder can be sent by telephone, voicemail, e-mail and/or beeper. the welder can communicate with the local system and/or remote system to schedule maintenance. the invention further provides for a expert component to facilitate welding diagnostics based, at least in part, upon the health status of the welder, welder data, an expert data store, a local service support data store, a remote expert data store and/or a remote service support data store. the expert component can employ various artificial intelligence technique(s) (e.g., bayesian model, probability tree network, fuzzy logic and/or neural network) to facilitate welding diagnostics based, at least in part, upon the health status received from the welder. the expert component can adaptively modify its modeling technique(s) based upon historical success (e.g., learn from success of previous welding diagnostics). the invention further provides for the welder, local system and/or remote system to initiate corrective action, at least temporarily, based, at least in part, upon the health status of the welder. dated 2004-09-21"
6798914,neural-network-based method of image compression,"a method for compressing images uses artificial intelligence and neural-network-based techniques to convert digital image data into symbolic data. the symbolic data is then further compressed using, for example, run-length-limited coding. the result is compressed data that represents the original image data at a high compression ratio. such high compression ratios are useful, for example, in medical diagnostic and high-definition television applications.",2004-09-28,"The title of the patent is neural-network-based method of image compression and its abstract is a method for compressing images uses artificial intelligence and neural-network-based techniques to convert digital image data into symbolic data. the symbolic data is then further compressed using, for example, run-length-limited coding. the result is compressed data that represents the original image data at a high compression ratio. such high compression ratios are useful, for example, in medical diagnostic and high-definition television applications. dated 2004-09-28"
6799117,predicting sample quality real time,"systems and methods for estimating properties of fluid samples pumped from a formation through a well are described. based upon input properties, an artificial neural network (ann) may predict a plurality of data points, and each data point may correspond to a predicted time sample of the property of the fluid sample. properties predicted by the ann include sample quality or pumping pressure differential.",2004-09-28,"The title of the patent is predicting sample quality real time and its abstract is systems and methods for estimating properties of fluid samples pumped from a formation through a well are described. based upon input properties, an artificial neural network (ann) may predict a plurality of data points, and each data point may correspond to a predicted time sample of the property of the fluid sample. properties predicted by the ann include sample quality or pumping pressure differential. dated 2004-09-28"
6799171,"applicator and method for combating pests, especially cockroaches","a neural network system including a plurality of tiers of interconnected computing elements. the plurality of tiers includes an input tier whereto a sequence of input speech vectors is applied at a first rate. two of the plurality of tiers are interconnected through a decimator configured to reduce the first rate of the sequence of input vectors. alternatively, two of the plurality of tiers are interconnected through an interpolator configured to increase the first rate of the sequence of input vectors.",2004-09-28,"The title of the patent is applicator and method for combating pests, especially cockroaches and its abstract is a neural network system including a plurality of tiers of interconnected computing elements. the plurality of tiers includes an input tier whereto a sequence of input speech vectors is applied at a first rate. two of the plurality of tiers are interconnected through a decimator configured to reduce the first rate of the sequence of input vectors. alternatively, two of the plurality of tiers are interconnected through an interpolator configured to increase the first rate of the sequence of input vectors. dated 2004-09-28"
6801655,spatial image processor,"a spatial image processor neural network for processing image data to discriminate between first and second spatial configurations of component objects includes a photo transducer input array for converting an input image to pixel data and sending the data to a localized gain network (lgn) module, a parallel memory processor and neuron array for receiving the pixel data and processing the pixel data into component recognition vectors and chaotic oscillators for receiving the recognition vectors and sending feedback data to the lgn module as attention activations. the network further includes a temporal spatial retina for receiving both the pixel data and temporal feedback activations and generating temporal spatial vectors, which are processed by a temporal parallel processor into temporal component recognition vectors. a spatial recognition vector array receives the temporal component recognition vectors and forms an object representation of the first configuration of component objects.",2004-10-05,"The title of the patent is spatial image processor and its abstract is a spatial image processor neural network for processing image data to discriminate between first and second spatial configurations of component objects includes a photo transducer input array for converting an input image to pixel data and sending the data to a localized gain network (lgn) module, a parallel memory processor and neuron array for receiving the pixel data and processing the pixel data into component recognition vectors and chaotic oscillators for receiving the recognition vectors and sending feedback data to the lgn module as attention activations. the network further includes a temporal spatial retina for receiving both the pixel data and temporal feedback activations and generating temporal spatial vectors, which are processed by a temporal parallel processor into temporal component recognition vectors. a spatial recognition vector array receives the temporal component recognition vectors and forms an object representation of the first configuration of component objects. dated 2004-10-05"
6803956,color recognition camera,"a color-recognition camera comprises a red-green-blue ccd-imaging device that provides an analog rgb-video signal. a set of three analog-to-digital converters convert the analog rgb-video signal into a digital rgb-video signal. a digital comparator tests the digital rgb-video signal pixel-by-pixel for a match against a color setpoint. if a match occurs, a pixel with a particular color represented by the color setpoint has been recognized and a &#8220;hit&#8221; is output. a pixel address counter provides a pixel address output each time a &#8220;hit&#8221; is registered. the number of hits per video frame are accumulated, and a color-match area magnitude value is output for each frame. alternatively, neural networks are used to indicate hits when a pixel in the video image comes close enough to the color setpoint value. just how close can be &#8220;learned&#8221; by the neural network.",2004-10-12,"The title of the patent is color recognition camera and its abstract is a color-recognition camera comprises a red-green-blue ccd-imaging device that provides an analog rgb-video signal. a set of three analog-to-digital converters convert the analog rgb-video signal into a digital rgb-video signal. a digital comparator tests the digital rgb-video signal pixel-by-pixel for a match against a color setpoint. if a match occurs, a pixel with a particular color represented by the color setpoint has been recognized and a &#8220;hit&#8221; is output. a pixel address counter provides a pixel address output each time a &#8220;hit&#8221; is registered. the number of hits per video frame are accumulated, and a color-match area magnitude value is output for each frame. alternatively, neural networks are used to indicate hits when a pixel in the video image comes close enough to the color setpoint value. just how close can be &#8220;learned&#8221; by the neural network. dated 2004-10-12"
6804390,computer-implemented neural network color matching formulation applications,"a method and apparatus for color matching are provided, in which paint recipe neural networks are utilized. the color of a standard is expressed as color values. the neural network includes an input layer having nodes for receiving input data related to paint bases. weighted connections connect to the nodes of the input layer and have coefficients for weighting the input data. an output layer having nodes are either directly or indirectly connected to the weighted connections and generates output data related to color values. the data to the input layer and the data from the output layer are interrelated through the neural network's nonlinear relationship. the paint color matching neural network can be used for, but not limited to, color formula correction, matching from scratch, effect pigment identification, selection of targets for color tools, searching existing formulas for the closest match, identification of formula mistakes, development of color tolerances and enhancing conversion routines.",2004-10-12,"The title of the patent is computer-implemented neural network color matching formulation applications and its abstract is a method and apparatus for color matching are provided, in which paint recipe neural networks are utilized. the color of a standard is expressed as color values. the neural network includes an input layer having nodes for receiving input data related to paint bases. weighted connections connect to the nodes of the input layer and have coefficients for weighting the input data. an output layer having nodes are either directly or indirectly connected to the weighted connections and generates output data related to color values. the data to the input layer and the data from the output layer are interrelated through the neural network's nonlinear relationship. the paint color matching neural network can be used for, but not limited to, color formula correction, matching from scratch, effect pigment identification, selection of targets for color tools, searching existing formulas for the closest match, identification of formula mistakes, development of color tolerances and enhancing conversion routines. dated 2004-10-12"
6805099,wavelet-based artificial neural net combustion sensing,"a method and apparatus for real-time measurement of combustion characteristics of each combustion event in each individual cylinder coupled with an ability to control the engine based upon the combustion characteristics are shown. the invention includes using selective sampling techniques and wavelet transforms to extract a critical signal feature from an ionization signal that is generated by an in-cylinder ion sensor, and then feeds that critical signal feature into an artificial neural network to determine a desired combustion characteristic of the combustion event. the desired combustion characteristic of the combustion event includes a location of peak pressure, an air/fuel ratio, or a percentage of mass-fraction burned, among others. the control system of the engine is then operable to control the engine based upon the combustion characteristic.",2004-10-19,"The title of the patent is wavelet-based artificial neural net combustion sensing and its abstract is a method and apparatus for real-time measurement of combustion characteristics of each combustion event in each individual cylinder coupled with an ability to control the engine based upon the combustion characteristics are shown. the invention includes using selective sampling techniques and wavelet transforms to extract a critical signal feature from an ionization signal that is generated by an in-cylinder ion sensor, and then feeds that critical signal feature into an artificial neural network to determine a desired combustion characteristic of the combustion event. the desired combustion characteristic of the combustion event includes a location of peak pressure, an air/fuel ratio, or a percentage of mass-fraction burned, among others. the control system of the engine is then operable to control the engine based upon the combustion characteristic. dated 2004-10-19"
6805668,system and method for processing patient polysomnograph data utilizing multiple neural network processing,"a system and method for processing patient polysomnograph data are provided. an abstractor obtains raw patient polysomnograph data and generates a subset of the data to include selected factors, including data clusters. the subset of the patient polysomnograph data is transferred to two or more neural networks that process the data and generate sleep classification data. an integrator obtains the sleep classification data from the two or more neural networks by integrating the sleep classification data from each neural network. a cumulative sleep stage score is generated including confidence values and accuracy estimations for review.",2004-10-19,"The title of the patent is system and method for processing patient polysomnograph data utilizing multiple neural network processing and its abstract is a system and method for processing patient polysomnograph data are provided. an abstractor obtains raw patient polysomnograph data and generates a subset of the data to include selected factors, including data clusters. the subset of the patient polysomnograph data is transferred to two or more neural networks that process the data and generate sleep classification data. an integrator obtains the sleep classification data from the two or more neural networks by integrating the sleep classification data from each neural network. a cumulative sleep stage score is generated including confidence values and accuracy estimations for review. dated 2004-10-19"
6807449,method for controlling and pre-setting a steelworks or parts of a steelworks,"method for controlling and preconfiguring a steelworks or parts of a steelworks, the rolling stand or the rolling mill train being controlled and preconfigured by means of a model of the rolling stand or the rolling mill train, the model having at least one neural network whose parameters are matched or adapted to the actual conditions in the rolling stand or in the rolling mill train, in particular to the properties of the strip, the rate at which the parameters are matched or adapted to the actual conditions in the rolling stand or in the rolling mill train, in particular to the properties of the strip, being varied.",2004-10-19,"The title of the patent is method for controlling and pre-setting a steelworks or parts of a steelworks and its abstract is method for controlling and preconfiguring a steelworks or parts of a steelworks, the rolling stand or the rolling mill train being controlled and preconfigured by means of a model of the rolling stand or the rolling mill train, the model having at least one neural network whose parameters are matched or adapted to the actual conditions in the rolling stand or in the rolling mill train, in particular to the properties of the strip, the rate at which the parameters are matched or adapted to the actual conditions in the rolling stand or in the rolling mill train, in particular to the properties of the strip, being varied. dated 2004-10-19"
6808630,system and method for ai controlling waste-water treatment by neural network and back-propagation algorithm,"a system and method for controlling treatment of the sewage/waste water. the method includes measuring attributes of inflow water flowing into a sewage/waste water treatment plant, attributes of an internal condition of a reaction tank having a first story and a second story, and fluid present values (pvs) of efficiency attributes of outflow water. the method also includes collecting data of the measured fluid pvs and operation-processing the data to convert the data into physical quantity data. the method also includes obtaining each optimum set point (sp) of each dissolved oxygen (do) and solids retention time (srt) of the first story and the second story of an exhalation tank by comparing pvs of the measured attributes using a neural network control program with a back-propagation algorithm. the method also includes converting each obtained optimum sp into an analog and digital control output value by comparing each obtained optimum sp with each pv of each do and srt of the first story and the second story of the exhalation tank. the method also includes controlling each air control valve of the first story and the second story of the exhalation tank and a pump for drawing sludge based on each obtained control output value.",2004-10-26,"The title of the patent is system and method for ai controlling waste-water treatment by neural network and back-propagation algorithm and its abstract is a system and method for controlling treatment of the sewage/waste water. the method includes measuring attributes of inflow water flowing into a sewage/waste water treatment plant, attributes of an internal condition of a reaction tank having a first story and a second story, and fluid present values (pvs) of efficiency attributes of outflow water. the method also includes collecting data of the measured fluid pvs and operation-processing the data to convert the data into physical quantity data. the method also includes obtaining each optimum set point (sp) of each dissolved oxygen (do) and solids retention time (srt) of the first story and the second story of an exhalation tank by comparing pvs of the measured attributes using a neural network control program with a back-propagation algorithm. the method also includes converting each obtained optimum sp into an analog and digital control output value by comparing each obtained optimum sp with each pv of each do and srt of the first story and the second story of the exhalation tank. the method also includes controlling each air control valve of the first story and the second story of the exhalation tank and a pump for drawing sludge based on each obtained control output value. dated 2004-10-26"
6819746,expert system for loop qualification of xdsl services,"a technique for qualification of loops for new digital subscriber line services (dsl) involves use of an expert system, such as a neural network. a database of loop characteristic information and performance data enables the expert system to train or learn how to predict performance for future loops. in response to data characterizing a new loop to be qualified, the trained expert system predicts digital subscriber line performance for the new loop. typically, the prediction enables classification of service capacity for the new loop into one of several classes corresponding to levels of dsl service offered through the network. the database for use by the expert system is updated as each newly qualified loop is brought into service and actual performance for that loop is known.",2004-11-16,"The title of the patent is expert system for loop qualification of xdsl services and its abstract is a technique for qualification of loops for new digital subscriber line services (dsl) involves use of an expert system, such as a neural network. a database of loop characteristic information and performance data enables the expert system to train or learn how to predict performance for future loops. in response to data characterizing a new loop to be qualified, the trained expert system predicts digital subscriber line performance for the new loop. typically, the prediction enables classification of service capacity for the new loop into one of several classes corresponding to levels of dsl service offered through the network. the database for use by the expert system is updated as each newly qualified loop is brought into service and actual performance for that loop is known. dated 2004-11-16"
6819790,massive training artificial neural network (mtann) for detecting abnormalities in medical images,"a method of training an artificial neural network (ann) involves receiving a likelihood distribution map as a teacher image, receiving a training image, moving a local window across sub-regions of the training image to obtain respective sub-region pixel sets, inputting the sub-region pixel sets to the ann so that it provides output pixel values that are compared to output pixel values of corresponding teacher image pixel values to determine an error, and training the ann to reduce the error. a method of detecting a target structure in an image involves scanning a local window across sub-regions of the image by moving the local window for each sub-region so as to obtain respective sub-region pixel sets, inputting the sub-region pixel sets to an ann so that it provides respective output pixel values that represent likelihoods that respective image pixels are part of a target structure, the output pixel values collectively constituting a likelihood distribution map. another method for detecting a target structure involves training n parallel anns on either (a) a same target structure and n mutually different non-target structures, or (b) a same non-target structure and n mutually different target structures, the anns outputting n respective indications of whether the image includes a target structure or a non-target structure, and combining the n indications to form a combined indication of whether the image includes a target structure or a non-target structure. the invention provides related apparatus and computer program products storing executable instructions to perform the methods.",2004-11-16,"The title of the patent is massive training artificial neural network (mtann) for detecting abnormalities in medical images and its abstract is a method of training an artificial neural network (ann) involves receiving a likelihood distribution map as a teacher image, receiving a training image, moving a local window across sub-regions of the training image to obtain respective sub-region pixel sets, inputting the sub-region pixel sets to the ann so that it provides output pixel values that are compared to output pixel values of corresponding teacher image pixel values to determine an error, and training the ann to reduce the error. a method of detecting a target structure in an image involves scanning a local window across sub-regions of the image by moving the local window for each sub-region so as to obtain respective sub-region pixel sets, inputting the sub-region pixel sets to an ann so that it provides respective output pixel values that represent likelihoods that respective image pixels are part of a target structure, the output pixel values collectively constituting a likelihood distribution map. another method for detecting a target structure involves training n parallel anns on either (a) a same target structure and n mutually different non-target structures, or (b) a same non-target structure and n mutually different target structures, the anns outputting n respective indications of whether the image includes a target structure or a non-target structure, and combining the n indications to form a combined indication of whether the image includes a target structure or a non-target structure. the invention provides related apparatus and computer program products storing executable instructions to perform the methods. dated 2004-11-16"
6820053,method and apparatus for suppressing audible noise in speech transmission,"method of suppressing audible noise in speech transmission by means of a multi-layer self-organizing fed-back neural network comprising a minima detection layer, a reaction layer, a diffusion layer and an integration layer, said layers defining a filter function f(f,t) for noise filtering.",2004-11-16,"The title of the patent is method and apparatus for suppressing audible noise in speech transmission and its abstract is method of suppressing audible noise in speech transmission by means of a multi-layer self-organizing fed-back neural network comprising a minima detection layer, a reaction layer, a diffusion layer and an integration layer, said layers defining a filter function f(f,t) for noise filtering. dated 2004-11-16"
6823296,method for forming an optimized neural network module intended to simulate the flow mode of a multiphase fluid stream,"a method for forming a module (hydrodynamic or thermodynamic for example) intended for real-time simulation of the flow mode, at any point of a pipe, of a multiphase fluid stream comprising at least a liquid phase and at least a gas phase. the method comprises using a modelling system based on non-linear neural networks each having inputs for structure parameters and physical quantities, outputs where quantities necessary for estimation of the flow mode are available, and at least one intermediate layer. the neural networks are determined iteratively to adjust to the values of a learning base with predetermined tables connecting various values obtained for the output data to the corresponding values of the input data. a learning base suited to the imposed operating conditions is used and optimized neural networks best adjusted to the imposed operating conditions are generated.",2004-11-23,"The title of the patent is method for forming an optimized neural network module intended to simulate the flow mode of a multiphase fluid stream and its abstract is a method for forming a module (hydrodynamic or thermodynamic for example) intended for real-time simulation of the flow mode, at any point of a pipe, of a multiphase fluid stream comprising at least a liquid phase and at least a gas phase. the method comprises using a modelling system based on non-linear neural networks each having inputs for structure parameters and physical quantities, outputs where quantities necessary for estimation of the flow mode are available, and at least one intermediate layer. the neural networks are determined iteratively to adjust to the values of a learning base with predetermined tables connecting various values obtained for the output data to the corresponding values of the input data. a learning base suited to the imposed operating conditions is used and optimized neural networks best adjusted to the imposed operating conditions are generated. dated 2004-11-23"
6823322,piecewise nonlinear mapper for digitals,a nonlinear signal mapper that can implement any continuous one-to-one nonlinear map of baseband or intermediate-frequency digital signals. the mapping method follows a &#8220;divide-and-conquer&#8221; approach in that a nonlinear map to be implemented is piecewise decomposed into a set of simpler nonlinear component maps. the component maps are implemented using code-enabled feed-forward neural networks (ff-nns). each code-enabled feed-forward neural network only operates on samples of a digital input signal that lie in a specified interval of the real-valued number line. code-enabled ff-nns are controlled by codewords produced by a scalar quantization encoder. the quantization encoder also controls a multiplexer that directs values produced by the ff-nns to the nonlinear mapper's output.,2004-11-23,The title of the patent is piecewise nonlinear mapper for digitals and its abstract is a nonlinear signal mapper that can implement any continuous one-to-one nonlinear map of baseband or intermediate-frequency digital signals. the mapping method follows a &#8220;divide-and-conquer&#8221; approach in that a nonlinear map to be implemented is piecewise decomposed into a set of simpler nonlinear component maps. the component maps are implemented using code-enabled feed-forward neural networks (ff-nns). each code-enabled feed-forward neural network only operates on samples of a digital input signal that lie in a specified interval of the real-valued number line. code-enabled ff-nns are controlled by codewords produced by a scalar quantization encoder. the quantization encoder also controls a multiplexer that directs values produced by the ff-nns to the nonlinear mapper's output. dated 2004-11-23
6826550,"method, system, and program for converting application program code to executable code using neural networks based on characteristics of the inputs","provided is a compiler to map application program code to object code capable of being executed on an operating system platform. a first neural network module is trained to generate characteristic output based on input information describing attributes of the application program. a second neural network module is trained to receive as input the application program code and the characteristic output and, in response, generate object code. the first and second neural network modules are used to convert the application program code to object code.",2004-11-30,"The title of the patent is method, system, and program for converting application program code to executable code using neural networks based on characteristics of the inputs and its abstract is provided is a compiler to map application program code to object code capable of being executed on an operating system platform. a first neural network module is trained to generate characteristic output based on input information describing attributes of the application program. a second neural network module is trained to receive as input the application program code and the characteristic output and, in response, generate object code. the first and second neural network modules are used to convert the application program code to object code. dated 2004-11-30"
6832214,"method, system, and program for converting code to executable code using neural networks implemented in a software program","disclosed is a system, method, and program for generating a compiler to map a code set to object code capable of being executed on an operating system platform. at least one neural network is trained to convert the code set to object code. the at least one trained neural network can then be used to convert the code set to the object code.",2004-12-14,"The title of the patent is method, system, and program for converting code to executable code using neural networks implemented in a software program and its abstract is disclosed is a system, method, and program for generating a compiler to map a code set to object code capable of being executed on an operating system platform. at least one neural network is trained to convert the code set to object code. the at least one trained neural network can then be used to convert the code set to the object code. dated 2004-12-14"
6836767,pipelined hardware implementation of a neural network circuit,"in a first aspect, a pipelined hardware implementation of a neural network circuit includes an input stage, two or more processing stages and an output stage. each processing stage includes one or more processing units. each processing unit includes storage for weighted values, a plurality of multipliers for multiplying input values by weighted values, an adder for adding products outputted from product multipliers, a function circuit for applying a non-linear function to the sum outputted by the adder, and a register for storing the output of the function circuit.",2004-12-28,"The title of the patent is pipelined hardware implementation of a neural network circuit and its abstract is in a first aspect, a pipelined hardware implementation of a neural network circuit includes an input stage, two or more processing stages and an output stage. each processing stage includes one or more processing units. each processing unit includes storage for weighted values, a plurality of multipliers for multiplying input values by weighted values, an adder for adding products outputted from product multipliers, a function circuit for applying a non-linear function to the sum outputted by the adder, and a register for storing the output of the function circuit. dated 2004-12-28"
6839581,method for detecting cheyne-stokes respiration in patients with congestive heart failure,the present invention provides a diagnostic tool for detection of cheyne-stokes respiration (csr). this invention also provides a method for development of the diagnostic tool. the method comprises the steps of performing overnight oximetry recordings in patients suspected of sleep disordered breathing. spectral analysis is performed on the oximetry recordings to obtain a set of parameters which can be used in the construction of a classification tree and a trained neural network. the diagnostic tools of the present invention can be used for classification of a patient as having csr or obstructive sleep apnea.,2005-01-04,The title of the patent is method for detecting cheyne-stokes respiration in patients with congestive heart failure and its abstract is the present invention provides a diagnostic tool for detection of cheyne-stokes respiration (csr). this invention also provides a method for development of the diagnostic tool. the method comprises the steps of performing overnight oximetry recordings in patients suspected of sleep disordered breathing. spectral analysis is performed on the oximetry recordings to obtain a set of parameters which can be used in the construction of a classification tree and a trained neural network. the diagnostic tools of the present invention can be used for classification of a patient as having csr or obstructive sleep apnea. dated 2005-01-04
6839608,hybrid model and method for determining mechanical properties and processing properties of an injection-molded part,"a method of predicting the properties (e.g., mechanical and/or processing properties) of an injection-molded article is disclosed. the method makes use of a hybrid model which includes at least one neural network. in order to forecast (or predict) properties with respect to the manufacture of a plastic molded article, a hybrid model is used in the present invention, which includes: one or more neural networks nn1, nn2, nn3, nn4, . . . , nnk; and optionally one or more rigorous models r1, r2, r3, r4, . . . , which are connected to one another. the rigorous models are used to map model elements which can be described in mathematical formulae. the neural networks are used to map processes whose relationship is present only in the form of data, as it is in effect impossible to model such processes rigorously. as a result, a forecast relating to properties including the mechanical, thermal and rheological processing properties and relating to the process time of a plastic molded article is obtained.",2005-01-04,"The title of the patent is hybrid model and method for determining mechanical properties and processing properties of an injection-molded part and its abstract is a method of predicting the properties (e.g., mechanical and/or processing properties) of an injection-molded article is disclosed. the method makes use of a hybrid model which includes at least one neural network. in order to forecast (or predict) properties with respect to the manufacture of a plastic molded article, a hybrid model is used in the present invention, which includes: one or more neural networks nn1, nn2, nn3, nn4, . . . , nnk; and optionally one or more rigorous models r1, r2, r3, r4, . . . , which are connected to one another. the rigorous models are used to map model elements which can be described in mathematical formulae. the neural networks are used to map processes whose relationship is present only in the form of data, as it is in effect impossible to model such processes rigorously. as a result, a forecast relating to properties including the mechanical, thermal and rheological processing properties and relating to the process time of a plastic molded article is obtained. dated 2005-01-04"
6842745,programmable chaos generator and process for use thereof,"a chaotic signal generator includes a set of elements connected together for generating chaotic signals. the connection scheme may correspond to the circuit generally referred to as chua's circuit, particularly when implemented as a cellular neural network. interposed in the connection scheme is at least one switch, such as a mos transistor. opening and closing of the switch causes variation in the chaotic dynamics of the generated signals. a command signal applied to the switch may correspond to a modulating signal for transmission on a channel, such as a high noise channel. the modulating signal may be a binary signal, and the command signal may be a switching signal having a frequency that increases or decreases depending on the logic level of the binary signal.",2005-01-11,"The title of the patent is programmable chaos generator and process for use thereof and its abstract is a chaotic signal generator includes a set of elements connected together for generating chaotic signals. the connection scheme may correspond to the circuit generally referred to as chua's circuit, particularly when implemented as a cellular neural network. interposed in the connection scheme is at least one switch, such as a mos transistor. opening and closing of the switch causes variation in the chaotic dynamics of the generated signals. a command signal applied to the switch may correspond to a modulating signal for transmission on a channel, such as a high noise channel. the modulating signal may be a binary signal, and the command signal may be a switching signal having a frequency that increases or decreases depending on the logic level of the binary signal. dated 2005-01-11"
6843564,three-dimensional image projection employing retro-reflective screens,"disclosed herein are three-dimensional projection systems and related methods employing two electronically controlled projectors and a retro-reflective screen. the retro-reflective screen produces a known non-linear light reflection pattern when images are projected thereon. image computational means are used to calculating flat image information for each projector based upon inputted stereopair images and information regarding the projectors and screen. in preferred embodiments of the present invention, the projection system uses an image computational device that employs a neural network feedback calculation to calculate the appropriate flat image information and appropriate images to be projected on the screen by the projectors at any given time. more than two projectors can be employed to produce multiple aspect views, to support multiple viewers, and the like. in another embodiment, the projection system includes a digital camera that provides feedback data to the image computational device.",2005-01-18,"The title of the patent is three-dimensional image projection employing retro-reflective screens and its abstract is disclosed herein are three-dimensional projection systems and related methods employing two electronically controlled projectors and a retro-reflective screen. the retro-reflective screen produces a known non-linear light reflection pattern when images are projected thereon. image computational means are used to calculating flat image information for each projector based upon inputted stereopair images and information regarding the projectors and screen. in preferred embodiments of the present invention, the projection system uses an image computational device that employs a neural network feedback calculation to calculate the appropriate flat image information and appropriate images to be projected on the screen by the projectors at any given time. more than two projectors can be employed to produce multiple aspect views, to support multiple viewers, and the like. in another embodiment, the projection system includes a digital camera that provides feedback data to the image computational device. dated 2005-01-18"
6844582,semiconductor device and learning method thereof,"a learning method of a semiconductor device of the present invention comprises a neuro device having a multiplier as a synapse in which a weight varies according to an input weight voltage, and functioning as a neural network system that processes analog data, comprising a step a of inputting predetermined input data to the neuro device and calculating an error between a target value of an output of the neuro device with respect to the input data and an actual output, a step b of calculating variation amount in the error by varying a weight of the multiplier thereafter, and a step c of varying the weight of the multiplier based on the variation amount in the error, wherein in the steps b and c, after inputting a reset voltage for setting the weight to a substantially constant value to the multiplier as the weight voltage, the weight is varied by inputting the weight voltage corresponding to the weight to be varied.",2005-01-18,"The title of the patent is semiconductor device and learning method thereof and its abstract is a learning method of a semiconductor device of the present invention comprises a neuro device having a multiplier as a synapse in which a weight varies according to an input weight voltage, and functioning as a neural network system that processes analog data, comprising a step a of inputting predetermined input data to the neuro device and calculating an error between a target value of an output of the neuro device with respect to the input data and an actual output, a step b of calculating variation amount in the error by varying a weight of the multiplier thereafter, and a step c of varying the weight of the multiplier based on the variation amount in the error, wherein in the steps b and c, after inputting a reset voltage for setting the weight to a substantially constant value to the multiplier as the weight voltage, the weight is varied by inputting the weight voltage corresponding to the weight to be varied. dated 2005-01-18"
6845289,hybrid model and method for determining manufacturing properties of an injection-molded part,"a method of determining properties relating to the manufacture of an injection-molded article is described. the method makes use of a hybrid model which includes at least one neural network and at least one rigorous model. in order to forecast (or predict) properties relating to the manufacture of a plastic molded part, a hybrid model is used which includes: one or more neural networks nn1, nn2, nn3, nn4, . . . , nnk; and one or more rigorous models r1, r2, r3, r4, . . . , which are connected to one another. the rigorous models are used to map model elements which can be described in mathematical formulae. the neural model elements are used to map processes whose relationship is present only in the form of data, as it is typically impossible to model such processes rigorously. as a result, a forecast (or prediction) relating to properties including, for example, the mechanical, thermal and rheological processing properties and relating to the cycle time of a plastic molded part can be made.",2005-01-18,"The title of the patent is hybrid model and method for determining manufacturing properties of an injection-molded part and its abstract is a method of determining properties relating to the manufacture of an injection-molded article is described. the method makes use of a hybrid model which includes at least one neural network and at least one rigorous model. in order to forecast (or predict) properties relating to the manufacture of a plastic molded part, a hybrid model is used which includes: one or more neural networks nn1, nn2, nn3, nn4, . . . , nnk; and one or more rigorous models r1, r2, r3, r4, . . . , which are connected to one another. the rigorous models are used to map model elements which can be described in mathematical formulae. the neural model elements are used to map processes whose relationship is present only in the form of data, as it is typically impossible to model such processes rigorously. as a result, a forecast (or prediction) relating to properties including, for example, the mechanical, thermal and rheological processing properties and relating to the cycle time of a plastic molded part can be made. dated 2005-01-18"
6847954,control-loop auto-tuner with nonlinear tuning rules estimators,"a system for tuning a process control loop includes a tuner module for receiving an error signal representative of the difference between a set point and a process variable, the module generating a first process control signal for controlling the process. the system further includes a controller module for receiving the error signal and a parameter signal from a nonlinear module to generate a second process control signal for controlling the process, wherein the nonlinear module applies a nonlinear procedure to generate the parameter signal. the system further includes a switching means coupled to the tuner module and the controller module to select the appropriate process control signal for controlling the process. the system provided uses nonlinear techniques in the nonlinear module to approximate the desired controller tuning parameters. the nonlinear techniques include neural network tuning, fuzzy logic tuning and nonlinear functions, including sigmoid tuning. a system also provides that the nonlinear module use nonlinear techniques to approximate the desired process model parameters. according to an embodiment of the present invention, the nonlinear module includes a process model identification module and a controller tuning module that provides controller parameters and model identification parameters using neural networks, fuzzy logic and nonlinear functions, including sigmoid tuning.",2005-01-25,"The title of the patent is control-loop auto-tuner with nonlinear tuning rules estimators and its abstract is a system for tuning a process control loop includes a tuner module for receiving an error signal representative of the difference between a set point and a process variable, the module generating a first process control signal for controlling the process. the system further includes a controller module for receiving the error signal and a parameter signal from a nonlinear module to generate a second process control signal for controlling the process, wherein the nonlinear module applies a nonlinear procedure to generate the parameter signal. the system further includes a switching means coupled to the tuner module and the controller module to select the appropriate process control signal for controlling the process. the system provided uses nonlinear techniques in the nonlinear module to approximate the desired controller tuning parameters. the nonlinear techniques include neural network tuning, fuzzy logic tuning and nonlinear functions, including sigmoid tuning. a system also provides that the nonlinear module use nonlinear techniques to approximate the desired process model parameters. according to an embodiment of the present invention, the nonlinear module includes a process model identification module and a controller tuning module that provides controller parameters and model identification parameters using neural networks, fuzzy logic and nonlinear functions, including sigmoid tuning. dated 2005-01-25"
6853889,vehicle dynamics production system and method,a vehicle dynamics prediction system for providing prediction of vehicle velocity for a predetermined future time period. the predictions utilize future vehicle control setting anticipated for the time period. an artificial intelligence database coupled to a processor that utilizes weighted values for neural network models representing dynamic performance of the vehicle forms part of the vehicle dynamics prediction system.,2005-02-08,The title of the patent is vehicle dynamics production system and method and its abstract is a vehicle dynamics prediction system for providing prediction of vehicle velocity for a predetermined future time period. the predictions utilize future vehicle control setting anticipated for the time period. an artificial intelligence database coupled to a processor that utilizes weighted values for neural network models representing dynamic performance of the vehicle forms part of the vehicle dynamics prediction system. dated 2005-02-08
6857112,method and apparatus for performing extraction using machine learning,"a system for using machine-learning to create a model for performing integrated circuit layout extraction is disclosed. the system of the present invention has two main phases: model creation and model application. the model creation phase comprises creating one or more extraction models using machine-learning techniques. first, a complex extraction problem is decomposed into smaller simpler extraction problems. then, each smaller extraction problem is then analyzed to identify a set of physical parameters that fully define the smaller extraction problem. next, models are created using machine learning techniques for all of the smaller simpler extraction problems. the machine learning is performed by first creating training data sets composed of the identified parameters from typical examples of the smaller extraction problem and the answers to those example extraction problems as solved using a highly accurate physics-based field solver. the training sets are then used to train the models. in one embodiment, neural networks are used to model the extraction problems. bayesian inference is employed by one embodiment in order to train the neural network models. bayesian inference may be implemented with normal monte carlo techniques or hybrid monte carlo techniques. after the creation of a set of models for each of the smaller simpler extraction problems, the machine-learning based models may be used for extraction.",2005-02-15,"The title of the patent is method and apparatus for performing extraction using machine learning and its abstract is a system for using machine-learning to create a model for performing integrated circuit layout extraction is disclosed. the system of the present invention has two main phases: model creation and model application. the model creation phase comprises creating one or more extraction models using machine-learning techniques. first, a complex extraction problem is decomposed into smaller simpler extraction problems. then, each smaller extraction problem is then analyzed to identify a set of physical parameters that fully define the smaller extraction problem. next, models are created using machine learning techniques for all of the smaller simpler extraction problems. the machine learning is performed by first creating training data sets composed of the identified parameters from typical examples of the smaller extraction problem and the answers to those example extraction problems as solved using a highly accurate physics-based field solver. the training sets are then used to train the models. in one embodiment, neural networks are used to model the extraction problems. bayesian inference is employed by one embodiment in order to train the neural network models. bayesian inference may be implemented with normal monte carlo techniques or hybrid monte carlo techniques. after the creation of a set of models for each of the smaller simpler extraction problems, the machine-learning based models may be used for extraction. dated 2005-02-15"
6857553,method and apparatus for in-process sensing of manufacturing quality,"a method for determining the quality of an examined weld joint comprising the steps of providing acoustical data from the examined weld joint, and performing a neural network operation on the acoustical data determine the quality of the examined weld joint produced by a friction weld process. the neural network may be trained by the steps of providing acoustical data and observable data from at least one test weld joint, and training the neural network based on the acoustical data and observable data to form a trained neural network so that the trained neural network is capable of determining the quality of a examined weld joint based on acoustical data from the examined weld joint. in addition, an apparatus having a housing, acoustical sensors mounted therein, and means for mounting the housing on a friction weld device so that the acoustical sensors do not contact the weld joint. the apparatus may sample the acoustical data necessary for the neural network to determine the quality of a weld joint.",2005-02-22,"The title of the patent is method and apparatus for in-process sensing of manufacturing quality and its abstract is a method for determining the quality of an examined weld joint comprising the steps of providing acoustical data from the examined weld joint, and performing a neural network operation on the acoustical data determine the quality of the examined weld joint produced by a friction weld process. the neural network may be trained by the steps of providing acoustical data and observable data from at least one test weld joint, and training the neural network based on the acoustical data and observable data to form a trained neural network so that the trained neural network is capable of determining the quality of a examined weld joint based on acoustical data from the examined weld joint. in addition, an apparatus having a housing, acoustical sensors mounted therein, and means for mounting the housing on a friction weld device so that the acoustical sensors do not contact the weld joint. the apparatus may sample the acoustical data necessary for the neural network to determine the quality of a weld joint. dated 2005-02-22"
6862253,sonic identification system and method,"a neural net sonic identification system includes an ultra-sonic emitter and receiver configured to detect characteristics of a subject by detecting reflections of the ultrasonic waves. the sonic reflections are processed by a neural network pattern recognition system that learns to recognize a particular subject and distinguish it from other subjects. the emitter/receiver is continuously optimized for phase, wavelength, spectral, and power frequencies, and the received sound “picture” is written-over the previously stored encrypted image data file for a particular individual, so as to continuously update the processor's retained identification information. in one embodiment, the emitter/receiver comprises a solid-state multi-layer tunable array of piezoelectric emitters and receivers that can be selectively “aimed” in different directions without physically moving by independently varying the wavelength and power of the various emitters/receivers.",2005-03-01,"The title of the patent is sonic identification system and method and its abstract is a neural net sonic identification system includes an ultra-sonic emitter and receiver configured to detect characteristics of a subject by detecting reflections of the ultrasonic waves. the sonic reflections are processed by a neural network pattern recognition system that learns to recognize a particular subject and distinguish it from other subjects. the emitter/receiver is continuously optimized for phase, wavelength, spectral, and power frequencies, and the received sound “picture” is written-over the previously stored encrypted image data file for a particular individual, so as to continuously update the processor's retained identification information. in one embodiment, the emitter/receiver comprises a solid-state multi-layer tunable array of piezoelectric emitters and receivers that can be selectively “aimed” in different directions without physically moving by independently varying the wavelength and power of the various emitters/receivers. dated 2005-03-01"
6869287,method of teaching reading,"a method for teaching reading is provided, reflecting the understanding that reading is a process, operating primarily implicitly and guided by a neural network built specifically for that purpose. during the excellent reading component, the student must produce excellent reading—made possible by enhancing the predictability of the text through listening repeatedly to it interspersed with reading it silently. the student's brain is seduced into utilizing prediction of meaning as the major reading strategy, which allows it to develop implicitly-operating prediction and confirmation strategies that consistently yield excellent reading. during the coached reading component, as the student reads the text aloud from unfamiliar text, a tutor gives feedback designed to assist the brain realize its current reading strategies are not producing the desired results and the erroneously-built neural network needs remodeling. during the independent reading component, the brain experiments with the strategy ideas formulated as a result of experiencing the other components.",2005-03-22,"The title of the patent is method of teaching reading and its abstract is a method for teaching reading is provided, reflecting the understanding that reading is a process, operating primarily implicitly and guided by a neural network built specifically for that purpose. during the excellent reading component, the student must produce excellent reading—made possible by enhancing the predictability of the text through listening repeatedly to it interspersed with reading it silently. the student's brain is seduced into utilizing prediction of meaning as the major reading strategy, which allows it to develop implicitly-operating prediction and confirmation strategies that consistently yield excellent reading. during the coached reading component, as the student reads the text aloud from unfamiliar text, a tutor gives feedback designed to assist the brain realize its current reading strategies are not producing the desired results and the erroneously-built neural network needs remodeling. during the independent reading component, the brain experiments with the strategy ideas formulated as a result of experiencing the other components. dated 2005-03-22"
6871194,interaction prediction system and method,"the invention provides an interaction prediction system comprising a memory in which is maintained a neural network trained on data retrieved from an interaction database of interaction data representing interactions between customers and merchants, retrieval means arranged to activate the neural network and to retrieve prediction data representing future interactions between customers and merchants, and display means arranged to display a representation of the prediction data. the invention also provides a neural network training system in which a neural network is arranged to receive input data representing the data retrieved from the interaction database and to output prediction data representing interaction data predicted by the neural network. the invention also provides computer programs and methods of interaction prediction and neural network training respectively.",2005-03-22,"The title of the patent is interaction prediction system and method and its abstract is the invention provides an interaction prediction system comprising a memory in which is maintained a neural network trained on data retrieved from an interaction database of interaction data representing interactions between customers and merchants, retrieval means arranged to activate the neural network and to retrieve prediction data representing future interactions between customers and merchants, and display means arranged to display a representation of the prediction data. the invention also provides a neural network training system in which a neural network is arranged to receive input data representing the data retrieved from the interaction database and to output prediction data representing interaction data predicted by the neural network. the invention also provides computer programs and methods of interaction prediction and neural network training respectively. dated 2005-03-22"
6873979,method of building predictive models on transactional data,"a method of building predictive statistical models provides a dedicated aggregation module for each transactional record source. each aggregation module aggregates the transactional records using a neural network function to produce a scalar output which can then be input to a traditional modeling function, which may employ either logistic regression, neural network, or radial basis function techniques. the output of the aggregation modules can be saved, and updated aggregation values can be updated by processing new transaction records and combining the new transaction values with the previous output values using a blending function. parameters of the neural network in the aggregation module may be calculated simultaneously with the parameters of the traditional modeling module.",2005-03-29,"The title of the patent is method of building predictive models on transactional data and its abstract is a method of building predictive statistical models provides a dedicated aggregation module for each transactional record source. each aggregation module aggregates the transactional records using a neural network function to produce a scalar output which can then be input to a traditional modeling function, which may employ either logistic regression, neural network, or radial basis function techniques. the output of the aggregation modules can be saved, and updated aggregation values can be updated by processing new transaction records and combining the new transaction values with the previous output values using a blending function. parameters of the neural network in the aggregation module may be calculated simultaneously with the parameters of the traditional modeling module. dated 2005-03-29"
6876758,methods and systems for improving a user's visual perception over a communications network,"methods and apparatus that provide proper identification of visual and neurological abilities related to defects in the neurological component of the brain, that improve visual and neurological performance, and that develop improved neural performance in the brain and nervous system. all of these methods are carried out over a communications network such as the internet.",2005-04-05,"The title of the patent is methods and systems for improving a user's visual perception over a communications network and its abstract is methods and apparatus that provide proper identification of visual and neurological abilities related to defects in the neurological component of the brain, that improve visual and neurological performance, and that develop improved neural performance in the brain and nervous system. all of these methods are carried out over a communications network such as the internet. dated 2005-04-05"
6876989,back-propagation neural network with enhanced neuron characteristics,a neural network system includes a feedforward network comprising at least one neuron circuit for producing an activation function and a first derivative of the activation function and a weight updating circuit for producing updated weights to the feedforward network. the system also includes an error back-propagation network for receiving the first derivative of the activation function and to provide weight change data information to the weight updating circuit.,2005-04-05,The title of the patent is back-propagation neural network with enhanced neuron characteristics and its abstract is a neural network system includes a feedforward network comprising at least one neuron circuit for producing an activation function and a first derivative of the activation function and a weight updating circuit for producing updated weights to the feedforward network. the system also includes an error back-propagation network for receiving the first derivative of the activation function and to provide weight change data information to the weight updating circuit. dated 2005-04-05
6878119,method for reducing artifacts in a spatial measurement,"a method for artifact reduction in sonomicrometer obtained intracranial pressure measurements generally comprises isolating a component of a sonomicrometer waveform attributable solely to changes in intracranial volume by using a neural network or other nonlinear engine to extract a heartbeat component from the sonomicrometer output. because the heartbeat is so characteristic, no actual measurement of the heartbeat as the forcing function is required to isolate the resulting changes in distance from the artifact induced changes in distance. the neural network then be utilized to directly map measured changes in skull distance over time to changes in intracranial pressure over a volume change, the inverse of compliance. the method is generally extendable to use with other volumetric based measurement techniques.",2005-04-12,"The title of the patent is method for reducing artifacts in a spatial measurement and its abstract is a method for artifact reduction in sonomicrometer obtained intracranial pressure measurements generally comprises isolating a component of a sonomicrometer waveform attributable solely to changes in intracranial volume by using a neural network or other nonlinear engine to extract a heartbeat component from the sonomicrometer output. because the heartbeat is so characteristic, no actual measurement of the heartbeat as the forcing function is required to isolate the resulting changes in distance from the artifact induced changes in distance. the neural network then be utilized to directly map measured changes in skull distance over time to changes in intracranial pressure over a volume change, the inverse of compliance. the method is generally extendable to use with other volumetric based measurement techniques. dated 2005-04-12"
6879885,rotor torque predictor,"a system is disclosed for determining the total torque required at the main and tail rotors of a helicopter for use in feed-forward rotor torque anticipation which includes a polynomial neural network adapted and configured to predict the aerodynamic torque at the main and tail rotors with the helicopter in motion based upon a plurality of pilot inputs and airframe inputs, a main rotor load map for determining the torque at the main rotor in hover out of ground effects based upon main rotor speed and collective stick position, a tail rotor load map for determining the torque at the tail rotor with the helicopter stationary based upon main rotor speed and pedal position, and a processor for calculating the required total torque at the main and tail rotors by summing the outputs of the polynomial neural network, and the main and tail rotor load maps.",2005-04-12,"The title of the patent is rotor torque predictor and its abstract is a system is disclosed for determining the total torque required at the main and tail rotors of a helicopter for use in feed-forward rotor torque anticipation which includes a polynomial neural network adapted and configured to predict the aerodynamic torque at the main and tail rotors with the helicopter in motion based upon a plurality of pilot inputs and airframe inputs, a main rotor load map for determining the torque at the main rotor in hover out of ground effects based upon main rotor speed and collective stick position, a tail rotor load map for determining the torque at the tail rotor with the helicopter stationary based upon main rotor speed and pedal position, and a processor for calculating the required total torque at the main and tail rotors by summing the outputs of the polynomial neural network, and the main and tail rotor load maps. dated 2005-04-12"
6879969,system and method for real-time recognition of driving patterns,"system and method for real-time, automatic, recognition of large time-scale driving patterns employs a statistical pattern recognition framework, implemented by means of feed-forward neural network utilizing models developed for recognizing, for example, four classes of driving environments, namely highway, main road, suburban traffic and city traffic, from vehicle performance data. a vehicle control application effects changes in vehicle performance aspects based on the recognized driving environment.",2005-04-12,"The title of the patent is system and method for real-time recognition of driving patterns and its abstract is system and method for real-time, automatic, recognition of large time-scale driving patterns employs a statistical pattern recognition framework, implemented by means of feed-forward neural network utilizing models developed for recognizing, for example, four classes of driving environments, namely highway, main road, suburban traffic and city traffic, from vehicle performance data. a vehicle control application effects changes in vehicle performance aspects based on the recognized driving environment. dated 2005-04-12"
6882919,shift control method and apparatus of an automatic transmission,"the shift control method of an automatic transmission calculates a shift-pattern-adjusting coefficient using a modular neural network. the shift-pattern adjusting coefficient is based on a plurality of input signals input from a plurality of detectors. it is then determined if shifting is required based on an adjusted shift-pattern. the adjusted shift-pattern being adjusted based on the shift-pattern adjusting coefficient. a target shift-speed when shifting is required is calculated based on the adjusted shift-pattern. finally, a shifting signal for shifting to the target shift-speed is generated.",2005-04-19,"The title of the patent is shift control method and apparatus of an automatic transmission and its abstract is the shift control method of an automatic transmission calculates a shift-pattern-adjusting coefficient using a modular neural network. the shift-pattern adjusting coefficient is based on a plurality of input signals input from a plurality of detectors. it is then determined if shifting is required based on an adjusted shift-pattern. the adjusted shift-pattern being adjusted based on the shift-pattern adjusting coefficient. a target shift-speed when shifting is required is calculated based on the adjusted shift-pattern. finally, a shifting signal for shifting to the target shift-speed is generated. dated 2005-04-19"
6885320,apparatus and method for selecting length of variable length coding bit stream using neural network,"an apparatus and method for selecting the length of a variable length code bitstream by using a neural network are provided. the apparatus for selecting the length of a variable length code bitstream includes a bitstream estimation length receiving unit which inputs a predetermined quantization dct coefficient block to a neural network whose training is finished, and receives the estimation length of a bitstream corresponding to the quantization dct coefficient block from the neural network; and a bitstream estimation length selection unit which receives user selection about an estimation length received by the bitstream estimation length receiving unit. according to the method and apparatus the length of a variable length code bit stream can be estimated such that a user can select a desired length of a bitstream in advance without performing variable length coding.",2005-04-26,"The title of the patent is apparatus and method for selecting length of variable length coding bit stream using neural network and its abstract is an apparatus and method for selecting the length of a variable length code bitstream by using a neural network are provided. the apparatus for selecting the length of a variable length code bitstream includes a bitstream estimation length receiving unit which inputs a predetermined quantization dct coefficient block to a neural network whose training is finished, and receives the estimation length of a bitstream corresponding to the quantization dct coefficient block from the neural network; and a bitstream estimation length selection unit which receives user selection about an estimation length received by the bitstream estimation length receiving unit. according to the method and apparatus the length of a variable length code bit stream can be estimated such that a user can select a desired length of a bitstream in advance without performing variable length coding. dated 2005-04-26"
6889216,physical neural network design incorporating nanotechnology,"a physical neural network based on nanotechnology, including methods thereof. such a physical neural network generally includes one or more neuron-like nodes, which are formed from a plurality of interconnected nanoconnections formed from nanoconductors. each neuron-like node sums one or more input signals and generates one or more output signals based on a threshold associated with the input signal. the physical neural network also includes a connection network formed from the interconnected nanoconnections, such that the interconnected nanoconnections used thereof by one or more of the neuron-like nodes are strengthened or weakened according to an application of an electric field.",2005-05-03,"The title of the patent is physical neural network design incorporating nanotechnology and its abstract is a physical neural network based on nanotechnology, including methods thereof. such a physical neural network generally includes one or more neuron-like nodes, which are formed from a plurality of interconnected nanoconnections formed from nanoconductors. each neuron-like node sums one or more input signals and generates one or more output signals based on a threshold associated with the input signal. the physical neural network also includes a connection network formed from the interconnected nanoconnections, such that the interconnected nanoconnections used thereof by one or more of the neuron-like nodes are strengthened or weakened according to an application of an electric field. dated 2005-05-03"
6889691,auto cpap,"a method for the detection and treatment of disordered breathing during sleep employs an artificial neural network (ann) in which data related to breathing gas flow are analyzed. a respiratory circuit is established by connecting the patient to a continuous positive airway pressure (cpap) system with pressurized breathing gas supply, the gas flow in the circuit is periodically sampled, one or several cepstrum parameters distinctive of various breathing patterns are periodically calculated; the parameter values are periodically fed to an ann trained to recognize breathing patterns characteristic of sleep disordered breathing and are analyzed in the network, the cpap pressurized breathing gas supply is controlled in response to the ann output. also disclosed is a corresponding apparatus.",2005-05-10,"The title of the patent is auto cpap and its abstract is a method for the detection and treatment of disordered breathing during sleep employs an artificial neural network (ann) in which data related to breathing gas flow are analyzed. a respiratory circuit is established by connecting the patient to a continuous positive airway pressure (cpap) system with pressurized breathing gas supply, the gas flow in the circuit is periodically sampled, one or several cepstrum parameters distinctive of various breathing patterns are periodically calculated; the parameter values are periodically fed to an ann trained to recognize breathing patterns characteristic of sleep disordered breathing and are analyzed in the network, the cpap pressurized breathing gas supply is controlled in response to the ann output. also disclosed is a corresponding apparatus. dated 2005-05-10"
6892194,system and method for organizing color values using an artificial intelligence based cluster model,"a system and method for organizing a plurality of sets color values into a plurality of color groups, such as paint, pigments, or dye formulations, is provided. the inputs to the system are the color values of a proposed paint, dye or colorant formulation and color measurement angles. the system includes an input device for entering a plurality of sets of color values and an artificial intelligence cluster model coupled to the input device. the cluster model produces an output signal indicative of the one color group to which a set of color values belongs. the artificial intelligence model may be embodied in a neural network. more specifically, the cluster model may be a self-organizing map neural network.",2005-05-10,"The title of the patent is system and method for organizing color values using an artificial intelligence based cluster model and its abstract is a system and method for organizing a plurality of sets color values into a plurality of color groups, such as paint, pigments, or dye formulations, is provided. the inputs to the system are the color values of a proposed paint, dye or colorant formulation and color measurement angles. the system includes an input device for entering a plurality of sets of color values and an artificial intelligence cluster model coupled to the input device. the cluster model produces an output signal indicative of the one color group to which a set of color values belongs. the artificial intelligence model may be embodied in a neural network. more specifically, the cluster model may be a self-organizing map neural network. dated 2005-05-10"
6895396,neural network methods to predict enzyme inhibitor or receptor ligand potency,"a new method to analyze and predict the binding energy for enzyme-transition state inhibitor interactions is presented. computational neural networks are employed to discovery quantum mechanical features of transition states and putative inhibitors necessary for binding. the method is able to generate its own relationship between the quantum mechanical structure of the inhibitor and the strength of binding. feed-forward neural networks with back propagation of error can be trained to recognize the quantum mechanical electrostatic potential at the entire van der waals surface, rather than a collapsed representation, of a group of training inhibitors and to predict the strength of interactions between the enzyme and a group of novel inhibitors. the experimental results show that the neural networks can predict with quantitative accuracy the binding strength of new inhibitors. the method is in fact able to predict the large binding free energy of the transition state, when trained with less tightly bound inhibitors. the present method is also applicable to prediction of the binding free energy of a ligand to a receptor. the application of this approach to the study of transition state inhibitors and ligands would permit evaluation of chemical libraries of potential inhibitory, agonistic, or antagonistic agents. the method is amenable to incorporation in a computer-readable medium accessible by general-purpose computers.",2005-05-17,"The title of the patent is neural network methods to predict enzyme inhibitor or receptor ligand potency and its abstract is a new method to analyze and predict the binding energy for enzyme-transition state inhibitor interactions is presented. computational neural networks are employed to discovery quantum mechanical features of transition states and putative inhibitors necessary for binding. the method is able to generate its own relationship between the quantum mechanical structure of the inhibitor and the strength of binding. feed-forward neural networks with back propagation of error can be trained to recognize the quantum mechanical electrostatic potential at the entire van der waals surface, rather than a collapsed representation, of a group of training inhibitors and to predict the strength of interactions between the enzyme and a group of novel inhibitors. the experimental results show that the neural networks can predict with quantitative accuracy the binding strength of new inhibitors. the method is in fact able to predict the large binding free energy of the transition state, when trained with less tightly bound inhibitors. the present method is also applicable to prediction of the binding free energy of a ligand to a receptor. the application of this approach to the study of transition state inhibitors and ligands would permit evaluation of chemical libraries of potential inhibitory, agonistic, or antagonistic agents. the method is amenable to incorporation in a computer-readable medium accessible by general-purpose computers. dated 2005-05-17"
6898583,"method and apparatus of creating application-specific, non-uniform wavelet transforms","a method and an apparatus of designing a set of wavelet basis trained to fit a particular problem. the method and apparatus include constructing a neural network of arbitrary complexity using a discrete and finite radon transform, feeding an input wavelet prototype through the neural network and its backpropagation to produce an output, and modifying the input wavelet prototype using the output.",2005-05-24,"The title of the patent is method and apparatus of creating application-specific, non-uniform wavelet transforms and its abstract is a method and an apparatus of designing a set of wavelet basis trained to fit a particular problem. the method and apparatus include constructing a neural network of arbitrary complexity using a discrete and finite radon transform, feeding an input wavelet prototype through the neural network and its backpropagation to produce an output, and modifying the input wavelet prototype using the output. dated 2005-05-24"
6901391,field/reservoir optimization utilizing neural networks,"a method of optimizing performance of a well system utilizes a neural network. in a described embodiment, the method includes the step of accumulating data indicative of the performance of the well system in response to variable influencing parameters. the data is used to train a neural network to model an output of the well system in response to the influencing parameters. an output of the neural network may then be input to a valuing model, e.g., to permit optimization of a value of the well system. the optimization process yields a set of prospective influencing parameters which may be incorporated into the well system to maximize its value.",2005-05-31,"The title of the patent is field/reservoir optimization utilizing neural networks and its abstract is a method of optimizing performance of a well system utilizes a neural network. in a described embodiment, the method includes the step of accumulating data indicative of the performance of the well system in response to variable influencing parameters. the data is used to train a neural network to model an output of the well system in response to the influencing parameters. an output of the neural network may then be input to a valuing model, e.g., to permit optimization of a value of the well system. the optimization process yields a set of prospective influencing parameters which may be incorporated into the well system to maximize its value. dated 2005-05-31"
6904422,adaptive control system having direct output feedback and related apparatuses and methods,"an adaptive control system (acs) uses direct output feedback to control a plant. the acs uses direct adaptive output feedback control developed for highly uncertain nonlinear systems, that does not rely on state estimation. the approach is also applicable to systems of unknown, but bounded dimension, whose output has known, but otherwise arbitrary relative degree. this includes systems with both parameter uncertainty and unmodeled dynamics. the result is achieved by extending the universal function approximation property of linearly parameterized neural networks to model unknown system dynamics from input/output data. the network weight adaptation rule is derived from lyapunov stability analysis, and guarantees that the adapted weight errors and the tracking error are bounded.",2005-06-07,"The title of the patent is adaptive control system having direct output feedback and related apparatuses and methods and its abstract is an adaptive control system (acs) uses direct output feedback to control a plant. the acs uses direct adaptive output feedback control developed for highly uncertain nonlinear systems, that does not rely on state estimation. the approach is also applicable to systems of unknown, but bounded dimension, whose output has known, but otherwise arbitrary relative degree. this includes systems with both parameter uncertainty and unmodeled dynamics. the result is achieved by extending the universal function approximation property of linearly parameterized neural networks to model unknown system dynamics from input/output data. the network weight adaptation rule is derived from lyapunov stability analysis, and guarantees that the adapted weight errors and the tracking error are bounded. dated 2005-06-07"
6907398,compressing hmm prototypes,a method is described for compressing the storage space required by hmm prototypes in an electronic memory. for this purpose prescribed hmm prototypes are mapped onto compressed hmm prototypes with the aid of a neural network (encoder). these can be stored with a smaller storage space than the uncompressed hmm prototypes. a second neural network (decoder) serves to reconstruct the hmm prototypes.,2005-06-14,The title of the patent is compressing hmm prototypes and its abstract is a method is described for compressing the storage space required by hmm prototypes in an electronic memory. for this purpose prescribed hmm prototypes are mapped onto compressed hmm prototypes with the aid of a neural network (encoder). these can be stored with a smaller storage space than the uncompressed hmm prototypes. a second neural network (decoder) serves to reconstruct the hmm prototypes. dated 2005-06-14
6907412,visualization and self-organization of multidimensional data through equalized orthogonal mapping,"the subject system provides reduced-dimension mapping of pattern data. mapping is applied through conventional single-hidden-layer feed-forward neural network with non-linear neurons. according to one aspect of the present invention, the system functions to equalize and orthogonalize lower dimensional output signals by reducing the covariance matrix of the output signals to the form of a diagonal matrix or constant times the identity matrix. the present invention allows for visualization of large bodies of complex multidimensional data in a relatively “topologically correct” low-dimension approximation, to reduce randomness associated with other methods of similar purposes, and to keep the mapping computationally efficient at the same time.",2005-06-14,"The title of the patent is visualization and self-organization of multidimensional data through equalized orthogonal mapping and its abstract is the subject system provides reduced-dimension mapping of pattern data. mapping is applied through conventional single-hidden-layer feed-forward neural network with non-linear neurons. according to one aspect of the present invention, the system functions to equalize and orthogonalize lower dimensional output signals by reducing the covariance matrix of the output signals to the form of a diagonal matrix or constant times the identity matrix. the present invention allows for visualization of large bodies of complex multidimensional data in a relatively “topologically correct” low-dimension approximation, to reduce randomness associated with other methods of similar purposes, and to keep the mapping computationally efficient at the same time. dated 2005-06-14"
6907591,method and apparatus for performing extraction using a neural network,"a system for using machine-learning to create a model for performing integrated circuit layout extraction is disclosed. the system of the present invention has two main phases: model creation and model application. the model creation phase comprises creating one or more extraction models using machine-learning techniques. first, a complex extraction problem is decomposed into smaller simpler extraction problems. then, each smaller extraction problem is then analyzed to identify a set of physical parameters that fully define the smaller extraction problem. next, models are created using machine learning techniques for all of the smaller simpler extraction problems. the machine learning is performed by first creating training data sets composed of the identified parameters from typical examples of the smaller extraction problem and the answers to those example extraction problems as solved using a highly accurate physics-based field solver. next, the system trains a set of neural networks using the training sets. in one embodiment, bayesian inference is used to train the neural networks that are used to model the extraction. after the creation the neural network based models for each of the smaller simpler extraction problems, the neural network based models may be used for extraction.",2005-06-14,"The title of the patent is method and apparatus for performing extraction using a neural network and its abstract is a system for using machine-learning to create a model for performing integrated circuit layout extraction is disclosed. the system of the present invention has two main phases: model creation and model application. the model creation phase comprises creating one or more extraction models using machine-learning techniques. first, a complex extraction problem is decomposed into smaller simpler extraction problems. then, each smaller extraction problem is then analyzed to identify a set of physical parameters that fully define the smaller extraction problem. next, models are created using machine learning techniques for all of the smaller simpler extraction problems. the machine learning is performed by first creating training data sets composed of the identified parameters from typical examples of the smaller extraction problem and the answers to those example extraction problems as solved using a highly accurate physics-based field solver. next, the system trains a set of neural networks using the training sets. in one embodiment, bayesian inference is used to train the neural networks that are used to model the extraction. after the creation the neural network based models for each of the smaller simpler extraction problems, the neural network based models may be used for extraction. dated 2005-06-14"
6910025,modeling behavior of an electrical circuit,"behavior of an electrical circuit can be modeled using a trained neural network. for example, using one or more neural networks, power consumption, including leakage power and switching energy, can be estimated. also, a profile of current versus time can be generated for the electrical circuit. a hierarchy of neural networks may be used to model the circuit at different levels. in one embodiment, a circuit behavior is modeled using one or more neural networks, cluster values, and cluster probabilities.",2005-06-21,"The title of the patent is modeling behavior of an electrical circuit and its abstract is behavior of an electrical circuit can be modeled using a trained neural network. for example, using one or more neural networks, power consumption, including leakage power and switching energy, can be estimated. also, a profile of current versus time can be generated for the electrical circuit. a hierarchy of neural networks may be used to model the circuit at different levels. in one embodiment, a circuit behavior is modeled using one or more neural networks, cluster values, and cluster probabilities. dated 2005-06-21"
6911006,pain inferring device and pain inferring method,"a pain inferring device for designing a shape giving a subject as little pain as possible when the subject touches the shape. the pain inferring device includes an input unit, an output unit, a main control unit, a learning storage unit, and a neural network. the neural network learns the relationship between the input value of shape data and the output value of the data on the degree of pain when a subject touches a shape. by the learning, an input/output function (namely, the coefficient of coupling of neurons in layers) representing the relationship between the input and output values is defined and stored in the learning storage unit. when the shape data is inputted through the input unit, the neural network infers the degree of pain by using the function stored in the learning storage unit.",2005-06-28,"The title of the patent is pain inferring device and pain inferring method and its abstract is a pain inferring device for designing a shape giving a subject as little pain as possible when the subject touches the shape. the pain inferring device includes an input unit, an output unit, a main control unit, a learning storage unit, and a neural network. the neural network learns the relationship between the input value of shape data and the output value of the data on the degree of pain when a subject touches a shape. by the learning, an input/output function (namely, the coefficient of coupling of neurons in layers) representing the relationship between the input and output values is defined and stored in the learning storage unit. when the shape data is inputted through the input unit, the neural network infers the degree of pain by using the function stored in the learning storage unit. dated 2005-06-28"
6912861,vehicle air conditioner,"a vehicle air conditioner has a front air conditioning unit for front seats of a vehicle and a rear air conditioning unit for rear seats of the vehicle. the front and rear air conditioning units are controlled by an air conditioning ecu. the ecu uses a non-linear model, such as a neural network, to determine a target blowout temperature, a blower voltage, and blowout port modes of the front air conditioning unit. the ecu uses a linear model to determine a target blowout temperature, a blower voltage, and blowout port modes of the rear air conditioning unit.",2005-07-05,"The title of the patent is vehicle air conditioner and its abstract is a vehicle air conditioner has a front air conditioning unit for front seats of a vehicle and a rear air conditioning unit for rear seats of the vehicle. the front and rear air conditioning units are controlled by an air conditioning ecu. the ecu uses a non-linear model, such as a neural network, to determine a target blowout temperature, a blower voltage, and blowout port modes of the front air conditioning unit. the ecu uses a linear model to determine a target blowout temperature, a blower voltage, and blowout port modes of the rear air conditioning unit. dated 2005-07-05"
6914961,speed binning by neural network,"a method of allocating subscriber lines in a telecommunications network into speed bins. with the method, more intelligent business actions can then be taken in the provision of high-speed data services over the subscriber lines. for example, only qualified lines might be used for high-speed data services, with the other lines being allocated to pots service. the lines are divided into speed bins using a pair of neural networks, with one predicting upstream speed and one predicting downstream speed. the combined predictions are then mapped to a speed bin, which is the basis for further business actions. the disclosure describes that the neural networks are created using conditional fuzzy logic to precondition the neural networks by line speed.",2005-07-05,"The title of the patent is speed binning by neural network and its abstract is a method of allocating subscriber lines in a telecommunications network into speed bins. with the method, more intelligent business actions can then be taken in the provision of high-speed data services over the subscriber lines. for example, only qualified lines might be used for high-speed data services, with the other lines being allocated to pots service. the lines are divided into speed bins using a pair of neural networks, with one predicting upstream speed and one predicting downstream speed. the combined predictions are then mapped to a speed bin, which is the basis for further business actions. the disclosure describes that the neural networks are created using conditional fuzzy logic to precondition the neural networks by line speed. dated 2005-07-05"
6915217,laser doppler vibrometer for remote assessment of structural components,"a method and system for remotely inspecting the integrity of a structure. this can be performed by a method creating a vibratory response in the structure from a remote location and then measuring the vibratory response of the structure remotely. alternatively, this can be performed by a system for remotely measuring the integrity of a structure using a vehicle and an artificial neural network, where the vehicle is equipped with a vibratory response device. the vibratory response can be produced by infrasonic and audio frequencies that can be produced by at least a vehicle, motor, or sound recording. the vibratory response can be measured with a laser vibrometer or an audio recording device.",2005-07-05,"The title of the patent is laser doppler vibrometer for remote assessment of structural components and its abstract is a method and system for remotely inspecting the integrity of a structure. this can be performed by a method creating a vibratory response in the structure from a remote location and then measuring the vibratory response of the structure remotely. alternatively, this can be performed by a system for remotely measuring the integrity of a structure using a vehicle and an artificial neural network, where the vehicle is equipped with a vibratory response device. the vibratory response can be produced by infrasonic and audio frequencies that can be produced by at least a vehicle, motor, or sound recording. the vibratory response can be measured with a laser vibrometer or an audio recording device. dated 2005-07-05"
6915283,"information processing apparatus and method, and recording medium","an information processing apparatus includes a first recurrent neural network (rnn) for performing processing which corresponds to a time-series and a second rnn for processing another correlated time-series. the difference between a context set output by the first rnn and a context set output by the second rnn is computed by a subtractor, and the obtained difference is used as a prediction error. backpropagation is performed based on the prediction error, thus determining a coefficient for each neuron of an output layer, an intermediate layer, and an input layer.",2005-07-05,"The title of the patent is information processing apparatus and method, and recording medium and its abstract is an information processing apparatus includes a first recurrent neural network (rnn) for performing processing which corresponds to a time-series and a second rnn for processing another correlated time-series. the difference between a context set output by the first rnn and a context set output by the second rnn is computed by a subtractor, and the obtained difference is used as a prediction error. backpropagation is performed based on the prediction error, thus determining a coefficient for each neuron of an output layer, an intermediate layer, and an input layer. dated 2005-07-05"
6917703,method and apparatus for image analysis of a gabor-wavelet transformed image using a neural network,"the present invention may be embodied in a method, and in a related apparatus, for classifying a feature in an image frame. in the method, an original image frame having an array of pixels is transformed using gabor-wavelet transformations to generate a transformed image frame. each pixel of the transformed image is associated with a respective pixel of the original image frame and is represented by a predetermined number of wavelet component values. a pixel of the transformed image frame associated with the feature is selected for analysis. a neural network is provided that has an output and a predetermined number of inputs. each input of the neural network is associated with a respective wavelet component value of the selected pixel. the neural network classifies the local feature based on the wavelet component values, and indicates a class of the feature at an output of the neural network.",2005-07-12,"The title of the patent is method and apparatus for image analysis of a gabor-wavelet transformed image using a neural network and its abstract is the present invention may be embodied in a method, and in a related apparatus, for classifying a feature in an image frame. in the method, an original image frame having an array of pixels is transformed using gabor-wavelet transformations to generate a transformed image frame. each pixel of the transformed image is associated with a respective pixel of the original image frame and is represented by a predetermined number of wavelet component values. a pixel of the transformed image frame associated with the feature is selected for analysis. a neural network is provided that has an output and a predetermined number of inputs. each input of the neural network is associated with a respective wavelet component value of the selected pixel. the neural network classifies the local feature based on the wavelet component values, and indicates a class of the feature at an output of the neural network. dated 2005-07-12"
6919842,hybrid navigation system using neural network,"a hybrid navigation system including a neural network is provided. the hybrid navigation system includes a global positioning system (gps) as a main system using satellites and a radio determination system when there are difficulties in reception of signals from the satellites. the hybrid navigation system comprises a gps signal processor, a tdoa signal processor and a neural network. the gps signal processor receives gps signals from the satellites to determine positions. the tdoa signal processor receives determination signals from mobile communication stations to determine positions. the neural network uses position values inputted from each of the gps signal processor and the tdoa signal processor to learn and predict the position of the mobile terminal.",2005-07-19,"The title of the patent is hybrid navigation system using neural network and its abstract is a hybrid navigation system including a neural network is provided. the hybrid navigation system includes a global positioning system (gps) as a main system using satellites and a radio determination system when there are difficulties in reception of signals from the satellites. the hybrid navigation system comprises a gps signal processor, a tdoa signal processor and a neural network. the gps signal processor receives gps signals from the satellites to determine positions. the tdoa signal processor receives determination signals from mobile communication stations to determine positions. the neural network uses position values inputted from each of the gps signal processor and the tdoa signal processor to learn and predict the position of the mobile terminal. dated 2005-07-19"
6920440,forming a signature of parameters extracted from information,"a method of storing information relating to the transmission of messages by an entity over a given time period comprises the step of creating a signature comprising a plurality of parameters related to the transmission of messages over that time period wherein the parameters comprise at least one parameter related to the transmission of messages over a portion of the period and also related to the position of the portion in the period, to enable output data to be derived from the stored information. the signature may be updated by a weighted averaging process with other more recent signatures. application in fraud detection where signature representing information in many call detail records from a particular subscriber is fed to a neural network.",2005-07-19,"The title of the patent is forming a signature of parameters extracted from information and its abstract is a method of storing information relating to the transmission of messages by an entity over a given time period comprises the step of creating a signature comprising a plurality of parameters related to the transmission of messages over that time period wherein the parameters comprise at least one parameter related to the transmission of messages over a portion of the period and also related to the position of the portion in the period, to enable output data to be derived from the stored information. the signature may be updated by a weighted averaging process with other more recent signatures. application in fraud detection where signature representing information in many call detail records from a particular subscriber is fed to a neural network. dated 2005-07-19"
6922680,method and apparatus for recommending an item of interest using a radial basis function to fuse a plurality of recommendation scores,"a method and apparatus are disclosed for recommending items of interest by fusing a plurality of recommendation scores from individual recommendation tools using one or more radial basis function neural networks. the radial basis function neural networks include n inputs and at least one output, interconnected by a plurality of hidden units in a hidden layer. a unique neural network can be used for each user, or a neural network can be shared by a plurality of users, such as a set of users having similar characteristics. a neural network training process initially trains each radial basis function neural network using data from a training data set. a neural network cross-validation process selects the radial basis function neural network that performs best on the cross-validation data set. a neural network program recommendation process uses the selected neural network(s) to recommend items of interest to a user.",2005-07-26,"The title of the patent is method and apparatus for recommending an item of interest using a radial basis function to fuse a plurality of recommendation scores and its abstract is a method and apparatus are disclosed for recommending items of interest by fusing a plurality of recommendation scores from individual recommendation tools using one or more radial basis function neural networks. the radial basis function neural networks include n inputs and at least one output, interconnected by a plurality of hidden units in a hidden layer. a unique neural network can be used for each user, or a neural network can be shared by a plurality of users, such as a set of users having similar characteristics. a neural network training process initially trains each radial basis function neural network using data from a training data set. a neural network cross-validation process selects the radial basis function neural network that performs best on the cross-validation data set. a neural network program recommendation process uses the selected neural network(s) to recommend items of interest to a user. dated 2005-07-26"
6924463,pyrometer calibrated wafer temperature estimator,"a wafer temperature estimator calibrates contact-type temperature sensor measurements that are used by a temperature controller to control substrate temperature in a high temperature processing chamber. wafer temperature estimator parameters provide an estimated wafer temperature from contact-type temperature sensor measurements. the estimator parameters are refined using non-contact-type temperature sensor measurements during periods when the substrate temperature is decreasing or the heaters are off. a corresponding temperature control system includes a heater, a contact-type temperature sensor in close proximity to the substrate, and an optical pyrometer placed to read temperature directly from the substrate. a wafer temperature estimator uses the estimator parameters and measurements from the contact-type sensor to determine an estimated wafer temperature. a temperature controller reads the estimated wafer temperature and makes changes to the heater power accordingly. the wafer temperature estimator has a nonlinear neural network system that is trained using inputs from the various sensors.",2005-08-02,"The title of the patent is pyrometer calibrated wafer temperature estimator and its abstract is a wafer temperature estimator calibrates contact-type temperature sensor measurements that are used by a temperature controller to control substrate temperature in a high temperature processing chamber. wafer temperature estimator parameters provide an estimated wafer temperature from contact-type temperature sensor measurements. the estimator parameters are refined using non-contact-type temperature sensor measurements during periods when the substrate temperature is decreasing or the heaters are off. a corresponding temperature control system includes a heater, a contact-type temperature sensor in close proximity to the substrate, and an optical pyrometer placed to read temperature directly from the substrate. a wafer temperature estimator uses the estimator parameters and measurements from the contact-type sensor to determine an estimated wafer temperature. a temperature controller reads the estimated wafer temperature and makes changes to the heater power accordingly. the wafer temperature estimator has a nonlinear neural network system that is trained using inputs from the various sensors. dated 2005-08-02"
6925361,distributed energy neural network integration system,a system that couples distributed power generators together as a collective unit for the purposes of selling or purchasing energy from the electrical power grid. the apparatus includes a charge/discharge controller and an adaptive controller. the charge/discharge controller transfers energy generated by the plurality of distributed power generators to the power grid. the adaptive controller directs when the charge/discharge controller transfers energy generated by at least one of the plurality of distributed power generators to the electrical grid.,2005-08-02,The title of the patent is distributed energy neural network integration system and its abstract is a system that couples distributed power generators together as a collective unit for the purposes of selling or purchasing energy from the electrical power grid. the apparatus includes a charge/discharge controller and an adaptive controller. the charge/discharge controller transfers energy generated by the plurality of distributed power generators to the power grid. the adaptive controller directs when the charge/discharge controller transfers energy generated by at least one of the plurality of distributed power generators to the electrical grid. dated 2005-08-02
6928371,monitoring system of vrla battery capacitance,"a method and apparatus for monitoring the capacity of a valve regulated lead acid battery comprising at least one battery monitor connected to the valve regulated lead acid battery; a centralized system connecting the battery monitor through an industry standard data system to a central office; and an alarm connected to the centralized system; wherein, a short-term discharge test is performed on the battery using the battery monitor which provides input parameters for a neural network and fuzzy logic network used in combination with a prediction algorithm to calculate the predicted capacity; and, wherein, the alarm is activated when said predicted capacity falls below eighty percent, when an individual cell voltage is reduced to 1.95 volts or less, or when a system failure occurs",2005-08-09,"The title of the patent is monitoring system of vrla battery capacitance and its abstract is a method and apparatus for monitoring the capacity of a valve regulated lead acid battery comprising at least one battery monitor connected to the valve regulated lead acid battery; a centralized system connecting the battery monitor through an industry standard data system to a central office; and an alarm connected to the centralized system; wherein, a short-term discharge test is performed on the battery using the battery monitor which provides input parameters for a neural network and fuzzy logic network used in combination with a prediction algorithm to calculate the predicted capacity; and, wherein, the alarm is activated when said predicted capacity falls below eighty percent, when an individual cell voltage is reduced to 1.95 volts or less, or when a system failure occurs dated 2005-08-09"
6931144,automated rip tide detection system,"a system substitutes digitized camera images for human vision, in determining the presence or absence of rip tides among sea water wave patterns at a public swimming beach. computer analysis of these images involves image pre-filtering that enhances the telltale signs of rip tides, before the digital data is processed for classification as normal or rip tide. the classification itself can proceed along by expert systems which mimic the manner in which a human observer performs the detection; or by building a neural network, that determines its own classification criteria for identifying rip tides.",2005-08-16,"The title of the patent is automated rip tide detection system and its abstract is a system substitutes digitized camera images for human vision, in determining the presence or absence of rip tides among sea water wave patterns at a public swimming beach. computer analysis of these images involves image pre-filtering that enhances the telltale signs of rip tides, before the digital data is processed for classification as normal or rip tide. the classification itself can proceed along by expert systems which mimic the manner in which a human observer performs the detection; or by building a neural network, that determines its own classification criteria for identifying rip tides. dated 2005-08-16"
6931269,multi-domain motion estimation and plethysmographic recognition using fuzzy neural-nets,"pulse oximetry is improved through classification of plethysmographic signals by processing the plethysmographic signals using a neural network that receives input coefficients from multiple signal domains including, for example, spectral, bispectral, cepstral and wavelet filtered signal domains. in one embodiment, a plethysmographic signal obtained from a patient is transformed (240) from a first domain to a plurality of different signal domains (242, 243, 244, 245) to obtain a corresponding plurality of transformed plethysmographic signals. a plurality of sets of coefficients derived from the transformed plethysmographic signals are selected and directed to an input layer (251) of a neural network (250). the plethysmographic signal is classified by an output layer (253) of the neural network (250) that is connected to the input layer (251) by one or more hidden layers (252).",2005-08-16,"The title of the patent is multi-domain motion estimation and plethysmographic recognition using fuzzy neural-nets and its abstract is pulse oximetry is improved through classification of plethysmographic signals by processing the plethysmographic signals using a neural network that receives input coefficients from multiple signal domains including, for example, spectral, bispectral, cepstral and wavelet filtered signal domains. in one embodiment, a plethysmographic signal obtained from a patient is transformed (240) from a first domain to a plurality of different signal domains (242, 243, 244, 245) to obtain a corresponding plurality of transformed plethysmographic signals. a plurality of sets of coefficients derived from the transformed plethysmographic signals are selected and directed to an input layer (251) of a neural network (250). the plethysmographic signal is classified by an output layer (253) of the neural network (250) that is connected to the input layer (251) by one or more hidden layers (252). dated 2005-08-16"
6933856,adaptive acoustic transmitter controller apparatus and method,"the invention describes a method and apparatus for effectively communicating data along the acoustic channel of a subterranean well. the method comprises optimally driving an acoustic transmitter with an adaptive transmitter controller. a data signal is transmitted along the acoustic channel and detected as a distorted signal along the acoustic channel. the distorted signal is input to the adaptive transmitter controller which, based on the detected signal, modifies later transmissions to counteract the distorting effects of the transmitter and acoustic channel. the adaptive transmitter controller preferably comprises a neural network. another receiver may be employed, at a point further from the transmitter, to receive the optimized signals.",2005-08-23,"The title of the patent is adaptive acoustic transmitter controller apparatus and method and its abstract is the invention describes a method and apparatus for effectively communicating data along the acoustic channel of a subterranean well. the method comprises optimally driving an acoustic transmitter with an adaptive transmitter controller. a data signal is transmitted along the acoustic channel and detected as a distorted signal along the acoustic channel. the distorted signal is input to the adaptive transmitter controller which, based on the detected signal, modifies later transmissions to counteract the distorting effects of the transmitter and acoustic channel. the adaptive transmitter controller preferably comprises a neural network. another receiver may be employed, at a point further from the transmitter, to receive the optimized signals. dated 2005-08-23"
6934217,countermeasure threat emulator and method,"a system and method is disclosed for a countermeasure threat emulator (cme) provided in a tubular housing that may be launched from a submarine or ship. the cme electronics include a cpu board for running software, communicating with a computer external to the housing and data recording. the external computer preferably incorporates a database having data representative of a plurality of both foreign and domestic countermeasures. the data may be downloaded to the cpu board as well as updated for reprogramming of the cpu board. a digital signal processing board utilizes a plurality of dsp processors for running software capable of producing a wide range of acoustic signal outputs. a neural network may be used for analyzing and identifying acoustic sounds from incoming threats and notifying the cpu board for selection of a preprogrammed response for transmission by a transducer stack.",2005-08-23,"The title of the patent is countermeasure threat emulator and method and its abstract is a system and method is disclosed for a countermeasure threat emulator (cme) provided in a tubular housing that may be launched from a submarine or ship. the cme electronics include a cpu board for running software, communicating with a computer external to the housing and data recording. the external computer preferably incorporates a database having data representative of a plurality of both foreign and domestic countermeasures. the data may be downloaded to the cpu board as well as updated for reprogramming of the cpu board. a digital signal processing board utilizes a plurality of dsp processors for running software capable of producing a wide range of acoustic signal outputs. a neural network may be used for analyzing and identifying acoustic sounds from incoming threats and notifying the cpu board for selection of a preprogrammed response for transmission by a transducer stack. dated 2005-08-23"
6940740,multilevel semiconductor memory device and method for driving the same as a neuron element in a neural network computer,"a semiconductor device includes: a control-voltage supply unit 110; an mos transistor including a gate electrode 109 and drain and source regions 103a and 103b; a dielectric capacitor 104; and a resistor 106. the dielectric capacitor 104 and the resistor 106 are disposed in parallel and interposed between the gate electrode 109 and the control-voltage supply unit 110. with this structure, a charge is accumulated in each of an intermediate electrode of the dielectric capacitor 104 and the gate electrode 109 upon the application of a voltage, thereby varying a threshold value of the mos transistor. in this manner, the history of input signals can be stored as a variation in a drain current in the mos transistor, thus allowing multilevel information to be held.",2005-09-06,"The title of the patent is multilevel semiconductor memory device and method for driving the same as a neuron element in a neural network computer and its abstract is a semiconductor device includes: a control-voltage supply unit 110; an mos transistor including a gate electrode 109 and drain and source regions 103a and 103b; a dielectric capacitor 104; and a resistor 106. the dielectric capacitor 104 and the resistor 106 are disposed in parallel and interposed between the gate electrode 109 and the control-voltage supply unit 110. with this structure, a charge is accumulated in each of an intermediate electrode of the dielectric capacitor 104 and the gate electrode 109 upon the application of a voltage, thereby varying a threshold value of the mos transistor. in this manner, the history of input signals can be stored as a variation in a drain current in the mos transistor, thus allowing multilevel information to be held. dated 2005-09-06"
6941254,method and system intended for real-time estimation of the flow mode of a multiphase fluid stream at all points of a pipe,"the invention is a method and system for real-time estimation of the flow mode, at all points of a pipe whose structure is defined by a certain number of structure parameters, of a multiphase fluid stream defined by several physical quantities providing simplified implementation of hydrodynamic modules that can be integrated in modelling tools. a non-linear neural network is formed with an input layer having as many inputs as there are structure parameters and physical quantities, an output layer with as many outputs as there are quantities necessary for estimation of the flow mode and at least one intermediate layer. a learning base is created with predetermined tables connecting various values obtained for the output data to the corresponding values of the input data, with iterative determination of the weighting factors of the activation function allowing to properly connect the values in the input and output data tables. in order to avoid singularities of the network output data likely to distort the determination of the weighting factors, a sorting procedure is used to eliminate non-pertinent data. the main advantages of the method are: modelling simplification and time saving.",2005-09-06,"The title of the patent is method and system intended for real-time estimation of the flow mode of a multiphase fluid stream at all points of a pipe and its abstract is the invention is a method and system for real-time estimation of the flow mode, at all points of a pipe whose structure is defined by a certain number of structure parameters, of a multiphase fluid stream defined by several physical quantities providing simplified implementation of hydrodynamic modules that can be integrated in modelling tools. a non-linear neural network is formed with an input layer having as many inputs as there are structure parameters and physical quantities, an output layer with as many outputs as there are quantities necessary for estimation of the flow mode and at least one intermediate layer. a learning base is created with predetermined tables connecting various values obtained for the output data to the corresponding values of the input data, with iterative determination of the weighting factors of the activation function allowing to properly connect the values in the input and output data tables. in order to avoid singularities of the network output data likely to distort the determination of the weighting factors, a sorting procedure is used to eliminate non-pertinent data. the main advantages of the method are: modelling simplification and time saving. dated 2005-09-06"
6941289,hybrid neural network generation system and method,"a computer-implemented method and system for building a neural network is disclosed. the neural network predicts at least one target based upon predictor variables defined in a state space. first, an input data set is retrieved that includes the predictor variables and at least one target associated with the predictor variables for each observation. in the state space, a number of points is inserted in the state space based upon the values of the predictor variables. the number of points is less than the number of observations. a statistical measure is determined that describes a relationship between the observations and the inserted points. weights and activation functions of the neural network are determined using the statistical measure.",2005-09-06,"The title of the patent is hybrid neural network generation system and method and its abstract is a computer-implemented method and system for building a neural network is disclosed. the neural network predicts at least one target based upon predictor variables defined in a state space. first, an input data set is retrieved that includes the predictor variables and at least one target associated with the predictor variables for each observation. in the state space, a number of points is inserted in the state space based upon the values of the predictor variables. the number of points is less than the number of observations. a statistical measure is determined that describes a relationship between the observations and the inserted points. weights and activation functions of the neural network are determined using the statistical measure. dated 2005-09-06"
6943358,method for developing a calibration algorithm for quantifying the hydrocarbon content of aqueous media,"a method for developing an algorithm for quantifying the hydrocarbon content of aqueous media includes: a) irradiating aqueous test samples containing hydrocarbons and particulates with light so that fluorescent emissions and scattered light signals are emitted from the test samples; b) detecting fluorescent emissions and the scattered light signals emitted from the test samples; c) generating first data signals representing the intensities of the fluorescent emissions, and second data signals representing the intensities and scatter angles of the second data signals; d) storing representations of the first and second data signals to create a data set; e) dividing the data set into training, test, and validation data sets; f) selecting input parameters from the data set; g) defining and training a neural network having hidden node using the training and the test data sets; and h) validating the neural network using the validation data set.",2005-09-13,"The title of the patent is method for developing a calibration algorithm for quantifying the hydrocarbon content of aqueous media and its abstract is a method for developing an algorithm for quantifying the hydrocarbon content of aqueous media includes: a) irradiating aqueous test samples containing hydrocarbons and particulates with light so that fluorescent emissions and scattered light signals are emitted from the test samples; b) detecting fluorescent emissions and the scattered light signals emitted from the test samples; c) generating first data signals representing the intensities of the fluorescent emissions, and second data signals representing the intensities and scatter angles of the second data signals; d) storing representations of the first and second data signals to create a data set; e) dividing the data set into training, test, and validation data sets; f) selecting input parameters from the data set; g) defining and training a neural network having hidden node using the training and the test data sets; and h) validating the neural network using the validation data set. dated 2005-09-13"
6943724,identification and tracking of moving objects in detected synthetic aperture imagery,"a method of tracking a moving object in an image created by use of a synthetic aperture includes identifying a plurality of receive phase centers for an image collector, obtaining a synthetic aperture image using the plurality of receive phase centers, and reading a signature of the moving object based on the synthetic aperture image, where the reading of the signature includes identifying, in the synthetic aperture image, as a function of image collection time, a presence of the moving object. the reading of the signature may also include identifying a changing position of the moving object, and may include associating a plurality of range difference values with respective ones of the plurality of phase centers. a signature of a scatterer may be formed using only its associated δr-versus-time profile. the presence of a mover in the image may be determined by filtering the image to detect all or part of a signature, or by using one or more signatures to train a neural network to observe the mover directly.",2005-09-13,"The title of the patent is identification and tracking of moving objects in detected synthetic aperture imagery and its abstract is a method of tracking a moving object in an image created by use of a synthetic aperture includes identifying a plurality of receive phase centers for an image collector, obtaining a synthetic aperture image using the plurality of receive phase centers, and reading a signature of the moving object based on the synthetic aperture image, where the reading of the signature includes identifying, in the synthetic aperture image, as a function of image collection time, a presence of the moving object. the reading of the signature may also include identifying a changing position of the moving object, and may include associating a plurality of range difference values with respective ones of the plurality of phase centers. a signature of a scatterer may be formed using only its associated δr-versus-time profile. the presence of a mover in the image may be determined by filtering the image to detect all or part of a signature, or by using one or more signatures to train a neural network to observe the mover directly. dated 2005-09-13"
6944319,pose-invariant face recognition system and process,"a face recognition system and process for identifying a person depicted in an input image and their face pose. this system and process entails locating and extracting face regions belonging to known people from a set of model images, and determining the face pose for each of the face regions extracted. all the extracted face regions are preprocessed by normalizing, cropping, categorizing and finally abstracting them. more specifically, the images are normalized and cropped to show only a person's face, categorized according to the face pose of the depicted person's face by assigning them to one of a series of face pose ranges, and abstracted preferably via an eigenface approach. the preprocessed face images are preferably used to train a neural network ensemble having a first stage made up of a bank of face recognition neural networks each of which is dedicated to a particular pose range, and a second stage constituting a single fusing neural network that is used to combine the outputs from each of the first stage neural networks. once trained, the input of a face region which has been extracted from an input image and preprocessed (i.e., normalized, cropped and abstracted) will cause just one of the output units of the fusing portion of the neural network ensemble to become active. the active output unit indicates either the identify of the person whose face was extracted from the input image and the associated face pose, or that the identity of the person is unknown to the system.",2005-09-13,"The title of the patent is pose-invariant face recognition system and process and its abstract is a face recognition system and process for identifying a person depicted in an input image and their face pose. this system and process entails locating and extracting face regions belonging to known people from a set of model images, and determining the face pose for each of the face regions extracted. all the extracted face regions are preprocessed by normalizing, cropping, categorizing and finally abstracting them. more specifically, the images are normalized and cropped to show only a person's face, categorized according to the face pose of the depicted person's face by assigning them to one of a series of face pose ranges, and abstracted preferably via an eigenface approach. the preprocessed face images are preferably used to train a neural network ensemble having a first stage made up of a bank of face recognition neural networks each of which is dedicated to a particular pose range, and a second stage constituting a single fusing neural network that is used to combine the outputs from each of the first stage neural networks. once trained, the input of a face region which has been extracted from an input image and preprocessed (i.e., normalized, cropped and abstracted) will cause just one of the output units of the fusing portion of the neural network ensemble to become active. the active output unit indicates either the identify of the person whose face was extracted from the input image and the associated face pose, or that the identity of the person is unknown to the system. dated 2005-09-13"
6947378,dynamic network resource allocation using multimedia content features and traffic features,a method for dynamically allocating network resources while transferring multimedia at variable bit-rates in a network extracts first content features from the multimedia to determine renegotiation points and observation periods. second content features and traffic features are extracted from the multimedia bit stream during the observation periods. the second content features and the traffic features are combined in a neural network to predict the network resources to be allocated at the renegotiation points.,2005-09-20,The title of the patent is dynamic network resource allocation using multimedia content features and traffic features and its abstract is a method for dynamically allocating network resources while transferring multimedia at variable bit-rates in a network extracts first content features from the multimedia to determine renegotiation points and observation periods. second content features and traffic features are extracted from the multimedia bit stream during the observation periods. the second content features and the traffic features are combined in a neural network to predict the network resources to be allocated at the renegotiation points. dated 2005-09-20
6947870,neural network model for electric submersible pump system,"an apparatus and method is disclosed for modeling an electric submersible pump using a neural network, data from a deterministic model, and, optionally, data obtained from a real world electric submersible pump. it is emphasized that this abstract is provided to comply with the rules requiring an abstract which will allow a searcher or other reader to quickly ascertain the subject matter of the technical disclosure. it is submitted with the understanding that it will not be used to interpret or limit the scope or meaning of the claims.",2005-09-20,"The title of the patent is neural network model for electric submersible pump system and its abstract is an apparatus and method is disclosed for modeling an electric submersible pump using a neural network, data from a deterministic model, and, optionally, data obtained from a real world electric submersible pump. it is emphasized that this abstract is provided to comply with the rules requiring an abstract which will allow a searcher or other reader to quickly ascertain the subject matter of the technical disclosure. it is submitted with the understanding that it will not be used to interpret or limit the scope or meaning of the claims. dated 2005-09-20"
6947891,efficient speech recognition system bases on an auditory model,"a speech recognition system that is insensitive to external noise and applicable to actual life includes an a/d converter that converts analog voice signals to digital signals. an fir filtering section employs powers-of-two conversion to filter the digital signals converted at the a/d converter into numbers of channels. a characteristic extraction section immediately extracts speech characteristics having strong noise-resistance from the output signals of the fir filtering section without using additional memories. a word boundary detection section discriminates the information of the start-point and the end-point of a voice signal on the basis of the characteristics extracted by the characteristic extraction section. finally, a normalization/recognition section codes and outputs the final result by carrying out a timing normalization and a classifying process using a radial basis function (rbf) neural network on the basis of the voice characteristics provided by the characteristic extraction section and the information for the start-point and the end-point of the voice signal from the word boundary detection section.",2005-09-20,"The title of the patent is efficient speech recognition system bases on an auditory model and its abstract is a speech recognition system that is insensitive to external noise and applicable to actual life includes an a/d converter that converts analog voice signals to digital signals. an fir filtering section employs powers-of-two conversion to filter the digital signals converted at the a/d converter into numbers of channels. a characteristic extraction section immediately extracts speech characteristics having strong noise-resistance from the output signals of the fir filtering section without using additional memories. a word boundary detection section discriminates the information of the start-point and the end-point of a voice signal on the basis of the characteristics extracted by the characteristic extraction section. finally, a normalization/recognition section codes and outputs the final result by carrying out a timing normalization and a classifying process using a radial basis function (rbf) neural network on the basis of the voice characteristics provided by the characteristic extraction section and the information for the start-point and the end-point of the voice signal from the word boundary detection section. dated 2005-09-20"
6947915,multiresolution learning paradigm and signal prediction,"a neural network learning process provides a trained network that has good generalization ability for even highly nonlinear dynamic systems, and is trained with approximations of a signal obtained, each at a different respective resolution, using wavelet transformation. approximations are used in order from low to high. the trained neural network is used to predict values. in a preferred embodiment of the invention, the trained neural network is used in predicting network traffic patterns.",2005-09-20,"The title of the patent is multiresolution learning paradigm and signal prediction and its abstract is a neural network learning process provides a trained network that has good generalization ability for even highly nonlinear dynamic systems, and is trained with approximations of a signal obtained, each at a different respective resolution, using wavelet transformation. approximations are used in order from low to high. the trained neural network is used to predict values. in a preferred embodiment of the invention, the trained neural network is used in predicting network traffic patterns. dated 2005-09-20"
6948102,predictive failure analysis for storage networks,"a method of predicting failures of data storage devices in a data storage network. in predicting failures, birth and/or health records of data storage devices in the data storage network are requested. the records are then scaled and thresholded and processed using a probabilistic neural network to classify the data storage devices. based on the classification, an action is taken to improve the reliability and availability of the data storage network.",2005-09-20,"The title of the patent is predictive failure analysis for storage networks and its abstract is a method of predicting failures of data storage devices in a data storage network. in predicting failures, birth and/or health records of data storage devices in the data storage network are requested. the records are then scaled and thresholded and processed using a probabilistic neural network to classify the data storage devices. based on the classification, an action is taken to improve the reliability and availability of the data storage network. dated 2005-09-20"
6950538,signature recognition system and method,"a method of authenticating a signature including the steps of sampling a signature and storing data representative of the signature, converting the data to high dimensions vectors, feeding the high dimension vectors to an unsupervised neural network, performing a high order principal component extraction process on the high dimensions vectors to thereby identifying clusters of high dimension points, and analyzing the clusters of high dimension points to determine, based on previously stored information, the authenticity of the signature. also an apparatus for such authentication including a sampling device for sampling a signature and storing data representative of the signature, a converting device connected downstream of the sampling device for converting the data to high dimension vectors, an unsupervised neural network for receiving the high dimension and performing a high order principal component extraction process on the high dimensions vectors to thereby identify clusters of high dimension points, and an analyzing device connected to the unsupervised neural network for analyzing the clusters of high dimension points to determine the authenticity of the signature.",2005-09-27,"The title of the patent is signature recognition system and method and its abstract is a method of authenticating a signature including the steps of sampling a signature and storing data representative of the signature, converting the data to high dimensions vectors, feeding the high dimension vectors to an unsupervised neural network, performing a high order principal component extraction process on the high dimensions vectors to thereby identifying clusters of high dimension points, and analyzing the clusters of high dimension points to determine, based on previously stored information, the authenticity of the signature. also an apparatus for such authentication including a sampling device for sampling a signature and storing data representative of the signature, a converting device connected downstream of the sampling device for converting the data to high dimension vectors, an unsupervised neural network for receiving the high dimension and performing a high order principal component extraction process on the high dimensions vectors to thereby identify clusters of high dimension points, and an analyzing device connected to the unsupervised neural network for analyzing the clusters of high dimension points to determine the authenticity of the signature. dated 2005-09-27"
6950755,genotype pattern recognition and classification,"interpreting data obtained by analysis of nucleic acids (dna) by obtaining nucleic acid data in a spatial domain, transforming the nucleic acid data from the spatial domain into a frequency domain, and obtaining sequence data of the nucleic acid data by executing a data mining process on the transformed nucleic acid data. the transformation may be performed by a hadamard transform, a fourier transform or a wavelet transform to obtain frequency coefficients, with less than all of the frequency coefficients being utilized in the data mining process. in addition, the frequency domain data may be normalized prior to the data mining process. the data mining process may be subjecting the frequency coefficients to a connectionist (neural network) algorithm or to a classification tree/rule induction (cart) algorithm.",2005-09-27,"The title of the patent is genotype pattern recognition and classification and its abstract is interpreting data obtained by analysis of nucleic acids (dna) by obtaining nucleic acid data in a spatial domain, transforming the nucleic acid data from the spatial domain into a frequency domain, and obtaining sequence data of the nucleic acid data by executing a data mining process on the transformed nucleic acid data. the transformation may be performed by a hadamard transform, a fourier transform or a wavelet transform to obtain frequency coefficients, with less than all of the frequency coefficients being utilized in the data mining process. in addition, the frequency domain data may be normalized prior to the data mining process. the data mining process may be subjecting the frequency coefficients to a connectionist (neural network) algorithm or to a classification tree/rule induction (cart) algorithm. dated 2005-09-27"
6953436,multi-modal cardiac diagnostic decision support system and method,"a method for extracting and evaluating features from cardiac acoustic signals includes the steps of obtaining a cardiac acoustic signal, extracting physiologically significant features from the cardiac acoustic signal using a neural network, analyzing the cardiac acoustic signal with a wavelist decomposition to extract time-frequency information, and identifying basic heart sounds using neutral networks applied to the extracted time-frequency information. a method for determining a status of heart murmurs includes the steps of obtaining a cardiac acoustic signal, detecting a murmur, if any, from the cardiac acoustic signal, and determining whether the murmur is one of functional and pathological based upon expert rules.",2005-10-11,"The title of the patent is multi-modal cardiac diagnostic decision support system and method and its abstract is a method for extracting and evaluating features from cardiac acoustic signals includes the steps of obtaining a cardiac acoustic signal, extracting physiologically significant features from the cardiac acoustic signal using a neural network, analyzing the cardiac acoustic signal with a wavelist decomposition to extract time-frequency information, and identifying basic heart sounds using neutral networks applied to the extracted time-frequency information. a method for determining a status of heart murmurs includes the steps of obtaining a cardiac acoustic signal, detecting a murmur, if any, from the cardiac acoustic signal, and determining whether the murmur is one of functional and pathological based upon expert rules. dated 2005-10-11"
6954678,artificial intelligence system for track defect problem solving,"a system and method facilitating lithography defect solution generation is provided. the invention includes a defect solution component and a defect alert component. the defect solution component provides potential solution(s) to a defect within the lithography process utilizing artificial intelligence technique(s) (e.g., bayesian learning methods that perform analysis over alternative dependent structures and apply a score, bayesian classifiers and other statistical classifiers, including decision tree learning methods, support vector machines, linear and non-linear regression and/or neural network).",2005-10-11,"The title of the patent is artificial intelligence system for track defect problem solving and its abstract is a system and method facilitating lithography defect solution generation is provided. the invention includes a defect solution component and a defect alert component. the defect solution component provides potential solution(s) to a defect within the lithography process utilizing artificial intelligence technique(s) (e.g., bayesian learning methods that perform analysis over alternative dependent structures and apply a score, bayesian classifiers and other statistical classifiers, including decision tree learning methods, support vector machines, linear and non-linear regression and/or neural network). dated 2005-10-11"
6954744,combinatorial approach for supervised neural network learning,"a technique for machine learning, such as supervised artificial neural network learning includes receiving data and checking the dimensionality of the read data and reducing the dimensionality to enhance machine learning performance using principal component analysis methodology. the technique further includes specifying the neural network architecture and initializing weights to establish a connection between read data including the reduced dimensionality and the predicted values. the technique also includes performing supervised machine learning using the specified neural network architecture, initialized weights, and the read data including the reduced dimensionality to predict values. predicted values are then compared to a normalized system error threshold value and the initialized weights are revised based on the outcome of the comparison to generate a learnt neural network having a reduced error in weight space. the learnt neural network is validated using known values and is then used for predicting values.",2005-10-11,"The title of the patent is combinatorial approach for supervised neural network learning and its abstract is a technique for machine learning, such as supervised artificial neural network learning includes receiving data and checking the dimensionality of the read data and reducing the dimensionality to enhance machine learning performance using principal component analysis methodology. the technique further includes specifying the neural network architecture and initializing weights to establish a connection between read data including the reduced dimensionality and the predicted values. the technique also includes performing supervised machine learning using the specified neural network architecture, initialized weights, and the read data including the reduced dimensionality to predict values. predicted values are then compared to a normalized system error threshold value and the initialized weights are revised based on the outcome of the comparison to generate a learnt neural network having a reduced error in weight space. the learnt neural network is validated using known values and is then used for predicting values. dated 2005-10-11"
6956506,method and arrangement for entering data in an electronic apparatus and an electronic apparatus,the invention relates to a method and arrangement for creating a virtual keyboard for a terminal (800) used in a cellular network. the virtual keyboard is generated using an ir transceiver arrangement (12) in which a reflection from an obstacle (15) placed in the field of ir transmitters is registered by discrete ir receivers. the received reflection data are processed in a neural network arrangement (33). the purpose of the data processing is to find out the virtual key position/function that the received reflection data corresponds to.,2005-10-18,The title of the patent is method and arrangement for entering data in an electronic apparatus and an electronic apparatus and its abstract is the invention relates to a method and arrangement for creating a virtual keyboard for a terminal (800) used in a cellular network. the virtual keyboard is generated using an ir transceiver arrangement (12) in which a reflection from an obstacle (15) placed in the field of ir transmitters is registered by discrete ir receivers. the received reflection data are processed in a neural network arrangement (33). the purpose of the data processing is to find out the virtual key position/function that the received reflection data corresponds to. dated 2005-10-18
6957203,method and apparatus for operating a neural network with missing and/or incomplete data,"a neural network system is provided that models the system in a system model (12) with the output thereof providing a predicted output. this predicted output is modified or controlled by an output control (14). input data is processed in a data preprocess step (10) to reconcile the data for input to the system model (12). additionally, the error resulted from the reconciliation is input to an uncertainty model to predict the uncertainty in the predicted output. this is input to a decision processor (20) which is utilized to control the output control (14). the output control (14) is controlled to either vary the predicted output or to inhibit the predicted output whenever the output of the uncertainty model (18) exceeds a predetermined decision threshold, input by a decision threshold block (22). additionally, a validity model (16) is also provided which represents the reliability or validity of the output as a function of the number of data points in a given data region during training of the system model (12). this predicts the confidence in the predicted output which is also input to the decision processor (20). the decision processor (20) therefore bases its decision on the predicted confidence and the predicted uncertainty. additionally, the uncertainty output by the data preprocess block (10) can be utilized to train the system model (12).",2005-10-18,"The title of the patent is method and apparatus for operating a neural network with missing and/or incomplete data and its abstract is a neural network system is provided that models the system in a system model (12) with the output thereof providing a predicted output. this predicted output is modified or controlled by an output control (14). input data is processed in a data preprocess step (10) to reconcile the data for input to the system model (12). additionally, the error resulted from the reconciliation is input to an uncertainty model to predict the uncertainty in the predicted output. this is input to a decision processor (20) which is utilized to control the output control (14). the output control (14) is controlled to either vary the predicted output or to inhibit the predicted output whenever the output of the uncertainty model (18) exceeds a predetermined decision threshold, input by a decision threshold block (22). additionally, a validity model (16) is also provided which represents the reliability or validity of the output as a function of the number of data points in a given data region during training of the system model (12). this predicts the confidence in the predicted output which is also input to the decision processor (20). the decision processor (20) therefore bases its decision on the predicted confidence and the predicted uncertainty. additionally, the uncertainty output by the data preprocess block (10) can be utilized to train the system model (12). dated 2005-10-18"
6959109,system and method for pose-angle estimation,"a system and method are disclosed for determining the pose angle of an object in an input image. in a preferred embodiment, the present system comprises a pose estimator having a prototype projector, a regression estimator, and an angle calculator. the prototype projector is preferably adapted to reduce the input image dimensionality for faster further processing by projecting the input pixels of the image onto a self-organizing map (som) neural network. the regression estimator is preferably implemented as a neural network and adapted to map the projections to a pattern unique to each pose. the angle calculator preferably includes a curve fitter and an error analyzer. the curve fitter is preferably adapted to estimate the pose angle from the mapping pattern. the error analyzer is preferably adapted to produce a confidence signal representing the likelihood of the input image being a face at the calculated pose. the system also preferably includes two network trainers responsible for synthesizing the neural networks.",2005-10-25,"The title of the patent is system and method for pose-angle estimation and its abstract is a system and method are disclosed for determining the pose angle of an object in an input image. in a preferred embodiment, the present system comprises a pose estimator having a prototype projector, a regression estimator, and an angle calculator. the prototype projector is preferably adapted to reduce the input image dimensionality for faster further processing by projecting the input pixels of the image onto a self-organizing map (som) neural network. the regression estimator is preferably implemented as a neural network and adapted to map the projections to a pattern unique to each pose. the angle calculator preferably includes a curve fitter and an error analyzer. the curve fitter is preferably adapted to estimate the pose angle from the mapping pattern. the error analyzer is preferably adapted to produce a confidence signal representing the likelihood of the input image being a face at the calculated pose. the system also preferably includes two network trainers responsible for synthesizing the neural networks. dated 2005-10-25"
6961719,hybrid neural network and support vector machine method for optimization,"system and method for optimization of a design associated with a response function, using a hybrid neural net and support vector machine (nn/svm) analysis to minimize or maximize an objective function, optionally subject to one or more constraints. as a first example, the nn/svm analysis is applied iteratively to design of an aerodynamic component, such as an airfoil shape, where the objective function measures deviation from a target pressure distribution on the perimeter of the aerodynamic component. as a second example, the nn/svm analysis is applied to data classification of a sequence of data points in a multidimensional space. the nn/svm analysis is also applied to data regression.",2005-11-01,"The title of the patent is hybrid neural network and support vector machine method for optimization and its abstract is system and method for optimization of a design associated with a response function, using a hybrid neural net and support vector machine (nn/svm) analysis to minimize or maximize an objective function, optionally subject to one or more constraints. as a first example, the nn/svm analysis is applied iteratively to design of an aerodynamic component, such as an airfoil shape, where the objective function measures deviation from a target pressure distribution on the perimeter of the aerodynamic component. as a second example, the nn/svm analysis is applied to data classification of a sequence of data points in a multidimensional space. the nn/svm analysis is also applied to data regression. dated 2005-11-01"
6961914,method and apparatus for selecting input points to train a machine learning model for extraction,"the present invention introduces novel methods generating training data for machine learning models that will be used for extraction. specifically, experimental design is employed to select a set of training points that provide the best information. in one embodiment, the training point set is created by creating a critical input spanning set, adding training points from critical regions in the input space, and adding training points from frequently encountered profile cases. the training point set then used to train a machine learning built model such as a neural network or support vector machine that will extract electrical characteristics.",2005-11-01,"The title of the patent is method and apparatus for selecting input points to train a machine learning model for extraction and its abstract is the present invention introduces novel methods generating training data for machine learning models that will be used for extraction. specifically, experimental design is employed to select a set of training points that provide the best information. in one embodiment, the training point set is created by creating a critical input spanning set, adding training points from critical regions in the input space, and adding training points from frequently encountered profile cases. the training point set then used to train a machine learning built model such as a neural network or support vector machine that will extract electrical characteristics. dated 2005-11-01"
6965736,"method for monitoring the transmission quality of an optical transmission system, in particular of an optical wavelength-division multiplex network","a method for monitoring the transmission quality of an optical transmission system, such as, for example, an optical wavelength division-multiplex network. an amplitude histogram of an optical signal (transmission signal) transmitted over the transmission system may be plotted and classified, with the assistance of a neural network, according to bit error rates and/or causes of faults. the need for setting requirements for transmission mode, transmission format and/or transmission timing cycle of the transmission system may be eliminated. the amplitude histogram may be implemented for any signal, and causes of faults, which are not able to be determined by a conventional bit rate classification, may be allocated.",2005-11-15,"The title of the patent is method for monitoring the transmission quality of an optical transmission system, in particular of an optical wavelength-division multiplex network and its abstract is a method for monitoring the transmission quality of an optical transmission system, such as, for example, an optical wavelength division-multiplex network. an amplitude histogram of an optical signal (transmission signal) transmitted over the transmission system may be plotted and classified, with the assistance of a neural network, according to bit error rates and/or causes of faults. the need for setting requirements for transmission mode, transmission format and/or transmission timing cycle of the transmission system may be eliminated. the amplitude histogram may be implemented for any signal, and causes of faults, which are not able to be determined by a conventional bit rate classification, may be allocated. dated 2005-11-15"
6968250,intelligent agent system and method for evaluating data integrity in process information databases,"a data integrity module and method for evaluating data in a process information database. a neural network generates statistical patterns for specifying patterns for the data being evaluated. a fuzzy expert rules base specifies rules for evaluating the data. a processor, responsive to the rules base and the statistical patterns, identifies suspect data in the process information database by evaluating the data according to the rules base and the statistical patterns. a modification system modifies the suspect data in the process information database.",2005-11-22,"The title of the patent is intelligent agent system and method for evaluating data integrity in process information databases and its abstract is a data integrity module and method for evaluating data in a process information database. a neural network generates statistical patterns for specifying patterns for the data being evaluated. a fuzzy expert rules base specifies rules for evaluating the data. a processor, responsive to the rules base and the statistical patterns, identifies suspect data in the process information database by evaluating the data according to the rules base and the statistical patterns. a modification system modifies the suspect data in the process information database. dated 2005-11-22"
6968327,method for training a neural network,"a method for training a neural network in order to optimize the structure of the neural network includes identifying and eliminating synapses that have no significant influence on the curve of the risk function. first and second sending neurons are selected that are connected to the same receiving neuron by respective first and second synapses. it is assumed that there is a correlation of response signals from the first and second sending neurons to the same receiving neuron. the first synapse is interrupted and a weight of the second synapse is adapted in its place. the output signals of the changed neural network are compared with the output signals of the unchanged neural network. if the comparison result does not exceed a predetermined level, the first synapse is eliminated, thereby simplifying the structure of the neural network.",2005-11-22,"The title of the patent is method for training a neural network and its abstract is a method for training a neural network in order to optimize the structure of the neural network includes identifying and eliminating synapses that have no significant influence on the curve of the risk function. first and second sending neurons are selected that are connected to the same receiving neuron by respective first and second synapses. it is assumed that there is a correlation of response signals from the first and second sending neurons to the same receiving neuron. the first synapse is interrupted and a weight of the second synapse is adapted in its place. the output signals of the changed neural network are compared with the output signals of the unchanged neural network. if the comparison result does not exceed a predetermined level, the first synapse is eliminated, thereby simplifying the structure of the neural network. dated 2005-11-22"
6976012,method and apparatus of using a neural network to train a neural network,"a method and apparatus of training a neural network. the method and apparatus include creating a model for a desired function as a multi-dimensional function, determining if the created model fits a simple finite geometry model, and generating a radon transform to fit the simple finite geometry model. the desired function is fed through the radon transform to generate weights. a multilayer perceptron of the neural network is trained using the weights.",2005-12-13,"The title of the patent is method and apparatus of using a neural network to train a neural network and its abstract is a method and apparatus of training a neural network. the method and apparatus include creating a model for a desired function as a multi-dimensional function, determining if the created model fits a simple finite geometry model, and generating a radon transform to fit the simple finite geometry model. the desired function is fed through the radon transform to generate weights. a multilayer perceptron of the neural network is trained using the weights. dated 2005-12-13"
6980894,method of managing interference during delay recovery on a train system,"the present invention provides methods for preventing low train voltages and managing interference, thereby improving the efficiency, reliability, and passenger comfort associated with commuter trains. an algorithm implementing neural network technology is used to predict low voltages before they occur. once voltages are predicted, then multiple trains can be controlled to prevent low voltage events. further, algorithms for managing inference are presented in the present invention. different types of interference problems are addressed in the present invention such as “interference during acceleration”, “interference near station stops”, and “interference during delay recovery.” managing such interference avoids unnecessary brake/acceleration cycles during acceleration, immediately before station stops, and after substantial delays. algorithms are demonstrated to avoid oscillatory brake/acceleration cycles due to interference and to smooth the trajectories of closely following trains. this is achieved by maintaining sufficient following distances to avoid unnecessary braking/accelerating. these methods generate smooth train trajectories, making for a more comfortable ride, and improve train motor reliability by avoiding unnecessary mode-changes between propulsion and braking. these algorithms can also have a favorable impact on traction power system requirements and energy consumption.",2005-12-27,"The title of the patent is method of managing interference during delay recovery on a train system and its abstract is the present invention provides methods for preventing low train voltages and managing interference, thereby improving the efficiency, reliability, and passenger comfort associated with commuter trains. an algorithm implementing neural network technology is used to predict low voltages before they occur. once voltages are predicted, then multiple trains can be controlled to prevent low voltage events. further, algorithms for managing inference are presented in the present invention. different types of interference problems are addressed in the present invention such as “interference during acceleration”, “interference near station stops”, and “interference during delay recovery.” managing such interference avoids unnecessary brake/acceleration cycles during acceleration, immediately before station stops, and after substantial delays. algorithms are demonstrated to avoid oscillatory brake/acceleration cycles due to interference and to smooth the trajectories of closely following trains. this is achieved by maintaining sufficient following distances to avoid unnecessary braking/accelerating. these methods generate smooth train trajectories, making for a more comfortable ride, and improve train motor reliability by avoiding unnecessary mode-changes between propulsion and braking. these algorithms can also have a favorable impact on traction power system requirements and energy consumption. dated 2005-12-27"
6983265,method to improve the data transfer rate between a computer and a neural network,"a method is described to improve the data transfer rate between a personal computer or a host computer and a neural network implemented in hardware by merging a plurality of input patterns into a single input pattern configured to globally represent the set of input patterns. a base consolidated vector (u′*n) representing the input pattern is defined to describe all the vectors (un, . . . , un+6) representing the input patterns derived thereof (u′n, . . . , u′n+6) by combining components having fixed and ‘don't care’ values. the base consolidated vector is provided only once with all the components of the vectors. an artificial neural network (ann) is then configured as a combination of sub-networks operating in parallel. in order to compute the distances with an adequate number of components, the prototypes are to include also components having a definite value and ‘don't care’ conditions. during the learning phase, the consolidated vectors are stored as prototypes. during the recognition phase, when a new base consolidated vector is provided to ann, each sub-network analyses a portion thereof after computing all the distances, they are sorted one sub-network at a time to obtain the distances associated to each vector.",2006-01-03,"The title of the patent is method to improve the data transfer rate between a computer and a neural network and its abstract is a method is described to improve the data transfer rate between a personal computer or a host computer and a neural network implemented in hardware by merging a plurality of input patterns into a single input pattern configured to globally represent the set of input patterns. a base consolidated vector (u′*n) representing the input pattern is defined to describe all the vectors (un, . . . , un+6) representing the input patterns derived thereof (u′n, . . . , u′n+6) by combining components having fixed and ‘don't care’ values. the base consolidated vector is provided only once with all the components of the vectors. an artificial neural network (ann) is then configured as a combination of sub-networks operating in parallel. in order to compute the distances with an adequate number of components, the prototypes are to include also components having a definite value and ‘don't care’ conditions. during the learning phase, the consolidated vectors are stored as prototypes. during the recognition phase, when a new base consolidated vector is provided to ann, each sub-network analyses a portion thereof after computing all the distances, they are sorted one sub-network at a time to obtain the distances associated to each vector. dated 2006-01-03"
6985290,visualization of three dimensional images and multi aspect imaging,"three-dimensional imaging without parallax barriers or specialized eye gear, and without attendant loss of resolution, is provided by a display that produces dynamic images for display on at least two stacked electronic transmissive displays to create a continuous 3-d image field in a large viewing area or in multiple viewing areas. the images on each display are derived from stereoscopic image sources corresponding to both eyes of a viewer, and the derived images act as a mask for each other causing 3-d perception. the derived images are processed by summing the predicted image data, comparing the predicted image data to the desired stereopair, and minimizing the error. in preferred embodiments, the processing can be performed by an artificial neural network. a viewer may be presented with different aspects of an image as their viewing position changes to allow the viewer to perceive various perspectives of an image in dynamic fashion.",2006-01-10,"The title of the patent is visualization of three dimensional images and multi aspect imaging and its abstract is three-dimensional imaging without parallax barriers or specialized eye gear, and without attendant loss of resolution, is provided by a display that produces dynamic images for display on at least two stacked electronic transmissive displays to create a continuous 3-d image field in a large viewing area or in multiple viewing areas. the images on each display are derived from stereoscopic image sources corresponding to both eyes of a viewer, and the derived images act as a mask for each other causing 3-d perception. the derived images are processed by summing the predicted image data, comparing the predicted image data to the desired stereopair, and minimizing the error. in preferred embodiments, the processing can be performed by an artificial neural network. a viewer may be presented with different aspects of an image as their viewing position changes to allow the viewer to perceive various perspectives of an image in dynamic fashion. dated 2006-01-10"
6985610,signature recognition system and method,"a method of authenticating a signature including the steps of sampling a signature and storing data representative of the signature, converting the data to high dimensions vectors, feeding the high dimension vectors to an unsupervised neural network, performing a high order principal component extraction process on the high dimensions vectors to thereby identifying clusters of high dimension points, and analyzing the clusters of high dimension points to determine, based on previously stored information, the authenticity of the signature. also an apparatus for such authentication including a sampling device for sampling a signature and storing data representative of the signature, a converting device connected downstream of the sampling device for converting the data to high dimension vectors, an unsupervised neural network for receiving the high dimension and performing a high order principal component extraction process on the high dimensions vectors to thereby identify clusters of high dimension points, and an analyzing device connected to the unsupervised neural network for analyzing the clusters of high dimension points to determine the authenticity of the signature.",2006-01-10,"The title of the patent is signature recognition system and method and its abstract is a method of authenticating a signature including the steps of sampling a signature and storing data representative of the signature, converting the data to high dimensions vectors, feeding the high dimension vectors to an unsupervised neural network, performing a high order principal component extraction process on the high dimensions vectors to thereby identifying clusters of high dimension points, and analyzing the clusters of high dimension points to determine, based on previously stored information, the authenticity of the signature. also an apparatus for such authentication including a sampling device for sampling a signature and storing data representative of the signature, a converting device connected downstream of the sampling device for converting the data to high dimension vectors, an unsupervised neural network for receiving the high dimension and performing a high order principal component extraction process on the high dimensions vectors to thereby identify clusters of high dimension points, and an analyzing device connected to the unsupervised neural network for analyzing the clusters of high dimension points to determine the authenticity of the signature. dated 2006-01-10"
6985781,residual activation neural network,a plant is operable to receive control inputs c(t) and provide an output y(t). the plant (72) has associated therewith state variables s(t) that are not variable. a control network (74) models the plant by providing a predicted output which is combined with a desired output to generate an error that is back propagated through an inverse control network to generate a control error signal that is input to a distributed control system to vary the control inputs to the plant in order to change the output y(t) to meet the desired output. the inverse model represents the dependencies of the plant output on the control variables parameterized by external influences to the plant.,2006-01-10,The title of the patent is residual activation neural network and its abstract is a plant is operable to receive control inputs c(t) and provide an output y(t). the plant (72) has associated therewith state variables s(t) that are not variable. a control network (74) models the plant by providing a predicted output which is combined with a desired output to generate an error that is back propagated through an inverse control network to generate a control error signal that is input to a distributed control system to vary the control inputs to the plant in order to change the output y(t) to meet the desired output. the inverse model represents the dependencies of the plant output on the control variables parameterized by external influences to the plant. dated 2006-01-10
6988995,method and system for detecting the effects of alzheimer's disease in the human retina,"a system for the in vivo detection of the effects of ad in the interior of an eye. a scanning polarimeter, including a residual retardance canceling system and an improved anterior segment retardance compensator, produces an optical analysis signal representing the birefringence of the retinal nerve fiber layer (rnfl) structures of the eye. the birefringence data is more accurate because of compensation for anterior segment birefringence and residual birefringence of optical components, such as, for example, the beam splitters, lenses, scanners and retarders. an electrical analysis signal representing a large (20 by 40 degrees) retardance map is produced and evaluated by an artificial neural network to produce an analysis classification signal representing the contribution of alzheimer's disease to the birefringence of the retinal layer corresponding to the relationship of the electrical analysis signal to an analysis signal database.",2006-01-24,"The title of the patent is method and system for detecting the effects of alzheimer's disease in the human retina and its abstract is a system for the in vivo detection of the effects of ad in the interior of an eye. a scanning polarimeter, including a residual retardance canceling system and an improved anterior segment retardance compensator, produces an optical analysis signal representing the birefringence of the retinal nerve fiber layer (rnfl) structures of the eye. the birefringence data is more accurate because of compensation for anterior segment birefringence and residual birefringence of optical components, such as, for example, the beam splitters, lenses, scanners and retarders. an electrical analysis signal representing a large (20 by 40 degrees) retardance map is produced and evaluated by an artificial neural network to produce an analysis classification signal representing the contribution of alzheimer's disease to the birefringence of the retinal layer corresponding to the relationship of the electrical analysis signal to an analysis signal database. dated 2006-01-24"
6990474,evaluation system,"an evaluation system (10) for evaluating media is described. the system is particularly suitable for evaluating banknotes to determine their suitability for use in an atm. the system comprises sensing means (12) for sensing properties of media (18) including the location of any imperfection in the media, and an evaluation module (16) for evaluating imperfections in the media(18). the evaluation module (16) includes a classifier (52) comprising an artificial neural network (60) and fuzzy logic (66). the evaluation module (16) may include a plurality of classifiers (52), and a second level classifier (56) for generating a suitability index (20) from the outputs of the first level classifiers (52). a method of evaluating media is also described.",2006-01-24,"The title of the patent is evaluation system and its abstract is an evaluation system (10) for evaluating media is described. the system is particularly suitable for evaluating banknotes to determine their suitability for use in an atm. the system comprises sensing means (12) for sensing properties of media (18) including the location of any imperfection in the media, and an evaluation module (16) for evaluating imperfections in the media(18). the evaluation module (16) includes a classifier (52) comprising an artificial neural network (60) and fuzzy logic (66). the evaluation module (16) may include a plurality of classifiers (52), and a second level classifier (56) for generating a suitability index (20) from the outputs of the first level classifiers (52). a method of evaluating media is also described. dated 2006-01-24"
6993423,method for adjusting vehicle cockpit devices,"a method for adjusting a plurality of vehicle cockpit devices via a two-part process that utilizes device position constraints to determine candidate arrangements and then, ultimately, recommended arrangements of the vehicle cockpit devices to determine a desired setting of the various devices. the position constraints are determined using positioning data obtained from an occupant. an exploratory search routine is used to determine the candidate arrangements with the cockpit devices being moved to each candidate arrangement so that the occupant can be queried concerning the desirability of each such arrangement. the occupant's responses are then stored for later retrieval. thereafter, a plurality of recommended arrangements of the cockpit devices are determined using a meta-heuristic pattern search along with a neural network search accelerator that permits screening of each recommended arrangement. the occupant can then select one of the recommended arrangements as a final positioning arrangement of the cockpit devices.",2006-01-31,"The title of the patent is method for adjusting vehicle cockpit devices and its abstract is a method for adjusting a plurality of vehicle cockpit devices via a two-part process that utilizes device position constraints to determine candidate arrangements and then, ultimately, recommended arrangements of the vehicle cockpit devices to determine a desired setting of the various devices. the position constraints are determined using positioning data obtained from an occupant. an exploratory search routine is used to determine the candidate arrangements with the cockpit devices being moved to each candidate arrangement so that the occupant can be queried concerning the desirability of each such arrangement. the occupant's responses are then stored for later retrieval. thereafter, a plurality of recommended arrangements of the cockpit devices are determined using a meta-heuristic pattern search along with a neural network search accelerator that permits screening of each recommended arrangement. the occupant can then select one of the recommended arrangements as a final positioning arrangement of the cockpit devices. dated 2006-01-31"
6993512,system and method for converting a color formula using an artificial intelligence based conversion model,"a system and method for converting a color formula from compositions such as paints, pigments, or dye formulations, is provided. the input to the system is a first color formula. the system includes an input device for entering a plurality of color formula values and an artificial intelligence conversion model coupled to the input device. the conversion model produces an output signal for communicating a second color formula. the artificial intelligence model may be embodied in a neural network. more specifically, the conversion model may be a back propagation neural network.",2006-01-31,"The title of the patent is system and method for converting a color formula using an artificial intelligence based conversion model and its abstract is a system and method for converting a color formula from compositions such as paints, pigments, or dye formulations, is provided. the input to the system is a first color formula. the system includes an input device for entering a plurality of color formula values and an artificial intelligence conversion model coupled to the input device. the conversion model produces an output signal for communicating a second color formula. the artificial intelligence model may be embodied in a neural network. more specifically, the conversion model may be a back propagation neural network. dated 2006-01-31"
6999950,n-tuple or ram based neural network classification system and method,"the invention relates to n-tuple or ram based neural network classification methods and systems and, more particularly, to n-tuple or ram based classification systems where the decision criteria applied to obtain the output sources and compare these output sources to obtain a classification are determined during a training process. accordingly, the invention relates to a system and a method of training a computer classification system which can be defined by a network comprising a number of n-tuples or look up tables (luts), with each n-tuple or lut comprising a number of rows corresponding to at least a subset of possible classes and comprising columns being addressed by signals or elements of sampled training input data examples.",2006-02-14,"The title of the patent is n-tuple or ram based neural network classification system and method and its abstract is the invention relates to n-tuple or ram based neural network classification methods and systems and, more particularly, to n-tuple or ram based classification systems where the decision criteria applied to obtain the output sources and compare these output sources to obtain a classification are determined during a training process. accordingly, the invention relates to a system and a method of training a computer classification system which can be defined by a network comprising a number of n-tuples or look up tables (luts), with each n-tuple or lut comprising a number of rows corresponding to at least a subset of possible classes and comprising columns being addressed by signals or elements of sampled training input data examples. dated 2006-02-14"
6999952,linear associative memory-based hardware architecture for fault tolerant asic/fpga work-around,"a programmable logic unit (e.g., an asic or fpga) having a feedforward linear associative memory (lam) neural network checking circuit which classifies input vectors to a faulty hardware block as either good or not good and, when a new input vector is classified as not good, blocks a corresponding output vector of the faulty hardware block, enables a software work-around for the new input vector, and accepts the software work-around input as the output vector of the programmable logic circuit. the feedforward lam neural network checking circuit has a weight matrix whose elements are based on a set of known bad input vectors for said faulty hardware block. the feedforward lam neural network checking circuit may update the weight matrix online using one or more additional bad input vectors. a discrete hopfield algorithm is used to calculate the weight matrix w. the feedforward lam neural network checking circuit calculates an output vector a(m) by multiplying the weight matrix w by the new input vector b(m), that is, a(m)=wb(m), adjusts elements of the output vector a(m) by respective thresholds, and processes the elements using a plurality of non-linear units to provide an output of 1 when a given adjusted element is positive, and provide an output of 0 when a given adjusted element is not positive. if a vector constructed of the outputs of these non-linear units matches with an entry in a content-addressable memory (cam) storing the set of known bad vectors (a cam hit), then the new input vector is classified as not good.",2006-02-14,"The title of the patent is linear associative memory-based hardware architecture for fault tolerant asic/fpga work-around and its abstract is a programmable logic unit (e.g., an asic or fpga) having a feedforward linear associative memory (lam) neural network checking circuit which classifies input vectors to a faulty hardware block as either good or not good and, when a new input vector is classified as not good, blocks a corresponding output vector of the faulty hardware block, enables a software work-around for the new input vector, and accepts the software work-around input as the output vector of the programmable logic circuit. the feedforward lam neural network checking circuit has a weight matrix whose elements are based on a set of known bad input vectors for said faulty hardware block. the feedforward lam neural network checking circuit may update the weight matrix online using one or more additional bad input vectors. a discrete hopfield algorithm is used to calculate the weight matrix w. the feedforward lam neural network checking circuit calculates an output vector a(m) by multiplying the weight matrix w by the new input vector b(m), that is, a(m)=wb(m), adjusts elements of the output vector a(m) by respective thresholds, and processes the elements using a plurality of non-linear units to provide an output of 1 when a given adjusted element is positive, and provide an output of 0 when a given adjusted element is not positive. if a vector constructed of the outputs of these non-linear units matches with an entry in a content-addressable memory (cam) storing the set of known bad vectors (a cam hit), then the new input vector is classified as not good. dated 2006-02-14"
7001243,neural network control of chemical mechanical planarization,"broadly speaking, a method for controlling a chemical mechanical planarization (cmp) process to obtain a desired result is provided. more specifically, the method incorporates a first neural network to estimate a cmp result and a second neural network to tune cmp control parameters used to obtain the cmp result. the second neural network tunes the cmp control parameters to minimize a difference between the cmp result and a desired cmp result.",2006-02-21,"The title of the patent is neural network control of chemical mechanical planarization and its abstract is broadly speaking, a method for controlling a chemical mechanical planarization (cmp) process to obtain a desired result is provided. more specifically, the method incorporates a first neural network to estimate a cmp result and a second neural network to tune cmp control parameters used to obtain the cmp result. the second neural network tunes the cmp control parameters to minimize a difference between the cmp result and a desired cmp result. dated 2006-02-21"
7003109,compact crypto-engine for random number and stream cipher generation,a compact dual function random number generator and stream cipher generator includes a crypto-engine has a controller for controlling the engine to operate in one or other of its functions. the crypto-engine incorporates a plurality of clipped hopfield neural network pairs.,2006-02-21,The title of the patent is compact crypto-engine for random number and stream cipher generation and its abstract is a compact dual function random number generator and stream cipher generator includes a crypto-engine has a controller for controlling the engine to operate in one or other of its functions. the crypto-engine incorporates a plurality of clipped hopfield neural network pairs. dated 2006-02-21
7003149,method and device for optically monitoring fabrication processes of finely structured surfaces in a semiconductor production,"a method for monitoring fabrication processes of finely structured surfaces in a semiconductor fabrication includes the steps of providing reference signatures of finely structured surfaces, measuring at least one signature of a test specimen surface, comparing the measured signature with the reference signatures, and classifying the test specimen surface by using the comparison results, wherein the measurement of the reference signatures is carried out by measuring the local distribution and/or intensity distribution of diffraction images on production prototypes having a specified quality. the classification is preferably carried out here with a neural network having a learning capability and/or a fuzzy logic. furthermore, a device for carrying out the method is provided.",2006-02-21,"The title of the patent is method and device for optically monitoring fabrication processes of finely structured surfaces in a semiconductor production and its abstract is a method for monitoring fabrication processes of finely structured surfaces in a semiconductor fabrication includes the steps of providing reference signatures of finely structured surfaces, measuring at least one signature of a test specimen surface, comparing the measured signature with the reference signatures, and classifying the test specimen surface by using the comparison results, wherein the measurement of the reference signatures is carried out by measuring the local distribution and/or intensity distribution of diffraction images on production prototypes having a specified quality. the classification is preferably carried out here with a neural network having a learning capability and/or a fuzzy logic. furthermore, a device for carrying out the method is provided. dated 2006-02-21"
7006205,method and system for event detection in plasma processes,"plasma events are detected by analyzing the spectral emissions of a plasma process of a substrate. a plasma is monitored by a spectrometer which produces plasma emission data which includes an intensity value for each individual wavelength which is a large quantity of information. the plasma emission data is processed with an algorithm or combination of algorithms to reduce the quantity of the plasma emission data. a peak finding algorithm which identifies the wavelengths of light which are associated with a plasma process allowing the other wavelengths to be ignored. a data reduction algorithm provides a single value representative of the intensity of the emitted light from each peak. a noise reduction algorithm removes noise from the spectral signal by eliminating signals at wavelengths which do not exceed a threshold intensity and do not exceed a threshold wavelength span. the data may also be processed with principal component analysis to further reduce the optical emission data. by reducing the optical emission data, the neural network can provide faster data analysis. the neural network of the inventive system identifies the plasma event after being trained and can control the processing equipment when plasma events are detected.",2006-02-28,"The title of the patent is method and system for event detection in plasma processes and its abstract is plasma events are detected by analyzing the spectral emissions of a plasma process of a substrate. a plasma is monitored by a spectrometer which produces plasma emission data which includes an intensity value for each individual wavelength which is a large quantity of information. the plasma emission data is processed with an algorithm or combination of algorithms to reduce the quantity of the plasma emission data. a peak finding algorithm which identifies the wavelengths of light which are associated with a plasma process allowing the other wavelengths to be ignored. a data reduction algorithm provides a single value representative of the intensity of the emitted light from each peak. a noise reduction algorithm removes noise from the spectral signal by eliminating signals at wavelengths which do not exceed a threshold intensity and do not exceed a threshold wavelength span. the data may also be processed with principal component analysis to further reduce the optical emission data. by reducing the optical emission data, the neural network can provide faster data analysis. the neural network of the inventive system identifies the plasma event after being trained and can control the processing equipment when plasma events are detected. dated 2006-02-28"
7006866,arrangement for predicting an abnormality of a system and for carrying out an action which counteracts the abnormality,"an arrangement and method are presented that enable a prediction of an abnormality and implement a suitable action opposing the abnormality. an information flow underlying a dynamic system is interpreted and a prediction quantity that comprises the abnormality as characterizing quantity of the dynamic system is determined from it. a neural network is trained with measured data of the system. after the training, the abnormality can be indicated on the basis of the prediction quantity before it occurs and the occurrence can be opposed with suitable measures.",2006-02-28,"The title of the patent is arrangement for predicting an abnormality of a system and for carrying out an action which counteracts the abnormality and its abstract is an arrangement and method are presented that enable a prediction of an abnormality and implement a suitable action opposing the abnormality. an information flow underlying a dynamic system is interpreted and a prediction quantity that comprises the abnormality as characterizing quantity of the dynamic system is determined from it. a neural network is trained with measured data of the system. after the training, the abnormality can be indicated on the basis of the prediction quantity before it occurs and the occurrence can be opposed with suitable measures. dated 2006-02-28"
7012893,adaptive control of data packet size in networks,"an adaptive packet mechanism and method for optimizing data packet transmission through a network connection between a sending node and a receiving node. current network conditions in the connection are periodically determined wherein the network conditions pertain to the latency and jitter of packet transmission between the sending node and receiving node. the measurements of latency and jitter are used to determine an optimum packet size and an optimum inter-packet interval for transmission of packet data between the sending node and the receiving node and are used in the transmission of data packets from the sending node to the receiving node. network conditions may be determined by transmission of monitor or data packets and may be determined at either or both of the sending or receiving nodes and the optimum packet size and inter-packet interval are determined by a fuzzy logic analyzer, a neural network analyzer or a combined fuzzy logic/neural network analyzer.",2006-03-14,"The title of the patent is adaptive control of data packet size in networks and its abstract is an adaptive packet mechanism and method for optimizing data packet transmission through a network connection between a sending node and a receiving node. current network conditions in the connection are periodically determined wherein the network conditions pertain to the latency and jitter of packet transmission between the sending node and receiving node. the measurements of latency and jitter are used to determine an optimum packet size and an optimum inter-packet interval for transmission of packet data between the sending node and the receiving node and are used in the transmission of data packets from the sending node to the receiving node. network conditions may be determined by transmission of monitor or data packets and may be determined at either or both of the sending or receiving nodes and the optimum packet size and inter-packet interval are determined by a fuzzy logic analyzer, a neural network analyzer or a combined fuzzy logic/neural network analyzer. dated 2006-03-14"
7013033,system for the automatic analysis of images such as dna microarray images,"the system can be used for the automatic analysis of images, including a matrix of spots, such as images of dna microarrays after hybridization. the system can be associated—and preferably integrated in a single monolithic component implementing vlsi cmos technology—to a sensor for acquiring the images. the system includes a circuit for processing the signals corresponding to the images, configured according to a cellular neural network (cnn) architecture for the parallel analogue processing of signals.",2006-03-14,"The title of the patent is system for the automatic analysis of images such as dna microarray images and its abstract is the system can be used for the automatic analysis of images, including a matrix of spots, such as images of dna microarrays after hybridization. the system can be associated—and preferably integrated in a single monolithic component implementing vlsi cmos technology—to a sensor for acquiring the images. the system includes a circuit for processing the signals corresponding to the images, configured according to a cellular neural network (cnn) architecture for the parallel analogue processing of signals. dated 2006-03-14"
7013245,method and apparatus for detecting concealed weapons,apparatus for classifying a ferromagnetic object within a sensing area may include a magnetic field sensor that produces magnetic field data. a signal processing system operatively associated with the magnetic field sensor includes a neural network. the neural network compares the magnetic field data with magnetic field data produced by known ferromagnetic objects to make a probabilistic determination as to the classification of the ferromagnetic object within the sensing area. a user interface operatively associated with the signal processing system produces a user-discernable output indicative of the probabilistic determination of the classification of the ferromagnetic object within a sensing area.,2006-03-14,The title of the patent is method and apparatus for detecting concealed weapons and its abstract is apparatus for classifying a ferromagnetic object within a sensing area may include a magnetic field sensor that produces magnetic field data. a signal processing system operatively associated with the magnetic field sensor includes a neural network. the neural network compares the magnetic field data with magnetic field data produced by known ferromagnetic objects to make a probabilistic determination as to the classification of the ferromagnetic object within the sensing area. a user interface operatively associated with the signal processing system produces a user-discernable output indicative of the probabilistic determination of the classification of the ferromagnetic object within a sensing area. dated 2006-03-14
7016529,system and method facilitating pattern recognition,a system and method facilitating pattern recognition is provided. the invention includes a pattern recognition system having a convolutional neural network employing feature extraction layer(s) and classifier layer(s). the feature extraction layer(s) comprises convolutional layers and the classifier layer(s) comprises fully connected layers. the pattern recognition system can be trained utilizing a calculated cross entropy error. the calculated cross entropy error is utilized to update trainable parameters of the pattern recognition system.,2006-03-21,The title of the patent is system and method facilitating pattern recognition and its abstract is a system and method facilitating pattern recognition is provided. the invention includes a pattern recognition system having a convolutional neural network employing feature extraction layer(s) and classifier layer(s). the feature extraction layer(s) comprises convolutional layers and the classifier layer(s) comprises fully connected layers. the pattern recognition system can be trained utilizing a calculated cross entropy error. the calculated cross entropy error is utilized to update trainable parameters of the pattern recognition system. dated 2006-03-21
7016815,performance assessment of data classifiers,"a method for operating a data classifier operable to generate an element of output data in response to an element of input data, such as a neural network, is disclosed. the method includes using the data classifier to generate elements of result output data in response to elements of test input data, generating a measure of difference between each element of test output data and each corresponding element of result output data, associating the measures of difference with categories corresponding to different values of measures of difference, and, based on the number of measures of difference associated with the categories, generating a performance measure of the data classifier.",2006-03-21,"The title of the patent is performance assessment of data classifiers and its abstract is a method for operating a data classifier operable to generate an element of output data in response to an element of input data, such as a neural network, is disclosed. the method includes using the data classifier to generate elements of result output data in response to elements of test input data, generating a measure of difference between each element of test output data and each corresponding element of result output data, associating the measures of difference with categories corresponding to different values of measures of difference, and, based on the number of measures of difference associated with the categories, generating a performance measure of the data classifier. dated 2006-03-21"
7024276,"legged mobile robot and its motion teaching method, and storage medium","learning-type motion control is performed using a hierarchical recurrent neural network. a motion pattern provided through human teaching work is automatically time-serially segmented with the hierarchical recurrent neural network, and the motion control of a machine body is carried out with a combination of the segmented data, whereby various motion patterns can be produced. with the time-serial segmentation, local time-serial patterns and an overall pattern as a combination of the local time-serial patterns are produced. for those motion patterns, indices for static stability and dynamic stability of the machine body, e.g., zmp stability criteria, are satisfied and hence control stability is ensured.",2006-04-04,"The title of the patent is legged mobile robot and its motion teaching method, and storage medium and its abstract is learning-type motion control is performed using a hierarchical recurrent neural network. a motion pattern provided through human teaching work is automatically time-serially segmented with the hierarchical recurrent neural network, and the motion control of a machine body is carried out with a combination of the segmented data, whereby various motion patterns can be produced. with the time-serial segmentation, local time-serial patterns and an overall pattern as a combination of the local time-serial patterns are produced. for those motion patterns, indices for static stability and dynamic stability of the machine body, e.g., zmp stability criteria, are satisfied and hence control stability is ensured. dated 2006-04-04"
7026946,method and apparatus for sensing seat occupancy,"method and apparatus for identifying and categorizing the weight and characteristics of the occupant currently occupying a vehicle seat. the method for identifying and categorizing the occupant or object involves measuring the deflection of the upper surface of the seat cushion at several points and therefore the weight distribution of the occupant. the system contains multiple weight sensors arrayed for detecting the distribution of the load causing the seat deflection. the system also includes a sensor to measure ambient temperature, preferably for temperature compensation due to the effects extreme temperatures may have on the compression properties of the seat cushion material and the weight sensors. a system processor interprets the data acquired by the sensors, and utilizes an algorithm and weight tables to simulate a neural network in providing an output a control signal indicative of the categorization of the occupant or object.",2006-04-11,"The title of the patent is method and apparatus for sensing seat occupancy and its abstract is method and apparatus for identifying and categorizing the weight and characteristics of the occupant currently occupying a vehicle seat. the method for identifying and categorizing the occupant or object involves measuring the deflection of the upper surface of the seat cushion at several points and therefore the weight distribution of the occupant. the system contains multiple weight sensors arrayed for detecting the distribution of the load causing the seat deflection. the system also includes a sensor to measure ambient temperature, preferably for temperature compensation due to the effects extreme temperatures may have on the compression properties of the seat cushion material and the weight sensors. a system processor interprets the data acquired by the sensors, and utilizes an algorithm and weight tables to simulate a neural network in providing an output a control signal indicative of the categorization of the occupant or object. dated 2006-04-11"
7031857,patient condition display,"data from a plurality of sensors representing a patient's condition, including measurement signals and also secondary parameters derived from the measurement signals, are displayed in a simple way by calculating a novelty index constituting a one-dimensional visualization space. the novelty index is based on the distance of the current data point in a multi-dimensional measurement space, whose coordinates are defined by the values of the measurement signals and secondary parameters, from a predefined normal point. this may be achieved by using a suitably trained artificial neural network to sum the distance between the current data point in the measurement space and a plurality of prototype points representing normality.",2006-04-18,"The title of the patent is patient condition display and its abstract is data from a plurality of sensors representing a patient's condition, including measurement signals and also secondary parameters derived from the measurement signals, are displayed in a simple way by calculating a novelty index constituting a one-dimensional visualization space. the novelty index is based on the distance of the current data point in a multi-dimensional measurement space, whose coordinates are defined by the values of the measurement signals and secondary parameters, from a predefined normal point. this may be achieved by using a suitably trained artificial neural network to sum the distance between the current data point in the measurement space and a plurality of prototype points representing normality. dated 2006-04-18"
7031950,method and apparatus for providing a virtual age estimation for remaining lifetime prediction of a system using neural networks,"a method for providing a virtual age estimation for predicting the remaining lifetime of a device of a given type, comprises the steps of monitoring a predetermined number of significant parameters of respective ones of a training set of devices of the given type, the parameters contributing respective wear increments, determining coefficients of a radial basis function neural network for modeling the wear increments determined from the training set operated to failure and whereof the respective virtual ages are normalized substantially to a desired norm value, deriving from the radial basis function neural network a formula for virtual age of a device of the given type, and applying the formula to the significant parameters from a further device of the given type for deriving wear increments for the further device.",2006-04-18,"The title of the patent is method and apparatus for providing a virtual age estimation for remaining lifetime prediction of a system using neural networks and its abstract is a method for providing a virtual age estimation for predicting the remaining lifetime of a device of a given type, comprises the steps of monitoring a predetermined number of significant parameters of respective ones of a training set of devices of the given type, the parameters contributing respective wear increments, determining coefficients of a radial basis function neural network for modeling the wear increments determined from the training set operated to failure and whereof the respective virtual ages are normalized substantially to a desired norm value, deriving from the radial basis function neural network a formula for virtual age of a device of the given type, and applying the formula to the significant parameters from a further device of the given type for deriving wear increments for the further device. dated 2006-04-18"
7034701,identification of fire signatures for shipboard multi-criteria fire detection systems,"a multi-criteria fire detection system, comprising a plurality of sensors, wherein each sensor is capable of detecting a signature characteristic of a presence of a fire and providing an output indicating the same. a processor for receiving each output of the plurality of sensors is also employed. the processor includes a probabilistic neural network for processing the sensor outputs. the probabilistic neural network comprises a nonlinear, nor-parametric pattern recognition algorithm that operates by defining a probability density function for a plurality of data sets that are each based on a training set data and an optimized kernel width parameter. the plurality of data sets includes a baseline, non-fire, fist data set; a second, fire data set, and a third, nuisance data set. the algorithm provides a decisional output indicative of the presence of a fire based on recognizing and discrimination between said data sets, and whether the outputs suffice to substantially indicate the presence of a fire, as opposed to a non-fire or nuisance situation.",2006-04-25,"The title of the patent is identification of fire signatures for shipboard multi-criteria fire detection systems and its abstract is a multi-criteria fire detection system, comprising a plurality of sensors, wherein each sensor is capable of detecting a signature characteristic of a presence of a fire and providing an output indicating the same. a processor for receiving each output of the plurality of sensors is also employed. the processor includes a probabilistic neural network for processing the sensor outputs. the probabilistic neural network comprises a nonlinear, nor-parametric pattern recognition algorithm that operates by defining a probability density function for a plurality of data sets that are each based on a training set data and an optimized kernel width parameter. the plurality of data sets includes a baseline, non-fire, fist data set; a second, fire data set, and a third, nuisance data set. the algorithm provides a decisional output indicative of the presence of a fire based on recognizing and discrimination between said data sets, and whether the outputs suffice to substantially indicate the presence of a fire, as opposed to a non-fire or nuisance situation. dated 2006-04-25"
7035740,artificial intelligence and global normalization methods for genotyping,"described herein are systems and methods for normalizing data without the use of external controls. also described herein are systems and methods for analyzing cluster data, such as genotyping data, using an artificial neural network.",2006-04-25,"The title of the patent is artificial intelligence and global normalization methods for genotyping and its abstract is described herein are systems and methods for normalizing data without the use of external controls. also described herein are systems and methods for analyzing cluster data, such as genotyping data, using an artificial neural network. dated 2006-04-25"
7035834,engine control system using a cascaded neural network,"a method, system and machine-readable storage medium for monitoring an engine using a cascaded neural network that includes a plurality of neural networks is disclosed. in operation, the method, system and machine-readable storage medium store data corresponding to the cascaded neural network. signals generated by a plurality of engine sensors are then inputted into the cascaded neural network. next, a second neural network is updated at a first rate, with an output of a first neural network, wherein the output is based on the inputted signals. in response, the second neural network outputs at a second rate, at least one engine control signal, wherein the second rate is faster than the first rate.",2006-04-25,"The title of the patent is engine control system using a cascaded neural network and its abstract is a method, system and machine-readable storage medium for monitoring an engine using a cascaded neural network that includes a plurality of neural networks is disclosed. in operation, the method, system and machine-readable storage medium store data corresponding to the cascaded neural network. signals generated by a plurality of engine sensors are then inputted into the cascaded neural network. next, a second neural network is updated at a first rate, with an output of a first neural network, wherein the output is based on the inputted signals. in response, the second neural network outputs at a second rate, at least one engine control signal, wherein the second rate is faster than the first rate. dated 2006-04-25"
7039473,system and method for adaptive control of uncertain nonlinear processes,"a computer system for controlling a nonlinear physical process. the computer system comprises a linear controller and a neural network. the linear controller receives a command signal for control of the nonlinear physical process and a measured output signal from the output of the nonlinear physical process. the linear controller generates a control signal based on the command signal, a measured output signal, and a fixed linear model for the process. the neural network receives the control signal from the linear controller and the measured output signal from the output of the nonlinear physical process. the neural network uses the measured output signal to modify the connection weights of the neural network. the neural network also generates a modified control signal supplied to the linear controller to iterate a fixed point solution for the modified control signal used to control the nonlinear physical process.",2006-05-02,"The title of the patent is system and method for adaptive control of uncertain nonlinear processes and its abstract is a computer system for controlling a nonlinear physical process. the computer system comprises a linear controller and a neural network. the linear controller receives a command signal for control of the nonlinear physical process and a measured output signal from the output of the nonlinear physical process. the linear controller generates a control signal based on the command signal, a measured output signal, and a fixed linear model for the process. the neural network receives the control signal from the linear controller and the measured output signal from the output of the nonlinear physical process. the neural network uses the measured output signal to modify the connection weights of the neural network. the neural network also generates a modified control signal supplied to the linear controller to iterate a fixed point solution for the modified control signal used to control the nonlinear physical process. dated 2006-05-02"
7039619,"utilized nanotechnology apparatus using a neutral network, a solution and a connection gap","an apparatus for maintaining components in neural network formed utilizing nanotechnology is described herein. a connection gap can be formed between two terminals. a solution comprising a melting point at approximately room temperature can be provided, wherein the solution is maintained in the connection gap and comprises a plurality of nanoparticles forming nanoconnections thereof having connection strengths thereof, wherein the solution and the connection gap are adapted for use with a neural network formed utilizing nanotechnology, such when power is removed from the neural network, the solution freezes, thereby locking into place the connection strengths.",2006-05-02,"The title of the patent is utilized nanotechnology apparatus using a neutral network, a solution and a connection gap and its abstract is an apparatus for maintaining components in neural network formed utilizing nanotechnology is described herein. a connection gap can be formed between two terminals. a solution comprising a melting point at approximately room temperature can be provided, wherein the solution is maintained in the connection gap and comprises a plurality of nanoparticles forming nanoconnections thereof having connection strengths thereof, wherein the solution and the connection gap are adapted for use with a neural network formed utilizing nanotechnology, such when power is removed from the neural network, the solution freezes, thereby locking into place the connection strengths. dated 2006-05-02"
7039621,"system, method, and computer program product for representing object relationships in a multidimensional space","a method and computer product is presented for mapping n-dimensional input patterns into an m-dimensional space so as to preserve relationships that may exist in the n-dimensional space. a subset of the input patterns is chosen and mapped into the m-dimensional space using an iterative nonlinear mapping process. a set of locally defined neural networks is created, then trained in accordance with the mapping produced by the iterative process. additional input patterns not in the subset are mapped into the m-dimensional space by using one of the local neural networks. in an alternative embodiment, the local neural networks are only used after training and use of a global neural network. the global neural network is trained in accordance with the mapping produced by the iterative process. input patterns are initially projected into the m-dimensional space using the global neural network. local neural networks are then used to refine the results of the global network.",2006-05-02,"The title of the patent is system, method, and computer program product for representing object relationships in a multidimensional space and its abstract is a method and computer product is presented for mapping n-dimensional input patterns into an m-dimensional space so as to preserve relationships that may exist in the n-dimensional space. a subset of the input patterns is chosen and mapped into the m-dimensional space using an iterative nonlinear mapping process. a set of locally defined neural networks is created, then trained in accordance with the mapping produced by the iterative process. additional input patterns not in the subset are mapped into the m-dimensional space by using one of the local neural networks. in an alternative embodiment, the local neural networks are only used after training and use of a global neural network. the global neural network is trained in accordance with the mapping produced by the iterative process. input patterns are initially projected into the m-dimensional space using the global neural network. local neural networks are then used to refine the results of the global network. dated 2006-05-02"
7043431,multilingual speech recognition system using text derived recognition models,"there is provided a novel approach for generating multilingual text-to-phoneme mappings for use in multilingual speech recognition systems. the multilingual mappings are based on the weighted output from a neural network text-to-phoneme model, trained on data mixed from several languages. the multilingual mappings used together with a branched grammar decoding scheme is able to capture both inter- and intra-language pronunciation variations which is ideal for multilingual speaker independent recognition systems. a significant improvement in overall system performance is obtained for a multilingual speaker independent name dialing task when applying multilingual instead of language dependent text-to-phoneme mapping.",2006-05-09,"The title of the patent is multilingual speech recognition system using text derived recognition models and its abstract is there is provided a novel approach for generating multilingual text-to-phoneme mappings for use in multilingual speech recognition systems. the multilingual mappings are based on the weighted output from a neural network text-to-phoneme model, trained on data mixed from several languages. the multilingual mappings used together with a branched grammar decoding scheme is able to capture both inter- and intra-language pronunciation variations which is ideal for multilingual speaker independent recognition systems. a significant improvement in overall system performance is obtained for a multilingual speaker independent name dialing task when applying multilingual instead of language dependent text-to-phoneme mapping. dated 2006-05-09"
7043466,neural network processing system using semiconductor memories,"the neural network processing system according to the present invention includes a memory circuit for storing neuron output values, connection weights, the desired values of outputs, and data necessary for learning; an input/output circuit for writing or reading data in or out of said memory circuit; a processing circuit for performing a processing for determining the neuron outputs such as the product, sum and nonlinear conversion of the data stored in said memory circuit, a comparison of the output value and its desired value, and a processing necessary for learning; and a control circuit for controlling the operations of said memory circuit, said input/output circuit and said processing circuit.",2006-05-09,"The title of the patent is neural network processing system using semiconductor memories and its abstract is the neural network processing system according to the present invention includes a memory circuit for storing neuron output values, connection weights, the desired values of outputs, and data necessary for learning; an input/output circuit for writing or reading data in or out of said memory circuit; a processing circuit for performing a processing for determining the neuron outputs such as the product, sum and nonlinear conversion of the data stored in said memory circuit, a comparison of the output value and its desired value, and a processing necessary for learning; and a control circuit for controlling the operations of said memory circuit, said input/output circuit and said processing circuit. dated 2006-05-09"
7051293,method and apparatus for creating an extraction model,"a system for using machine-learning to create a model for performing integrated circuit layout extraction is disclosed. the system of the present invention has two main phases: model creation and model application. the model creation phase comprises creating one or more extraction models using machine-learning techniques. first, a complex extraction problem is decomposed into smaller simpler extraction problems. then, each smaller extraction problem is then analyzed to identify a set of physical parameters that fully define the smaller extraction problem. next, models are created using machine learning techniques for all of the smaller simpler extraction problems. the machine learning is performed by first creating training data sets composed of the identified parameters from typical examples of the smaller extraction problem and the answers to those example extraction problems as solved using a highly accurate physics-based field solver. the training sets are then used to train the models. in one embodiment, neural networks are used to model the extraction problems. to train the neural network models. bayesian inference is used in one embodiment. bayesian inference may be implemented with normal monte carlo techniques or hybrid monte carlo techniques. after the creation of a set of models for each of the smaller simpler extraction problems, the machine-learning based models may be used for extraction.",2006-05-23,"The title of the patent is method and apparatus for creating an extraction model and its abstract is a system for using machine-learning to create a model for performing integrated circuit layout extraction is disclosed. the system of the present invention has two main phases: model creation and model application. the model creation phase comprises creating one or more extraction models using machine-learning techniques. first, a complex extraction problem is decomposed into smaller simpler extraction problems. then, each smaller extraction problem is then analyzed to identify a set of physical parameters that fully define the smaller extraction problem. next, models are created using machine learning techniques for all of the smaller simpler extraction problems. the machine learning is performed by first creating training data sets composed of the identified parameters from typical examples of the smaller extraction problem and the answers to those example extraction problems as solved using a highly accurate physics-based field solver. the training sets are then used to train the models. in one embodiment, neural networks are used to model the extraction problems. to train the neural network models. bayesian inference is used in one embodiment. bayesian inference may be implemented with normal monte carlo techniques or hybrid monte carlo techniques. after the creation of a set of models for each of the smaller simpler extraction problems, the machine-learning based models may be used for extraction. dated 2006-05-23"
7058455,interface for making spatially resolved electrical contact to neural cells in a biological neural network,"an interface for selective excitation or sensing of neural cells in a biological neural network is provided. the interface includes a membrane with a number of channels passing through the membrane. each channel has at least one electrode within it. neural cells in the biological neural network grow or migrate into the channels, thereby coming into close proximity to the electrodes.once one or more neural cells have grown or migrated into a channel, a voltage applied to the electrode within the channel selectively excites the neural cell (or cells) in that channel. the excitation of these neural cell(s) will then transmit throughout the neural network (i.e. cells and axons) that is associated with the neural cell(s) stimulated in the channel.",2006-06-06,"The title of the patent is interface for making spatially resolved electrical contact to neural cells in a biological neural network and its abstract is an interface for selective excitation or sensing of neural cells in a biological neural network is provided. the interface includes a membrane with a number of channels passing through the membrane. each channel has at least one electrode within it. neural cells in the biological neural network grow or migrate into the channels, thereby coming into close proximity to the electrodes.once one or more neural cells have grown or migrated into a channel, a voltage applied to the electrode within the channel selectively excites the neural cell (or cells) in that channel. the excitation of these neural cell(s) will then transmit throughout the neural network (i.e. cells and axons) that is associated with the neural cell(s) stimulated in the channel. dated 2006-06-06"
7058616,method and system for predicting resistance of a disease to a therapeutic agent using a neural network,"a method and system for predicting the resistance of a disease to a therapeutic agent is provided. further provided is a method and system for designing a therapeutic treatment agent for a patient afflicted with a disease. specifically, the methods use a trained neural network to interpret genotypic information obtained from the disease. the trained neural network is trained using a database of known or determined genotypic mutations that are correlated with phenotypic therapeutic agent resistance. the present invention also provides methods and systems for predicting the probability of a patient developing a genetic disease. a trained neural network for making such predictions is also provided.",2006-06-06,"The title of the patent is method and system for predicting resistance of a disease to a therapeutic agent using a neural network and its abstract is a method and system for predicting the resistance of a disease to a therapeutic agent is provided. further provided is a method and system for designing a therapeutic treatment agent for a patient afflicted with a disease. specifically, the methods use a trained neural network to interpret genotypic information obtained from the disease. the trained neural network is trained using a database of known or determined genotypic mutations that are correlated with phenotypic therapeutic agent resistance. the present invention also provides methods and systems for predicting the probability of a patient developing a genetic disease. a trained neural network for making such predictions is also provided. dated 2006-06-06"
7058618,method for establishing stress/strain curves by means of spline interpolation on the basis of characteristic points and with the use of neural networks,"a stress/strain curve is established by means of neural networks 1 to n and 4. to that end, parameters are input into the input 50, from which the neural networks 1 to n respectively establish the principal components of characteristic points. the curve type is selected on the basis of the output of the neural network 4. the principal components of the characteristic points of the corresponding curve type are then inverse-transformed. the stress/strain curve is then calculated by the generator 59 on the basis of the inverse transformation.",2006-06-06,"The title of the patent is method for establishing stress/strain curves by means of spline interpolation on the basis of characteristic points and with the use of neural networks and its abstract is a stress/strain curve is established by means of neural networks 1 to n and 4. to that end, parameters are input into the input 50, from which the neural networks 1 to n respectively establish the principal components of characteristic points. the curve type is selected on the basis of the output of the neural network 4. the principal components of the characteristic points of the corresponding curve type are then inverse-transformed. the stress/strain curve is then calculated by the generator 59 on the basis of the inverse transformation. dated 2006-06-06"
7062476,student neural network,"a student neural network that is capable of receiving a series of tutoring inputs from one or more teacher networks to generate a student network output that is similar to the output of the one or more teacher networks. the tutoring inputs are repeatedly processed by the student until, using a suitable method such as back propagation of errors, the outputs of the student approximate the outputs of the teachers within a predefined range. once the desired outputs are obtained, the weights of the student network are set. using this weight set the student is now capable of solving all of the problems of the teacher networks without the need for adjustment of its internal weights. if the user desires to use the student to solve a different series of problems, the user only needs to retrain the student by supplying a different series of tutoring inputs.",2006-06-13,"The title of the patent is student neural network and its abstract is a student neural network that is capable of receiving a series of tutoring inputs from one or more teacher networks to generate a student network output that is similar to the output of the one or more teacher networks. the tutoring inputs are repeatedly processed by the student until, using a suitable method such as back propagation of errors, the outputs of the student approximate the outputs of the teachers within a predefined range. once the desired outputs are obtained, the weights of the student network are set. using this weight set the student is now capable of solving all of the problems of the teacher networks without the need for adjustment of its internal weights. if the user desires to use the student to solve a different series of problems, the user only needs to retrain the student by supplying a different series of tutoring inputs. dated 2006-06-13"
7067993,light source control system,"a system and method for controlling an optical light source is provided. a current source drives the light source, while the voltage across and the current through the light source is measured. the voltage and current are converted to digital signals and sent to a neural network, which generates a modeled optical output power of the light source and a modeled value of the optical wavelength. a control circuit receives the modeled optical output power and wavelength and sends a control signal to the current source to minimize the difference between the desired power output and the modeled output power. in addition, a control signal is sent to a peltier driver to control the temperature of a peltier cooler in order to increase or decrease the wavelength emitted by a laser diode.",2006-06-27,"The title of the patent is light source control system and its abstract is a system and method for controlling an optical light source is provided. a current source drives the light source, while the voltage across and the current through the light source is measured. the voltage and current are converted to digital signals and sent to a neural network, which generates a modeled optical output power of the light source and a modeled value of the optical wavelength. a control circuit receives the modeled optical output power and wavelength and sends a control signal to the current source to minimize the difference between the desired power output and the modeled output power. in addition, a control signal is sent to a peltier driver to control the temperature of a peltier cooler in order to increase or decrease the wavelength emitted by a laser diode. dated 2006-06-27"
7069256,neural network module for data mining,"a system, software module, and computer program product for performing neural network based data mining that improved performance in model building, good integration with the various databases throughout the enterprise, flexible specification and adjustment of the models being built, and flexible model arrangement and export capability. the software module for performing neural network based data mining in an electronic data processing system comprises: a model setup block operable to receive client input including information specifying a setup of a neural network data mining models, generate the model setup, generate parameters for the model setup based on the received information, a modeling algorithms block operable to select and initialize a neural network modeling algorithm based on the generated model setup, a model building block operable to receive training data and build a neural network model using the training data and the selected neural network modeling algorithm and a model scoring block operable to receive scoring data and generate predictions and/or recommendations using the scoring data and the neural network model.",2006-06-27,"The title of the patent is neural network module for data mining and its abstract is a system, software module, and computer program product for performing neural network based data mining that improved performance in model building, good integration with the various databases throughout the enterprise, flexible specification and adjustment of the models being built, and flexible model arrangement and export capability. the software module for performing neural network based data mining in an electronic data processing system comprises: a model setup block operable to receive client input including information specifying a setup of a neural network data mining models, generate the model setup, generate parameters for the model setup based on the received information, a modeling algorithms block operable to select and initialize a neural network modeling algorithm based on the generated model setup, a model building block operable to receive training data and build a neural network model using the training data and the selected neural network modeling algorithm and a model scoring block operable to receive scoring data and generate predictions and/or recommendations using the scoring data and the neural network model. dated 2006-06-27"
7069257,pattern recognition method for reducing classification errors,"a rbf pattern recognition method for reducing classification errors is provided. an optimum rbf training approach is obtained for reducing an error calculated by an error function. the invention continuously generates the updated differences of parameters in the learning process of recognizing training samples. the modified parameters are employed to stepwise adjust the rbf neural network. the invention can distinguish different degrees of importance and learning contributions among the training samples and evaluate the learning contribution of each training sample for obtaining differences of the parameters of the training samples. when the learning contribution is larger, the updated difference is larger to speed up the learning. thus, the difference of the parameters is zero when the training samples are classified as the correct pattern type.",2006-06-27,"The title of the patent is pattern recognition method for reducing classification errors and its abstract is a rbf pattern recognition method for reducing classification errors is provided. an optimum rbf training approach is obtained for reducing an error calculated by an error function. the invention continuously generates the updated differences of parameters in the learning process of recognizing training samples. the modified parameters are employed to stepwise adjust the rbf neural network. the invention can distinguish different degrees of importance and learning contributions among the training samples and evaluate the learning contribution of each training sample for obtaining differences of the parameters of the training samples. when the learning contribution is larger, the updated difference is larger to speed up the learning. thus, the difference of the parameters is zero when the training samples are classified as the correct pattern type. dated 2006-06-27"
7072741,robot control algorithm constructing apparatus,"a control algorithm constructing device that constructs a control algorithm controlling the motion of a robot, and a controller that controls the motion of the robot in accordance with the constructed control algorithm, with the purpose of reducing the cost and time taken to create the control algorithm as compared with the conventional method such as an mzp method to solve a mechanical equation, in which the control algorithm is constructed by a recurrent neural network (rnn) including a neuron generating an output with an analogue lag with respect to an input, the coefficients of the rnn are determined in succession from the term of lower degree to the term of higher degree.",2006-07-04,"The title of the patent is robot control algorithm constructing apparatus and its abstract is a control algorithm constructing device that constructs a control algorithm controlling the motion of a robot, and a controller that controls the motion of the robot in accordance with the constructed control algorithm, with the purpose of reducing the cost and time taken to create the control algorithm as compared with the conventional method such as an mzp method to solve a mechanical equation, in which the control algorithm is constructed by a recurrent neural network (rnn) including a neuron generating an output with an analogue lag with respect to an input, the coefficients of the rnn are determined in succession from the term of lower degree to the term of higher degree. dated 2006-07-04"
7072872,representation and retrieval of images using context vectors derived from image information elements,"image features are generated by performing wavelet transformations at sample points on images stored in electronic form. multiple wavelet transformations at a point are combined to form an image feature vector. a prototypical set of feature vectors, or atoms, is derived from the set of feature vectors to form an “atomic vocabulary.” the prototypical feature vectors are derived using a vector quantization method, e.g., using neural network self-organization techniques, in which a vector quantization network is also generated. the atomic vocabulary is used to define new images. meaning is established between atoms in the atomic vocabulary. high-dimensional context vectors are assigned to each atom. the context vectors are then trained as a function of the proximity and co-occurrence of each atom to other atoms in the image. after training, the context vectors associated with the atoms that comprise an image are combined to form a summary vector for the image. images are retrieved using a number of query methods, e.g., images, image portions, vocabulary atoms, index terms. the user's query is converted into a query context vector. a dot product is calculated between the query vector and the summary vectors to locate images having the closest meaning. the invention is also applicable to video or temporally related images, and can also be used in conjunction with other context vector data domains such as text or audio, thereby linking images to such data domains.",2006-07-04,"The title of the patent is representation and retrieval of images using context vectors derived from image information elements and its abstract is image features are generated by performing wavelet transformations at sample points on images stored in electronic form. multiple wavelet transformations at a point are combined to form an image feature vector. a prototypical set of feature vectors, or atoms, is derived from the set of feature vectors to form an “atomic vocabulary.” the prototypical feature vectors are derived using a vector quantization method, e.g., using neural network self-organization techniques, in which a vector quantization network is also generated. the atomic vocabulary is used to define new images. meaning is established between atoms in the atomic vocabulary. high-dimensional context vectors are assigned to each atom. the context vectors are then trained as a function of the proximity and co-occurrence of each atom to other atoms in the image. after training, the context vectors associated with the atoms that comprise an image are combined to form a summary vector for the image. images are retrieved using a number of query methods, e.g., images, image portions, vocabulary atoms, index terms. the user's query is converted into a query context vector. a dot product is calculated between the query vector and the summary vectors to locate images having the closest meaning. the invention is also applicable to video or temporally related images, and can also be used in conjunction with other context vector data domains such as text or audio, thereby linking images to such data domains. dated 2006-07-04"
7072874,optimization of training sets for neural-net processing of characteristic patterns from vibrating solids,an artificial neural network is disclosed that processes holography generated characteristic patterns of vibrating structures along with finite-element models. the present invention provides for a folding operation for conditioning training sets for optimally training forward-neural networks to process characteristic fringe patterns. the folding pattern increases the sensitivity of the feed-forward network for detecting changes in the characteristic pattern. the folding routine manipulates input pixels so as to be scaled according to the location in an intensity range rather than the position in the characteristic pattern.,2006-07-04,The title of the patent is optimization of training sets for neural-net processing of characteristic patterns from vibrating solids and its abstract is an artificial neural network is disclosed that processes holography generated characteristic patterns of vibrating structures along with finite-element models. the present invention provides for a folding operation for conditioning training sets for optimally training forward-neural networks to process characteristic fringe patterns. the folding pattern increases the sensitivity of the feed-forward network for detecting changes in the characteristic pattern. the folding routine manipulates input pixels so as to be scaled according to the location in an intensity range rather than the position in the characteristic pattern. dated 2006-07-04
7072875,"information processing apparatus and method, and recording medium","an information processing apparatus includes a first recurrent neural network (rnn) for performing processing which corresponds to a time-series and a second rnn for processing another correlated time-series. the difference between a context set output by the first rnn and a context set output by the second rnn is computed by a subtractor, and the obtained difference is used as a prediction error. backpropagation is performed based on the prediction error, thus determining a coefficient for each neuron of an output layer, an intermediate layer, and an input layer.",2006-07-04,"The title of the patent is information processing apparatus and method, and recording medium and its abstract is an information processing apparatus includes a first recurrent neural network (rnn) for performing processing which corresponds to a time-series and a second rnn for processing another correlated time-series. the difference between a context set output by the first rnn and a context set output by the second rnn is computed by a subtractor, and the obtained difference is used as a prediction error. backpropagation is performed based on the prediction error, thus determining a coefficient for each neuron of an output layer, an intermediate layer, and an input layer. dated 2006-07-04"
7076371,non-invasive diagnostic and monitoring method and apparatus based on odor detection,"a set of volatile markers are determined which are characteristic of a particular condition or disease, and which will be found in the exhaled breath of a person or odor from other parts of a body or from an entity. these markers are detected in the breath odor or gaseous emanations from the body or entity noninvasively using a volatile substance detector of sufficient sensitivity, such as an artificial olfactory system. the detected marker data is processed in an artificial neural network/fuzzy filter system with an algorithm that intelligently adapts to the individual body or entity and also optionally (if necessary) with a correction algorithm to eliminate environmental and other erroneous contributions to the markers. any number of markers may be used, depending on how well they correlate with the condition and how accurate a result is desired, i.e. general screening or accurate diagnosis and monitoring.",2006-07-11,"The title of the patent is non-invasive diagnostic and monitoring method and apparatus based on odor detection and its abstract is a set of volatile markers are determined which are characteristic of a particular condition or disease, and which will be found in the exhaled breath of a person or odor from other parts of a body or from an entity. these markers are detected in the breath odor or gaseous emanations from the body or entity noninvasively using a volatile substance detector of sufficient sensitivity, such as an artificial olfactory system. the detected marker data is processed in an artificial neural network/fuzzy filter system with an algorithm that intelligently adapts to the individual body or entity and also optionally (if necessary) with a correction algorithm to eliminate environmental and other erroneous contributions to the markers. any number of markers may be used, depending on how well they correlate with the condition and how accurate a result is desired, i.e. general screening or accurate diagnosis and monitoring. dated 2006-07-11"
7078680,ion mobility spectrometer using ion beam modulation and wavelet decomposition,"the present invention is directed to a system and a method for analyzing and identifying an unknown sample using ion mobility spectrometry. the method pulses an ion gate using a temporally spaced pattern of ion admitting periods and ion repelling periods to achieve an admission duty cycle of about 50% of the total scan time. ions passing through the drift tube strike an ion detector, generating a time dependent mobility spectrum. a combination of wavelet decomposition and statistical evaluators are used on the mobility spectrum to produce a distinct signature associated with the sample. signatures are also generated for a number of known agents, and the known agent signatures are used to program a neural network. the sample signature is then compared to the signatures for the known agents using a fuzzy decision maker.",2006-07-18,"The title of the patent is ion mobility spectrometer using ion beam modulation and wavelet decomposition and its abstract is the present invention is directed to a system and a method for analyzing and identifying an unknown sample using ion mobility spectrometry. the method pulses an ion gate using a temporally spaced pattern of ion admitting periods and ion repelling periods to achieve an admission duty cycle of about 50% of the total scan time. ions passing through the drift tube strike an ion detector, generating a time dependent mobility spectrum. a combination of wavelet decomposition and statistical evaluators are used on the mobility spectrum to produce a distinct signature associated with the sample. signatures are also generated for a number of known agents, and the known agent signatures are used to program a neural network. the sample signature is then compared to the signatures for the known agents using a fuzzy decision maker. dated 2006-07-18"
7079235,reticle design inspection system,"a method of reticle inspection, comprising generating a test reticle comprising a plurality of test pattern-features thereon; manufacturing a wafer using the reticle; and determining a transfer of at least one of said plurality of pattern features from said reticle to said wafer. preferably, a neural network is trained using the determination. preferably, a reticle is inspected by running detected defects through the neural network to determine if the detected defect has a consequence.",2006-07-18,"The title of the patent is reticle design inspection system and its abstract is a method of reticle inspection, comprising generating a test reticle comprising a plurality of test pattern-features thereon; manufacturing a wafer using the reticle; and determining a transfer of at least one of said plurality of pattern features from said reticle to said wafer. preferably, a neural network is trained using the determination. preferably, a reticle is inspected by running detected defects through the neural network to determine if the detected defect has a consequence. dated 2006-07-18"
7080053,neural network device for evolving appropriate connections,"a method for evolving appropriate connections among units in a neural network includes a) calculating weight changes at each existing connection and incipient connections between units for each training example; and b) determining a k ratio using the weight changes, wherein said k ratio comprises the weight change of existing connections, and wherein if the k ratio the weight change of incipient connections exceeds a threshold, further including b1) increasing a weight of the existing connection; b2) creating new connections at the incipient connections. the method further includes c) pruning weak connections between the units.",2006-07-18,"The title of the patent is neural network device for evolving appropriate connections and its abstract is a method for evolving appropriate connections among units in a neural network includes a) calculating weight changes at each existing connection and incipient connections between units for each training example; and b) determining a k ratio using the weight changes, wherein said k ratio comprises the weight change of existing connections, and wherein if the k ratio the weight change of incipient connections exceeds a threshold, further including b1) increasing a weight of the existing connection; b2) creating new connections at the incipient connections. the method further includes c) pruning weak connections between the units. dated 2006-07-18"
7080055,backlash compensation with filtered prediction in discrete time nonlinear systems by dynamic inversion using neural networks,"methods and apparatuses for backlash compensation. a dynamics inversion compensation scheme is designed for control of nonlinear discrete-time systems with input backlash. the techniques of this disclosure extend the dynamic inversion technique to discrete-time systems by using a filtered prediction, and shows how to use a neural network (nn) for inverting the backlash nonlinearity in the feedforward path. the techniques provide a general procedure for using nn to determine the dynamics preinverse of an invertible discrete time dynamical system. a discrete-time tuning algorithm is given for the nn weights so that the backlash compensation scheme guarantees bounded tracking and backlash errors, and also bounded parameter estimates. a rigorous proof of stability and performance is given and a simulation example verifies performance. unlike standard discrete-time adaptive control techniques, no certainty equivalence (ce) or linear-in-the-parameters (lip) assumptions are needed.",2006-07-18,"The title of the patent is backlash compensation with filtered prediction in discrete time nonlinear systems by dynamic inversion using neural networks and its abstract is methods and apparatuses for backlash compensation. a dynamics inversion compensation scheme is designed for control of nonlinear discrete-time systems with input backlash. the techniques of this disclosure extend the dynamic inversion technique to discrete-time systems by using a filtered prediction, and shows how to use a neural network (nn) for inverting the backlash nonlinearity in the feedforward path. the techniques provide a general procedure for using nn to determine the dynamics preinverse of an invertible discrete time dynamical system. a discrete-time tuning algorithm is given for the nn weights so that the backlash compensation scheme guarantees bounded tracking and backlash errors, and also bounded parameter estimates. a rigorous proof of stability and performance is given and a simulation example verifies performance. unlike standard discrete-time adaptive control techniques, no certainty equivalence (ce) or linear-in-the-parameters (lip) assumptions are needed. dated 2006-07-18"
7082419,neural processing element for use in a neural network,"a neural processing element for use in a modular neural network is provided. one embodiment provides a neural network comprising an array of autonomous modules (300). the modules (300) can be arranged in a variety of configurations to form neural networks with various topologies, for example, with a hierarchical modular structure. each module (300) contains sufficient neurons (100) to enable it to do useful work as a stand alone system, with the advantage that many modules (300) can be connected together to create a wide variety of configurations and network sizes. this modular approach results in a scaleable system that meets increased workload with an increase in parallelism and thereby avoids the usually extensive increases in training times associated with unitary implementations.",2006-07-25,"The title of the patent is neural processing element for use in a neural network and its abstract is a neural processing element for use in a modular neural network is provided. one embodiment provides a neural network comprising an array of autonomous modules (300). the modules (300) can be arranged in a variety of configurations to form neural networks with various topologies, for example, with a hierarchical modular structure. each module (300) contains sufficient neurons (100) to enable it to do useful work as a stand alone system, with the advantage that many modules (300) can be connected together to create a wide variety of configurations and network sizes. this modular approach results in a scaleable system that meets increased workload with an increase in parallelism and thereby avoids the usually extensive increases in training times associated with unitary implementations. dated 2006-07-25"
7082421,"information processing apparatus and method, and recording medium","an information processing apparatus includes a first recurrent neural network (rnn) for performing processing which corresponds to a time-series and a second rnn for processing another correlated time-series. the difference between a context set output by the first rnn and a context set output by the second rnn is computed by a subtractor, and the obtained difference is used as a prediction error. backpropagation is performed based on the prediction error, thus determining a coefficient for each neuron of an output layer, an intermediate layer, and an input layer.",2006-07-25,"The title of the patent is information processing apparatus and method, and recording medium and its abstract is an information processing apparatus includes a first recurrent neural network (rnn) for performing processing which corresponds to a time-series and a second rnn for processing another correlated time-series. the difference between a context set output by the first rnn and a context set output by the second rnn is computed by a subtractor, and the obtained difference is used as a prediction error. backpropagation is performed based on the prediction error, thus determining a coefficient for each neuron of an output layer, an intermediate layer, and an input layer. dated 2006-07-25"
7085655,method and device for detecting defects of at least one rotary wing aircraft rotor,"a method and apparatus for detecting the defects of at least one rotor of a rotary wing aircraft, in particular a helicopter, in order to adjust the rotor. in the method, in a preliminary step, a reference aircraft is used corresponding to a particular type of aircraft and having a rotor (6, 10) without defect and adjusted so that the level of vibration is at a minimum, a series of measurements are taken on the aircraft (1), and a neural network is deduced therefrom illustrating the relationships between accelerations representative of vibration, and defects, and adjustment parameters. thereafter, in a later step, for a particular aircraft (1) of the same type, measurements are taken on the particular aircraft (1) and on the basis of said measurements and on the basis of the neural network, the defects, if any, are detected and values for the adjustment parameters are determined that will enable the level of vibration of the aircraft (1) to be minimized, which parameters are applied to the rotor (6, 10).",2006-08-01,"The title of the patent is method and device for detecting defects of at least one rotary wing aircraft rotor and its abstract is a method and apparatus for detecting the defects of at least one rotor of a rotary wing aircraft, in particular a helicopter, in order to adjust the rotor. in the method, in a preliminary step, a reference aircraft is used corresponding to a particular type of aircraft and having a rotor (6, 10) without defect and adjusted so that the level of vibration is at a minimum, a series of measurements are taken on the aircraft (1), and a neural network is deduced therefrom illustrating the relationships between accelerations representative of vibration, and defects, and adjustment parameters. thereafter, in a later step, for a particular aircraft (1) of the same type, measurements are taken on the particular aircraft (1) and on the basis of said measurements and on the basis of the neural network, the defects, if any, are detected and values for the adjustment parameters are determined that will enable the level of vibration of the aircraft (1) to be minimized, which parameters are applied to the rotor (6, 10). dated 2006-08-01"
7085749,"pulse signal circuit, parallel processing circuit, pattern recognition system, and image input system","a synaptic connection element for connecting neuron elements inputs a plurality of pulsed signals from different neuron elements n1 through n4, effects a common modulation (time window integration or pulse phase/width modulation) on a plurality of predetermined signals among the plurality of pulse signals, and outputs the modulated pulse signals to different signal lines to a neuron element m1. a neural network for representing and processing pattern information by the pulse modulation is thereby downsized in scale.",2006-08-01,"The title of the patent is pulse signal circuit, parallel processing circuit, pattern recognition system, and image input system and its abstract is a synaptic connection element for connecting neuron elements inputs a plurality of pulsed signals from different neuron elements n1 through n4, effects a common modulation (time window integration or pulse phase/width modulation) on a plurality of predetermined signals among the plurality of pulse signals, and outputs the modulated pulse signals to different signal lines to a neuron element m1. a neural network for representing and processing pattern information by the pulse modulation is thereby downsized in scale. dated 2006-08-01"
7089217,adaptive learning system and method,"a neural network module including an input layer having one or more input nodes arranged to receive input data, a rule base layer having one or more rule nodes, an output layer having one or more output nodes, and an adaptive component arranged to aggregate selected two or more rule nodes in the rule base layer based on the input data, an adaptive learning system having one or more of the neural network modules, related methods of implementing the neural network module and an adaptive learning system, and a neural network program.",2006-08-08,"The title of the patent is adaptive learning system and method and its abstract is a neural network module including an input layer having one or more input nodes arranged to receive input data, a rule base layer having one or more rule nodes, an output layer having one or more output nodes, and an adaptive component arranged to aggregate selected two or more rule nodes in the rule base layer based on the input data, an adaptive learning system having one or more of the neural network modules, related methods of implementing the neural network module and an adaptive learning system, and a neural network program. dated 2006-08-08"
7089219,"information processing apparatus and method, and recording medium","an information processing apparatus includes a first recurrent neural network (rnn) for performing processing which corresponds to a time-series and a second rnn for processing another correlated time-series. the difference between a context set output by the first rnn and a context set output by the second rnn is computed by a subtractor, and the obtained difference is used as a prediction error. backpropagation is performed based on the prediction error, thus determining a coefficient for each neuron of an output layer, an intermediate layer, and an input layer.",2006-08-08,"The title of the patent is information processing apparatus and method, and recording medium and its abstract is an information processing apparatus includes a first recurrent neural network (rnn) for performing processing which corresponds to a time-series and a second rnn for processing another correlated time-series. the difference between a context set output by the first rnn and a context set output by the second rnn is computed by a subtractor, and the obtained difference is used as a prediction error. backpropagation is performed based on the prediction error, thus determining a coefficient for each neuron of an output layer, an intermediate layer, and an input layer. dated 2006-08-08"
7089499,personalizing user interfaces across operating systems,"a user-centered interface agent learns user preferences and typical behaviors and, based on what is learned, predicts the user's preferred user interface for different types of host computers. the interface agent consists of a learning program which operates on the user's primary computer, a shadow program which is installed on a personal digital assistant (pda), and a remote program which operates on host computers. the pda transfers data between the primary and remote machines, and can also be used as the user's primary computer. on the primary computer, the agent learns a user's preferences automatically by observing the user's actions, requiring minimal initialization by the user. the learning algorithm may be statistical, rule-based, case-based, neural network, or employ any other technique for reasoning under uncertainty. the automated personalizing of a user interface configuration has a particular advantage for individuals with disabilities who require configuration before they can use a new computer system.",2006-08-08,"The title of the patent is personalizing user interfaces across operating systems and its abstract is a user-centered interface agent learns user preferences and typical behaviors and, based on what is learned, predicts the user's preferred user interface for different types of host computers. the interface agent consists of a learning program which operates on the user's primary computer, a shadow program which is installed on a personal digital assistant (pda), and a remote program which operates on host computers. the pda transfers data between the primary and remote machines, and can also be used as the user's primary computer. on the primary computer, the agent learns a user's preferences automatically by observing the user's actions, requiring minimal initialization by the user. the learning algorithm may be statistical, rule-based, case-based, neural network, or employ any other technique for reasoning under uncertainty. the automated personalizing of a user interface configuration has a particular advantage for individuals with disabilities who require configuration before they can use a new computer system. dated 2006-08-08"
7089592,systems and methods for dynamic detection and prevention of electronic fraud,"the present invention provides systems and methods for dynamic detection and prevention of electronic fraud and network intrusion using an integrated set of intelligent technologies. the intelligent technologies include neural networks, multi-agents, data mining, case-based reasoning, rule-based reasoning, fuzzy logic, constraint programming, and genetic algorithms. the systems and methods of the present invention involve a fraud detection and prevention model that successfully detects and prevents electronic fraud and network intrusion in real-time. the model is not sensitive to known or unknown different types of fraud or network intrusion attacks, and can be used to detect and prevent fraud and network intrusion across multiple networks and industries.",2006-08-08,"The title of the patent is systems and methods for dynamic detection and prevention of electronic fraud and its abstract is the present invention provides systems and methods for dynamic detection and prevention of electronic fraud and network intrusion using an integrated set of intelligent technologies. the intelligent technologies include neural networks, multi-agents, data mining, case-based reasoning, rule-based reasoning, fuzzy logic, constraint programming, and genetic algorithms. the systems and methods of the present invention involve a fraud detection and prevention model that successfully detects and prevents electronic fraud and network intrusion in real-time. the model is not sensitive to known or unknown different types of fraud or network intrusion attacks, and can be used to detect and prevent fraud and network intrusion across multiple networks and industries. dated 2006-08-08"
7092777,intelligent agent system and method for evaluating data integrity in process information databases,"a data integrity module and method for evaluating data in a process information database. a neural network generates statistical patterns for specifying patterns for the data being evaluated. a fuzzy expert rules base specifies rules for evaluating the data. a processor, responsive to the rules base and the statistical patterns, identifies suspect data in the process information database by evaluating the data according to the rules base and the statistical patterns. a modification system modifies the suspect data in the process information database.",2006-08-15,"The title of the patent is intelligent agent system and method for evaluating data integrity in process information databases and its abstract is a data integrity module and method for evaluating data in a process information database. a neural network generates statistical patterns for specifying patterns for the data being evaluated. a fuzzy expert rules base specifies rules for evaluating the data. a processor, responsive to the rules base and the statistical patterns, identifies suspect data in the process information database by evaluating the data according to the rules base and the statistical patterns. a modification system modifies the suspect data in the process information database. dated 2006-08-15"
7092857,neural network for computer-aided knowledge management,"the invention relates to a method and a neural network for computer-assisted knowledge management, based on a neural network (1) that is formed by a computer in its memory location. the invention method and neural network are especially for use in a decentralized, computer-assisted patent system that can be used via the internet system, in the broad sense. the neural network (1) forms a system of artificial intelligence (ki), covering a fundamental knowledge base in the form of computer-readable texts. the neural network (1) consists of elements (2) that are associated with each other and weighted in relation to each other so that the sets of knowledge available can be managed and analyzed in relation to each other by computer means.",2006-08-15,"The title of the patent is neural network for computer-aided knowledge management and its abstract is the invention relates to a method and a neural network for computer-assisted knowledge management, based on a neural network (1) that is formed by a computer in its memory location. the invention method and neural network are especially for use in a decentralized, computer-assisted patent system that can be used via the internet system, in the broad sense. the neural network (1) forms a system of artificial intelligence (ki), covering a fundamental knowledge base in the form of computer-readable texts. the neural network (1) consists of elements (2) that are associated with each other and weighted in relation to each other so that the sets of knowledge available can be managed and analyzed in relation to each other by computer means. dated 2006-08-15"
7092881,parametric speech codec for representing synthetic speech in the presence of background noise,"a system and method are provided for processing audio and speech signals using a pitch and voicing dependent spectral estimation algorithm (voicing algorithm) to accurately represent voiced speech, unvoiced speech, and mixed speech in the presence of background noise, and background noise with a single model. the present invention also modifies the synthesis model based on an estimate of the current input signal to improve the perceptual quality of the speech and background noise under a variety of input conditions. the present invention also improves the voicing dependent spectral estimation algorithm robustness by introducing the use of a multi-layer neural network in the estimation process. the voicing dependent spectral estimation algorithm provides an accurate and robust estimate of the voicing probability under a variety of background noise conditions. this is essential to providing high quality intelligible speech in the presence of background noise.",2006-08-15,"The title of the patent is parametric speech codec for representing synthetic speech in the presence of background noise and its abstract is a system and method are provided for processing audio and speech signals using a pitch and voicing dependent spectral estimation algorithm (voicing algorithm) to accurately represent voiced speech, unvoiced speech, and mixed speech in the presence of background noise, and background noise with a single model. the present invention also modifies the synthesis model based on an estimate of the current input signal to improve the perceptual quality of the speech and background noise under a variety of input conditions. the present invention also improves the voicing dependent spectral estimation algorithm robustness by introducing the use of a multi-layer neural network in the estimation process. the voicing dependent spectral estimation algorithm provides an accurate and robust estimate of the voicing probability under a variety of background noise conditions. this is essential to providing high quality intelligible speech in the presence of background noise. dated 2006-08-15"
7092923,synapse element with learning function and semiconductor integrated circuit device including the synapse element,"a synapse configured of an a-mos transistor has a learning function and can implement high integration similar to that of a dram because of its simplified circuit configuration and compact circuit size. with the presently cutting-edge technology (0.15 μm cmos), approximately 1g synapses can be integrated on one chip. accordingly, it is possible to implement a neural network with approximately 30,000 neurons all coupled together on one chip. this corresponds to a network scale capable of associatively storing approximately 5,000 patterns.",2006-08-15,"The title of the patent is synapse element with learning function and semiconductor integrated circuit device including the synapse element and its abstract is a synapse configured of an a-mos transistor has a learning function and can implement high integration similar to that of a dram because of its simplified circuit configuration and compact circuit size. with the presently cutting-edge technology (0.15 μm cmos), approximately 1g synapses can be integrated on one chip. accordingly, it is possible to implement a neural network with approximately 30,000 neurons all coupled together on one chip. this corresponds to a network scale capable of associatively storing approximately 5,000 patterns. dated 2006-08-15"
7095872,automated digital watermarking methods using neural networks,"embodiments of the present invention provide digital watermarking methods that embed a digital watermark in both the low and high frequencies of an image or other production, providing a digital watermark that is resistant to a variety of attacks. the digital watermarking methods of the present invention optimize the strength of the embedded digital watermark such that it is as powerful as possible without being perceptible to the human eye. the digital watermarking methods of the present invention do this relatively quickly, in real-time, and in an automated fashion using an intelligent system, such as a neural network. the digital watermarking methods of the present invention may also be used in a variety of new applications, such as the digital watermarking of sensitive aircraft parts and military equipment.",2006-08-22,"The title of the patent is automated digital watermarking methods using neural networks and its abstract is embodiments of the present invention provide digital watermarking methods that embed a digital watermark in both the low and high frequencies of an image or other production, providing a digital watermark that is resistant to a variety of attacks. the digital watermarking methods of the present invention optimize the strength of the embedded digital watermark such that it is as powerful as possible without being perceptible to the human eye. the digital watermarking methods of the present invention do this relatively quickly, in real-time, and in an automated fashion using an intelligent system, such as a neural network. the digital watermarking methods of the present invention may also be used in a variety of new applications, such as the digital watermarking of sensitive aircraft parts and military equipment. dated 2006-08-22"
7099852,detection of pump cavitation/blockage and seal failure via current signature analysis,"a system and method is provided for monitoring the operating condition of a pump by evaluating fault data encoded in the instantaneous current of the motor driving the pump. the data is converted to a frequency spectrum which is analyzed to create a fault signature having fault attributes relating to various fault conditions associated with the pump. the fault signature is then input to a neural network that operates in conjunction with a preprocessing and post processing module to perform decisions and output those decisions to a user interface. a stand alone module is also provided that includes an adaptive preprocessing module, a one-shot unsupervised neural network and a fuzzy based expert system to provide a decision making module that operates with limited human supervision.",2006-08-29,"The title of the patent is detection of pump cavitation/blockage and seal failure via current signature analysis and its abstract is a system and method is provided for monitoring the operating condition of a pump by evaluating fault data encoded in the instantaneous current of the motor driving the pump. the data is converted to a frequency spectrum which is analyzed to create a fault signature having fault attributes relating to various fault conditions associated with the pump. the fault signature is then input to a neural network that operates in conjunction with a preprocessing and post processing module to perform decisions and output those decisions to a user interface. a stand alone module is also provided that includes an adaptive preprocessing module, a one-shot unsupervised neural network and a fuzzy based expert system to provide a decision making module that operates with limited human supervision. dated 2006-08-29"
7099880,system and method of using data mining prediction methodology,"a system and associated method for data mining prediction is presented according to which the user selects a database table by means of a graphical user interface. some records in the table are complete, while other records are incomplete. a subset of records of the database table is determined wherein each record of the subset contains a data value in the column selected for prediction. this subset of records is used to generate a model by means of a data mining algorithm, such as linear regression, radial basis function, decision tree or neural network methods. the resulting model is then utilized to predict the empty data fields in the column. after completing the prediction, the predicted values are entered into the column for display to the user.",2006-08-29,"The title of the patent is system and method of using data mining prediction methodology and its abstract is a system and associated method for data mining prediction is presented according to which the user selects a database table by means of a graphical user interface. some records in the table are complete, while other records are incomplete. a subset of records of the database table is determined wherein each record of the subset contains a data value in the column selected for prediction. this subset of records is used to generate a model by means of a data mining algorithm, such as linear regression, radial basis function, decision tree or neural network methods. the resulting model is then utilized to predict the empty data fields in the column. after completing the prediction, the predicted values are entered into the column for display to the user. dated 2006-08-29"
7103452,method and system for targeting and monitoring the energy performance of manufacturing facilities,"an on-line neural network based software application that enables manufacturing facilities to meaningfully determine their energy performance, no matter how complex, with respect to the production rates and ambient conditions. this is achieved by the generation of three levels of targets; facility overall performance, departmental key performance indicators, and key operating parameters that impact the facility's energy consumption and over which, the operators have control. a unit and cost gap analysis of actual versus target is executed for overall and departmental energy performance. causes of statistically significant deviations are diagnosed and corrective actions highlighted. the software application is designed to be updated dynamically so that users can effectively manage performance on the basis of current information.",2006-09-05,"The title of the patent is method and system for targeting and monitoring the energy performance of manufacturing facilities and its abstract is an on-line neural network based software application that enables manufacturing facilities to meaningfully determine their energy performance, no matter how complex, with respect to the production rates and ambient conditions. this is achieved by the generation of three levels of targets; facility overall performance, departmental key performance indicators, and key operating parameters that impact the facility's energy consumption and over which, the operators have control. a unit and cost gap analysis of actual versus target is executed for overall and departmental energy performance. causes of statistically significant deviations are diagnosed and corrective actions highlighted. the software application is designed to be updated dynamically so that users can effectively manage performance on the basis of current information. dated 2006-09-05"
7103460,system and method for vehicle diagnostics,"method and system for diagnosing whether vehicular components are operating abnormally based on data obtained from sensors arranged on a vehicle. in a training stage, output from the sensors during normal operation of the components is obtained, each component is adjusted to induce abnormal operation thereof and output from the sensors is obtained during the induced abnormal operation. a determination is made as to which sensors provide data about abnormal operation of each component based on analysis of the output from the sensors during normal operation and during induced abnormal operation of the components. during operation of the vehicle, the output from the sensors is obtained and analyzed, e.g., by inputting it into a pattern recognition algorithm or neural network generated during the training stage, in order to output an indication of abnormal operation of any components being diagnosed.",2006-09-05,"The title of the patent is system and method for vehicle diagnostics and its abstract is method and system for diagnosing whether vehicular components are operating abnormally based on data obtained from sensors arranged on a vehicle. in a training stage, output from the sensors during normal operation of the components is obtained, each component is adjusted to induce abnormal operation thereof and output from the sensors is obtained during the induced abnormal operation. a determination is made as to which sensors provide data about abnormal operation of each component based on analysis of the output from the sensors during normal operation and during induced abnormal operation of the components. during operation of the vehicle, the output from the sensors is obtained and analyzed, e.g., by inputting it into a pattern recognition algorithm or neural network generated during the training stage, in order to output an indication of abnormal operation of any components being diagnosed. dated 2006-09-05"
7103801,method and apparatus for analyzing alarms coming from a communications network,"the invention enables alarms (ma) coming from a network (4), such as a communications network, to be analyzed by a step of transforming the detected alarms (ma) into a signal (v(t)) which expresses variation in time of a numerical value representative of all of the detected alarms. this transformation can be implemented using a learning system, in particular a neural network (6). the signal can then be analyzed using various techniques, in particular time/frequency analysis techniques, in order to make a diagnosis.",2006-09-05,"The title of the patent is method and apparatus for analyzing alarms coming from a communications network and its abstract is the invention enables alarms (ma) coming from a network (4), such as a communications network, to be analyzed by a step of transforming the detected alarms (ma) into a signal (v(t)) which expresses variation in time of a numerical value representative of all of the detected alarms. this transformation can be implemented using a learning system, in particular a neural network (6). the signal can then be analyzed using various techniques, in particular time/frequency analysis techniques, in order to make a diagnosis. dated 2006-09-05"
7107252,pattern recognition utilizing a nanotechnology-based neural network,"a pattern recognition system, comprising a neural network formed utilizing nanotechnology and a pattern input unit, which communicates with the neural network, wherein the neural network processes data input via the pattern input unit in order to recognize data patterns thereof. such a pattern recognition system can be implemented in the context of a speech recognition system and/or other pattern recognition systems, such as visual and/or imaging recognition systems.",2006-09-12,"The title of the patent is pattern recognition utilizing a nanotechnology-based neural network and its abstract is a pattern recognition system, comprising a neural network formed utilizing nanotechnology and a pattern input unit, which communicates with the neural network, wherein the neural network processes data input via the pattern input unit in order to recognize data patterns thereof. such a pattern recognition system can be implemented in the context of a speech recognition system and/or other pattern recognition systems, such as visual and/or imaging recognition systems. dated 2006-09-12"
7110526,neural network for controlling calls in a telephone switch,"a method and apparatus are provided for processing calls in an automatic call distributor. the method includes the steps of learning a set of desired resource relationships for servicing a plurality of call processing load conditions in the automatic call distributor and, afterwards, distributing resources of the automatic call distributor based upon call processor loading and the learned desired set of resource relationships.",2006-09-19,"The title of the patent is neural network for controlling calls in a telephone switch and its abstract is a method and apparatus are provided for processing calls in an automatic call distributor. the method includes the steps of learning a set of desired resource relationships for servicing a plurality of call processing load conditions in the automatic call distributor and, afterwards, distributing resources of the automatic call distributor based upon call processor loading and the learned desired set of resource relationships. dated 2006-09-19"
7110571,smart optical sensor for airbag systems,"a sensor having an array of photo sensitive elements for acquiring images of the passenger compartment in a motor vehicle and a circuit for processing the signals corresponding to the images generated by said photo sensitive elements. the processing circuit is configured according to a cellular neural network processing architecture of the image signals and can generate an output signal indicating the decision on whether to deploy an airbag to which the sensor is associated or to control the explosion of the airbag. preferably, the photo sensitive array and the processing circuit are comprised on a single integrated component, preferably implementing cmos technology.",2006-09-19,"The title of the patent is smart optical sensor for airbag systems and its abstract is a sensor having an array of photo sensitive elements for acquiring images of the passenger compartment in a motor vehicle and a circuit for processing the signals corresponding to the images generated by said photo sensitive elements. the processing circuit is configured according to a cellular neural network processing architecture of the image signals and can generate an output signal indicating the decision on whether to deploy an airbag to which the sensor is associated or to control the explosion of the airbag. preferably, the photo sensitive array and the processing circuit are comprised on a single integrated component, preferably implementing cmos technology. dated 2006-09-19"
7111469,process for refrigerant charge level detection using a neural net,"the invention is a process for determining the charge level of a vapor cycle environmental control system, having a condenser, evaporator, and an expansion valve, comprising the steps of providing a neural network having four input neurons, two hidden neurons and three output neurons; determining the number of degrees below the saturation temperature of the liquid refrigerant: exiting the condenser and providing this measurement to the first input neuron; sensing the condenser sink temperature and providing the measurement to the second input neuron; sensing either the refrigerant outlet temperature from the condenser or the evaporator exhaust air temperature and providing the measurement to the third input neuron, sensing the evaporator inlet temperature and providing the measurement to the fourth input neuron; and using the trained neural network to monitor the charge level in the system.",2006-09-26,"The title of the patent is process for refrigerant charge level detection using a neural net and its abstract is the invention is a process for determining the charge level of a vapor cycle environmental control system, having a condenser, evaporator, and an expansion valve, comprising the steps of providing a neural network having four input neurons, two hidden neurons and three output neurons; determining the number of degrees below the saturation temperature of the liquid refrigerant: exiting the condenser and providing this measurement to the first input neuron; sensing the condenser sink temperature and providing the measurement to the second input neuron; sensing either the refrigerant outlet temperature from the condenser or the evaporator exhaust air temperature and providing the measurement to the third input neuron, sensing the evaporator inlet temperature and providing the measurement to the fourth input neuron; and using the trained neural network to monitor the charge level in the system. dated 2006-09-26"
7113819,method and apparatus for monitoring the condition of a fetus,"a method and apparatus for monitoring the condition of a fetus to assess a degree of risk of developing a permanent neurological condition is provided. a signal indicative of a fetal heart is processed to derive a degree of risk of developing a permanent neurological condition. the data indicative of the degree of risk of developing a permanent neurological condition indicates a likelihood that the condition of the fetus belongs to a class in a group of classes, where each class in the group of classes is associated with a pre-defined fetal condition. data indicative of the degree of risk of developing a permanent neurological condition is then released. in one example, a neural network is used to obtain data indicating the likelihood that the condition of the fetus belongs to a class in the group of classes.",2006-09-26,"The title of the patent is method and apparatus for monitoring the condition of a fetus and its abstract is a method and apparatus for monitoring the condition of a fetus to assess a degree of risk of developing a permanent neurological condition is provided. a signal indicative of a fetal heart is processed to derive a degree of risk of developing a permanent neurological condition. the data indicative of the degree of risk of developing a permanent neurological condition indicates a likelihood that the condition of the fetus belongs to a class in a group of classes, where each class in the group of classes is associated with a pre-defined fetal condition. data indicative of the degree of risk of developing a permanent neurological condition is then released. in one example, a neural network is used to obtain data indicating the likelihood that the condition of the fetus belongs to a class in the group of classes. dated 2006-09-26"
7113931,method of fabricating a fractal structure for constructing complex neural networks,"a fracture structure is grown from a plurality of starting points. a fractal structure, grown from respective starting points and interconnected by interactive growths, forms a neural network. a growth speed originated at a specific starting point is determined by the probability of a material reaching a grown portion from a remote location by means of a diffusion process and the probability of a growth promoting factor reaching a grown portion by means of a diffusion process from a portion grown from a starting point other than the specific one. anisotropy is introduced into a space in which a fractal structure is to be grown, as required.",2006-09-26,"The title of the patent is method of fabricating a fractal structure for constructing complex neural networks and its abstract is a fracture structure is grown from a plurality of starting points. a fractal structure, grown from respective starting points and interconnected by interactive growths, forms a neural network. a growth speed originated at a specific starting point is determined by the probability of a material reaching a grown portion from a remote location by means of a diffusion process and the probability of a growth promoting factor reaching a grown portion by means of a diffusion process from a portion grown from a starting point other than the specific one. anisotropy is introduced into a space in which a fractal structure is to be grown, as required. dated 2006-09-26"
7113932,artificial intelligence trending system,"a data processing system program to develop, train, and implement a neural network for identifying customers who represent a bad debt risk is disclosed. a feature vector is applied to a neural network to generate outputs that approximate the relative likelihood that customers who are the subjects of the records used to generate the feature vector will be a bad debt risk. statistical values relating categorical attributes of the customers to the likelihood of their becoming a bad debt risk are substituted for the categorical attributes, and the attributes are normalized before the feature vector is applied to the network. in one embodiment the customers are customers of a long distance service provider.",2006-09-26,"The title of the patent is artificial intelligence trending system and its abstract is a data processing system program to develop, train, and implement a neural network for identifying customers who represent a bad debt risk is disclosed. a feature vector is applied to a neural network to generate outputs that approximate the relative likelihood that customers who are the subjects of the records used to generate the feature vector will be a bad debt risk. statistical values relating categorical attributes of the customers to the likelihood of their becoming a bad debt risk are substituted for the categorical attributes, and the attributes are normalized before the feature vector is applied to the network. in one embodiment the customers are customers of a long distance service provider. dated 2006-09-26"
7117045,combined proportional plus integral (pi) and neural network (nn) controller,"a neural network controller in parallel with a proportional-plus-integral (pi) feedback controller in a control system. at least one input port of the neural network for receiving an input signal representing a condition of a process is included. a first set of data is obtained that includes a plurality of output values of the neural network obtained during a training period thereof using a plurality of first inputs representing a plurality of conditions of the process. the process/plant condition signals generally define the process/plant, and may include one set-point as well as signals generated using measured systems variables/parameters. in operation, the neural network contributes to an output of the pi controller only upon detection of at least one triggering event, at which time a value of the first set of data corresponding with the condition deviation is added-in thus, contributing to the proportional-plus-integral feedback controller. the triggering event can be characterized as (a) a change in any one of the input signals greater-than a preselected amount, or (b) a detectable process condition deviation greater-than a preselected magnitude, for which an adjustment is needed to the process/plant being controlled. also a method for controlling a process with a neural network controller operating in parallel with a ip controller is included.",2006-10-03,"The title of the patent is combined proportional plus integral (pi) and neural network (nn) controller and its abstract is a neural network controller in parallel with a proportional-plus-integral (pi) feedback controller in a control system. at least one input port of the neural network for receiving an input signal representing a condition of a process is included. a first set of data is obtained that includes a plurality of output values of the neural network obtained during a training period thereof using a plurality of first inputs representing a plurality of conditions of the process. the process/plant condition signals generally define the process/plant, and may include one set-point as well as signals generated using measured systems variables/parameters. in operation, the neural network contributes to an output of the pi controller only upon detection of at least one triggering event, at which time a value of the first set of data corresponding with the condition deviation is added-in thus, contributing to the proportional-plus-integral feedback controller. the triggering event can be characterized as (a) a change in any one of the input signals greater-than a preselected amount, or (b) a detectable process condition deviation greater-than a preselected magnitude, for which an adjustment is needed to the process/plant being controlled. also a method for controlling a process with a neural network controller operating in parallel with a ip controller is included. dated 2006-10-03"
7117056,cnc control unit with learning ability for machining centers,"a computer numerical control unit with learning ability solves the problem of automatic and intelligent generating of numerical control programs for computer numerical control machining centers for milling, drilling and similar operations. the key module of the computer numerical control unit is a neural network (nn) device that learns to generate the numerical control programs through an neural network teaching module. upon completion of learning process the neural network device can generate automatically, without any intervention of the operator, merely on the basis of the cad 2d, 2,5d or 3d part models, taken from a conventional cad/cam system, various new numerical control programs for different parts, which have not been in the machining process before. the computer numerical control control unit with learning ability is suitable especially for machining centers intended for milling, including face milling (rough), contour milling (rough), final milling following the contour and in z-plane, final contour 3d milling, contour final milling, milling in z-plane, final contour milling on the equidistant, and milling of pockets; drilling, including normal drilling, deep drilling, and center drilling; and reaming, sinking and threading.",2006-10-03,"The title of the patent is cnc control unit with learning ability for machining centers and its abstract is a computer numerical control unit with learning ability solves the problem of automatic and intelligent generating of numerical control programs for computer numerical control machining centers for milling, drilling and similar operations. the key module of the computer numerical control unit is a neural network (nn) device that learns to generate the numerical control programs through an neural network teaching module. upon completion of learning process the neural network device can generate automatically, without any intervention of the operator, merely on the basis of the cad 2d, 2,5d or 3d part models, taken from a conventional cad/cam system, various new numerical control programs for different parts, which have not been in the machining process before. the computer numerical control control unit with learning ability is suitable especially for machining centers intended for milling, including face milling (rough), contour milling (rough), final milling following the contour and in z-plane, final contour 3d milling, contour final milling, milling in z-plane, final contour milling on the equidistant, and milling of pockets; drilling, including normal drilling, deep drilling, and center drilling; and reaming, sinking and threading. dated 2006-10-03"
7117187,"method, system and computer program product for non-linear mapping of multi-dimensional data","a method, system and computer program product are provided for scaling, or dimensionally reducing, multi-dimensional data sets that scale well for large data sets. the invention scales multi-dimensional data sets by determining one or more non-linear functions between a sample of points from the multi-dimensional data set and a corresponding set of dimensionally reduced points. thereafter, these one or more non-linear functions are used to non-linearly map additional points. the additional points may be members of the original multi-dimensional data set or may be new, previously unseen points. in an embodiment, the determination of the non-linear relationship between the sample of points from the multi-dimensional data set and the corresponding set of dimensionally reduced points is performed by a self-learning system such as a neural network. the additional points are mapped using the self-learning system in a feed-forward/predictive manner.",2006-10-03,"The title of the patent is method, system and computer program product for non-linear mapping of multi-dimensional data and its abstract is a method, system and computer program product are provided for scaling, or dimensionally reducing, multi-dimensional data sets that scale well for large data sets. the invention scales multi-dimensional data sets by determining one or more non-linear functions between a sample of points from the multi-dimensional data set and a corresponding set of dimensionally reduced points. thereafter, these one or more non-linear functions are used to non-linearly map additional points. the additional points may be members of the original multi-dimensional data set or may be new, previously unseen points. in an embodiment, the determination of the non-linear relationship between the sample of points from the multi-dimensional data set and the corresponding set of dimensionally reduced points is performed by a self-learning system such as a neural network. the additional points are mapped using the self-learning system in a feed-forward/predictive manner. dated 2006-10-03"
7120291,method and apparatus for analyzing input information,"a method and an apparatus operate like a human neural network to analyze and store input information and form patterns according to the input and stored information. the apparatus has a preprocessing unit (3), an activity computation unit (5), a mutual repression unit (6), and a composition unit (7). the apparatus receives an input pattern, calculates the similarity and activity levels of each stored pattern with respect to the input pattern, and repeats a predetermined number of times the activity calculation of each stored pattern according to the calculated activity level (a(i)), a negative repression coefficient, and the activity levels of the other stored patterns. the apparatus applies final activity levels to cell values of the stored patterns, totals the cell values through the stored patterns, and generates a new pattern according to the totaled cell values.",2006-10-10,"The title of the patent is method and apparatus for analyzing input information and its abstract is a method and an apparatus operate like a human neural network to analyze and store input information and form patterns according to the input and stored information. the apparatus has a preprocessing unit (3), an activity computation unit (5), a mutual repression unit (6), and a composition unit (7). the apparatus receives an input pattern, calculates the similarity and activity levels of each stored pattern with respect to the input pattern, and repeats a predetermined number of times the activity calculation of each stored pattern according to the calculated activity level (a(i)), a negative repression coefficient, and the activity levels of the other stored patterns. the apparatus applies final activity levels to cell values of the stored patterns, totals the cell values through the stored patterns, and generates a new pattern according to the totaled cell values. dated 2006-10-10"
7120615,neural network system and method for controlling information output based on user feedback,"a system and method for controlling information output based on user feedback about the information is provided that comprises a plurality of information sources providing information. the information sources may be electronic mail providers, chat participants, or page links. at least one neural network module selects one or more of a plurality of objects to receive information from the plurality of information sources based at least in part on a plurality of inputs and a plurality of weight values during that epoch. at least one server, associated with the neural network module, provides one or more of the objects to a plurality of recipients. the objects may comprise electronic mail messages, chat participants viewers, or slots within a link directory page. the recipients provide feedback about the information during an epoch. at the conclusion of an epoch, the neural network takes all of the feedback that has been provided from the recipients and generates a rating value for each of the plurality of objects. based on the rating value and the selections made, the neural network redetermines the weight values within the network. the neural network then selects the objects to receive information during a subsequent epoch using the redetermined weight values and the inputs for that subsequent epoch.",2006-10-10,"The title of the patent is neural network system and method for controlling information output based on user feedback and its abstract is a system and method for controlling information output based on user feedback about the information is provided that comprises a plurality of information sources providing information. the information sources may be electronic mail providers, chat participants, or page links. at least one neural network module selects one or more of a plurality of objects to receive information from the plurality of information sources based at least in part on a plurality of inputs and a plurality of weight values during that epoch. at least one server, associated with the neural network module, provides one or more of the objects to a plurality of recipients. the objects may comprise electronic mail messages, chat participants viewers, or slots within a link directory page. the recipients provide feedback about the information during an epoch. at the conclusion of an epoch, the neural network takes all of the feedback that has been provided from the recipients and generates a rating value for each of the plurality of objects. based on the rating value and the selections made, the neural network redetermines the weight values within the network. the neural network then selects the objects to receive information during a subsequent epoch using the redetermined weight values and the inputs for that subsequent epoch. dated 2006-10-10"
7127087,pose-invariant face recognition system and process,"a face recognition system and process for identifying a person depicted in an input image and their face pose. this system and process entails locating and extracting face regions belonging to known people from a set of model images, and determining the face pose for each of the face regions extracted. all the extracted face regions are preprocessed by normalizing, cropping, categorizing and finally abstracting them. more specifically, the images are normalized and cropped to show only a persons face, categorized according to the face pose of the depicted person's face by assigning them to one of a series of face pose ranges, and abstracted preferably via an eigenface approach. the preprocessed face images are preferably used to train a neural network ensemble having a first stage made up of a bank of face recognition neural networks each of which is dedicated to a particular pose range, and a second stage constituting a single fusing neural network that is used to combine the outputs from each of the first stage neural networks. once trained, the input of a face region which has been extracted from an input image and preprocessed (i.e., normalized, cropped and abstracted) will cause just one of the output units of the fusing portion of the neural network ensemble to become active. the active output unit indicates either the identify of the person whose face was extracted from the input image and the associated face pose, or that the identity of the person is unknown to the system.",2006-10-24,"The title of the patent is pose-invariant face recognition system and process and its abstract is a face recognition system and process for identifying a person depicted in an input image and their face pose. this system and process entails locating and extracting face regions belonging to known people from a set of model images, and determining the face pose for each of the face regions extracted. all the extracted face regions are preprocessed by normalizing, cropping, categorizing and finally abstracting them. more specifically, the images are normalized and cropped to show only a persons face, categorized according to the face pose of the depicted person's face by assigning them to one of a series of face pose ranges, and abstracted preferably via an eigenface approach. the preprocessed face images are preferably used to train a neural network ensemble having a first stage made up of a bank of face recognition neural networks each of which is dedicated to a particular pose range, and a second stage constituting a single fusing neural network that is used to combine the outputs from each of the first stage neural networks. once trained, the input of a face region which has been extracted from an input image and preprocessed (i.e., normalized, cropped and abstracted) will cause just one of the output units of the fusing portion of the neural network ensemble to become active. the active output unit indicates either the identify of the person whose face was extracted from the input image and the associated face pose, or that the identity of the person is unknown to the system. dated 2006-10-24"
7127435,distribution theory based enrichment of sparse data for machine learning,a technique for enriching sparse data for machine learning techniques such as supervised artificial neural network includes receiving the sparse data and enriching the received data around a deviation of the mean of the received data using a predetermined distribution. the technique further includes outputting the enriched data for unbiased and increased performance during the machine learning.,2006-10-24,The title of the patent is distribution theory based enrichment of sparse data for machine learning and its abstract is a technique for enriching sparse data for machine learning techniques such as supervised artificial neural network includes receiving the sparse data and enriching the received data around a deviation of the mean of the received data using a predetermined distribution. the technique further includes outputting the enriched data for unbiased and increased performance during the machine learning. dated 2006-10-24
7130834,identification system and method for determining the geographic origin of a fresh commodity,"the detection method includes generating a plurality of neural network models. each model has as a training set a data set from a plurality of samples of a commodity of known origins. each sample has been analyzed for a plurality of elemental concentrations. each neural network model is presented for classification a test data set from a plurality of samples of a commodity of unknown origins. as with the training set, the samples have been analyzed for the same plurality of elemental concentrations. next a bootstrap aggregating strategy is employed to combine the results of the classifications for each sample in the test data set made by each neural network model. finally, a determination is made from the bootstrap aggregating strategy as to a final classification of each sample in the test data set. this final classification is indicative of the geographical origin of the commodity. the system includes software for generating the neural network models and a software routine for performing the bootstrap aggregating strategy.",2006-10-31,"The title of the patent is identification system and method for determining the geographic origin of a fresh commodity and its abstract is the detection method includes generating a plurality of neural network models. each model has as a training set a data set from a plurality of samples of a commodity of known origins. each sample has been analyzed for a plurality of elemental concentrations. each neural network model is presented for classification a test data set from a plurality of samples of a commodity of unknown origins. as with the training set, the samples have been analyzed for the same plurality of elemental concentrations. next a bootstrap aggregating strategy is employed to combine the results of the classifications for each sample in the test data set made by each neural network model. finally, a determination is made from the bootstrap aggregating strategy as to a final classification of each sample in the test data set. this final classification is indicative of the geographical origin of the commodity. the system includes software for generating the neural network models and a software routine for performing the bootstrap aggregating strategy. dated 2006-10-31"
7130850,rating and controlling access to emails,"computer-implemented methods are described for, first, characterizing a specific category of information content—pornography, for example—and then accurately identifying instances of that category of content within a real-time media stream, such as a web page, e-mail or other digital dataset. this content-recognition technology enables a new class of highly scalable applications to manage such content, including filtering, classifying, prioritizing, tracking, etc. an illustrative application of the invention is a software product for use in conjunction with web-browser client software for screening access to web pages that contain pornography or other potentially harmful or offensive content. a target attribute set of regular expression, such as natural language words and/or phrases, is formed by statistical analysis of a number of samples of datasets characterized as “containing,” and another set of samples characterized as “not containing,” the selected category of information content. this list of expressions is refined by applying correlation analysis to the samples or “training data.” neural-network feed-forward techniques are then applied, again using a substantial training dataset, for adaptively assigning relative weights to each of the expressions in the target attribute set, thereby forming an awaited list that is highly predictive of the information content category of interest.",2006-10-31,"The title of the patent is rating and controlling access to emails and its abstract is computer-implemented methods are described for, first, characterizing a specific category of information content—pornography, for example—and then accurately identifying instances of that category of content within a real-time media stream, such as a web page, e-mail or other digital dataset. this content-recognition technology enables a new class of highly scalable applications to manage such content, including filtering, classifying, prioritizing, tracking, etc. an illustrative application of the invention is a software product for use in conjunction with web-browser client software for screening access to web pages that contain pornography or other potentially harmful or offensive content. a target attribute set of regular expression, such as natural language words and/or phrases, is formed by statistical analysis of a number of samples of datasets characterized as “containing,” and another set of samples characterized as “not containing,” the selected category of information content. this list of expressions is refined by applying correlation analysis to the samples or “training data.” neural-network feed-forward techniques are then applied, again using a substantial training dataset, for adaptively assigning relative weights to each of the expressions in the target attribute set, thereby forming an awaited list that is highly predictive of the information content category of interest. dated 2006-10-31"
7132617,method and system for assessing quality of spot welds,"a system and method for assessing the quality of spot weld joints between pieces of metal includes an ultrasound transducer probing a spot weld joint. the ultrasound transducer transmits ultrasonic radiation into the spot weld joint, receives corresponding echoes, and transforms the echoes into electrical signals. an image reconstructor connected to the ultrasound transducer transforms the electrical signals into numerical data representing an ultrasound image. a neural network connected to the image reconstructor analyzes the numerical data and an output system presents information representing the quality of the spot weld joint. the system is trained to assess the quality of spot weld joints by scanning a spot weld joint with an ultrasound transducer to produce the data set representing the joint; then physically deconstructing the joint to assess the joint quality.",2006-11-07,"The title of the patent is method and system for assessing quality of spot welds and its abstract is a system and method for assessing the quality of spot weld joints between pieces of metal includes an ultrasound transducer probing a spot weld joint. the ultrasound transducer transmits ultrasonic radiation into the spot weld joint, receives corresponding echoes, and transforms the echoes into electrical signals. an image reconstructor connected to the ultrasound transducer transforms the electrical signals into numerical data representing an ultrasound image. a neural network connected to the image reconstructor analyzes the numerical data and an output system presents information representing the quality of the spot weld joint. the system is trained to assess the quality of spot weld joints by scanning a spot weld joint with an ultrasound transducer to produce the data set representing the joint; then physically deconstructing the joint to assess the joint quality. dated 2006-11-07"
7134291,process for refrigerant charge level detection using a neural net having one output neuron,"the invention is a process for determining the charge level of a vapor cycle environmental control system, having a condenser, evaporator, and an expansion valve, comprising the steps of providing a neural network having four input neurons, two hidden neurons and one output neurons; determining the number of degrees below the saturation temperature of the liquid refrigerant exiting the condenser and providing this measurement to the first input neuron; sensing the condenser sink temperature and providing the measurement to the second input neuron; sensing either the refrigerant outlet temperature from the condenser or the evaporator exhaust air temperature and providing the measurement to the third input neuron, sensing the evaporator inlet temperature and providing the measurement to the fourth input neuron; and using the trained neural network to monitor the charge level in the system.",2006-11-14,"The title of the patent is process for refrigerant charge level detection using a neural net having one output neuron and its abstract is the invention is a process for determining the charge level of a vapor cycle environmental control system, having a condenser, evaporator, and an expansion valve, comprising the steps of providing a neural network having four input neurons, two hidden neurons and one output neurons; determining the number of degrees below the saturation temperature of the liquid refrigerant exiting the condenser and providing this measurement to the first input neuron; sensing the condenser sink temperature and providing the measurement to the second input neuron; sensing either the refrigerant outlet temperature from the condenser or the evaporator exhaust air temperature and providing the measurement to the third input neuron, sensing the evaporator inlet temperature and providing the measurement to the fourth input neuron; and using the trained neural network to monitor the charge level in the system. dated 2006-11-14"
7136759,method for enhanced accuracy in predicting peptides using liquid separations or chromatography,"a method for predicting the elution time of a peptide in chromatographic and electrophoretic separations by first providing a data set of known elution times of known peptides, then creating a plurality of vectors, each vector having a plurality of dimensions, and each dimension representing the elution time of amino acids present in each of these known peptides from the data set. the elution time of any protein is then be predicted by first creating a vector by assigning dimensional values for the elution time of amino acids of at least one hypothetical peptide and then calculating a predicted elution time for the vector by performing a multivariate regression of the dimensional values of the hypothetical peptide using the dimensional values of the known peptides. preferably, the multivariate regression is accomplished by the use of an artificial neural network and the elution times are first normalized using a transfer function.",2006-11-14,"The title of the patent is method for enhanced accuracy in predicting peptides using liquid separations or chromatography and its abstract is a method for predicting the elution time of a peptide in chromatographic and electrophoretic separations by first providing a data set of known elution times of known peptides, then creating a plurality of vectors, each vector having a plurality of dimensions, and each dimension representing the elution time of amino acids present in each of these known peptides from the data set. the elution time of any protein is then be predicted by first creating a vector by assigning dimensional values for the elution time of amino acids of at least one hypothetical peptide and then calculating a predicted elution time for the vector by performing a multivariate regression of the dimensional values of the hypothetical peptide using the dimensional values of the known peptides. preferably, the multivariate regression is accomplished by the use of an artificial neural network and the elution times are first normalized using a transfer function. dated 2006-11-14"
7136802,method and apparatus for detecting prosodic phrase break in a text to speech (tts) system,"methods for processing speech data are described herein. in one aspect of the invention, an exemplary method includes receiving a text sentence comprising a plurality of words, each of the plurality of words having a part of speech (pos) tag, generating a pos sequence based on the pos tag of each of the plurality of words, detecting a prosodic phrase break through a recurrent neural network (rnn), based on the pos sequence, and generating a prosodic phrases boundary based on the prosodic phrase break. other methods and apparatuses are also described.",2006-11-14,"The title of the patent is method and apparatus for detecting prosodic phrase break in a text to speech (tts) system and its abstract is methods for processing speech data are described herein. in one aspect of the invention, an exemplary method includes receiving a text sentence comprising a plurality of words, each of the plurality of words having a part of speech (pos) tag, generating a pos sequence based on the pos tag of each of the plurality of words, detecting a prosodic phrase break through a recurrent neural network (rnn), based on the pos sequence, and generating a prosodic phrases boundary based on the prosodic phrase break. other methods and apparatuses are also described. dated 2006-11-14"
7142387,method and apparatus positioning a read head to follow a track in a hard disk drive,"read positioning method includes adjusting at least one read head position when accessing read track on at least one rotating disk surface based upon burst correction value of nearest write track, when pes burst patterns of read track and nearest write track match. apparatus supporting read positioning method may include means for at least partly performing each step. at least one means may include at least one instance of at least one of following: computer, finite state machine, neural network and inferential engine. at least one read method for the hard disk drive included. these read methods may be used during initialization and/or normal hard disk drive operation. hard disk drive may include servo controller driving voice coil actuator and, preferably further driving micro-actuator. the hard disk drive may include more than one rotating disk surface and more than one rotating disk surface.",2006-11-28,"The title of the patent is method and apparatus positioning a read head to follow a track in a hard disk drive and its abstract is read positioning method includes adjusting at least one read head position when accessing read track on at least one rotating disk surface based upon burst correction value of nearest write track, when pes burst patterns of read track and nearest write track match. apparatus supporting read positioning method may include means for at least partly performing each step. at least one means may include at least one instance of at least one of following: computer, finite state machine, neural network and inferential engine. at least one read method for the hard disk drive included. these read methods may be used during initialization and/or normal hard disk drive operation. hard disk drive may include servo controller driving voice coil actuator and, preferably further driving micro-actuator. the hard disk drive may include more than one rotating disk surface and more than one rotating disk surface. dated 2006-11-28"
7142697,pose-invariant face recognition system and process,"a face recognition system and process for identifying a person depicted in an input image and their face pose. this system and process entails locating and extracting face regions belonging to known people from a set of model images, and determining the face pose for each of the face regions extracted. all the extracted face regions are preprocessed by normalizing, cropping, categorizing and finally abstracting them. more specifically, the images are normalized and cropped to show only a persons face, categorized according to the face pose of the depicted person's face by assigning them to one of a series of face pose ranges, and abstracted preferably via an eigenface approach. the preprocessed face images are preferably used to train a neural network ensemble having a first stage made up of a bank of face recognition neural networks each of which is dedicated to a particular pose range, and a second stage constituting a single fusing neural network that is used to combine the outputs from each of the first stage neural networks. once trained, the input of a face region which has been extracted from an input image and preprocessed (i.e., normalized, cropped and abstracted) will cause just one of the output units of the fusing portion of the neural network ensemble to become active. the active output unit indicates either the identify of the person whose face was extracted from the input image and the associated face pose, or that the identity of the person is unknown to the system.",2006-11-28,"The title of the patent is pose-invariant face recognition system and process and its abstract is a face recognition system and process for identifying a person depicted in an input image and their face pose. this system and process entails locating and extracting face regions belonging to known people from a set of model images, and determining the face pose for each of the face regions extracted. all the extracted face regions are preprocessed by normalizing, cropping, categorizing and finally abstracting them. more specifically, the images are normalized and cropped to show only a persons face, categorized according to the face pose of the depicted person's face by assigning them to one of a series of face pose ranges, and abstracted preferably via an eigenface approach. the preprocessed face images are preferably used to train a neural network ensemble having a first stage made up of a bank of face recognition neural networks each of which is dedicated to a particular pose range, and a second stage constituting a single fusing neural network that is used to combine the outputs from each of the first stage neural networks. once trained, the input of a face region which has been extracted from an input image and preprocessed (i.e., normalized, cropped and abstracted) will cause just one of the output units of the fusing portion of the neural network ensemble to become active. the active output unit indicates either the identify of the person whose face was extracted from the input image and the associated face pose, or that the identity of the person is unknown to the system. dated 2006-11-28"
7143071,method for changing cpu frequency under control of neural network,"a method for changing the cpu frequency under control of a neural network. the neural network has m basis functions and n basis points that are connected together. using the learning capability of the neural network to deduce basis weights based on dummy environmental parameters and a dummy output vector. in an application procedure, environmental parameters are input to the basis points and basis vectors are calculated based on the basis functions. integrating the multiplication of each basis vector and its corresponding basis weight, an output vector can be generated to determine a control signal so that the cpu can be controlled to raise or lower its operating frequency. in addition, if the user has to change the parameters due to behavior, a fast learning function of a radial neural network can be used for complying with each user's behavior.",2006-11-28,"The title of the patent is method for changing cpu frequency under control of neural network and its abstract is a method for changing the cpu frequency under control of a neural network. the neural network has m basis functions and n basis points that are connected together. using the learning capability of the neural network to deduce basis weights based on dummy environmental parameters and a dummy output vector. in an application procedure, environmental parameters are input to the basis points and basis vectors are calculated based on the basis functions. integrating the multiplication of each basis vector and its corresponding basis weight, an output vector can be generated to determine a control signal so that the cpu can be controlled to raise or lower its operating frequency. in addition, if the user has to change the parameters due to behavior, a fast learning function of a radial neural network can be used for complying with each user's behavior. dated 2006-11-28"
7143072,method and a system for calculating the values of the neurons of a neural network,a neural network having layers of neurons divided into sublayers of neurons. the values of target neurons in one layer are calculated from sublayers of source neurons in a second underlying layer. it is therefore always possible to use for this calculation the same group of weights to be multiplied by respective source neurons related thereto and situated in the underlying layer of the neural network.,2006-11-28,The title of the patent is method and a system for calculating the values of the neurons of a neural network and its abstract is a neural network having layers of neurons divided into sublayers of neurons. the values of target neurons in one layer are calculated from sublayers of source neurons in a second underlying layer. it is therefore always possible to use for this calculation the same group of weights to be multiplied by respective source neurons related thereto and situated in the underlying layer of the neural network. dated 2006-11-28
7146037,vlsi neural fuzzy classifier for handwriting recognition,a handwriting recognition device using fuzzy logic and cellular neural network for unconstrained handwritten numeral classification is provided. the current mode vlsi classifier has a i/o circuit for inputting and outputting a plurality of membership functions. an extraction unit comprising a ccd extractor with a cnn structure and a compression unit receives a to-be-recognized character having a plurality of input features for generating a plurality of features values after compression. a membership function generator stores the plurality of membership functions and receives the plurality of features values to generate a plurality of current-type membership degrees. a plurality of switched-current integrators receives the plurality of current-type membership degrees for generating a plurality of synthesis membership degrees. a k-wta circuit is provided for comparing the plurality of synthesis membership degrees and output the plurality of synthesis membership degrees as well as the corresponding characters in an order of magnitude.,2006-12-05,The title of the patent is vlsi neural fuzzy classifier for handwriting recognition and its abstract is a handwriting recognition device using fuzzy logic and cellular neural network for unconstrained handwritten numeral classification is provided. the current mode vlsi classifier has a i/o circuit for inputting and outputting a plurality of membership functions. an extraction unit comprising a ccd extractor with a cnn structure and a compression unit receives a to-be-recognized character having a plurality of input features for generating a plurality of features values after compression. a membership function generator stores the plurality of membership functions and receives the plurality of features values to generate a plurality of current-type membership degrees. a plurality of switched-current integrators receives the plurality of current-type membership degrees for generating a plurality of synthesis membership degrees. a k-wta circuit is provided for comparing the plurality of synthesis membership degrees and output the plurality of synthesis membership degrees as well as the corresponding characters in an order of magnitude. dated 2006-12-05
7152051,intelligent control with hierarchical stacked neural networks,"an intelligent control system based on an explicit model of cognitive development (table 1) performs high-level functions. it comprises up to o hierarchically stacked neural networks, nm, . . . , nm+(o−1), where m denotes the stage/order tasks performed in the first neural network, nm, and o denotes the highest stage/order tasks performed in the highest-level neural network. the type of processing actions performed in a network, nm, corresponds to the complexity for stage/order m. thus n1 performs tasks at the level corresponding to stage/order 1. n5 processes information at the level corresponding to stage/order 5. stacked neural networks begin and end at any stage/order, but information must be processed by each stage in ascending order sequence. stages/orders cannot be skipped. each neural network in a stack may use different architectures, interconnections, algorithms, and training methods, depending on the stage/order of the neural network and the type of intelligent control system implemented.",2006-12-19,"The title of the patent is intelligent control with hierarchical stacked neural networks and its abstract is an intelligent control system based on an explicit model of cognitive development (table 1) performs high-level functions. it comprises up to o hierarchically stacked neural networks, nm, . . . , nm+(o−1), where m denotes the stage/order tasks performed in the first neural network, nm, and o denotes the highest stage/order tasks performed in the highest-level neural network. the type of processing actions performed in a network, nm, corresponds to the complexity for stage/order m. thus n1 performs tasks at the level corresponding to stage/order 1. n5 processes information at the level corresponding to stage/order 5. stacked neural networks begin and end at any stage/order, but information must be processed by each stage in ascending order sequence. stages/orders cannot be skipped. each neural network in a stack may use different architectures, interconnections, algorithms, and training methods, depending on the stage/order of the neural network and the type of intelligent control system implemented. dated 2006-12-19"
7155401,automatic sales promotion selection system and method,"an automated sales promotion selection system uses neural networks to identify promising sales promotions based on recent customer purchases. the system includes a customer information device that receives customer data relating to customer purchases of items from an inventory of items, a central processing unit having a sales promotion neural network and a storage unit containing a plurality of item identifiers comprising potential customer purchases of additional items from the inventory, wherein the sales opportunity neural network responds to customer data received from the customer information device by determining if one or more of the item identifiers in the storage unit corresponds to an item likely to be purchased by one of the customers, and an output device that receives the item identifiers of the likely purchases determined by the sales promotion neural network and produces a sales promotion relating to at least one of the item identifiers.",2006-12-26,"The title of the patent is automatic sales promotion selection system and method and its abstract is an automated sales promotion selection system uses neural networks to identify promising sales promotions based on recent customer purchases. the system includes a customer information device that receives customer data relating to customer purchases of items from an inventory of items, a central processing unit having a sales promotion neural network and a storage unit containing a plurality of item identifiers comprising potential customer purchases of additional items from the inventory, wherein the sales opportunity neural network responds to customer data received from the customer information device by determining if one or more of the item identifiers in the storage unit corresponds to an item likely to be purchased by one of the customers, and an output device that receives the item identifiers of the likely purchases determined by the sales promotion neural network and produces a sales promotion relating to at least one of the item identifiers. dated 2006-12-26"
7162461,hybrid neural network generation system and method,"a computer-implemented method and system for building a neural network is disclosed. the neural network predicts at least one target based upon predictor variables defined in a state space. first, an input data set is retrieved that includes the predictor variables and at least one target associated with the predictor variables for each observation. in the state space, a number of points is inserted in the state space based upon the values of the predictor variables. the number of points is less than the number of observations. a statistical measure is determined that describes a relationship between the observations and the inserted points. weights and activation functions of the neural network are determined using the statistical measure.",2007-01-09,"The title of the patent is hybrid neural network generation system and method and its abstract is a computer-implemented method and system for building a neural network is disclosed. the neural network predicts at least one target based upon predictor variables defined in a state space. first, an input data set is retrieved that includes the predictor variables and at least one target associated with the predictor variables for each observation. in the state space, a number of points is inserted in the state space based upon the values of the predictor variables. the number of points is less than the number of observations. a statistical measure is determined that describes a relationship between the observations and the inserted points. weights and activation functions of the neural network are determined using the statistical measure. dated 2007-01-09"
7164468,"lidar system controlled by computer for smoke identification applied, in particular, to early stage forest fire detection","this invention relates to a method and an active system for detection and localization of early stage forest fires using lidar. in the simplest configuration the system includes a lidar and a control computer that operates the beam-scanning system and performs automatic recognition of the smoke signature in the lidar signal on the basis of a neural-network algorithm. the scanning procedure is optimized for the given topography and other characteristics of the area under surveillance. the neural network is simulated or implemented as a co-processor. to cover wider areas, several lidar stations may be linked together in a network, which allows simultaneous scanning of the suspicious areas by several neighboring lidars in order to guarantee maximum efficiency and false alarm reduction. the system allows the detection and localization of fires earlier and farther than passive systems, whose sensitivity is lower.",2007-01-16,"The title of the patent is lidar system controlled by computer for smoke identification applied, in particular, to early stage forest fire detection and its abstract is this invention relates to a method and an active system for detection and localization of early stage forest fires using lidar. in the simplest configuration the system includes a lidar and a control computer that operates the beam-scanning system and performs automatic recognition of the smoke signature in the lidar signal on the basis of a neural-network algorithm. the scanning procedure is optimized for the given topography and other characteristics of the area under surveillance. the neural network is simulated or implemented as a co-processor. to cover wider areas, several lidar stations may be linked together in a network, which allows simultaneous scanning of the suspicious areas by several neighboring lidars in order to guarantee maximum efficiency and false alarm reduction. the system allows the detection and localization of fires earlier and farther than passive systems, whose sensitivity is lower. dated 2007-01-16"
7164771,process and system for objective audio quality measurement,"a process and system for providing objective quality measurement of a target audio signal. reference and target signals are processed by a peripheral ear processor, and compared to provide a basilar degradation signal. a cognitive processor employing a neural network then determines an objective quality measure from the basilar degradation signal by calculating certain key cognitive model components.",2007-01-16,"The title of the patent is process and system for objective audio quality measurement and its abstract is a process and system for providing objective quality measurement of a target audio signal. reference and target signals are processed by a peripheral ear processor, and compared to provide a basilar degradation signal. a cognitive processor employing a neural network then determines an objective quality measure from the basilar degradation signal by calculating certain key cognitive model components. dated 2007-01-16"
7164794,unconstrained handwriting recognition,"methods and systems of the present invention may be used to recognize digital image data arranged in rows and columns. exemplary embodiments may include a feature extractor for extracting feature information from data representing the rows and columns of the digital image data, a feature compressor for compressing the extracted feature information, and a neural network for classifying the digital image data from the compressed, extracted feature information.",2007-01-16,"The title of the patent is unconstrained handwriting recognition and its abstract is methods and systems of the present invention may be used to recognize digital image data arranged in rows and columns. exemplary embodiments may include a feature extractor for extracting feature information from data representing the rows and columns of the digital image data, a feature compressor for compressing the extracted feature information, and a neural network for classifying the digital image data from the compressed, extracted feature information. dated 2007-01-16"
7165655,neural network detection of obstructions within and motion toward elevator doors,"a camera (26), with suitable illumination (such as ir) provides images to a processing card (33) which converts the images to numerical vectors and applies them to a neural network (35) which is capable of providing a door-open signal (38) in response to something, either moving or still, in the paths of the doors (29), or anything moving in a manner to indicate intent to enter the elevator in the landing adjacent to the elevator (27). the door open signal is provided to the elevator door controller (39) to cause the doors to become or remain open in response to anything moving toward the elevator or anything disposed in the door pathway.",2007-01-23,"The title of the patent is neural network detection of obstructions within and motion toward elevator doors and its abstract is a camera (26), with suitable illumination (such as ir) provides images to a processing card (33) which converts the images to numerical vectors and applies them to a neural network (35) which is capable of providing a door-open signal (38) in response to something, either moving or still, in the paths of the doors (29), or anything moving in a manner to indicate intent to enter the elevator in the landing adjacent to the elevator (27). the door open signal is provided to the elevator door controller (39) to cause the doors to become or remain open in response to anything moving toward the elevator or anything disposed in the door pathway. dated 2007-01-23"
7167123,object detection method and apparatus,"method and apparatus for detecting objects. in one embodiment, a person entering a secured zone is illuminated with low-power polarized radio waves. differently polarized waves which are reflected back from the person are collected. concealed weapons are detected by measuring various parameters of the reflected signals and then calculating various selected differences between them. these differences create patterns when plotted as a function of time. preferably a trained neural network pattern recognition program is then used to evaluate these patterns and autonomously render a decision on the presence of a weapon.",2007-01-23,"The title of the patent is object detection method and apparatus and its abstract is method and apparatus for detecting objects. in one embodiment, a person entering a secured zone is illuminated with low-power polarized radio waves. differently polarized waves which are reflected back from the person are collected. concealed weapons are detected by measuring various parameters of the reflected signals and then calculating various selected differences between them. these differences create patterns when plotted as a function of time. preferably a trained neural network pattern recognition program is then used to evaluate these patterns and autonomously render a decision on the presence of a weapon. dated 2007-01-23"
7169197,pyrolysis processing for solid waste resource recovery,"solid waste resource recovery in space is effected by pyrolysis processing, to produce light gases as the main products (ch4, h2, co2, co, h2o, nh3) and a reactive carbon-rich char as the main byproduct. significant amounts of liquid products are formed under less severe pyrolysis conditions, and are cracked almost completely to gases as the temperature is raised. a primary pyrolysis model for the composite mixture is based on an existing model for whole biomass materials, and an artificial neural network models the changes in gas composition with the severity of pyrolysis conditions.",2007-01-30,"The title of the patent is pyrolysis processing for solid waste resource recovery and its abstract is solid waste resource recovery in space is effected by pyrolysis processing, to produce light gases as the main products (ch4, h2, co2, co, h2o, nh3) and a reactive carbon-rich char as the main byproduct. significant amounts of liquid products are formed under less severe pyrolysis conditions, and are cracked almost completely to gases as the temperature is raised. a primary pyrolysis model for the composite mixture is based on an existing model for whole biomass materials, and an artificial neural network models the changes in gas composition with the severity of pyrolysis conditions. dated 2007-01-30"
7170418,probabilistic neural network for multi-criteria event detector,"a multi-criteria event detection system, comprising a plurality of sensors, wherein each sensor is capable of detecting a signature characteristic of a presence of an event and providing an output indicating the same. a processor for receiving each output of the plurality of sensors is also employed. the processor includes a probabilistic neural network for processing the sensor outputs. the probabilistic neural network comprises a nonlinear, nor-parametric pattern recognition algorithm that operates by defining a probability density function for a plurality of data sets that are each based on a training set data and an optimized kernel width parameter. the plurality of data sets includes a baseline, non-event, first data set; a second, event data set; and a third, nuisance data set. the algorithm provides a decisional output indicative of the presence of a fire based on recognizing and discrimination between said data sets, and whether the outputs suffice to substantially indicate the presence of an event, as opposed to a non-event or nuisance situation.",2007-01-30,"The title of the patent is probabilistic neural network for multi-criteria event detector and its abstract is a multi-criteria event detection system, comprising a plurality of sensors, wherein each sensor is capable of detecting a signature characteristic of a presence of an event and providing an output indicating the same. a processor for receiving each output of the plurality of sensors is also employed. the processor includes a probabilistic neural network for processing the sensor outputs. the probabilistic neural network comprises a nonlinear, nor-parametric pattern recognition algorithm that operates by defining a probability density function for a plurality of data sets that are each based on a training set data and an optimized kernel width parameter. the plurality of data sets includes a baseline, non-event, first data set; a second, event data set; and a third, nuisance data set. the algorithm provides a decisional output indicative of the presence of a fire based on recognizing and discrimination between said data sets, and whether the outputs suffice to substantially indicate the presence of an event, as opposed to a non-event or nuisance situation. dated 2007-01-30"
7171394,global paint process optimization,"the present invention provides a method of optimizing a painting process for applying a paint layer on an article. the method comprises defining a functional relationship paint processing parameters and a paint layer property (i.e., the average paint layer thickness) using a neural network. this functional relationship is then used in a paint optimization function that measures a combination of quality control parameters and efficiency parameters. finally, the paint optimization function is optimized by adjusting the paint processing parameters utilizing the functional relationship formed by the neural network. the invention also provides a system that implements the methods of the invention.",2007-01-30,"The title of the patent is global paint process optimization and its abstract is the present invention provides a method of optimizing a painting process for applying a paint layer on an article. the method comprises defining a functional relationship paint processing parameters and a paint layer property (i.e., the average paint layer thickness) using a neural network. this functional relationship is then used in a paint optimization function that measures a combination of quality control parameters and efficiency parameters. finally, the paint optimization function is optimized by adjusting the paint processing parameters utilizing the functional relationship formed by the neural network. the invention also provides a system that implements the methods of the invention. dated 2007-01-30"
7173560,land mine detector,"a forwarding looking ground penetrating mine detection apparatus includes a radiation source for irradiating a sample of ground suspected of containing at least one mine with a plurality of frequency swept ground penetrating radar signals. a detector receives target signals backscattered from the ground responsive to the radar signal. the detector includes a time-frequency analyzer which transforms the target signals into a time-frequency image representation (tfr). in a preferred embodiment, the detector can include a wavelet packet transformer (wpt) for extracting time-frequency localized information from the tfr in the form of feature set constructed from a wavelet table. the apparatus can also include a data dimensionality reducer for selecting features to form a feature subset from the feature set, preferably based on reference to a training data set. a multilayer neural network classifier can be based on the feature subset, and be adaptable to the surrounding environment through learning.",2007-02-06,"The title of the patent is land mine detector and its abstract is a forwarding looking ground penetrating mine detection apparatus includes a radiation source for irradiating a sample of ground suspected of containing at least one mine with a plurality of frequency swept ground penetrating radar signals. a detector receives target signals backscattered from the ground responsive to the radar signal. the detector includes a time-frequency analyzer which transforms the target signals into a time-frequency image representation (tfr). in a preferred embodiment, the detector can include a wavelet packet transformer (wpt) for extracting time-frequency localized information from the tfr in the form of feature set constructed from a wavelet table. the apparatus can also include a data dimensionality reducer for selecting features to form a feature subset from the feature set, preferably based on reference to a training data set. a multilayer neural network classifier can be based on the feature subset, and be adaptable to the surrounding environment through learning. dated 2007-02-06"
7174324,crimping connection design system using multilayer feedforward neural networks,"estimation sections which have beforehand learned a relationship between known connection data pertaining to connection design and unknown connection data pertaining to connection design for the known connection data estimate the unknown connection data for the known connection data in accordance wit an input of the known connection data, on the basis of the result of learning. the respective estimation sections are formed from a multilayer feedforward neural network in which layers constituted of a plurality of neurons are coupled together in a direction in which the layer runs from an input layer to an output layer by way of an intermediate layer.",2007-02-06,"The title of the patent is crimping connection design system using multilayer feedforward neural networks and its abstract is estimation sections which have beforehand learned a relationship between known connection data pertaining to connection design and unknown connection data pertaining to connection design for the known connection data estimate the unknown connection data for the known connection data in accordance wit an input of the known connection data, on the basis of the result of learning. the respective estimation sections are formed from a multilayer feedforward neural network in which layers constituted of a plurality of neurons are coupled together in a direction in which the layer runs from an input layer to an output layer by way of an intermediate layer. dated 2007-02-06"
7177653,mobile user position locating system,"a method and system for locating a mobile user of one or more mobile users while on a call in a wireless telecommunication network utilizes a decision network that determines if the call signal is propagating through a repeater station. when the call is propagating through a repeater station, the decision network alters the signal measurement parameters to correspond with the location of the mobile user relative to the repeater station co-ordinates. by revising signal measurement parameters to compensate for repeater stations, the present invention provides for improved position estimates of the location of the mobile user in the network. for multiple cell soft handoff conditions, the decision network utilizes trained neural networks.",2007-02-13,"The title of the patent is mobile user position locating system and its abstract is a method and system for locating a mobile user of one or more mobile users while on a call in a wireless telecommunication network utilizes a decision network that determines if the call signal is propagating through a repeater station. when the call is propagating through a repeater station, the decision network alters the signal measurement parameters to correspond with the location of the mobile user relative to the repeater station co-ordinates. by revising signal measurement parameters to compensate for repeater stations, the present invention provides for improved position estimates of the location of the mobile user in the network. for multiple cell soft handoff conditions, the decision network utilizes trained neural networks. dated 2007-02-13"
7177710,system and method for adaptive control of uncertain nonlinear processes,"a computer system for controlling a nonlinear physical process. the computer system comprises a linear controller and a neural network. the linear controller receives a command signal for control of the nonlinear physical process and a measured output signal from the output of the nonlinear physical process. the linear controller generates a control signal based on the command signal, a measured output signal, and a fixed linear model for the process. the neural network receives the control signal from the linear controller and the measured output signal from the output of the nonlinear physical process. the neural network uses the measured output signal to modify the connection weights of the neural network. the neural network also generates a modified control signal supplied to the linear controller to iterate a fixed point solution for the modified control signal used to control the nonlinear physical process.",2007-02-13,"The title of the patent is system and method for adaptive control of uncertain nonlinear processes and its abstract is a computer system for controlling a nonlinear physical process. the computer system comprises a linear controller and a neural network. the linear controller receives a command signal for control of the nonlinear physical process and a measured output signal from the output of the nonlinear physical process. the linear controller generates a control signal based on the command signal, a measured output signal, and a fixed linear model for the process. the neural network receives the control signal from the linear controller and the measured output signal from the output of the nonlinear physical process. the neural network uses the measured output signal to modify the connection weights of the neural network. the neural network also generates a modified control signal supplied to the linear controller to iterate a fixed point solution for the modified control signal used to control the nonlinear physical process. dated 2007-02-13"
7177743,vehicle control system having an adaptive controller,"the vehicle control system having an adaptive controller is provided that accomplishes unsupervised learning such that no prior extensive training is needed for every situation. the inventive controller system is based on a neural network evolved with genetic algorithm. the genetic algorithm will determine the parameters of the neurons, the connections between the neurons and the associated weights to yield the best results. the genetic algorithm evaluates current candidate structures for accomplishing the desired result and develops new candidate structures by reproducing prior candidate structures with modification that replaces the least fit former candidate structures until the system is well satisfied. the vehicle control system is well satisfied when the desired result is met or some failure condition is triggered for vehicle control system whereby the action is never repeated.",2007-02-13,"The title of the patent is vehicle control system having an adaptive controller and its abstract is the vehicle control system having an adaptive controller is provided that accomplishes unsupervised learning such that no prior extensive training is needed for every situation. the inventive controller system is based on a neural network evolved with genetic algorithm. the genetic algorithm will determine the parameters of the neurons, the connections between the neurons and the associated weights to yield the best results. the genetic algorithm evaluates current candidate structures for accomplishing the desired result and develops new candidate structures by reproducing prior candidate structures with modification that replaces the least fit former candidate structures until the system is well satisfied. the vehicle control system is well satisfied when the desired result is met or some failure condition is triggered for vehicle control system whereby the action is never repeated. dated 2007-02-13"
7180629,methods and apparatus for color device characterization,"methods and apparatus for color correction of color image data obtained by an electronic camera determine a correction to data representative of the color image based upon an estimated illuminant using a neural network, multilayer perceptron models and/or coactive neuro-fuzzy inference system models, and apply the correction to the data representative of the color image. data representative of the color corrected data may be recorded or transmitted. a method of recording image data obtained by an electronic camera captures a color image, generates data representative of the image, estimates an illuminant for the captured color image, generates data representative of the estimated illuminant and records the data representative of the image with the data representative of the estimated illuminant. a method of transmitting image data obtained by an electronic camera captures a color image, generates data representative of the image, estimates an illuminant for the captured color image, generates data representative of the estimated illuminant and transmits the data representative of the image with the data representative of the estimated illuminant.",2007-02-20,"The title of the patent is methods and apparatus for color device characterization and its abstract is methods and apparatus for color correction of color image data obtained by an electronic camera determine a correction to data representative of the color image based upon an estimated illuminant using a neural network, multilayer perceptron models and/or coactive neuro-fuzzy inference system models, and apply the correction to the data representative of the color image. data representative of the color corrected data may be recorded or transmitted. a method of recording image data obtained by an electronic camera captures a color image, generates data representative of the image, estimates an illuminant for the captured color image, generates data representative of the estimated illuminant and records the data representative of the image with the data representative of the estimated illuminant. a method of transmitting image data obtained by an electronic camera captures a color image, generates data representative of the image, estimates an illuminant for the captured color image, generates data representative of the estimated illuminant and transmits the data representative of the image with the data representative of the estimated illuminant. dated 2007-02-20"
7181334,method and apparatus to diagnose intake airflow,"an engine control system including an engine controller, a plurality of sensors coupled to the engine controller, and a neural network operating in the engine controller, and where upon failure of at least one of the plurality sensors, the neural network generates a representative value of the failed sensor.",2007-02-20,"The title of the patent is method and apparatus to diagnose intake airflow and its abstract is an engine control system including an engine controller, a plurality of sensors coupled to the engine controller, and a neural network operating in the engine controller, and where upon failure of at least one of the plurality sensors, the neural network generates a representative value of the failed sensor. dated 2007-02-20"
7181768,computer intrusion detection system and method based on application monitoring,"an intrusion detection system (ids) that uses application monitors for detecting application-based attacks against computer systems. the ids implements application monitors in the form of a software program to learn and monitor the behavior of system programs in order to detect attacks against computer hosts. the application monitors implement machine learning algorithms to provide a mechanism for learning from previously observed behavior in order to recognize future attacks that it has not seen before. the application monitors include temporal locality algorithms to increased the accuracy of the ids. the ids of the present invention may comprise a string-matching program, a neural network, or a time series prediction algorithm for learning normal application behavior and for detecting anomalies.",2007-02-20,"The title of the patent is computer intrusion detection system and method based on application monitoring and its abstract is an intrusion detection system (ids) that uses application monitors for detecting application-based attacks against computer systems. the ids implements application monitors in the form of a software program to learn and monitor the behavior of system programs in order to detect attacks against computer hosts. the application monitors implement machine learning algorithms to provide a mechanism for learning from previously observed behavior in order to recognize future attacks that it has not seen before. the application monitors include temporal locality algorithms to increased the accuracy of the ids. the ids of the present invention may comprise a string-matching program, a neural network, or a time series prediction algorithm for learning normal application behavior and for detecting anomalies. dated 2007-02-20"
7184080,automatic white balancing via illuminant scoring,"automatic white balancing and/or autoexposure as useful in a digital camera extracts color channel gains from comparisons of image colors with reference colors under various color temperature illuminants and/or extracts exposure settings from illuminance mean, illuminance variance, illuminance minimum, and illuminance maximum in areas of an image with a trained neural network.",2007-02-27,"The title of the patent is automatic white balancing via illuminant scoring and its abstract is automatic white balancing and/or autoexposure as useful in a digital camera extracts color channel gains from comparisons of image colors with reference colors under various color temperature illuminants and/or extracts exposure settings from illuminance mean, illuminance variance, illuminance minimum, and illuminance maximum in areas of an image with a trained neural network. dated 2007-02-27"
7187778,neurofuzzy based device for programmable hearing aids,"a neurofuzzy device is described that provides a fuzzy logic based user-machine interface for optimal fitting of programmable hearing prosthesis using a neural network that generates targets to be matched by the hearing prosthesis based on individual audiometric and other relevant data to the specific impairment and on the neural network accumulated learning from previous successful fittings. the incorporated learning process can occur on or off line and implements fitting rationales that can satisfy the needs of a general or specific clientele. the parameters of the programmable prosthetic device are set as a group in order to achieve optimal matching to the targets. the user-machine interface realized by a fuzzy logic system deciphers the commends/responses of the user while listening to various stimuli and modifies the targets accordingly thus, providing a closed loop system for in-situ interactive fitting.",2007-03-06,"The title of the patent is neurofuzzy based device for programmable hearing aids and its abstract is a neurofuzzy device is described that provides a fuzzy logic based user-machine interface for optimal fitting of programmable hearing prosthesis using a neural network that generates targets to be matched by the hearing prosthesis based on individual audiometric and other relevant data to the specific impairment and on the neural network accumulated learning from previous successful fittings. the incorporated learning process can occur on or off line and implements fitting rationales that can satisfy the needs of a general or specific clientele. the parameters of the programmable prosthetic device are set as a group in order to achieve optimal matching to the targets. the user-machine interface realized by a fuzzy logic system deciphers the commends/responses of the user while listening to various stimuli and modifies the targets accordingly thus, providing a closed loop system for in-situ interactive fitting. dated 2007-03-06"
7187811,method for image resolution enhancement,"the invention provides a method for image resolution enhancement, which utilizes a fuzzy analysis system simulating the human vision system and uses the neural network as a basis for digital image interpolation. after an original image is inputted in, the image analysis module will divide and classify the original image, and then each of the image being classified will be processed by either the bilinear interpolation or the neural network interpolation. because the fuzzy analysis system is configured according to the human vision system, and because the neural network is a model obtained from learning real natural images, the vision effect of the image enlarged through the processing method of the invention is very close to the real natural image.",2007-03-06,"The title of the patent is method for image resolution enhancement and its abstract is the invention provides a method for image resolution enhancement, which utilizes a fuzzy analysis system simulating the human vision system and uses the neural network as a basis for digital image interpolation. after an original image is inputted in, the image analysis module will divide and classify the original image, and then each of the image being classified will be processed by either the bilinear interpolation or the neural network interpolation. because the fuzzy analysis system is configured according to the human vision system, and because the neural network is a model obtained from learning real natural images, the vision effect of the image enlarged through the processing method of the invention is very close to the real natural image. dated 2007-03-06"
7191150,enhancing delinquent debt collection using statistical models of debt historical information and account events,"a predictive model, for example, a neural network, evaluates individual debt holder accounts and predicts the amount that will be collected on each account based on learned relationships among known variables. the predictive model is generated using historical data of delinquent debt accounts, the collection methods used to collect the debts in the accounts, and the success of the collection methods. in one embodiment, the predictive model is generated using profiles of delinquent debt accounts summarizing patterns of events in the accounts, and the success of the collection effort in each account. in another embodiment, the predictive model includes a mathematical representation of the collector's notes created during the collection period for each account.",2007-03-13,"The title of the patent is enhancing delinquent debt collection using statistical models of debt historical information and account events and its abstract is a predictive model, for example, a neural network, evaluates individual debt holder accounts and predicts the amount that will be collected on each account based on learned relationships among known variables. the predictive model is generated using historical data of delinquent debt accounts, the collection methods used to collect the debts in the accounts, and the success of the collection methods. in one embodiment, the predictive model is generated using profiles of delinquent debt accounts summarizing patterns of events in the accounts, and the success of the collection effort in each account. in another embodiment, the predictive model includes a mathematical representation of the collector's notes created during the collection period for each account. dated 2007-03-13"
7191161,method for constructing composite response surfaces by combining neural networks with polynominal interpolation or estimation techniques,"a method and system for data modeling that incorporates the advantages of both traditional response surface methodology (rsm) and neural networks is disclosed. the invention partitions the parameters into a first set of s simple parameters, where observable data are expressible as low order polynomials, and c complex parameters that reflect more complicated variation of the observed data. variation of the data with the simple parameters is modeled using polynomials; and variation of the data with the complex parameters at each vertex is analyzed using a neural network. variations with the simple parameters and with the complex parameters are expressed using a first sequence of shape functions and a second sequence of neural network functions. the first and second sequences are multiplicatively combined to form a composite response surface, dependent upon the parameter values, that can be used to identify an accurate model.",2007-03-13,"The title of the patent is method for constructing composite response surfaces by combining neural networks with polynominal interpolation or estimation techniques and its abstract is a method and system for data modeling that incorporates the advantages of both traditional response surface methodology (rsm) and neural networks is disclosed. the invention partitions the parameters into a first set of s simple parameters, where observable data are expressible as low order polynomials, and c complex parameters that reflect more complicated variation of the observed data. variation of the data with the simple parameters is modeled using polynomials; and variation of the data with the complex parameters at each vertex is analyzed using a neural network. variations with the simple parameters and with the complex parameters are expressed using a first sequence of shape functions and a second sequence of neural network functions. the first and second sequences are multiplicatively combined to form a composite response surface, dependent upon the parameter values, that can be used to identify an accurate model. dated 2007-03-13"
7194383,vibration analysis system and method for a machine,"a system and method for detecting and analyzing anomalies in a machine during operation. the system and method includes at least one sensor configured to detect characteristics of the machine indicative of machine vibration, at least one other sensor configured to detect characteristics of the machine indicative of other than machine vibration, a plurality of neural networks to receive input data, at least one neural network receiving vibration data from the at least one sensor, and at least one other neural network receiving non-vibration data from the at least one other sensor, and an expert system to receive output data from the neural networks and responsively analyze machine operation for anomalies.",2007-03-20,"The title of the patent is vibration analysis system and method for a machine and its abstract is a system and method for detecting and analyzing anomalies in a machine during operation. the system and method includes at least one sensor configured to detect characteristics of the machine indicative of machine vibration, at least one other sensor configured to detect characteristics of the machine indicative of other than machine vibration, a plurality of neural networks to receive input data, at least one neural network receiving vibration data from the at least one sensor, and at least one other neural network receiving non-vibration data from the at least one other sensor, and an expert system to receive output data from the neural networks and responsively analyze machine operation for anomalies. dated 2007-03-20"
7196786,method and apparatus for a tunable diode laser spectrometer for analysis of hydrocarbon samples,"the present invention provides an down hole apparatus and method for ultrahigh resolution spectroscopy using a tunable diode laser (tdl) for analyzing a formation fluid sample downhole or at the surface to determine formation fluid parameters. in addition to absorption spectroscopy, the present invention can perform raman spectroscopy on the fluid, by sweeping the wavelength of the tdl and detecting the raman-scattered light using a narrow-band detector at a fixed wavelength. the spectrometer analyzes a pressurized well bore fluid sample that is collected downhole. the analysis is performed either downhole or at the surface onsite. near infrared, mid-infrared and visible light analysis is also performed on the sample to provide an onsite surface or downhole analysis of sample properties and contamination level. the onsite and downhole analysis comprises determination of aromatics, olefins, saturates, gas oil ratio, api gravity and various other parameters which can be estimated by correlation, a trained neural network or a chemometric equation.",2007-03-27,"The title of the patent is method and apparatus for a tunable diode laser spectrometer for analysis of hydrocarbon samples and its abstract is the present invention provides an down hole apparatus and method for ultrahigh resolution spectroscopy using a tunable diode laser (tdl) for analyzing a formation fluid sample downhole or at the surface to determine formation fluid parameters. in addition to absorption spectroscopy, the present invention can perform raman spectroscopy on the fluid, by sweeping the wavelength of the tdl and detecting the raman-scattered light using a narrow-band detector at a fixed wavelength. the spectrometer analyzes a pressurized well bore fluid sample that is collected downhole. the analysis is performed either downhole or at the surface onsite. near infrared, mid-infrared and visible light analysis is also performed on the sample to provide an onsite surface or downhole analysis of sample properties and contamination level. the onsite and downhole analysis comprises determination of aromatics, olefins, saturates, gas oil ratio, api gravity and various other parameters which can be estimated by correlation, a trained neural network or a chemometric equation. dated 2007-03-27"
7198630,"method and apparatus for controlling a surgical robot to mimic, harmonize and enhance the natural neurophysiological behavior of a surgeon","the present invention was developed by a neurosurgeon and seeks to mimic the results of primate neurological research which is indicative of a human's actual neurological control structures and logic. specifically, the motor proprioceptive and tactile neurophysiology functioning of the surgeon's hands and internal hand control system from the muscular level through the intrafusal fiber system of the neural network is considered in creating the robot and method of operation of the present invention. therefore, the surgery is not slowed down as in the art, because the surgeon is in conscious and subsconscious natural agreement and harmonization with the robotically actuated surgical instruments based on neurological mimicing of the surgeon's behavior with the functioning of the robot. therefore, the robot can enhance the surgeon's humanly limited senses while not introducing disruptive variables to the surgeon's naturally occurring operation of his neurophysiology. this is therefore also a new field, neurophysiological symbiotic robotics.",2007-04-03,"The title of the patent is method and apparatus for controlling a surgical robot to mimic, harmonize and enhance the natural neurophysiological behavior of a surgeon and its abstract is the present invention was developed by a neurosurgeon and seeks to mimic the results of primate neurological research which is indicative of a human's actual neurological control structures and logic. specifically, the motor proprioceptive and tactile neurophysiology functioning of the surgeon's hands and internal hand control system from the muscular level through the intrafusal fiber system of the neural network is considered in creating the robot and method of operation of the present invention. therefore, the surgery is not slowed down as in the art, because the surgeon is in conscious and subsconscious natural agreement and harmonization with the robotically actuated surgical instruments based on neurological mimicing of the surgeon's behavior with the functioning of the robot. therefore, the robot can enhance the surgeon's humanly limited senses while not introducing disruptive variables to the surgeon's naturally occurring operation of his neurophysiology. this is therefore also a new field, neurophysiological symbiotic robotics. dated 2007-04-03"
7200435,neural network based learning engine to adapt therapies,"a cardiac device system for implementing a cardiac device having adaptive treatment therapies utilizing a neural network based learning engine includes an implantable cardiac device module and an external data processing system for specifying the operating characteristics of the cardiac device module. both the cardiac device module and the external processing system possess an artificial neural network to specify the operation of the cardiac device module as it provides adaptive treatment therapies. the external data processing system includes a complete neural network module that trains and validates the operation of the neural network to match the optimal treatment options with a received set of collected patient data. in contrast, a runtime neural network module that only provides real time operation of the neural network using collected patient data is located within the cardiac device module. the cardiac device module and the external processing module communicate with each other to pass collected patient data from the cardiac device module to the external processing system when the operation of the neural network is to be updated. the cardiac device module and the external processing module also communicate with each other to pass operating coefficients for the neural network back from the external processing system to the cardiac device module once these coefficients are updated.",2007-04-03,"The title of the patent is neural network based learning engine to adapt therapies and its abstract is a cardiac device system for implementing a cardiac device having adaptive treatment therapies utilizing a neural network based learning engine includes an implantable cardiac device module and an external data processing system for specifying the operating characteristics of the cardiac device module. both the cardiac device module and the external processing system possess an artificial neural network to specify the operation of the cardiac device module as it provides adaptive treatment therapies. the external data processing system includes a complete neural network module that trains and validates the operation of the neural network to match the optimal treatment options with a received set of collected patient data. in contrast, a runtime neural network module that only provides real time operation of the neural network using collected patient data is located within the cardiac device module. the cardiac device module and the external processing module communicate with each other to pass collected patient data from the cardiac device module to the external processing system when the operation of the neural network is to be updated. the cardiac device module and the external processing module also communicate with each other to pass operating coefficients for the neural network back from the external processing system to the cardiac device module once these coefficients are updated. dated 2007-04-03"
7200475,methods and devices for identifying the type of occupancy of a supporting surface,"a method and device for identifying the type of occupancy of a supporting surface, particularly of a motor vehicle seat (2), with the aid of force sensor-assisted signals. the sensor signals si, which are recorded at predetermined instants ti, or quantities derived therefrom, are continuously stored in a memory (4) in such a manner that, the sensor signals from the recent past, or the quantities derived therefrom are available for analysis at any time, and that the type of occupancy is derived from these stored values using at least two independent calculation methods or using a neural network.",2007-04-03,"The title of the patent is methods and devices for identifying the type of occupancy of a supporting surface and its abstract is a method and device for identifying the type of occupancy of a supporting surface, particularly of a motor vehicle seat (2), with the aid of force sensor-assisted signals. the sensor signals si, which are recorded at predetermined instants ti, or quantities derived therefrom, are continuously stored in a memory (4) in such a manner that, the sensor signals from the recent past, or the quantities derived therefrom are available for analysis at any time, and that the type of occupancy is derived from these stored values using at least two independent calculation methods or using a neural network. dated 2007-04-03"
7202794,flame detection system,a flame detection system includes a plurality of sensors for generating a plurality of respective sensor signals. the plurality of sensors includes a set of discrete optical radiation sensors responsive to flame as well as non-flame emissions. an artificial neural network may be applied in processing the sensor signals to provide an output corresponding to a flame condition.,2007-04-10,The title of the patent is flame detection system and its abstract is a flame detection system includes a plurality of sensors for generating a plurality of respective sensor signals. the plurality of sensors includes a set of discrete optical radiation sensors responsive to flame as well as non-flame emissions. an artificial neural network may be applied in processing the sensor signals to provide an output corresponding to a flame condition. dated 2007-04-10
7213008,"method and apparatus of creating application-specific, non-uniform wavelet transforms","a method and an apparatus of designing a set of wavelet basis trained to fit a particular problem. the method and apparatus include constructing a neural network of arbitrary complexity using a discrete and finite radon transform, feeding an input wavelet prototype through the neural network and its backpropagation to produce an output, and modifying the input wavelet prototype using the output.",2007-05-01,"The title of the patent is method and apparatus of creating application-specific, non-uniform wavelet transforms and its abstract is a method and an apparatus of designing a set of wavelet basis trained to fit a particular problem. the method and apparatus include constructing a neural network of arbitrary complexity using a discrete and finite radon transform, feeding an input wavelet prototype through the neural network and its backpropagation to produce an output, and modifying the input wavelet prototype using the output. dated 2007-05-01"
7216071,hybrid gas turbine engine state variable model,"the present invention relates to a system and a method for developing an engine model. the system broadly comprises a module for generating a state variable model of an engine, which module receives a plurality of inputs to an engine representative of a particular flight condition and generates a set of estimated engine parameters representative of the model. the system further comprises a comparator for comparing the set of estimated engine parameters to a set of measured engine parameters for generating a set of residuals and an artificial neural network module to be trained and to be used in an implementation configuration after training has been completed. the artificial neural network receives the set of residuals and the engine inputs during a training phase and generates a set of estimated residuals representative of the engine condition.",2007-05-08,"The title of the patent is hybrid gas turbine engine state variable model and its abstract is the present invention relates to a system and a method for developing an engine model. the system broadly comprises a module for generating a state variable model of an engine, which module receives a plurality of inputs to an engine representative of a particular flight condition and generates a set of estimated engine parameters representative of the model. the system further comprises a comparator for comparing the set of estimated engine parameters to a set of measured engine parameters for generating a set of residuals and an artificial neural network module to be trained and to be used in an implementation configuration after training has been completed. the artificial neural network receives the set of residuals and the engine inputs during a training phase and generates a set of estimated residuals representative of the engine condition. dated 2007-05-08"
7216112,"memory system, memory method, and robotic apparatus","a memory system and a method as well as robotic apparatus are strong against noise and excellent in memory capacity, volume of calculation, quantity of physical memory, and memory responsiveness. it is designed to store, in the frame form, the first information on a symbol as well as the second information on a symbol supplied separately from a variety of inputs in relation to competitive neurons corresponding to the symbol in a way to strengthen the connection between relevant input neurons and competitive neurons in response to the input patterns of a variety of inputs for each symbol with the use of a competitive neural network having a set of input layers composed of a plurality of input neurons and a set of competitive layers composed of a plurality of competitive neurons.",2007-05-08,"The title of the patent is memory system, memory method, and robotic apparatus and its abstract is a memory system and a method as well as robotic apparatus are strong against noise and excellent in memory capacity, volume of calculation, quantity of physical memory, and memory responsiveness. it is designed to store, in the frame form, the first information on a symbol as well as the second information on a symbol supplied separately from a variety of inputs in relation to competitive neurons corresponding to the symbol in a way to strengthen the connection between relevant input neurons and competitive neurons in response to the input patterns of a variety of inputs for each symbol with the use of a competitive neural network having a set of input layers composed of a plurality of input neurons and a set of competitive layers composed of a plurality of competitive neurons. dated 2007-05-08"
7218775,method and apparatus for identifying and quantifying characteristics of seeds and other small objects,"the invention provides a method for identifying or quantifying characteristics of interest of unknown objects, comprising training a single neural network model with training sets of known objects having known values for the characteristics; validating the optimal neural network model; and analyzing unknown objects having unknown values of the characteristics by imaging them to obtain a digital image comprising pixels representing the unknown objects, background and any debris; processing the image to identify, separate, and retain pixels representing the unknown objects from pixels and to eliminate background and debris; analyzing the pixels representing each of the unknown objects to generate data representative of image parameters; providing the data to the flash code deployed from the candidate neural network model; analyzing the data through the flash code; and receiving output data (the unknown values of the characteristics of interest of the unknown objects) from the flash code in a predetermined format.",2007-05-15,"The title of the patent is method and apparatus for identifying and quantifying characteristics of seeds and other small objects and its abstract is the invention provides a method for identifying or quantifying characteristics of interest of unknown objects, comprising training a single neural network model with training sets of known objects having known values for the characteristics; validating the optimal neural network model; and analyzing unknown objects having unknown values of the characteristics by imaging them to obtain a digital image comprising pixels representing the unknown objects, background and any debris; processing the image to identify, separate, and retain pixels representing the unknown objects from pixels and to eliminate background and debris; analyzing the pixels representing each of the unknown objects to generate data representative of image parameters; providing the data to the flash code deployed from the candidate neural network model; analyzing the data through the flash code; and receiving output data (the unknown values of the characteristics of interest of the unknown objects) from the flash code in a predetermined format. dated 2007-05-15"
7219061,method for detecting the time sequences of a fundamental frequency of an audio response unit to be synthesized,"predetermined macrosegments of the fundamental frequency are determined by a neural network, and these predefined macrosegments are reproduced by fundamental-frequency sequences stored in a database. the fundamental frequency is generated on the basis of a relatively large text section which is analyzed by the neural network. microstructures from the database are received in the fundamental frequency. the fundamental frequency thus formed is thus optimized both with regard to its macrostructure and to its microstructure. as a result, an extremely natural sound is achieved.",2007-05-15,"The title of the patent is method for detecting the time sequences of a fundamental frequency of an audio response unit to be synthesized and its abstract is predetermined macrosegments of the fundamental frequency are determined by a neural network, and these predefined macrosegments are reproduced by fundamental-frequency sequences stored in a database. the fundamental frequency is generated on the basis of a relatively large text section which is analyzed by the neural network. microstructures from the database are received in the fundamental frequency. the fundamental frequency thus formed is thus optimized both with regard to its macrostructure and to its microstructure. as a result, an extremely natural sound is achieved. dated 2007-05-15"
7219087,soft computing optimizer of intelligent control system structures,"the present invention involves a soft computing (sc) optimizer for designing a knowledge base (kb) to be used in a control system for controlling a plant such as, for example, an internal combustion engine or an automobile suspension system. the sc optimizer includes a fuzzy inference engine based on a fuzzy neural network (fnn). the sc optimizer provides fuzzy inference system (fis) structure selection, fis structure optimization method selection, and teaching signal selection and generation. the user selects a fuzzy model, including one or more of: the number of input and/or output variables; the type of fuzzy inference model (e.g., mamdani, sugeno, tsukamoto, etc.); and the preliminary type of membership functions. a genetic algorithm (ga) is used to optimize linguistic variable parameters and the input-output training patterns. a ga is also used to optimize the rule base, using the fuzzy model, optimal linguistic variable parameters, and a teaching signal. the ga produces a near-optimal fnn. the near-optimal fnn can be improved using classical derivative-based optimization procedures. the fis structure found by the ga is optimized with a fitness function based on a response of the actual plant model of the controlled plant. the sc optimizer produces a robust kb that is typically smaller that the kb produced by prior art methods.",2007-05-15,"The title of the patent is soft computing optimizer of intelligent control system structures and its abstract is the present invention involves a soft computing (sc) optimizer for designing a knowledge base (kb) to be used in a control system for controlling a plant such as, for example, an internal combustion engine or an automobile suspension system. the sc optimizer includes a fuzzy inference engine based on a fuzzy neural network (fnn). the sc optimizer provides fuzzy inference system (fis) structure selection, fis structure optimization method selection, and teaching signal selection and generation. the user selects a fuzzy model, including one or more of: the number of input and/or output variables; the type of fuzzy inference model (e.g., mamdani, sugeno, tsukamoto, etc.); and the preliminary type of membership functions. a genetic algorithm (ga) is used to optimize linguistic variable parameters and the input-output training patterns. a ga is also used to optimize the rule base, using the fuzzy model, optimal linguistic variable parameters, and a teaching signal. the ga produces a near-optimal fnn. the near-optimal fnn can be improved using classical derivative-based optimization procedures. the fis structure found by the ga is optimized with a fitness function based on a response of the actual plant model of the controlled plant. the sc optimizer produces a robust kb that is typically smaller that the kb produced by prior art methods. dated 2007-05-15"
7221462,method and device for optical detection of the position of an object,"the present invention relates to a method and to a device enabling data to be input, and also to an optical system for detecting the position of an article or a member and suitable for being used (or incorporated) in such a method (or device). the technical field of the invention is that of making keyboards and similar devices enabling manual input of data for processing by a digital computer. the method of the invention for determining the position of an article in a zone uses the steps of structuring a neural network by training, applying data to the input of the structured neural network, the data being the result of converting signals delivered by a plurality of detectors sensitive to illumination of said zone, and determining the position of the article in the zone as a function of at least one data item output by the structured neural network.",2007-05-22,"The title of the patent is method and device for optical detection of the position of an object and its abstract is the present invention relates to a method and to a device enabling data to be input, and also to an optical system for detecting the position of an article or a member and suitable for being used (or incorporated) in such a method (or device). the technical field of the invention is that of making keyboards and similar devices enabling manual input of data for processing by a digital computer. the method of the invention for determining the position of an article in a zone uses the steps of structuring a neural network by training, applying data to the input of the structured neural network, the data being the result of converting signals delivered by a plurality of detectors sensitive to illumination of said zone, and determining the position of the article in the zone as a function of at least one data item output by the structured neural network. dated 2007-05-22"
7221621,system and method for identification and quantification of sonar targets in a liquid medium,a system and method for identifying and quantifying targets within a liquid medium. a raw sidescan sonar image is collected. a region of interest is separated from the image. an image transformation is performed using an extraction algorithm. salient image characteristics are calculated. spurious pixels are removed from the image to obtain an extracted region of interest. particle analysis is performed on the extracted region of interest to generate a feature vector which is presented to a neural network for classification.,2007-05-22,The title of the patent is system and method for identification and quantification of sonar targets in a liquid medium and its abstract is a system and method for identifying and quantifying targets within a liquid medium. a raw sidescan sonar image is collected. a region of interest is separated from the image. an image transformation is performed using an extraction algorithm. salient image characteristics are calculated. spurious pixels are removed from the image to obtain an extracted region of interest. particle analysis is performed on the extracted region of interest to generate a feature vector which is presented to a neural network for classification. dated 2007-05-22
7221807,methods and systems for digital image characteristic adjustment using a neural network,embodiments of the present invention comprise methods and systems for automatically adjusting images to conform to preference data.,2007-05-22,The title of the patent is methods and systems for digital image characteristic adjustment using a neural network and its abstract is embodiments of the present invention comprise methods and systems for automatically adjusting images to conform to preference data. dated 2007-05-22
7222002,vibration engine monitoring neural network object monitoring,"the present invention provides an aircraft engine vibration system that provides information about engine health. embodiments of the present invention monitor for excessive vibration, monitor for bird strike, monitor for ice build up on the fan section, and monitor general engine health. an embodiment of the present invention utilizes neural network architecture for the detection of excessive vibration and ice detection build-up on the fan section of a turbo-fan engine and to monitor engine health through the high-pressure turbine section of the engine.",2007-05-22,"The title of the patent is vibration engine monitoring neural network object monitoring and its abstract is the present invention provides an aircraft engine vibration system that provides information about engine health. embodiments of the present invention monitor for excessive vibration, monitor for bird strike, monitor for ice build up on the fan section, and monitor general engine health. an embodiment of the present invention utilizes neural network architecture for the detection of excessive vibration and ice detection build-up on the fan section of a turbo-fan engine and to monitor engine health through the high-pressure turbine section of the engine. dated 2007-05-22"
7222112,engine control system using a cascaded neural network,"a method, system and machine-readable storage medium for monitoring an engine using a cascaded neural network that includes a plurality of neural networks is disclosed. in operation, the method, system and machine-readable storage medium store data corresponding to the cascaded neural network. signals generated by a plurality of engine sensors are then inputted into the cascaded neural network. next, a second neural network is updated at a first rate, with an output of a first neural network, wherein the output is based on the inputted signals. in response, the second neural network outputs at a second rate, at least one engine control signal, wherein the second rate is faster than the first rate.",2007-05-22,"The title of the patent is engine control system using a cascaded neural network and its abstract is a method, system and machine-readable storage medium for monitoring an engine using a cascaded neural network that includes a plurality of neural networks is disclosed. in operation, the method, system and machine-readable storage medium store data corresponding to the cascaded neural network. signals generated by a plurality of engine sensors are then inputted into the cascaded neural network. next, a second neural network is updated at a first rate, with an output of a first neural network, wherein the output is based on the inputted signals. in response, the second neural network outputs at a second rate, at least one engine control signal, wherein the second rate is faster than the first rate. dated 2007-05-22"
7224526,three-dimensional free space image projection employing fresnel lenses,"disclosed herein are three-dimensional free space imaging systems and related methods employing a dynamic stereoscopic image projection system in combination with a optic module comprising a doublet of fresnel lenses. the dynamic projector systems calculating derived flat image information for each projector based upon inputted stereopair images and information regarding the projector elements and optic module. in preferred embodiments of the present invention, the projection system uses an image computational device that employs a neural network feedback calculation to calculate the appropriate flat image information and appropriate images to be projected on the screen by the projectors at any given time.",2007-05-29,"The title of the patent is three-dimensional free space image projection employing fresnel lenses and its abstract is disclosed herein are three-dimensional free space imaging systems and related methods employing a dynamic stereoscopic image projection system in combination with a optic module comprising a doublet of fresnel lenses. the dynamic projector systems calculating derived flat image information for each projector based upon inputted stereopair images and information regarding the projector elements and optic module. in preferred embodiments of the present invention, the projection system uses an image computational device that employs a neural network feedback calculation to calculate the appropriate flat image information and appropriate images to be projected on the screen by the projectors at any given time. dated 2007-05-29"
7225172,method and apparatus for multivariable analysis of biological measurements,"in a method and apparatus for analyzing multivariable data sets, a general computerized platform is provided for evaluating the relationship between large number of measurements of sets of variables characterizing components of complex states of a system under induced stimulation or controlled conditions. the linked responses of variables and their temporal relations tell about the network of interactions and their hierarchy. processing of data sets by a simple neural network gives a matrix of weight parameters, that allow to identify fingerprints of complex states characterized by patterns of measured variable and estimate the interactions between the components characterized by the measured variables. the results are provided numerically and by color-coded presentation indicating dominating relations between variables and strongly responding variables. when applied to dynamic responses of a system, the analysis can construct a schematic hierarchical architecture of the network of interaction between the components of the studied system. applications in biology include analysis of measurements characterizing responses of molecular components in cells under changes induced by stimuli (e.g. drugs, growth factors, hormones, mutations or forced expression of a proteins), and identification of complex cellular states (e.g. proliferation, differentiation, transformation, starvation, necrosis, apoptosis, and the time dependencies of the above effects).",2007-05-29,"The title of the patent is method and apparatus for multivariable analysis of biological measurements and its abstract is in a method and apparatus for analyzing multivariable data sets, a general computerized platform is provided for evaluating the relationship between large number of measurements of sets of variables characterizing components of complex states of a system under induced stimulation or controlled conditions. the linked responses of variables and their temporal relations tell about the network of interactions and their hierarchy. processing of data sets by a simple neural network gives a matrix of weight parameters, that allow to identify fingerprints of complex states characterized by patterns of measured variable and estimate the interactions between the components characterized by the measured variables. the results are provided numerically and by color-coded presentation indicating dominating relations between variables and strongly responding variables. when applied to dynamic responses of a system, the analysis can construct a schematic hierarchical architecture of the network of interaction between the components of the studied system. applications in biology include analysis of measurements characterizing responses of molecular components in cells under changes induced by stimuli (e.g. drugs, growth factors, hormones, mutations or forced expression of a proteins), and identification of complex cellular states (e.g. proliferation, differentiation, transformation, starvation, necrosis, apoptosis, and the time dependencies of the above effects). dated 2007-05-29"
7225173,apparatus and method for efficient adaptation of finite element meshes for numerical solutions of partial differential equations,"a device for calculating numerical solutions for partial differential equations in successive intervals using adaptive meshes, comprises: a neural network part for producing predictions of gradients at a following interval based on gradients available from previous intervals, and a mesh adaptation part, associated with the neural network part, configured for adapting a mesh over a domain of a respective partial differential equation using the predictions, such that the mesh adaptively refines itself about emerging regions of complexity as the partial differential equation progresses over the successive intervals. the neural network part succeeds in its predictions since its use herein is equivalent to using time series function fitting techniques.",2007-05-29,"The title of the patent is apparatus and method for efficient adaptation of finite element meshes for numerical solutions of partial differential equations and its abstract is a device for calculating numerical solutions for partial differential equations in successive intervals using adaptive meshes, comprises: a neural network part for producing predictions of gradients at a following interval based on gradients available from previous intervals, and a mesh adaptation part, associated with the neural network part, configured for adapting a mesh over a domain of a respective partial differential equation using the predictions, such that the mesh adaptively refines itself about emerging regions of complexity as the partial differential equation progresses over the successive intervals. the neural network part succeeds in its predictions since its use herein is equivalent to using time series function fitting techniques. dated 2007-05-29"
7233932,fault detection system and method using approximate null space base fault signature classification,"a system and method for fault detection is provided. the fault detection system provides the ability to detect symptoms of fault in turbine engines and other mechanical systems that have nonlinear relationships between two or more variables. the fault detection system uses a neural network to perform feature extraction from data for representation of faulty or normal conditions. the values of extracted features, referred to herein as scores, are then used to determine the likelihood of fault in the system. specifically, the lower order scores, referred to herein as “approximate null space” scores can be classified into one or more clusters, where some clusters represent types of faults in the turbine engine. classification based on the approximate null space scores provides the ability to classify faulty or nominal conditions that could not be reliably classified using higher order scores.",2007-06-19,"The title of the patent is fault detection system and method using approximate null space base fault signature classification and its abstract is a system and method for fault detection is provided. the fault detection system provides the ability to detect symptoms of fault in turbine engines and other mechanical systems that have nonlinear relationships between two or more variables. the fault detection system uses a neural network to perform feature extraction from data for representation of faulty or normal conditions. the values of extracted features, referred to herein as scores, are then used to determine the likelihood of fault in the system. specifically, the lower order scores, referred to herein as “approximate null space” scores can be classified into one or more clusters, where some clusters represent types of faults in the turbine engine. classification based on the approximate null space scores provides the ability to classify faulty or nominal conditions that could not be reliably classified using higher order scores. dated 2007-06-19"
7236615,synergistic face detection and pose estimation with energy-based models,"a method for human face detection that detects faces independently of their particular poses and simultaneously estimates those poses. our method exhibits an immunity to variations in skin color, eyeglasses, facial hair, lighting, scale and facial expressions, and others. in operation, we train a convolutional neural network to map face images to points on a face manifold, and non-face images to points far away from that manifold, wherein that manifold is parameterized by facial pose. conceptually, we view a pose parameter as a latent variable, which may be inferred through an energy-minimization process. to train systems based upon our inventive method, we derive a new type of discriminative loss function that is tailored to such detection tasks. our method enables a multi-view detector that can detect faces in a variety of poses, for example, looking left or right (yaw axis), up or down (pitch axis), or tilting left or right (roll axis). systems employing our method are highly-reliable, run at near real time (5 frames per second on conventional hardware), and is robust against variations in yaw (±90°), roll(±45°), and pitch(±60°).",2007-06-26,"The title of the patent is synergistic face detection and pose estimation with energy-based models and its abstract is a method for human face detection that detects faces independently of their particular poses and simultaneously estimates those poses. our method exhibits an immunity to variations in skin color, eyeglasses, facial hair, lighting, scale and facial expressions, and others. in operation, we train a convolutional neural network to map face images to points on a face manifold, and non-face images to points far away from that manifold, wherein that manifold is parameterized by facial pose. conceptually, we view a pose parameter as a latent variable, which may be inferred through an energy-minimization process. to train systems based upon our inventive method, we derive a new type of discriminative loss function that is tailored to such detection tasks. our method enables a multi-view detector that can detect faces in a variety of poses, for example, looking left or right (yaw axis), up or down (pitch axis), or tilting left or right (roll axis). systems employing our method are highly-reliable, run at near real time (5 frames per second on conventional hardware), and is robust against variations in yaw (±90°), roll(±45°), and pitch(±60°). dated 2007-06-26"
7241323,pyrolysis process for producing fuel gas,"solid waste resource recovery in space is effected by pyrolysis processing, to produce light gases as the main products (ch4, h2, co2, co, h2o, nh3) and a reactive carbon-rich char as the main byproduct. significant amounts of liquid products are formed under less severe pyrolysis conditions, and are cracked almost completely to gases as the temperature is raised. a primary pyrolysis model for the composite mixture is based on an existing model for whole biomass materials, and an artificial neural network models the changes in gas composition with the severity of pyrolysis conditions.",2007-07-10,"The title of the patent is pyrolysis process for producing fuel gas and its abstract is solid waste resource recovery in space is effected by pyrolysis processing, to produce light gases as the main products (ch4, h2, co2, co, h2o, nh3) and a reactive carbon-rich char as the main byproduct. significant amounts of liquid products are formed under less severe pyrolysis conditions, and are cracked almost completely to gases as the temperature is raised. a primary pyrolysis model for the composite mixture is based on an existing model for whole biomass materials, and an artificial neural network models the changes in gas composition with the severity of pyrolysis conditions. dated 2007-07-10"
7246079,"method of predicting initial input of new product, system for predicting initial input of new product, and recording medium","based on numerical values with respect to factors influencing shares of existing products and a new product evaluated by more than one people, a structured neural network calculates predictive shares of the new product predicted by the respective persons. comprehensive evaluations on the respective products and the new product are calculated for each person, based on the numerical values with respect to the respective factors. correlation coefficients between the comprehensive evaluations on the respective products by the respective persons and the actual shares are calculated. the predictive shares calculated by the structural neural network are layered out in accordance with the correlation coefficients for the respective person. average values of the predictive shares and confidence intervals are calculated for the respective layers, and based on them and the calculation result obtained by the structured neural network, a share of the new product is predicted.",2007-07-17,"The title of the patent is method of predicting initial input of new product, system for predicting initial input of new product, and recording medium and its abstract is based on numerical values with respect to factors influencing shares of existing products and a new product evaluated by more than one people, a structured neural network calculates predictive shares of the new product predicted by the respective persons. comprehensive evaluations on the respective products and the new product are calculated for each person, based on the numerical values with respect to the respective factors. correlation coefficients between the comprehensive evaluations on the respective products by the respective persons and the actual shares are calculated. the predictive shares calculated by the structural neural network are layered out in accordance with the correlation coefficients for the respective person. average values of the predictive shares and confidence intervals are calculated for the respective layers, and based on them and the calculation result obtained by the structured neural network, a share of the new product is predicted. dated 2007-07-17"
7249115,network modelling,"according to a first aspect of the present invention there is provided a method of modelling a network comprising operating the network as a neural network and executing a neural network modelling algorithm on the network, whereby the network models its own response to a requested action.",2007-07-24,"The title of the patent is network modelling and its abstract is according to a first aspect of the present invention there is provided a method of modelling a network comprising operating the network as a neural network and executing a neural network modelling algorithm on the network, whereby the network models its own response to a requested action. dated 2007-07-24"
7251637,context vector generation and retrieval,"a system and method for generating context vectors for use in storage and retrieval of documents and other information items. context vectors represent conceptual relationships among information items by quantitative means. a neural network operates on a training corpus of records to develop relationship-based context vectors based on word proximity and co-importance using a technique of “windowed co-occurrence”. relationships among context vectors are deterministic, so that a context vector set has one logical solution, although it may have a plurality of physical solutions. no human knowledge, thesaurus, synonym list, knowledge base, or conceptual hierarchy, is required. summary vectors of records may be clustered to reduce searching time, by forming a tree of clustered nodes. once the context vectors are determined, records may be retrieved using a query interface that allows a user to specify content terms, boolean terms, and/or document feedback. the present invention further facilitates visualization of textual information by translating context vectors into visual and graphical representations. thus, a user can explore visual representations of meaning, and can apply human visual pattern recognition skills to document searches.",2007-07-31,"The title of the patent is context vector generation and retrieval and its abstract is a system and method for generating context vectors for use in storage and retrieval of documents and other information items. context vectors represent conceptual relationships among information items by quantitative means. a neural network operates on a training corpus of records to develop relationship-based context vectors based on word proximity and co-importance using a technique of “windowed co-occurrence”. relationships among context vectors are deterministic, so that a context vector set has one logical solution, although it may have a plurality of physical solutions. no human knowledge, thesaurus, synonym list, knowledge base, or conceptual hierarchy, is required. summary vectors of records may be clustered to reduce searching time, by forming a tree of clustered nodes. once the context vectors are determined, records may be retrieved using a query interface that allows a user to specify content terms, boolean terms, and/or document feedback. the present invention further facilitates visualization of textual information by translating context vectors into visual and graphical representations. thus, a user can explore visual representations of meaning, and can apply human visual pattern recognition skills to document searches. dated 2007-07-31"
7251638,intelligent robust control system for motorcycle using soft computing optimizer,"a soft computing (sc) optimizer for designing a knowledge base (kb) to be used in a control system for controlling a motorcycle is described. in one embodiment, a simulation model of the motorcycle and rider control is used. in one embodiment, the simulation model includes a feedforward rider model. the sc optimizer includes a fuzzy inference engine based on a fuzzy neural network (fnn). the sc optimizer provides fuzzy inference system (fis) structure selection, fis structure optimization method selection, and teaching signal selection and generation. the user selects a fuzzy model, including one or more of: the number of input and/or output variables; the type of fuzzy inference; and the preliminary type of membership functions. a genetic algorithm (ga) is used to optimize linguistic variable parameters and the input-output training patterns. a ga is also used to optimize the rule base, using the fuzzy model, optimal linguistic variable parameters, and a teaching signal. the ga produces a near-optimal fnn. the near-optimal fnn can be improved using classical derivative-based optimization procedures. the fis structure found by the ga is optimized with a fitness function based on a response of the actual plant model of the controlled plant. the sc optimizer produces a robust kb that is typically smaller that the kb produced by prior art methods.",2007-07-31,"The title of the patent is intelligent robust control system for motorcycle using soft computing optimizer and its abstract is a soft computing (sc) optimizer for designing a knowledge base (kb) to be used in a control system for controlling a motorcycle is described. in one embodiment, a simulation model of the motorcycle and rider control is used. in one embodiment, the simulation model includes a feedforward rider model. the sc optimizer includes a fuzzy inference engine based on a fuzzy neural network (fnn). the sc optimizer provides fuzzy inference system (fis) structure selection, fis structure optimization method selection, and teaching signal selection and generation. the user selects a fuzzy model, including one or more of: the number of input and/or output variables; the type of fuzzy inference; and the preliminary type of membership functions. a genetic algorithm (ga) is used to optimize linguistic variable parameters and the input-output training patterns. a ga is also used to optimize the rule base, using the fuzzy model, optimal linguistic variable parameters, and a teaching signal. the ga produces a near-optimal fnn. the near-optimal fnn can be improved using classical derivative-based optimization procedures. the fis structure found by the ga is optimized with a fitness function based on a response of the actual plant model of the controlled plant. the sc optimizer produces a robust kb that is typically smaller that the kb produced by prior art methods. dated 2007-07-31"
7252090,selection of neurostimulator parameter configurations using neural network,"in general, the invention is directed to a technique for selection of parameter configurations for a neurostimulator using neural networks. the technique may be employed by a programming device to allow a clinician to select parameter configurations, and then program an implantable neurostimulator to deliver therapy using the selected parameter configurations. the parameter configurations may include one or more of a variety of parameters, such as electrode configurations defining electrode combinations and polarities for an electrode set implanted in a patient. the electrode set may be carried by one or more implanted leads that are electrically coupled to the neurostimulator. in operation, the programming device executes a parameter configuration search algorithm to guide the clinician in the selection of parameter configurations. the search algorithm relies on a neural network that identifies potential optimum parameter configurations.",2007-08-07,"The title of the patent is selection of neurostimulator parameter configurations using neural network and its abstract is in general, the invention is directed to a technique for selection of parameter configurations for a neurostimulator using neural networks. the technique may be employed by a programming device to allow a clinician to select parameter configurations, and then program an implantable neurostimulator to deliver therapy using the selected parameter configurations. the parameter configurations may include one or more of a variety of parameters, such as electrode configurations defining electrode combinations and polarities for an electrode set implanted in a patient. the electrode set may be carried by one or more implanted leads that are electrically coupled to the neurostimulator. in operation, the programming device executes a parameter configuration search algorithm to guide the clinician in the selection of parameter configurations. the search algorithm relies on a neural network that identifies potential optimum parameter configurations. dated 2007-08-07"
7254564,neural network based predication and optimization for groundwater/surface water system,"the present invention relates to a method and apparatus, based on the use of a neural network (nn), for (a) predicting important groundwater/surface water output/state variables, (b) optimizing groundwater/surface water control variables, and/or (c) sensitivity analysis, to identify physical relationships between input and output/state variables used to model the groundwater/surface water system or to analyze the performance parameters of the neural network.",2007-08-07,"The title of the patent is neural network based predication and optimization for groundwater/surface water system and its abstract is the present invention relates to a method and apparatus, based on the use of a neural network (nn), for (a) predicting important groundwater/surface water output/state variables, (b) optimizing groundwater/surface water control variables, and/or (c) sensitivity analysis, to identify physical relationships between input and output/state variables used to model the groundwater/surface water system or to analyze the performance parameters of the neural network. dated 2007-08-07"
7254565,method and circuits to virtually increase the number of prototypes in artificial neural networks,"an improved artificial neural network (ann) is disclosed that comprises a conventional ann, a database block, and a compare and update circuit. the conventional ann is formed by a plurality of neurons, each neuron having a prototype memory dedicated to store a prototype and a distance evaluator to evaluate the distance between the input pattern presented to the ann and the prototype stored therein. the database block has: all the prototypes arranged in slices, each slice being capable to store up to a maximum number of prototypes; the input patterns or queries to be presented to the ann; and the distances resulting of the evaluation performed during the recognition/classification phase. the compare and update circuit compares the distance with the distance previously found for the same input pattern updates or not the distance previously stored.",2007-08-07,"The title of the patent is method and circuits to virtually increase the number of prototypes in artificial neural networks and its abstract is an improved artificial neural network (ann) is disclosed that comprises a conventional ann, a database block, and a compare and update circuit. the conventional ann is formed by a plurality of neurons, each neuron having a prototype memory dedicated to store a prototype and a distance evaluator to evaluate the distance between the input pattern presented to the ann and the prototype stored therein. the database block has: all the prototypes arranged in slices, each slice being capable to store up to a maximum number of prototypes; the input patterns or queries to be presented to the ann; and the distances resulting of the evaluation performed during the recognition/classification phase. the compare and update circuit compares the distance with the distance previously found for the same input pattern updates or not the distance previously stored. dated 2007-08-07"
7255166,imbibition well stimulation via neural network design,a method for stimulation of hydrocarbon production via imbibition by utilization of surfactants. the method includes use of fuzzy logic and neural network architecture constructs to determine surfactant use.,2007-08-14,The title of the patent is imbibition well stimulation via neural network design and its abstract is a method for stimulation of hydrocarbon production via imbibition by utilization of surfactants. the method includes use of fuzzy logic and neural network architecture constructs to determine surfactant use. dated 2007-08-14
7256681,asset tracking using wireless lan infrastructure,transponders capable of providing identification information and possibly additional information are detected from wireless access points of a computer network as a substitute for closed radio frequency identification (rfid) systems while providing numerous additional functionalities and applications. total asset visibility or responses to more limited queries are provided by inclusion of a geographic information system software application. location reporting of proximity of devices/transponders to access points can be enhanced to a fine-grained level by triangulation or other algorithms including neural networks.,2007-08-14,The title of the patent is asset tracking using wireless lan infrastructure and its abstract is transponders capable of providing identification information and possibly additional information are detected from wireless access points of a computer network as a substitute for closed radio frequency identification (rfid) systems while providing numerous additional functionalities and applications. total asset visibility or responses to more limited queries are provided by inclusion of a geographic information system software application. location reporting of proximity of devices/transponders to access points can be enhanced to a fine-grained level by triangulation or other algorithms including neural networks. dated 2007-08-14
7257535,parametric speech codec for representing synthetic speech in the presence of background noise,"a system and method are provided for processing audio and speech signals using a pitch and voicing dependent spectral estimation algorithm (voicing algorithm) to accurately represent voiced speech, unvoiced speech, and mixed speech in the presence of background noise, and background noise with a single model. the present invention also modifies the synthesis model based on an estimate of the current input signal to improve the perceptual quality of the speech and background noise under a variety of input conditions. the present invention also improves the voicing dependent spectral estimation algorithm robustness by introducing the use of a multi-layer neural network in the estimation process. the voicing dependent spectral estimation algorithm provides an accurate and robust estimate of the voicing probability under a variety of background noise conditions. this is essential to providing high quality intelligible speech in the presence of background noise.",2007-08-14,"The title of the patent is parametric speech codec for representing synthetic speech in the presence of background noise and its abstract is a system and method are provided for processing audio and speech signals using a pitch and voicing dependent spectral estimation algorithm (voicing algorithm) to accurately represent voiced speech, unvoiced speech, and mixed speech in the presence of background noise, and background noise with a single model. the present invention also modifies the synthesis model based on an estimate of the current input signal to improve the perceptual quality of the speech and background noise under a variety of input conditions. the present invention also improves the voicing dependent spectral estimation algorithm robustness by introducing the use of a multi-layer neural network in the estimation process. the voicing dependent spectral estimation algorithm provides an accurate and robust estimate of the voicing probability under a variety of background noise conditions. this is essential to providing high quality intelligible speech in the presence of background noise. dated 2007-08-14"
7262414,thermal luminescence surface contamination detection system,"a thermal luminescent (tl) spectroscopy system and method for remote sensing and detection of surface chemical contamination involves irradiation of a target surface with energy from a near infrared pump beam, and measurement of tl liberated by that surface within a middle infrared (mir) region. fundamental molecular vibration modes of target contaminants present are briefly activated after the surface has been driven out of thermal equilibrium. an emissivity contrast between strata and target contaminant develops, peaks, and then subsides during a finite thermal window of detection opportunity in which detection of fingerprint identifiers for target contaminants is most probable. target contaminant identification employs neural network models trained and tested against known molecular absorption frequencies of target contaminants. the use of a pump beam that radiates energy outside the mir spectra of received tl reduces possible interference with the very weak mir signals given off by target contaminants.",2007-08-28,"The title of the patent is thermal luminescence surface contamination detection system and its abstract is a thermal luminescent (tl) spectroscopy system and method for remote sensing and detection of surface chemical contamination involves irradiation of a target surface with energy from a near infrared pump beam, and measurement of tl liberated by that surface within a middle infrared (mir) region. fundamental molecular vibration modes of target contaminants present are briefly activated after the surface has been driven out of thermal equilibrium. an emissivity contrast between strata and target contaminant develops, peaks, and then subsides during a finite thermal window of detection opportunity in which detection of fingerprint identifiers for target contaminants is most probable. target contaminant identification employs neural network models trained and tested against known molecular absorption frequencies of target contaminants. the use of a pump beam that radiates energy outside the mir spectra of received tl reduces possible interference with the very weak mir signals given off by target contaminants. dated 2007-08-28"
7266460,nox software sensor,"methods and apparatus for determining the nox content in a combustion flue gas. several parameters of the combustion process are measured in the combustion furnace. these parameters are then input to a neural network, which then, based upon the input data, generates output data representative of the concentration of nox in the flue gas. this result is determined without the use of physical nox sensors.",2007-09-04,"The title of the patent is nox software sensor and its abstract is methods and apparatus for determining the nox content in a combustion flue gas. several parameters of the combustion process are measured in the combustion furnace. these parameters are then input to a neural network, which then, based upon the input data, generates output data representative of the concentration of nox in the flue gas. this result is determined without the use of physical nox sensors. dated 2007-09-04"
7266532,reconfigurable autonomous device networks,"the present invention describes the use of autonomous devices, which can be arranged in networks, such as neural networks, to better identify, track, and acquire sources of signals present in an environment. the environment may be a physical environment, such as a battlefield, or a more abstract environment, such as a communication network. the devices may be mobile, in the form of vehicles with sensors, or may be information agents, and may also interact with one another, thus allowing for a great deal of flexibility in carrying out a task. in some cases, the devices may be in the form of autonomous vehicles which can collaboratively sense, identify, or classify a number of sources or targets concurrently. the autonomous devices may function as mobile agents or attractors in a network, such as a neural network. the devices may also be aggregated to form a network of networks and provide scalability to a system in which the autonomous devices are operating.",2007-09-04,"The title of the patent is reconfigurable autonomous device networks and its abstract is the present invention describes the use of autonomous devices, which can be arranged in networks, such as neural networks, to better identify, track, and acquire sources of signals present in an environment. the environment may be a physical environment, such as a battlefield, or a more abstract environment, such as a communication network. the devices may be mobile, in the form of vehicles with sensors, or may be information agents, and may also interact with one another, thus allowing for a great deal of flexibility in carrying out a task. in some cases, the devices may be in the form of autonomous vehicles which can collaboratively sense, identify, or classify a number of sources or targets concurrently. the autonomous devices may function as mobile agents or attractors in a network, such as a neural network. the devices may also be aggregated to form a network of networks and provide scalability to a system in which the autonomous devices are operating. dated 2007-09-04"
7267439,"optometric apparatus, optometric method, and optometric server","an optometric apparatus and an optometric method includes the steps of acquiring subject's attributes and an orientation selected by the subject on an astigmatic axis determination chart displayed on the computer screen, displaying vision measurement charts in the acquired orientation and the orientation perpendicular thereto to acquire visual recognition limits selected by the subject, calculating far point distances based on the acquired visual recognition limits and the acquired subject's attributes, and calculating a refractive power based on the acquired orientation and the calculated two far point distances. the far point distance is calculated using a neural network that has been determined by a number of subjects in advance. the astigmatic axis determination chart has four groups of a plurality of parallel lines, each group having lines arranged in their respective orientation, and the vision measurement chart has a plurality of light and dark line images of different sizes, thereby reducing the risk of presenting an erroneous refractive power.",2007-09-11,"The title of the patent is optometric apparatus, optometric method, and optometric server and its abstract is an optometric apparatus and an optometric method includes the steps of acquiring subject's attributes and an orientation selected by the subject on an astigmatic axis determination chart displayed on the computer screen, displaying vision measurement charts in the acquired orientation and the orientation perpendicular thereto to acquire visual recognition limits selected by the subject, calculating far point distances based on the acquired visual recognition limits and the acquired subject's attributes, and calculating a refractive power based on the acquired orientation and the calculated two far point distances. the far point distance is calculated using a neural network that has been determined by a number of subjects in advance. the astigmatic axis determination chart has four groups of a plurality of parallel lines, each group having lines arranged in their respective orientation, and the vision measurement chart has a plurality of light and dark line images of different sizes, thereby reducing the risk of presenting an erroneous refractive power. dated 2007-09-11"
7272261,method and system for classifying scanned-media,"a method for automatically classifying a printed image, includes scanning the printed image; selecting an n by n block of pixels from the scanned image; calculating an array of dct coefficients of the pixel block, wherein the array of calculated dct coefficients are representative of spatial frequency and spatial orientation of the pixel block; comparing the dct coefficients with an array of predetermined values, wherein the array of predetermined values are indicative of different image marking processes used to produce printed images; and determining an image marking process used to create the printed image based on the comparison of the dct coefficients with the array of predetermined values. the array of dct coefficients may be sampled into a feature set and the feature set provided to a neural network to output the determined image marking process.",2007-09-18,"The title of the patent is method and system for classifying scanned-media and its abstract is a method for automatically classifying a printed image, includes scanning the printed image; selecting an n by n block of pixels from the scanned image; calculating an array of dct coefficients of the pixel block, wherein the array of calculated dct coefficients are representative of spatial frequency and spatial orientation of the pixel block; comparing the dct coefficients with an array of predetermined values, wherein the array of predetermined values are indicative of different image marking processes used to produce printed images; and determining an image marking process used to create the printed image based on the comparison of the dct coefficients with the array of predetermined values. the array of dct coefficients may be sampled into a feature set and the feature set provided to a neural network to output the determined image marking process. dated 2007-09-18"
7272515,digital signal processor implementation of high impedance fault algorithms,"a digital signal processor implementation of three algorithms used to detect high impedance faults. the algorithms can be wavelet based, higher order statistics based and neural network based. the algorithms are modified to process one second of data instead of ten seconds of data and a double buffered acquisition is connected to the output of the algorithms.",2007-09-18,"The title of the patent is digital signal processor implementation of high impedance fault algorithms and its abstract is a digital signal processor implementation of three algorithms used to detect high impedance faults. the algorithms can be wavelet based, higher order statistics based and neural network based. the algorithms are modified to process one second of data instead of ten seconds of data and a double buffered acquisition is connected to the output of the algorithms. dated 2007-09-18"
7274992,method for predicting pore pressure,"a method for predicting pore pressure is described which initially involves obtaining, for one point in a volume of earth, a value of pore pressure and one or more seismic attributes. a relationship is determined between the value of pore pressure and the seismic attribute, for example by use of a neural network. seismic data is then obtained for the volume of earth and the same attributes as selected above when determining relationship are extracted from the seismic data for another point in the volume. the extracted seismic attributes are then used as inputs to the previously determined relationship to produce as an output, a prediction of pore pressure at the other point. the seismic attributes are frequency related seismic attributes and include for example instantaneous frequency, weighted mean frequency, instantaneous pseudo-quality factor, instantaneous dominant frequency, instantaneous bandwidth, instantaneous phase, effective bandwidth, peak frequency, envelope and energy half time.",2007-09-25,"The title of the patent is method for predicting pore pressure and its abstract is a method for predicting pore pressure is described which initially involves obtaining, for one point in a volume of earth, a value of pore pressure and one or more seismic attributes. a relationship is determined between the value of pore pressure and the seismic attribute, for example by use of a neural network. seismic data is then obtained for the volume of earth and the same attributes as selected above when determining relationship are extracted from the seismic data for another point in the volume. the extracted seismic attributes are then used as inputs to the previously determined relationship to produce as an output, a prediction of pore pressure at the other point. the seismic attributes are frequency related seismic attributes and include for example instantaneous frequency, weighted mean frequency, instantaneous pseudo-quality factor, instantaneous dominant frequency, instantaneous bandwidth, instantaneous phase, effective bandwidth, peak frequency, envelope and energy half time. dated 2007-09-25"
7276031,system and method for classifying patient's breathing using artificial neural network,"described is a method and system for analyzing a patient's breaths. the arrangement may include a sensor and a processor. the sensor detects data corresponding to a patient's breathing patterns over a plurality of breaths. the processor separates the detected data into data segments corresponding to individual breaths. then, the processor analyzes the data segments using a pretrained artificial neural network to classify the breaths based on a likelihood that individual ones of the breaths include an abnormal flow limitation.",2007-10-02,"The title of the patent is system and method for classifying patient's breathing using artificial neural network and its abstract is described is a method and system for analyzing a patient's breaths. the arrangement may include a sensor and a processor. the sensor detects data corresponding to a patient's breathing patterns over a plurality of breaths. the processor separates the detected data into data segments corresponding to individual breaths. then, the processor analyzes the data segments using a pretrained artificial neural network to classify the breaths based on a likelihood that individual ones of the breaths include an abnormal flow limitation. dated 2007-10-02"
7277823,method and system of monitoring and prognostics,a neural network learns the operating modes of a system being monitored under normal operating conditions. anomalies can be automatically detected and learned. a control command can be issued or an alert can be issued in response thereto.,2007-10-02,The title of the patent is method and system of monitoring and prognostics and its abstract is a neural network learns the operating modes of a system being monitored under normal operating conditions. anomalies can be automatically detected and learned. a control command can be issued or an alert can be issued in response thereto. dated 2007-10-02
7277831,method for detecting time dependent modes of dynamic systems,"in a method for detecting the modes of a dynamic system with a large number of modes that each have a set α (t) of characteristic system parameters, a time series of at least one system variable x(t) is subjected to modeling, for example switch segmentation, so that in each time segment of a predetermined minimum length a predetermined prediction model, for example a neural network, for a system mode is detected for each system variable x(t), whereby modeling of the time series is followed by drift segmentation in which, in each time segment in which there is transition of the system from a first system mode to a second system mode, a series of mixed prediction models is detected produced by linear, paired superimposition of the prediction models of the two system modes.",2007-10-02,"The title of the patent is method for detecting time dependent modes of dynamic systems and its abstract is in a method for detecting the modes of a dynamic system with a large number of modes that each have a set α (t) of characteristic system parameters, a time series of at least one system variable x(t) is subjected to modeling, for example switch segmentation, so that in each time segment of a predetermined minimum length a predetermined prediction model, for example a neural network, for a system mode is detected for each system variable x(t), whereby modeling of the time series is followed by drift segmentation in which, in each time segment in which there is transition of the system from a first system mode to a second system mode, a series of mixed prediction models is detected produced by linear, paired superimposition of the prediction models of the two system modes. dated 2007-10-02"
7280214,method and apparatus for a high resolution downhole spectrometer,"the present invention provides a simple, robust, and versatile high-resolution spectrometer that is suitable for downhole use. the present invention provides a method and apparatus incorporating a spinning, oscillating or stepping optical interference filter to change the angle at which light passes through the filters after passing through a sample under analysis downhole. as each filter is tilted, the color or wavelength of light passed by the filter changes. black plates are placed between the filters to isolate each filter's photodiode. the spectrometer of the present invention is suitable for use with a wire line formation tester, such as the baker atlas reservation characterization instrument to provide supplemental analysis and monitoring of sample clean up. the present invention is also suitable for deployment in a monitoring while drilling environment. the present invention provides a high resolution spectometer which enables quantification of a crude oil's percentage of aromatics, olefins, and saturates to estimate a sample's gas oil ratio (gor). gases such as co2 are also detectable. the percentage of oil-based mud filtrate contamination in a crude oil sample can be estimated with the present invention by using a suitable training set and chemometrics, a neural network, or other type of correlation method.",2007-10-09,"The title of the patent is method and apparatus for a high resolution downhole spectrometer and its abstract is the present invention provides a simple, robust, and versatile high-resolution spectrometer that is suitable for downhole use. the present invention provides a method and apparatus incorporating a spinning, oscillating or stepping optical interference filter to change the angle at which light passes through the filters after passing through a sample under analysis downhole. as each filter is tilted, the color or wavelength of light passed by the filter changes. black plates are placed between the filters to isolate each filter's photodiode. the spectrometer of the present invention is suitable for use with a wire line formation tester, such as the baker atlas reservation characterization instrument to provide supplemental analysis and monitoring of sample clean up. the present invention is also suitable for deployment in a monitoring while drilling environment. the present invention provides a high resolution spectometer which enables quantification of a crude oil's percentage of aromatics, olefins, and saturates to estimate a sample's gas oil ratio (gor). gases such as co2 are also detectable. the percentage of oil-based mud filtrate contamination in a crude oil sample can be estimated with the present invention by using a suitable training set and chemometrics, a neural network, or other type of correlation method. dated 2007-10-09"
7280987,genetic algorithm based selection of neural network ensemble for processing well logging data,"a system and method for generating a neural network ensemble. conventional algorithms are used to train a number of neural networks having error diversity, for example by having a different number of hidden nodes in each network. a genetic algorithm having a multi-objective fitness function is used to select one or more ensembles. the fitness function includes a negative error correlation objective to insure diversity among the ensemble members. a genetic algorithm may be used to select weighting factors for the multi-objective function. in one application, a trained model may be used to produce synthetic open hole logs in response to inputs of cased hole log data.",2007-10-09,"The title of the patent is genetic algorithm based selection of neural network ensemble for processing well logging data and its abstract is a system and method for generating a neural network ensemble. conventional algorithms are used to train a number of neural networks having error diversity, for example by having a different number of hidden nodes in each network. a genetic algorithm having a multi-objective fitness function is used to select one or more ensembles. the fitness function includes a negative error correlation objective to insure diversity among the ensemble members. a genetic algorithm may be used to select weighting factors for the multi-objective function. in one application, a trained model may be used to produce synthetic open hole logs in response to inputs of cased hole log data. dated 2007-10-09"
7280989,phase-locked loop oscillatory neurocomputer,"a neural network computer (20) includes a weighting network (21) coupled to a plurality of phase-locked loop circuits (251-25n). the weighting network (21) has a plurality of weighting circuits (c11, . . . , cnn) having output terminals connected to a plurality of adder circuits (311-31n). a single weighting element (ckj) typically has a plurality of output terminals coupled to a corresponding adder circuit (31k). each adder circuit (31k) is coupled to a corresponding bandpass filter circuit (31k) which is in turn coupled to a corresponding phase-locked loop circuit (25k). the weighting elements (c1,1, . . . , cn,n) are programmed with connection strengths, wherein the connection strengths have phase-encoded weights. the phase relationships are used to recognize an incoming pattern.",2007-10-09,"The title of the patent is phase-locked loop oscillatory neurocomputer and its abstract is a neural network computer (20) includes a weighting network (21) coupled to a plurality of phase-locked loop circuits (251-25n). the weighting network (21) has a plurality of weighting circuits (c11, . . . , cnn) having output terminals connected to a plurality of adder circuits (311-31n). a single weighting element (ckj) typically has a plurality of output terminals coupled to a corresponding adder circuit (31k). each adder circuit (31k) is coupled to a corresponding bandpass filter circuit (31k) which is in turn coupled to a corresponding phase-locked loop circuit (25k). the weighting elements (c1,1, . . . , cn,n) are programmed with connection strengths, wherein the connection strengths have phase-encoded weights. the phase relationships are used to recognize an incoming pattern. dated 2007-10-09"
7284769,method and apparatus for sensing a vehicle crash,"method and apparatus for sensing a crash of a vehicle in which acceleration and angular motion of the vehicle are measured and analyzed by a processor to determine whether a crash is occurring. a deployable occupant protection device is deployed by a processor when deployment criteria are satisfied. the processor may embody a pattern recognition technique for analyzing the measured acceleration and angular motion of the vehicle and determine whether deployment of the occupant protection device is beneficial. for example, the pattern recognition technique may be a neural network trained to determine whether the vehicle is experiencing a crash such as a rollover based on the measured acceleration and angular motion of the vehicle.",2007-10-23,"The title of the patent is method and apparatus for sensing a vehicle crash and its abstract is method and apparatus for sensing a crash of a vehicle in which acceleration and angular motion of the vehicle are measured and analyzed by a processor to determine whether a crash is occurring. a deployable occupant protection device is deployed by a processor when deployment criteria are satisfied. the processor may embody a pattern recognition technique for analyzing the measured acceleration and angular motion of the vehicle and determine whether deployment of the occupant protection device is beneficial. for example, the pattern recognition technique may be a neural network trained to determine whether the vehicle is experiencing a crash such as a rollover based on the measured acceleration and angular motion of the vehicle. dated 2007-10-23"
7286699,system and method facilitating pattern recognition,a system and method facilitating pattern recognition is provided. the invention includes a pattern recognition system having a convolutional neural network employing feature extraction layer(s) and classifier layer(s). the feature extraction layer(s) comprises convolutional layers and the classifier layer(s) comprises fully connected layers. the pattern recognition system can be trained utilizing a calculated cross entropy error. the calculated cross entropy error is utilized to update trainable parameters of the pattern recognition system.,2007-10-23,The title of the patent is system and method facilitating pattern recognition and its abstract is a system and method facilitating pattern recognition is provided. the invention includes a pattern recognition system having a convolutional neural network employing feature extraction layer(s) and classifier layer(s). the feature extraction layer(s) comprises convolutional layers and the classifier layer(s) comprises fully connected layers. the pattern recognition system can be trained utilizing a calculated cross entropy error. the calculated cross entropy error is utilized to update trainable parameters of the pattern recognition system. dated 2007-10-23
7286964,methods for monitoring structural health conditions,"methods and recordable media for monitoring structural health conditions. the present invention provides a method for interrogating a damage of a host structure using a diagnostic network patch (dnp) system having patches. an interrogation module partitions the plurality of patched in subgroups and measures the sensor signals generated and received by actuator and sensor patches, respectively. then, a process module loads sensor signal data to identify lamb wave modes, determine the time of arrival of the modes and generate a tomographic image. it also determines distribution of other structural condition indices to generate tomographic images of the host structure. a set of tomographic images can be stacked to generate a hyperspectral tomography cube. a classification module generates codebook based on k-mean/learning vector quantization algorithm and uses a neural-fuzzy-inference system to determine the type of damages of the host structure.",2007-10-23,"The title of the patent is methods for monitoring structural health conditions and its abstract is methods and recordable media for monitoring structural health conditions. the present invention provides a method for interrogating a damage of a host structure using a diagnostic network patch (dnp) system having patches. an interrogation module partitions the plurality of patched in subgroups and measures the sensor signals generated and received by actuator and sensor patches, respectively. then, a process module loads sensor signal data to identify lamb wave modes, determine the time of arrival of the modes and generate a tomographic image. it also determines distribution of other structural condition indices to generate tomographic images of the host structure. a set of tomographic images can be stacked to generate a hyperspectral tomography cube. a classification module generates codebook based on k-mean/learning vector quantization algorithm and uses a neural-fuzzy-inference system to determine the type of damages of the host structure. dated 2007-10-23"
7287014,plausible neural network with supervised and unsupervised cluster analysis,"a plausible neural network (plann) is an artificial neural network with weight connection given by mutual information, which has the capability of inference and learning, and yet retains many characteristics of a biological neural network. the learning algorithm is based on statistical estimation, which is faster than the gradient decent approach currently used. the network after training becomes a fuzzy/belief network; the inference and weight are exchangeable, and as a result, knowledge extraction becomes simple. plann performs associative memory, supervised, semi-supervised, unsupervised learning and function/relation approximation in a single network architecture. this network architecture can easily be implemented by analog vlsi circuit design.",2007-10-23,"The title of the patent is plausible neural network with supervised and unsupervised cluster analysis and its abstract is a plausible neural network (plann) is an artificial neural network with weight connection given by mutual information, which has the capability of inference and learning, and yet retains many characteristics of a biological neural network. the learning algorithm is based on statistical estimation, which is faster than the gradient decent approach currently used. the network after training becomes a fuzzy/belief network; the inference and weight are exchangeable, and as a result, knowledge extraction becomes simple. plann performs associative memory, supervised, semi-supervised, unsupervised learning and function/relation approximation in a single network architecture. this network architecture can easily be implemented by analog vlsi circuit design. dated 2007-10-23"
7293001,hybrid neural network and support vector machine method for optimization,"system and method for optimization of a design associated with a response function, using a hybrid neural net and support vector machine (nn/svm) analysis to minimize or maximize an objective function, optionally subject to one or more constraints. as a first example, the nn/svm analysis is applied iteratively to design of an aerodynamic component, such as an airfoil shape, where the objective function measures deviation from a target pressure distribution on the perimeter of the aerodynamic component. as a second example, the nn/svm analysis is applied to data classification of a sequence of data points in a multidimensional space. the nn/svm analysis is also applied to data regression.",2007-11-06,"The title of the patent is hybrid neural network and support vector machine method for optimization and its abstract is system and method for optimization of a design associated with a response function, using a hybrid neural net and support vector machine (nn/svm) analysis to minimize or maximize an objective function, optionally subject to one or more constraints. as a first example, the nn/svm analysis is applied iteratively to design of an aerodynamic component, such as an airfoil shape, where the objective function measures deviation from a target pressure distribution on the perimeter of the aerodynamic component. as a second example, the nn/svm analysis is applied to data classification of a sequence of data points in a multidimensional space. the nn/svm analysis is also applied to data regression. dated 2007-11-06"
7293002,self-organizing data driven learning hardware with local interconnections,"a method for organizing processors to perform artificial neural network tasks is provided. the method provides a computer executable methodology for organizing processors in a self-organizing, data driven, learning hardware with local interconnections. a training data is processed substantially in parallel by the locally interconnected processors. the local processors determine local interconnections between the processors based on the training data. the local processors then determine, substantially in parallel, transformation functions and/or entropy based thresholds for the processors based on the training data.",2007-11-06,"The title of the patent is self-organizing data driven learning hardware with local interconnections and its abstract is a method for organizing processors to perform artificial neural network tasks is provided. the method provides a computer executable methodology for organizing processors in a self-organizing, data driven, learning hardware with local interconnections. a training data is processed substantially in parallel by the locally interconnected processors. the local processors determine local interconnections between the processors based on the training data. the local processors then determine, substantially in parallel, transformation functions and/or entropy based thresholds for the processors based on the training data. dated 2007-11-06"
7295687,face recognition method using artificial neural network and apparatus thereof,"a face recognition method using artificial neural network and an apparatus thereof are provided. the apparatus comprises an eigenpaxel selection unit which generates eigenpaxels indicating characteristic patterns of a face and selects a predetermined number of eigenpaxels among the generated eigenpaxels; an eigenfiltering unit which filters an input image with the selected eigenpaxels; a predetermined number of neural networks, each of which corresponds to one of the selected eigenpaxels, receives an image signal which is filtered by the corresponding eigenpaxel, and output a face recognition result; and a determination unit which receives the recognition result from each of the neural networks and outputs a final face recognition result of the input image.",2007-11-13,"The title of the patent is face recognition method using artificial neural network and apparatus thereof and its abstract is a face recognition method using artificial neural network and an apparatus thereof are provided. the apparatus comprises an eigenpaxel selection unit which generates eigenpaxels indicating characteristic patterns of a face and selects a predetermined number of eigenpaxels among the generated eigenpaxels; an eigenfiltering unit which filters an input image with the selected eigenpaxels; a predetermined number of neural networks, each of which corresponds to one of the selected eigenpaxels, receives an image signal which is filtered by the corresponding eigenpaxel, and output a face recognition result; and a determination unit which receives the recognition result from each of the neural networks and outputs a final face recognition result of the input image. dated 2007-11-13"
7295977,extracting classifying data in music from an audio bitstream,"the method of the present invention utilizes machine-learning techniques, particularly support vector machines in combination with a neural network, to process a unique machine-learning enabled representation of the audio bitstream. using this method, a classifying machine is able to autonomously detect characteristics of a piece of music, such as the artist or genre, and classify it accordingly. the method includes transforming digital time-domain representation of music into a frequency-domain representation, then dividing that frequency data into time slices, and compressing it into frequency bands to form multiple learning representations of each song. the learning representations that result are processed by a group of support vector machines, then by a neural network, both previously trained to distinguish among a given set of characteristics, to determine the classification.",2007-11-13,"The title of the patent is extracting classifying data in music from an audio bitstream and its abstract is the method of the present invention utilizes machine-learning techniques, particularly support vector machines in combination with a neural network, to process a unique machine-learning enabled representation of the audio bitstream. using this method, a classifying machine is able to autonomously detect characteristics of a piece of music, such as the artist or genre, and classify it accordingly. the method includes transforming digital time-domain representation of music into a frequency-domain representation, then dividing that frequency data into time slices, and compressing it into frequency bands to form multiple learning representations of each song. the learning representations that result are processed by a group of support vector machines, then by a neural network, both previously trained to distinguish among a given set of characteristics, to determine the classification. dated 2007-11-13"
7296009,search system,"a search engine and system for data, such as internet web pages, including a query analyser for processing a query to assign respective weights to terms of the query and to generate a query vector including the weights, and an index network responsive to the query vector to output at least one index to data in response to the query. the index network is a self-generating neural network built using training examples derived from a feature extractor. the feature extractor is used during both the search and training phase. a clusterer is used to group search results.",2007-11-13,"The title of the patent is search system and its abstract is a search engine and system for data, such as internet web pages, including a query analyser for processing a query to assign respective weights to terms of the query and to generate a query vector including the weights, and an index network responsive to the query vector to output at least one index to data in response to the query. the index network is a self-generating neural network built using training examples derived from a feature extractor. the feature extractor is used during both the search and training phase. a clusterer is used to group search results. dated 2007-11-13"
7297479,dna-based analog neural networks,"this invention is an oligomer-based analog neural network (ann) comprising weight and saturation oligomers, the concentrations of which are selected such that activation of the ann by a set of input oligomers generates a set of output oligomers, the sequences and relative concentrations of which are dependent on the sequences and relative concentrations of the input oligomers. the invention further includes methods for using such an ann for solving any problems amenable to solution by a trained neural network. a preferred embodiment of the claimed invention is a dna-based ann that accepts cdna molecules as inputs and analyzes the gene expression profile of the cells from which the cdna is derived. the dna-based ann is typically trained with a computer to identify the weights giving accurate mapping of the inputs to the outputs; and the concentrations of weight oligomers of the dna-based ann are then selected accordingly.",2007-11-20,"The title of the patent is dna-based analog neural networks and its abstract is this invention is an oligomer-based analog neural network (ann) comprising weight and saturation oligomers, the concentrations of which are selected such that activation of the ann by a set of input oligomers generates a set of output oligomers, the sequences and relative concentrations of which are dependent on the sequences and relative concentrations of the input oligomers. the invention further includes methods for using such an ann for solving any problems amenable to solution by a trained neural network. a preferred embodiment of the claimed invention is a dna-based ann that accepts cdna molecules as inputs and analyzes the gene expression profile of the cells from which the cdna is derived. the dna-based ann is typically trained with a computer to identify the weights giving accurate mapping of the inputs to the outputs; and the concentrations of weight oligomers of the dna-based ann are then selected accordingly. dated 2007-11-20"
7297860,system and method for determining genre of audio,"both the beat and genre of music are used to select or generate a dance robot object for display of a robot icon on a computer monitor or for establishing the movements of a three-dimensional robot. the detected genre can define the type of dance performed by the robot. also, the detected genre can be used to sort music. the genre can be detected using a neural network or by correlating the compressibility of the music to a genre.",2007-11-20,"The title of the patent is system and method for determining genre of audio and its abstract is both the beat and genre of music are used to select or generate a dance robot object for display of a robot icon on a computer monitor or for establishing the movements of a three-dimensional robot. the detected genre can define the type of dance performed by the robot. also, the detected genre can be used to sort music. the genre can be detected using a neural network or by correlating the compressibility of the music to a genre. dated 2007-11-20"
7302229,enabling desired wireless connectivity in a high frequency wireless local area network,"the present invention provides a method and an apparatus to enable desired wireless connectivity in a high frequency and/or high speed wireless local area network for providing mobile communications to a user of a wireless communication device which may be otherwise unable to connect to the high frequency and/or speed wireless local area network in response to a request for a wireless service. by using a chipset disposed in another wireless communication device, a neural network may be formed for enabling such wireless connectivity. in one embodiment, availability of wireless connectivity may be determined to a first user of a wireless service at a first wireless communication device to communicate with an access point associated with a wi-fi wireless network that offers the wireless service. absent such wireless connectivity at the first wireless communication device, additional bandwidth available at a second wireless communication device may be used to connect the first user at the first wireless communication device over another network that offers the wireless service, for example, a wide area network capable of communicating mobile or cellular data. accordingly, a user that desires use of a wireless service in a wireless network of sparsely populated wi-fi access points may obtain desired wireless connectivity for mobile communications across a relatively longer range and/or at much higher transfer speeds than otherwise available.",2007-11-27,"The title of the patent is enabling desired wireless connectivity in a high frequency wireless local area network and its abstract is the present invention provides a method and an apparatus to enable desired wireless connectivity in a high frequency and/or high speed wireless local area network for providing mobile communications to a user of a wireless communication device which may be otherwise unable to connect to the high frequency and/or speed wireless local area network in response to a request for a wireless service. by using a chipset disposed in another wireless communication device, a neural network may be formed for enabling such wireless connectivity. in one embodiment, availability of wireless connectivity may be determined to a first user of a wireless service at a first wireless communication device to communicate with an access point associated with a wi-fi wireless network that offers the wireless service. absent such wireless connectivity at the first wireless communication device, additional bandwidth available at a second wireless communication device may be used to connect the first user at the first wireless communication device over another network that offers the wireless service, for example, a wide area network capable of communicating mobile or cellular data. accordingly, a user that desires use of a wireless service in a wireless network of sparsely populated wi-fi access points may obtain desired wireless connectivity for mobile communications across a relatively longer range and/or at much higher transfer speeds than otherwise available. dated 2007-11-27"
7305369,method and apparatus for producing three dimensional shapes,a method and system for automatically producing data representative of a modified head shape from data representative of a deformed head is provided. the method includes a step of processing captured data for the deformed head utilizing principal component analysis (pca) to generate pca data representative of the deformed head. the method also includes the steps of providing the pca data as input to a neural network; and utilizing the neural network to process the pca data to provide data representative of a corresponding modified head shape.,2007-12-04,The title of the patent is method and apparatus for producing three dimensional shapes and its abstract is a method and system for automatically producing data representative of a modified head shape from data representative of a deformed head is provided. the method includes a step of processing captured data for the deformed head utilizing principal component analysis (pca) to generate pca data representative of the deformed head. the method also includes the steps of providing the pca data as input to a neural network; and utilizing the neural network to process the pca data to provide data representative of a corresponding modified head shape. dated 2007-12-04
7305370,neural cortex,a neural network system includes a random access memory (ram); and an index-based weightless neural network with a columnar topography; wherein patterns of binary connections and values of output nodes' activities are stored in the ram. information is processed by pattern recognition using the neural network by storing a plurality of output patterns to be recognized in a pattern index; accepting an input pattern and dividing the input pattern into a plurality of components; and processing each component according to the pattern index to identify a recognized output pattern corresponding to the input pattern.,2007-12-04,The title of the patent is neural cortex and its abstract is a neural network system includes a random access memory (ram); and an index-based weightless neural network with a columnar topography; wherein patterns of binary connections and values of output nodes' activities are stored in the ram. information is processed by pattern recognition using the neural network by storing a plurality of output patterns to be recognized in a pattern index; accepting an input pattern and dividing the input pattern into a plurality of components; and processing each component according to the pattern index to identify a recognized output pattern corresponding to the input pattern. dated 2007-12-04
7308134,pattern recognition with hierarchical networks,"within the frameworks of hierarchical neural feed-forward architectures for performing real-world 3d invariant object recognition a technique is proposed that shares components like weight-sharing (2), and pooling stages (3, 5) with earlier approaches, but focuses on new methods for determining optimal feature-detecting units in intermediate stages (4) of the hierarchical network. a new approach for training the hierarchical network is proposed which uses statistical means for (incrementally) learning new feature detection stages and significantly reduces the training effort for complex pattern recognition tasks, compared to the prior art. the incremental learning is based on detecting increasingly statistically independent features in higher stages of the processing hierarchy. since this learning is unsupervised, no teacher signal is necessary and the recognition architecture can be pre-structured for a certain recognition scenario. only a final classification step must be trained with supervised learning, which reduces significantly the effort for adaptation to a recognition task.",2007-12-11,"The title of the patent is pattern recognition with hierarchical networks and its abstract is within the frameworks of hierarchical neural feed-forward architectures for performing real-world 3d invariant object recognition a technique is proposed that shares components like weight-sharing (2), and pooling stages (3, 5) with earlier approaches, but focuses on new methods for determining optimal feature-detecting units in intermediate stages (4) of the hierarchical network. a new approach for training the hierarchical network is proposed which uses statistical means for (incrementally) learning new feature detection stages and significantly reduces the training effort for complex pattern recognition tasks, compared to the prior art. the incremental learning is based on detecting increasingly statistically independent features in higher stages of the processing hierarchy. since this learning is unsupervised, no teacher signal is necessary and the recognition architecture can be pre-structured for a certain recognition scenario. only a final classification step must be trained with supervised learning, which reduces significantly the effort for adaptation to a recognition task. dated 2007-12-11"
7308322,"motorized system integrated control and diagnostics using vibration, pressure, temperature, speed, and/or current analysis","systems and methods are disclosed for controlling, diagnosing and prognosing the health of a motorized system. the systems may comprise a diagnostics system, a prognostic system and a controller, wherein the diagnostics system and/or prognostic system employs a neural network, an expert system, and/or a data fusion component in order to assess and/or prognose the health of the motorized system according to one or more attributes associated therewith. the controller may operate the motorized system in accordance with a setpoint and/or a diagnostics signal from the diagnostics system and/or prognostic information. also disclosed are methodologies for controlling, diagnosing and prognosing the health of a motorized system, comprising operating a motor in the motorized system in a controlled fashion, diagnosing and/or prognosing the health of the motorized system according to a measured attribute associated with the motorized system, wherein the motor may be operated according to a setpoint and/or the diagnostics signal and/or prognosis.",2007-12-11,"The title of the patent is motorized system integrated control and diagnostics using vibration, pressure, temperature, speed, and/or current analysis and its abstract is systems and methods are disclosed for controlling, diagnosing and prognosing the health of a motorized system. the systems may comprise a diagnostics system, a prognostic system and a controller, wherein the diagnostics system and/or prognostic system employs a neural network, an expert system, and/or a data fusion component in order to assess and/or prognose the health of the motorized system according to one or more attributes associated therewith. the controller may operate the motorized system in accordance with a setpoint and/or a diagnostics signal from the diagnostics system and/or prognostic information. also disclosed are methodologies for controlling, diagnosing and prognosing the health of a motorized system, comprising operating a motor in the motorized system in a controlled fashion, diagnosing and/or prognosing the health of the motorized system according to a measured attribute associated with the motorized system, wherein the motor may be operated according to a setpoint and/or the diagnostics signal and/or prognosis. dated 2007-12-11"
7308432,vehicle motion model generating device and method for generating vehicle motion model,"there are equipped a first recurrent neural network formed by connecting plural nodes so as to have a loop in which the output from one node is input to another node in accordance with a predetermined coupling weight coefficient. meanwhile, the output of at least one node is fed back to the node concerned or another node, and an optimizing unit for determining the optimum solution of the coupling weight coefficient in the first recurrent neural network based on a learning rule using a hereditary algorithm. in this case, the first recurrent neural network outputs a first parameter indicating a motion state of a vehicle based on predetermined input information, thereby functioning as a vehicle motion model.",2007-12-11,"The title of the patent is vehicle motion model generating device and method for generating vehicle motion model and its abstract is there are equipped a first recurrent neural network formed by connecting plural nodes so as to have a loop in which the output from one node is input to another node in accordance with a predetermined coupling weight coefficient. meanwhile, the output of at least one node is fed back to the node concerned or another node, and an optimizing unit for determining the optimum solution of the coupling weight coefficient in the first recurrent neural network based on a learning rule using a hereditary algorithm. in this case, the first recurrent neural network outputs a first parameter indicating a motion state of a vehicle based on predetermined input information, thereby functioning as a vehicle motion model. dated 2007-12-11"
7313550,performance of artificial neural network models in the presence of instrumental noise and measurement errors,"a method is described for improving the prediction accuracy and generalization performance of artificial neural network models in presence of input-output example data containing instrumental noise and/or measurement errors, the presence of noise and/or errors in the input-output example data used for training the network models create difficulties in learning accurately the nonlinear relationships existing between the inputs and the outputs, to effectively learn the noisy relationships, the methodology envisages creation of a large-sized noise-superimposed sample input-output dataset using computer simulations, here, a specific amount of gaussian noise is added to each input/output variable in the example set and the enlarged sample data set created thereby is used as the training set for constructing the artificial neural network model, the amount of noise to be added is specific to an input/output variable and its optimal value is determined using a stochastic search and optimization technique, namely, genetic algorithms, the network trained on the noise-superimposed enlarged training set shows significant improvements in its prediction accuracy and generalization performance, the invented methodology is illustrated by its successful application to the example data comprising instrumental errors and/or measurement noise from an industrial polymerization reactor and a continuous stirred tank reactor (cstr).",2007-12-25,"The title of the patent is performance of artificial neural network models in the presence of instrumental noise and measurement errors and its abstract is a method is described for improving the prediction accuracy and generalization performance of artificial neural network models in presence of input-output example data containing instrumental noise and/or measurement errors, the presence of noise and/or errors in the input-output example data used for training the network models create difficulties in learning accurately the nonlinear relationships existing between the inputs and the outputs, to effectively learn the noisy relationships, the methodology envisages creation of a large-sized noise-superimposed sample input-output dataset using computer simulations, here, a specific amount of gaussian noise is added to each input/output variable in the example set and the enlarged sample data set created thereby is used as the training set for constructing the artificial neural network model, the amount of noise to be added is specific to an input/output variable and its optimal value is determined using a stochastic search and optimization technique, namely, genetic algorithms, the network trained on the noise-superimposed enlarged training set shows significant improvements in its prediction accuracy and generalization performance, the invented methodology is illustrated by its successful application to the example data comprising instrumental errors and/or measurement noise from an industrial polymerization reactor and a continuous stirred tank reactor (cstr). dated 2007-12-25"
7321796,method and system for training a visual prosthesis,"a method for training a visual prosthesis includes presenting a non-visual reference stimulus corresponding to a reference image to a visual prosthesis patient. the visual prosthesis including a plurality of electrodes. training data sets are generated by presenting a series of stimulation patterns to the patient through the visual prosthesis. each stimulation pattern in the series, after the first, is determined at least in part on a previous subjective patient selection of a preferred stimulation pattern among stimulation patterns previously presented in the series and a fitness function optimization algorithm. the presented stimulation patterns and the selections of the patient are stored and presented to a neural network off-line to determine a vision solution.",2008-01-22,"The title of the patent is method and system for training a visual prosthesis and its abstract is a method for training a visual prosthesis includes presenting a non-visual reference stimulus corresponding to a reference image to a visual prosthesis patient. the visual prosthesis including a plurality of electrodes. training data sets are generated by presenting a series of stimulation patterns to the patient through the visual prosthesis. each stimulation pattern in the series, after the first, is determined at least in part on a previous subjective patient selection of a preferred stimulation pattern among stimulation patterns previously presented in the series and a fitness function optimization algorithm. the presented stimulation patterns and the selections of the patient are stored and presented to a neural network off-line to determine a vision solution. dated 2008-01-22"
7321809,methods and systems for analyzing engine unbalance conditions,"methods and systems for analyzing engine unbalance conditions are disclosed. in one embodiment, a method includes receiving vibrational data from a plurality of locations distributed over an engine and a surrounding engine support structure, and inputting the vibrational data into a neural network inverse model. the neural network inverse model establishes a relationship between the vibrational data and an unbalance condition of the engine, and outputs diagnostic information indicating the unbalance condition of the engine. in a further embodiment, a method further includes subjecting the vibrational data to a fast fourier transformation to extract a desired once per revolution vibrational data prior to input to the neural network inverse model.",2008-01-22,"The title of the patent is methods and systems for analyzing engine unbalance conditions and its abstract is methods and systems for analyzing engine unbalance conditions are disclosed. in one embodiment, a method includes receiving vibrational data from a plurality of locations distributed over an engine and a surrounding engine support structure, and inputting the vibrational data into a neural network inverse model. the neural network inverse model establishes a relationship between the vibrational data and an unbalance condition of the engine, and outputs diagnostic information indicating the unbalance condition of the engine. in a further embodiment, a method further includes subjecting the vibrational data to a fast fourier transformation to extract a desired once per revolution vibrational data prior to input to the neural network inverse model. dated 2008-01-22"
7321881,methods and systems for predicting occurrence of an event,"embodiments of the present invention are directed to methods and systems for training a neural network having weighted connections for classification of data, as well as embodiments corresponding to the use of such a neural network for the classification of data, including, for example, prediction of an event (e.g., disease). the method may include inputting input training data into the neural network, processing, by the neural network, the input training data to produce an output, determining an error between the output and a desired output corresponding to the input training data, rating the performance neural network using an objective function, wherein the objective function comprises a function c substantially in accordance with an approximation of the concordance index and adapting the weighted connections of the neural network based upon results of the objective function.",2008-01-22,"The title of the patent is methods and systems for predicting occurrence of an event and its abstract is embodiments of the present invention are directed to methods and systems for training a neural network having weighted connections for classification of data, as well as embodiments corresponding to the use of such a neural network for the classification of data, including, for example, prediction of an event (e.g., disease). the method may include inputting input training data into the neural network, processing, by the neural network, the input training data to produce an output, determining an error between the output and a desired output corresponding to the input training data, rating the performance neural network using an objective function, wherein the objective function comprises a function c substantially in accordance with an approximation of the concordance index and adapting the weighted connections of the neural network based upon results of the objective function. dated 2008-01-22"
7321882,method for supervised teaching of a recurrent artificial neural network,"a method for the supervised teaching of a recurrent neutral network (rnn) is disclosed. a typical embodiment of the method utilizes a large (50 units or more), randomly initialized rnn with a globally stable dynamics. during the training period, the output units of this rnn are teacher-forced to follow the desired output signal. during this period, activations from all hidden units are recorded. at the end of the teaching period, these recorded data are used as input for a method which computes new weights of those connections that feed into the output units. the method is distinguished from existing training methods for rnns through the following characteristics: (1) only the weights of connections to output units are changed by learning—existing methods for teaching recurrent networks adjust all network weights. (2) the internal dynamics of large networks are used as a “reservoir” of dynamical components which are not changed, but only newly combined by the learning procedure—existing methods use small networks, whose internal dynamics are themselves completely re-shaped through learning.",2008-01-22,"The title of the patent is method for supervised teaching of a recurrent artificial neural network and its abstract is a method for the supervised teaching of a recurrent neutral network (rnn) is disclosed. a typical embodiment of the method utilizes a large (50 units or more), randomly initialized rnn with a globally stable dynamics. during the training period, the output units of this rnn are teacher-forced to follow the desired output signal. during this period, activations from all hidden units are recorded. at the end of the teaching period, these recorded data are used as input for a method which computes new weights of those connections that feed into the output units. the method is distinguished from existing training methods for rnns through the following characteristics: (1) only the weights of connections to output units are changed by learning—existing methods for teaching recurrent networks adjust all network weights. (2) the internal dynamics of large networks are used as a “reservoir” of dynamical components which are not changed, but only newly combined by the learning procedure—existing methods use small networks, whose internal dynamics are themselves completely re-shaped through learning. dated 2008-01-22"
7324979,genetically adaptive neural network classification systems and methods,"genetically adaptive neural network systems and methods provide environmentally adaptable classification algorithms for use, among other things, in multi-static active sonar classification. classification training occurs in-situ with data acquired at the onset of data collection to improve the classification of sonar energy detections in difficult littoral environments. accordingly, in-situ training sets are developed while the training process is supervised and refined. candidate weights vectors evolve through genetic-based search procedures, and the fitness of candidate weight vectors is evaluated. feature vectors of interest may be classified using multiple neural networks and statistical averaging techniques to provide accurate and reliable signal classification.",2008-01-29,"The title of the patent is genetically adaptive neural network classification systems and methods and its abstract is genetically adaptive neural network systems and methods provide environmentally adaptable classification algorithms for use, among other things, in multi-static active sonar classification. classification training occurs in-situ with data acquired at the onset of data collection to improve the classification of sonar energy detections in difficult littoral environments. accordingly, in-situ training sets are developed while the training process is supervised and refined. candidate weights vectors evolve through genetic-based search procedures, and the fitness of candidate weight vectors is evaluated. feature vectors of interest may be classified using multiple neural networks and statistical averaging techniques to provide accurate and reliable signal classification. dated 2008-01-29"
7324980,information processing apparatus and method,"this invention relates to an information processing device and method that enable generation of an unlearned new pattern. data xt corresponding to a predetermined time series pattern is inputted to an input layer (11) of a recurrent neural network (1), and a prediction value x*t+1 is acquired from an output layer 13. a difference between teacher data xt+1 and the prediction value x*t+1 is learned by a back propagation method, and a weighting coefficient of an intermediate layer 12 is set at a predetermined value. after the recurrent neural network is caused to learn plural time series patterns, a parameter having a different value from the value in learning is inputted to parametric bias nodes (11-2), and an unlearned time series pattern corresponding to the parameter is generated from the output layer (13). this invention can be applied to a robot.",2008-01-29,"The title of the patent is information processing apparatus and method and its abstract is this invention relates to an information processing device and method that enable generation of an unlearned new pattern. data xt corresponding to a predetermined time series pattern is inputted to an input layer (11) of a recurrent neural network (1), and a prediction value x*t+1 is acquired from an output layer 13. a difference between teacher data xt+1 and the prediction value x*t+1 is learned by a back propagation method, and a weighting coefficient of an intermediate layer 12 is set at a predetermined value. after the recurrent neural network is caused to learn plural time series patterns, a parameter having a different value from the value in learning is inputted to parametric bias nodes (11-2), and an unlearned time series pattern corresponding to the parameter is generated from the output layer (13). this invention can be applied to a robot. dated 2008-01-29"
7328197,identifying a state of a data storage drive using an artificial neural network generated model,"the state or condition of a data storage drive, or a subsystem within a drive, may be evaluated by comparing a set of selected parameter values, converted into a trial vector, with a number of model or exemplar vectors, each of which was represents a particular state or condition of a sample drive. examples of such conditions may include “good”, “marginal”, “unacceptable”, “worn”, “defective”, or other general or specific conditions. sets of parameter values from the drive are converted into input vectors. unprocessed vectors are then processed against the input vectors in an artificial neural network to generate the exemplar vectors. the exemplar vectors are stored in a memory of an operational drive. during operation of the drive, the trial vector is compared with the exemplar vectors. the exemplar vector which is closest to the trial vector represents a state which most closely represents the current state of the drive. thus, a high similarity between the trial vector and an exemplar vector which represent a “good” drive is likely to have come from a “good” drive.",2008-02-05,"The title of the patent is identifying a state of a data storage drive using an artificial neural network generated model and its abstract is the state or condition of a data storage drive, or a subsystem within a drive, may be evaluated by comparing a set of selected parameter values, converted into a trial vector, with a number of model or exemplar vectors, each of which was represents a particular state or condition of a sample drive. examples of such conditions may include “good”, “marginal”, “unacceptable”, “worn”, “defective”, or other general or specific conditions. sets of parameter values from the drive are converted into input vectors. unprocessed vectors are then processed against the input vectors in an artificial neural network to generate the exemplar vectors. the exemplar vectors are stored in a memory of an operational drive. during operation of the drive, the trial vector is compared with the exemplar vectors. the exemplar vector which is closest to the trial vector represents a state which most closely represents the current state of the drive. thus, a high similarity between the trial vector and an exemplar vector which represent a “good” drive is likely to have come from a “good” drive. dated 2008-02-05"
7333850,maternal-fetal monitoring system,"a maternal-fetal monitoring system for use during all stages of pregnancy, including antepartum and intrapartum stages. the maternal-fetal monitoring system of the subject invention comprises (1) a set of sensors; (2) an amplifying/filtering means; (3) a computing means; and (4) a graphical user interface. accurate clinical data, which can be extracted and provided to the user in real-time using the system of the invention, include without limitation, maternal electrocardiogram (ecg) signals, maternal uterine activity signals (ehg), maternal heart rate, fetal ecg signals, and fetal heart rate. in a preferred embodiment, the maternal-fetal monitoring system of the invention includes an intelligence means, such as a neural network system, to analyze and interpret clinical data for use in clinical diagnosis antepartum, intrapartum and postpartum, as well as delivery strategy.",2008-02-19,"The title of the patent is maternal-fetal monitoring system and its abstract is a maternal-fetal monitoring system for use during all stages of pregnancy, including antepartum and intrapartum stages. the maternal-fetal monitoring system of the subject invention comprises (1) a set of sensors; (2) an amplifying/filtering means; (3) a computing means; and (4) a graphical user interface. accurate clinical data, which can be extracted and provided to the user in real-time using the system of the invention, include without limitation, maternal electrocardiogram (ecg) signals, maternal uterine activity signals (ehg), maternal heart rate, fetal ecg signals, and fetal heart rate. in a preferred embodiment, the maternal-fetal monitoring system of the invention includes an intelligence means, such as a neural network system, to analyze and interpret clinical data for use in clinical diagnosis antepartum, intrapartum and postpartum, as well as delivery strategy. dated 2008-02-19"
7333963,cognitive memory and auto-associative neural network based search engine for computer and network located images and photographs,"designs for cognitive memory systems storing input data, images, or patterns, and retrieving it without knowledge of where stored when cognitive memory is prompted by query pattern that is related to sought stored pattern. retrieval system of cognitive memory uses autoassociative neural networks and techniques for pre-processing query pattern to establish relationship between query pattern and sought stored pattern, to locate sought pattern, and to retrieve it and ancillary data. cognitive memory, when connected to computer or information appliance introduces computational architecture that applies to systems and methods for navigation, location and recognition of objects in images, character recognition, facial recognition, medical analysis and diagnosis, video image analysis, and to photographic search engines that when prompted with a query photograph containing faces and objects will retrieve related photographs stored in computer or other information appliance, and will identify url's of related photographs and documents stored on the world wide web.",2008-02-19,"The title of the patent is cognitive memory and auto-associative neural network based search engine for computer and network located images and photographs and its abstract is designs for cognitive memory systems storing input data, images, or patterns, and retrieving it without knowledge of where stored when cognitive memory is prompted by query pattern that is related to sought stored pattern. retrieval system of cognitive memory uses autoassociative neural networks and techniques for pre-processing query pattern to establish relationship between query pattern and sought stored pattern, to locate sought pattern, and to retrieve it and ancillary data. cognitive memory, when connected to computer or information appliance introduces computational architecture that applies to systems and methods for navigation, location and recognition of objects in images, character recognition, facial recognition, medical analysis and diagnosis, video image analysis, and to photographic search engines that when prompted with a query photograph containing faces and objects will retrieve related photographs stored in computer or other information appliance, and will identify url's of related photographs and documents stored on the world wide web. dated 2008-02-19"
7337025,neural network based method for exponent coding in a transform coder for high quality audio,a method and apparatus for assigning an exponent coding strategy in a digital audio transform coder. different coding strategies having different differential coding limits may be assigned to different set of transform exponents according to the frequency domain characteristics of the audio signal. a neural network processing system is utilised to perform an efficient mapping of each exponent set to an appropriate coding strategy.,2008-02-26,The title of the patent is neural network based method for exponent coding in a transform coder for high quality audio and its abstract is a method and apparatus for assigning an exponent coding strategy in a digital audio transform coder. different coding strategies having different differential coding limits may be assigned to different set of transform exponents according to the frequency domain characteristics of the audio signal. a neural network processing system is utilised to perform an efficient mapping of each exponent set to an appropriate coding strategy. dated 2008-02-26
7337155,communication analysis apparatus,"to deal with user network communication activity which cannot easily and clearly be determined as problematic behavior, a behavior analysis apparatus 14 monitors communication between each user pc 16 in a domain 10 and internet 20 via a gateway apparatus 12. for example, when there is a monitored item related to information leakage of the user in the detected communication, a weight value corresponding to the monitored item is added to a score concerning a possibility of the user leaking information. subsequently, the scores are totaled and recorded for each unit of time. the behavior analysis apparatus 14 inputs data of time-series transition of the total value to a neural network which has performed learning for prediction processing, and predicts the possibility of the user's information leak at a time in the near future. when an increasing risk of leakage is predicted, the behavior analysis apparatus 14 communicates an alarm to a security manager.",2008-02-26,"The title of the patent is communication analysis apparatus and its abstract is to deal with user network communication activity which cannot easily and clearly be determined as problematic behavior, a behavior analysis apparatus 14 monitors communication between each user pc 16 in a domain 10 and internet 20 via a gateway apparatus 12. for example, when there is a monitored item related to information leakage of the user in the detected communication, a weight value corresponding to the monitored item is added to a score concerning a possibility of the user leaking information. subsequently, the scores are totaled and recorded for each unit of time. the behavior analysis apparatus 14 inputs data of time-series transition of the total value to a neural network which has performed learning for prediction processing, and predicts the possibility of the user's information leak at a time in the near future. when an increasing risk of leakage is predicted, the behavior analysis apparatus 14 communicates an alarm to a security manager. dated 2008-02-26"
7340408,method for evaluating customer valve to guide loyalty and retention programs,"a method and apparatus for training a neural network to compute hazard functions for customers and analyzing hazard functions, both for an individual customer, and for set of customers to focus marketing techniques. the hazard function represents the likelihood of churn for a particular customer. the gain in lifetime value is also calculated for each customer which incorporates the present value of the customer with the future value of the customer if a new contract is entered. the overall shape of the hazard function, combined with the gain in lifetime value, specifies what marketing techniques are to be applied together with what additional incentives are to be offered to the customer in order prevent churn.",2008-03-04,"The title of the patent is method for evaluating customer valve to guide loyalty and retention programs and its abstract is a method and apparatus for training a neural network to compute hazard functions for customers and analyzing hazard functions, both for an individual customer, and for set of customers to focus marketing techniques. the hazard function represents the likelihood of churn for a particular customer. the gain in lifetime value is also calculated for each customer which incorporates the present value of the customer with the future value of the customer if a new contract is entered. the overall shape of the hazard function, combined with the gain in lifetime value, specifies what marketing techniques are to be applied together with what additional incentives are to be offered to the customer in order prevent churn. dated 2008-03-04"
7340440,hybrid neural network generation system and method,"a computer-implemented method and system for building a neural network is disclosed. the neural network predicts at least one target based upon predictor variables defined in a state space. first, an input data set is retrieved that includes the predictor variables and at least one target associated with the predictor variables for each observation. in the state space, a number of points is inserted in the state space based upon the values of the predictor variables. the number of points is less than the number of observations. a statistical measure is determined that describes a relationship between the observations and the inserted points. weights and activation functions of the neural network are determined using the statistical measure.",2008-03-04,"The title of the patent is hybrid neural network generation system and method and its abstract is a computer-implemented method and system for building a neural network is disclosed. the neural network predicts at least one target based upon predictor variables defined in a state space. first, an input data set is retrieved that includes the predictor variables and at least one target associated with the predictor variables for each observation. in the state space, a number of points is inserted in the state space based upon the values of the predictor variables. the number of points is less than the number of observations. a statistical measure is determined that describes a relationship between the observations and the inserted points. weights and activation functions of the neural network are determined using the statistical measure. dated 2008-03-04"
7343289,system and method for audio/video speaker detection,"a system and method for detecting speech utilizing audio and video inputs. in one aspect, the invention collects audio data generated from a microphone device. in another aspect, the invention collects video data and processes the data to determine a mouth location for a given speaker. the audio and video are inputted into a time-delay neural network that processes the data to determine which target is speaking. the neural network processing is based upon a correlation to detected mouth movement from the video data and audio sounds detected by the microphone.",2008-03-11,"The title of the patent is system and method for audio/video speaker detection and its abstract is a system and method for detecting speech utilizing audio and video inputs. in one aspect, the invention collects audio data generated from a microphone device. in another aspect, the invention collects video data and processes the data to determine a mouth location for a given speaker. the audio and video are inputted into a time-delay neural network that processes the data to determine which target is speaking. the neural network processing is based upon a correlation to detected mouth movement from the video data and audio sounds detected by the microphone. dated 2008-03-11"
7345691,method of image processing and electronic device utilizing the same,"a method of image processing using neural networks. images to be processed and displayed are received. each image is divided into sections, each comprising pixels represented by color values. a neural network training procedure is executed for the sections to obtain color value tables and mapping tables. each color value table comprises colors represented by the color values. the mapping tables record the relations between each pixel and the colors in the color value tables. the color value tables and the mapping tables are recorded in an electronic device for image display.",2008-03-18,"The title of the patent is method of image processing and electronic device utilizing the same and its abstract is a method of image processing using neural networks. images to be processed and displayed are received. each image is divided into sections, each comprising pixels represented by color values. a neural network training procedure is executed for the sections to obtain color value tables and mapping tables. each color value table comprises colors represented by the color values. the mapping tables record the relations between each pixel and the colors in the color value tables. the color value tables and the mapping tables are recorded in an electronic device for image display. dated 2008-03-18"
7346208,image artifact reduction using a neural network,"a neural network is trained and used to reduce artifacts in spatial domain representations of images that were compressed by a transform method and then decompressed. for example, the neural network can be trained and used to reduce artifacts such as blocking and ringing artifacts in jpeg images.",2008-03-18,"The title of the patent is image artifact reduction using a neural network and its abstract is a neural network is trained and used to reduce artifacts in spatial domain representations of images that were compressed by a transform method and then decompressed. for example, the neural network can be trained and used to reduce artifacts such as blocking and ringing artifacts in jpeg images. dated 2008-03-18"
7346489,system and method of determining phrasing in text,"a system analyzes text, determines phrasing and, in an exemplary embodiment, reformats the text to establish optimal spacing and related features for readability, reader comprehension and publishing economies. a neural network uses a library of text data to analyze text and determine phrases. formatting emphasizes phrases using one or more of a plurality of techniques including word spacing, text darkness and controlling line breaks.",2008-03-18,"The title of the patent is system and method of determining phrasing in text and its abstract is a system analyzes text, determines phrasing and, in an exemplary embodiment, reformats the text to establish optimal spacing and related features for readability, reader comprehension and publishing economies. a neural network uses a library of text data to analyze text and determine phrases. formatting emphasizes phrases using one or more of a plurality of techniques including word spacing, text darkness and controlling line breaks. dated 2008-03-18"
7346497,high-order entropy error functions for neural classifiers,"an automatic speech recognition system comprising a speech decoder to resolve phone and word level information, a vector generator to generate information vectors on which a confidence measure is based by a neural network classifier (ann). an error signal is designed which is not subject to false saturation or over specialization. the error signal is integrated into an error function which is back propagated through the ann.",2008-03-18,"The title of the patent is high-order entropy error functions for neural classifiers and its abstract is an automatic speech recognition system comprising a speech decoder to resolve phone and word level information, a vector generator to generate information vectors on which a confidence measure is based by a neural network classifier (ann). an error signal is designed which is not subject to false saturation or over specialization. the error signal is integrated into an error function which is back propagated through the ann. dated 2008-03-18"
7349806,system and method for extracting optical properties from environmental parameters in water,"a method for predicting water clarity at a plurality of water depths for a location including providing training data to a neural network, the training data representative of water measurements at the location, thereafter receiving inputs including temperature, salinity, tidal information, water depth, and sediment data, and generating values for optical attenuation at a wavelength at a plurality of depths. in one embodiment, a default cloudy day algorithm operates at all times and a clear sky algorithm operates only when clear satellite images are available.",2008-03-25,"The title of the patent is system and method for extracting optical properties from environmental parameters in water and its abstract is a method for predicting water clarity at a plurality of water depths for a location including providing training data to a neural network, the training data representative of water measurements at the location, thereafter receiving inputs including temperature, salinity, tidal information, water depth, and sediment data, and generating values for optical attenuation at a wavelength at a plurality of depths. in one embodiment, a default cloudy day algorithm operates at all times and a clear sky algorithm operates only when clear satellite images are available. dated 2008-03-25"
7352918,method and circuits for scaling images using neural networks,"an artificial neural network (ann) based system that is adapted to process an input pattern to generate an output pattern related thereto having a different number of components than the input pattern. the system (26) is comprised of an ann (27) and a memory (28), such as a dram memory, that are serially connected. the input pattern (23) is applied to a processor (22), where it can be processed or not (the most general case), before it is applied to the ann and stored therein as a prototype (if learned). a category is associated with each stored prototype. the processor computes the coefficients that allow the determination of the estimated values of the output pattern, these coefficients are the components of a so-called intermediate pattern (24). assuming the ann has already learned a number of input patterns, when a new input pattern is presented to the ann in the recognition phase, the category of the closest prototype is output therefrom and is used as a pointer to the memory. in turn, the memory outputs the corresponding intermediate pattern. the input pattern and the intermediate pattern are applied to the processor to construct the output pattern (25) using the coefficients. typically, the input pattern is a block of pixels in the field of scaling images.",2008-04-01,"The title of the patent is method and circuits for scaling images using neural networks and its abstract is an artificial neural network (ann) based system that is adapted to process an input pattern to generate an output pattern related thereto having a different number of components than the input pattern. the system (26) is comprised of an ann (27) and a memory (28), such as a dram memory, that are serially connected. the input pattern (23) is applied to a processor (22), where it can be processed or not (the most general case), before it is applied to the ann and stored therein as a prototype (if learned). a category is associated with each stored prototype. the processor computes the coefficients that allow the determination of the estimated values of the output pattern, these coefficients are the components of a so-called intermediate pattern (24). assuming the ann has already learned a number of input patterns, when a new input pattern is presented to the ann in the recognition phase, the category of the closest prototype is output therefrom and is used as a pointer to the memory. in turn, the memory outputs the corresponding intermediate pattern. the input pattern and the intermediate pattern are applied to the processor to construct the output pattern (25) using the coefficients. typically, the input pattern is a block of pixels in the field of scaling images. dated 2008-04-01"
7353088,system and method for detecting presence of a human in a vehicle,"a system for detecting the presence of a human in a vehicle is provided. the system includes a vibration sensor that is configured to detect vibrations of the vehicle, and to output signals related to the sensed vibrations. a processor is configured to receive the signals output from the vibration sensor. the processor also operates a neural network that has a plurality of nodes, at least some of which are recurrent. the use of the recurrent nodes allows the output of a recurrent node to be fed back into itself, or another node. in addition, the output that is fed back can be combined with other inputs entering the node. in this way, the neural network can quickly learn to distinguish between various conditions, including an occupied state and an unoccupied state of the vehicle. the neural network provides an output indicating whether the vehicle is occupied.",2008-04-01,"The title of the patent is system and method for detecting presence of a human in a vehicle and its abstract is a system for detecting the presence of a human in a vehicle is provided. the system includes a vibration sensor that is configured to detect vibrations of the vehicle, and to output signals related to the sensed vibrations. a processor is configured to receive the signals output from the vibration sensor. the processor also operates a neural network that has a plurality of nodes, at least some of which are recurrent. the use of the recurrent nodes allows the output of a recurrent node to be fed back into itself, or another node. in addition, the output that is fed back can be combined with other inputs entering the node. in this way, the neural network can quickly learn to distinguish between various conditions, including an occupied state and an unoccupied state of the vehicle. the neural network provides an output indicating whether the vehicle is occupied. dated 2008-04-01"
7359537,dna microarray image analysis system,"in a microarray image analysis system, when one of a plurality of statuses is set for a spot of a microarray by the user, the status of a similar spot is automatically determined. in a microarray image, the user determines a status of a spot, the pixel value matrix of an image in a spot region is learned by a neural network, a vertically and horizontally symmetrical image and an image rotated about the center of the region are formed and are learned by the neural network, and the neural network formed by repeating these steps is used for automatically recognizing the status of an undecided spot.",2008-04-15,"The title of the patent is dna microarray image analysis system and its abstract is in a microarray image analysis system, when one of a plurality of statuses is set for a spot of a microarray by the user, the status of a similar spot is automatically determined. in a microarray image, the user determines a status of a spot, the pixel value matrix of an image in a spot region is learned by a neural network, a vertically and horizontally symmetrical image and an image rotated about the center of the region are formed and are learned by the neural network, and the neural network formed by repeating these steps is used for automatically recognizing the status of an undecided spot. dated 2008-04-15"
7359888,molecular-junction-nanowire-crossbar-based neural network,"a method for configuring nanoscale neural network circuits using molecular-junction-nanowire crossbars, and nanoscale neural networks produced by this method. summing of weighted inputs within a neural-network node is implemented using variable-resistance resistors selectively configured at molecular-junction-nanowire-crossbar junctions. thresholding functions for neural network nodes are implemented using pfet and nfet components selectively configured at molecular-junction-nanowire-crossbar junctions to provide an inverter. the output of one level of neural network nodes is directed, through selectively configured connections, to the resistor elements of a second level of neural network nodes via circuits created in the molecular-junction-nanowire crossbar. an arbitrary number of inputs, outputs, neural network node levels, nodes, weighting functions, and thresholding functions for any desired neural network are readily obtained by the methods of the present invention.",2008-04-15,"The title of the patent is molecular-junction-nanowire-crossbar-based neural network and its abstract is a method for configuring nanoscale neural network circuits using molecular-junction-nanowire crossbars, and nanoscale neural networks produced by this method. summing of weighted inputs within a neural-network node is implemented using variable-resistance resistors selectively configured at molecular-junction-nanowire-crossbar junctions. thresholding functions for neural network nodes are implemented using pfet and nfet components selectively configured at molecular-junction-nanowire-crossbar junctions to provide an inverter. the output of one level of neural network nodes is directed, through selectively configured connections, to the resistor elements of a second level of neural network nodes via circuits created in the molecular-junction-nanowire crossbar. an arbitrary number of inputs, outputs, neural network node levels, nodes, weighting functions, and thresholding functions for any desired neural network are readily obtained by the methods of the present invention. dated 2008-04-15"
7362422,method and apparatus for a downhole spectrometer based on electronically tunable optical filters,"the present invention provides an apparatus and method for high resolution spectroscopy (approximately 10 picometer wavelength resolution) using a tunable optical filter (tof) for analyzing a formation fluid sample downhole and at the surface to determine formation fluid parameters. the analysis comprises determination of gas oil ratio, api gravity and various other fluid parameters which can be estimated after developing correlations to a training set of samples using a neural network or a chemometric equation.",2008-04-22,"The title of the patent is method and apparatus for a downhole spectrometer based on electronically tunable optical filters and its abstract is the present invention provides an apparatus and method for high resolution spectroscopy (approximately 10 picometer wavelength resolution) using a tunable optical filter (tof) for analyzing a formation fluid sample downhole and at the surface to determine formation fluid parameters. the analysis comprises determination of gas oil ratio, api gravity and various other fluid parameters which can be estimated after developing correlations to a training set of samples using a neural network or a chemometric equation. dated 2008-04-22"
7363111,methods and systems for analyzing engine unbalance conditions,"methods and systems for analyzing engine unbalance conditions are disclosed. in one embodiment, a method includes receiving vibrational data from a plurality of locations distributed over an engine and a surrounding engine support structure, and inputting the vibrational data into a neural network inverse model. the neural network inverse model establishes a relationship between the vibrational data and an unbalance condition of the engine, and outputs diagnostic information indicating the unbalance condition of the engine. in a further embodiment, a method further includes subjecting the vibrational data to a fast fourier transformation to extract a desired once per revolution vibrational data prior to input to the neural network inverse model.",2008-04-22,"The title of the patent is methods and systems for analyzing engine unbalance conditions and its abstract is methods and systems for analyzing engine unbalance conditions are disclosed. in one embodiment, a method includes receiving vibrational data from a plurality of locations distributed over an engine and a surrounding engine support structure, and inputting the vibrational data into a neural network inverse model. the neural network inverse model establishes a relationship between the vibrational data and an unbalance condition of the engine, and outputs diagnostic information indicating the unbalance condition of the engine. in a further embodiment, a method further includes subjecting the vibrational data to a fast fourier transformation to extract a desired once per revolution vibrational data prior to input to the neural network inverse model. dated 2008-04-22"
7363120,method of adjusting at least one defective rotor of a rotorcraft,"the present invention relates to a method of adjusting at least one defective, main or anti-torque rotor of a particular rotorcraft. the method uses a neural network representing the relationships between firstly accelerations representative of vibration generated on at least a portion of a reference rotorcraft, and secondly defects and adjustment parameters. after determining the defects, if any, of a defective rotor, an adjustment value α is defined for at least one of the adjustment parameters, advantageously by minimizing the following relationship:",2008-04-22,"The title of the patent is method of adjusting at least one defective rotor of a rotorcraft and its abstract is the present invention relates to a method of adjusting at least one defective, main or anti-torque rotor of a particular rotorcraft. the method uses a neural network representing the relationships between firstly accelerations representative of vibration generated on at least a portion of a reference rotorcraft, and secondly defects and adjustment parameters. after determining the defects, if any, of a defective rotor, an adjustment value α is defined for at least one of the adjustment parameters, advantageously by minimizing the following relationship: dated 2008-04-22"
7363185,fluxgate sensor for calibrating azimuth at slope and calibration method thereof,"a two-axis fluxgate sensor has a driving pulse generating circuit which generates pulse signal and outputs as a driving signal, and x-axis and y-axis fluxgates which are in proportional relation with each other. the two-axis fluxgate sensor generates voltage values of x-axis and y-axis fluxgates corresponding to the magnetism which is generated from the driving signal, and a memory stores therein a neural network weight matrix. when the voltage values of the x-axis and y-axis fluxgates are measured, a control unit compensates for the voltage values by using the neural network weight matrix which is stored in the memory, and computes an azimuth angle by using the compensated voltage values. an accurate azimuth angle can be obtained even at slopes.",2008-04-22,"The title of the patent is fluxgate sensor for calibrating azimuth at slope and calibration method thereof and its abstract is a two-axis fluxgate sensor has a driving pulse generating circuit which generates pulse signal and outputs as a driving signal, and x-axis and y-axis fluxgates which are in proportional relation with each other. the two-axis fluxgate sensor generates voltage values of x-axis and y-axis fluxgates corresponding to the magnetism which is generated from the driving signal, and a memory stores therein a neural network weight matrix. when the voltage values of the x-axis and y-axis fluxgates are measured, a control unit compensates for the voltage values by using the neural network weight matrix which is stored in the memory, and computes an azimuth angle by using the compensated voltage values. an accurate azimuth angle can be obtained even at slopes. dated 2008-04-22"
7363243,system and method for predicting external events from electronic posting activity,"a system and method for collecting and analyzing electronic discussion messages to categorize the message communications and the identify trends and patterns in pre-determined markets. electronic messages are collected and analyzed according to characteristics and data inherent in the messages. objective data is collected by the system for use in analyzing the electronic discussion data against real-world events to facilitate trend analysis and event forecasting based on the volume, nature and content of messages posted to electronic discussion forums. the message posting activity of a plurality of posters can be tracked to identify opinion leaders, and patterns in the messages can be identified using at least one of the following techniques: chaos theory; complexity theory; social network theory; and neural network theory.",2008-04-22,"The title of the patent is system and method for predicting external events from electronic posting activity and its abstract is a system and method for collecting and analyzing electronic discussion messages to categorize the message communications and the identify trends and patterns in pre-determined markets. electronic messages are collected and analyzed according to characteristics and data inherent in the messages. objective data is collected by the system for use in analyzing the electronic discussion data against real-world events to facilitate trend analysis and event forecasting based on the volume, nature and content of messages posted to electronic discussion forums. the message posting activity of a plurality of posters can be tracked to identify opinion leaders, and patterns in the messages can be identified using at least one of the following techniques: chaos theory; complexity theory; social network theory; and neural network theory. dated 2008-04-22"
7363281,reduction of fitness evaluations using clustering techniques and neural network ensembles,"the invention relates to an evolutionary optimization method. first, an initial population of individuals is set up and an original fitness function is applied. then the offspring individuals having a high evaluated quality value as parents are selected. in a third step, the parents are reproduced to create a plurality of offspring individuals. the quality of the offspring individuals is evaluated selectively using an original fitness function or an approximate fitness function. finally, the method returns to the selection step until a termination condition is met. the step of evaluating the quality of the offspring individuals includes grouping all offspring individuals in clusters, selecting for each cluster one or a plurality of offspring individuals, resulting in altogether selected offspring individuals, evaluating the selected offspring individuals by the original fitness function, and evaluating the remaining offspring individuals by means of the approximate fitness function.",2008-04-22,"The title of the patent is reduction of fitness evaluations using clustering techniques and neural network ensembles and its abstract is the invention relates to an evolutionary optimization method. first, an initial population of individuals is set up and an original fitness function is applied. then the offspring individuals having a high evaluated quality value as parents are selected. in a third step, the parents are reproduced to create a plurality of offspring individuals. the quality of the offspring individuals is evaluated selectively using an original fitness function or an approximate fitness function. finally, the method returns to the selection step until a termination condition is met. the step of evaluating the quality of the offspring individuals includes grouping all offspring individuals in clusters, selecting for each cluster one or a plurality of offspring individuals, resulting in altogether selected offspring individuals, evaluating the selected offspring individuals by the original fitness function, and evaluating the remaining offspring individuals by means of the approximate fitness function. dated 2008-04-22"
7366704,system and method for deconvoluting the effect of topography on scanning probe microscopy measurements,"a method for using a neural network to deconvolute the effects due to surface topography from the effects due to the other physical property being measured in a scanning probe microscopy (spm) or atomic force microscopy (afm) image. in the case of a thermal spm, the spm probe is scanned across the surface of a sample having known uniform thermal properties, measuring both the surface topography and thermal properties of the sample. the data thus collected forms a training data set. several training data sets can be collected, preferably on samples having different surface topographies. a neural network is applied to the training data sets, such that the neural network learns how to deconvolute the effects dues to surface topography from the effects dues to the variations in thermal properties of a sample.",2008-04-29,"The title of the patent is system and method for deconvoluting the effect of topography on scanning probe microscopy measurements and its abstract is a method for using a neural network to deconvolute the effects due to surface topography from the effects due to the other physical property being measured in a scanning probe microscopy (spm) or atomic force microscopy (afm) image. in the case of a thermal spm, the spm probe is scanned across the surface of a sample having known uniform thermal properties, measuring both the surface topography and thermal properties of the sample. the data thus collected forms a training data set. several training data sets can be collected, preferably on samples having different surface topographies. a neural network is applied to the training data sets, such that the neural network learns how to deconvolute the effects dues to surface topography from the effects dues to the variations in thermal properties of a sample. dated 2008-04-29"
7369702,template-based cursive handwriting recognition,"input handwritten characters are classified as print or cursive based upon numerical feature values calculated from the shape of an input character. the feature values are applied to inputs of an artificial neural network which outputs a probability of the input character being print or cursive. if a character is classified as print, it is analyzed by a print character recognizer. if a character is classified as cursive, it is analyzed using a cursive character recognizer. the cursive character recognizer compares the input character to multiple prototype characters using a dynamic time warping (dtw) algorithm.",2008-05-06,"The title of the patent is template-based cursive handwriting recognition and its abstract is input handwritten characters are classified as print or cursive based upon numerical feature values calculated from the shape of an input character. the feature values are applied to inputs of an artificial neural network which outputs a probability of the input character being print or cursive. if a character is classified as print, it is analyzed by a print character recognizer. if a character is classified as cursive, it is analyzed using a cursive character recognizer. the cursive character recognizer compares the input character to multiple prototype characters using a dynamic time warping (dtw) algorithm. dated 2008-05-06"
7369976,"method of designing tire, optimization analyzer and storage medium on which optimization analysis program is recorded","a design of a tire can be facilitated. an optimization apparatus 30 inputs known design parameters of the shape, structure, and pattern of a tire, and performances thereof by an experimental data input unit 40 and learns, as a conversion system of a neural network, a correlation between design parameters of the shape, structure, and pattern of the tire, and performances thereof. ranges which constrain performances of the tire and design parameters of the shape, structure, and pattern of the tire, which are to be optimized, are inputted in an optimization item input unit 42, and the performances of the tire are predicted in an optimization calculation unit 34 from the design parameters of the shape, structure, and pattern of the tire by using the optimization item and models of the calculation unit 32, and an objective function is optimized until the objective function which is the performances of the tire is converged. the optimized design parameters of the shape, structure, and pattern of the tire are outputted from an optimization result output unit 44.",2008-05-06,"The title of the patent is method of designing tire, optimization analyzer and storage medium on which optimization analysis program is recorded and its abstract is a design of a tire can be facilitated. an optimization apparatus 30 inputs known design parameters of the shape, structure, and pattern of a tire, and performances thereof by an experimental data input unit 40 and learns, as a conversion system of a neural network, a correlation between design parameters of the shape, structure, and pattern of the tire, and performances thereof. ranges which constrain performances of the tire and design parameters of the shape, structure, and pattern of the tire, which are to be optimized, are inputted in an optimization item input unit 42, and the performances of the tire are predicted in an optimization calculation unit 34 from the design parameters of the shape, structure, and pattern of the tire by using the optimization item and models of the calculation unit 32, and an objective function is optimized until the objective function which is the performances of the tire is converged. the optimized design parameters of the shape, structure, and pattern of the tire are outputted from an optimization result output unit 44. dated 2008-05-06"
7370020,sequence generator,"apparatus for generating sequences of elements including at least one task unit, each of which has an upper and a lower neural network connected in a hierarchical relationship and is operable to output a sequence of elements. each of the upper and lower neural networks is a class of temporal neural networks having an infinite number of internal states.",2008-05-06,"The title of the patent is sequence generator and its abstract is apparatus for generating sequences of elements including at least one task unit, each of which has an upper and a lower neural network connected in a hierarchical relationship and is operable to output a sequence of elements. each of the upper and lower neural networks is a class of temporal neural networks having an infinite number of internal states. dated 2008-05-06"
7370021,medical applications of adaptive learning systems using gene expression data,a neural network module is provided. it comprises an input layer comprising one or more input nodes configured to receive gene expression data. it also has a rule base layer comprising one or more rule nodes and an output layer comprising one or more output nodes configured to output one or more conditions. it also comprises an adaptive component configured to extract one or more rules from the rule base layer representing relationships between the gene expression data and the one or more conditions. methods and systems using the module are disclosed as well as specific profiles utilising the system.,2008-05-06,The title of the patent is medical applications of adaptive learning systems using gene expression data and its abstract is a neural network module is provided. it comprises an input layer comprising one or more input nodes configured to receive gene expression data. it also has a rule base layer comprising one or more rule nodes and an output layer comprising one or more output nodes configured to output one or more conditions. it also comprises an adaptive component configured to extract one or more rules from the rule base layer representing relationships between the gene expression data and the one or more conditions. methods and systems using the module are disclosed as well as specific profiles utilising the system. dated 2008-05-06
7370969,corneal topography analysis system,"a corneal topography analysis system includes: an input unit for inputting corneal curvature data; and an analysis unit that determines plural indexes characterizing topography of the cornea based on the input corneal curvature data, the analysis unit further judges corneal topography from features inherent in predetermined classifications of corneal topography using the determined indexes and a neural network so as to judge at least one of normal cornea, myopic refractive surgery, hyperopic refractive surgery, corneal astigmatism, penetrating keratoplasty, keratoconus, keratoconus suspect, pellucid marginal degeneration, or other classification of corneal topography.",2008-05-13,"The title of the patent is corneal topography analysis system and its abstract is a corneal topography analysis system includes: an input unit for inputting corneal curvature data; and an analysis unit that determines plural indexes characterizing topography of the cornea based on the input corneal curvature data, the analysis unit further judges corneal topography from features inherent in predetermined classifications of corneal topography using the determined indexes and a neural network so as to judge at least one of normal cornea, myopic refractive surgery, hyperopic refractive surgery, corneal astigmatism, penetrating keratoplasty, keratoconus, keratoconus suspect, pellucid marginal degeneration, or other classification of corneal topography. dated 2008-05-13"
7372952,telephony control system with intelligent call routing,"a communications system and method, for analyzing at least two characteristics for each of at least a portion of a plurality of communications, selected from the group comprising artificial neural network parameters, stochastic parameters, cost parameters, and utility parameters, and determining in dependence thereon an optimum routing of the plurality of communications, based on the at least two characteristics for each of the at least a portion of the communications.",2008-05-13,"The title of the patent is telephony control system with intelligent call routing and its abstract is a communications system and method, for analyzing at least two characteristics for each of at least a portion of a plurality of communications, selected from the group comprising artificial neural network parameters, stochastic parameters, cost parameters, and utility parameters, and determining in dependence thereon an optimum routing of the plurality of communications, based on the at least two characteristics for each of the at least a portion of the communications. dated 2008-05-13"
7373333,"information processing apparatus and method, program storage medium and program","an information processing method and an information processing apparatus in which the learning efficiency may be improved and the scale may be extended readily. an integrated module 42 is formed by a movement pattern learning module by a local expression scheme. the local modules 43-1 to 43-3 of the integrated module 42 are each formed by a recurrent neural network as a movement pattern learning model by a distributed expression scheme. the local modules 43-1 to 43-3 are caused to learn plural movement patterns. outputs from the local modules 43-1 to 43-3, supplied with preset parameters, as inputs, are multiplied by gates 44-1 to 44-3 with coefficients w1 to w3, respectively, and the resulting products are summed together and output.",2008-05-13,"The title of the patent is information processing apparatus and method, program storage medium and program and its abstract is an information processing method and an information processing apparatus in which the learning efficiency may be improved and the scale may be extended readily. an integrated module 42 is formed by a movement pattern learning module by a local expression scheme. the local modules 43-1 to 43-3 of the integrated module 42 are each formed by a recurrent neural network as a movement pattern learning model by a distributed expression scheme. the local modules 43-1 to 43-3 are caused to learn plural movement patterns. outputs from the local modules 43-1 to 43-3, supplied with preset parameters, as inputs, are multiplied by gates 44-1 to 44-3 with coefficients w1 to w3, respectively, and the resulting products are summed together and output. dated 2008-05-13"
7375523,system and method for fast mr coil sensitivity mapping,"a system and method for mapping the sensitivity of mr coils includes a neural network or other computer intelligence trained from sample mr data to determine coil sensitivity profiles or sensitivity normalizations. once the network is trained, subsequent coil mapping determinations may include fewer mapping acquisitions per coil. the resulting sensitivity map can be used in compensating for b1 inhomogeneities, parallel imaging reconstruction, generating tailored excitation currents for each individual coil, rf shimming, or other processes.",2008-05-20,"The title of the patent is system and method for fast mr coil sensitivity mapping and its abstract is a system and method for mapping the sensitivity of mr coils includes a neural network or other computer intelligence trained from sample mr data to determine coil sensitivity profiles or sensitivity normalizations. once the network is trained, subsequent coil mapping determinations may include fewer mapping acquisitions per coil. the resulting sensitivity map can be used in compensating for b1 inhomogeneities, parallel imaging reconstruction, generating tailored excitation currents for each individual coil, rf shimming, or other processes. dated 2008-05-20"
7376661,xml-based symbolic language and interpreter,"an xml-based symbolic computer language, interpreter, and corresponding execution environments are disclosed. the xml-based symbolic computer language, called “olin” (one language intelligent network) enables a computer program to be written as an xml-compliant document. olin programs are interpreted by an interpreter by parsing the xml content to extract symbolic expressions embedded therein, and evaluating those symbolic expressions. the xml-based symbolic computer language is object-oriented and is based on inherent principles of the lisp programming language. accordingly, code and data are treated the same. the language provides built-in structures and functions for implementing neural networks and genetic programming. olin programs may be executed by a single computer, or via a multiprocessing environment, including distributing processing environments and multiprocessor computers.",2008-05-20,"The title of the patent is xml-based symbolic language and interpreter and its abstract is an xml-based symbolic computer language, interpreter, and corresponding execution environments are disclosed. the xml-based symbolic computer language, called “olin” (one language intelligent network) enables a computer program to be written as an xml-compliant document. olin programs are interpreted by an interpreter by parsing the xml content to extract symbolic expressions embedded therein, and evaluating those symbolic expressions. the xml-based symbolic computer language is object-oriented and is based on inherent principles of the lisp programming language. accordingly, code and data are treated the same. the language provides built-in structures and functions for implementing neural networks and genetic programming. olin programs may be executed by a single computer, or via a multiprocessing environment, including distributing processing environments and multiprocessor computers. dated 2008-05-20"
7379839,multi-function air data probes employing neural networks for determining local air data parameters,"an air data sensing probe or mfp includes a barrel having multiple pressure sensing ports for sensing multiple pressures. instrumentation coupled to the pressure sensing ports provides electrical signals related to the multiple pressures. a neural network, coupled to the instrumentation, receives as inputs the electrical signals related to the multiple pressures, and in response, the neural network provides, as an output, electrical signals indicative of at least one local air data parameter for the air data sensing probe.",2008-05-27,"The title of the patent is multi-function air data probes employing neural networks for determining local air data parameters and its abstract is an air data sensing probe or mfp includes a barrel having multiple pressure sensing ports for sensing multiple pressures. instrumentation coupled to the pressure sensing ports provides electrical signals related to the multiple pressures. a neural network, coupled to the instrumentation, receives as inputs the electrical signals related to the multiple pressures, and in response, the neural network provides, as an output, electrical signals indicative of at least one local air data parameter for the air data sensing probe. dated 2008-05-27"
7383235,method and hardware architecture for controlling a process or for processing data based on quantum soft computing,a method of controlling a process driven by a control signal for producing a corresponding output includes producing an error signal as a function of a state of the process and of a reference signal. a control signal is generated as a function of the error signal and of a parameter adjustment signal. the control signal is applied to the process. a derived signal representative of a quantity to be minimized is calculated by processing paired values of the state of the process and the control signal. a correction signal is calculated from a set of several different values of the control signal that minimizes the derived signal. the parameter adjustment signal is calculated by a neural network and fuzzy logic processor from the error signal and the correction signal. the correction signal is periodically calculated by a quantum genetic search algorithm that results from a merging of a genetic algorithm and a quantum search algorithm.,2008-06-03,The title of the patent is method and hardware architecture for controlling a process or for processing data based on quantum soft computing and its abstract is a method of controlling a process driven by a control signal for producing a corresponding output includes producing an error signal as a function of a state of the process and of a reference signal. a control signal is generated as a function of the error signal and of a parameter adjustment signal. the control signal is applied to the process. a derived signal representative of a quantity to be minimized is calculated by processing paired values of the state of the process and the control signal. a correction signal is calculated from a set of several different values of the control signal that minimizes the derived signal. the parameter adjustment signal is calculated by a neural network and fuzzy logic processor from the error signal and the correction signal. the correction signal is periodically calculated by a quantum genetic search algorithm that results from a merging of a genetic algorithm and a quantum search algorithm. dated 2008-06-03
7386388,"air-fuel ratio control system and method for internal combustion engine, and engine control unit","an air-fuel ratio control system for an internal combustion engine, which is capable of accurately estimating an exhaust gas state parameter according to the properties of fuel, thereby making it possible to properly control the air-fuel ratio of a mixture. the air-fuel ratio control system 1 estimates an exhaust gas state parameter indicative of a state of exhaust gases, as an estimated exhaust gas state parameter (af13 nn) by inputting a detected combustion state parameter (dcadlyig) indicative of a combustion state of the mixture in the engine 3, and detected operating state parameters (ne, tw, pba, iglog, tout) indicative of operating states of the engine 3, to a neural network (nn) configured as a network to which are input the combustion state parameter (dcadlyig) and the operating state parameters (ne, tw, pba, iglog, tout), and in which the exhaust gas state parameter is used as a teacher signal (step 1), and controls the air-fuel ratio based on the estimated exhaust gas state parameter (af_nn) (steps 3, 4, and 24 to 28).",2008-06-10,"The title of the patent is air-fuel ratio control system and method for internal combustion engine, and engine control unit and its abstract is an air-fuel ratio control system for an internal combustion engine, which is capable of accurately estimating an exhaust gas state parameter according to the properties of fuel, thereby making it possible to properly control the air-fuel ratio of a mixture. the air-fuel ratio control system 1 estimates an exhaust gas state parameter indicative of a state of exhaust gases, as an estimated exhaust gas state parameter (af13 nn) by inputting a detected combustion state parameter (dcadlyig) indicative of a combustion state of the mixture in the engine 3, and detected operating state parameters (ne, tw, pba, iglog, tout) indicative of operating states of the engine 3, to a neural network (nn) configured as a network to which are input the combustion state parameter (dcadlyig) and the operating state parameters (ne, tw, pba, iglog, tout), and in which the exhaust gas state parameter is used as a teacher signal (step 1), and controls the air-fuel ratio based on the estimated exhaust gas state parameter (af_nn) (steps 3, 4, and 24 to 28). dated 2008-06-10"
7386389,method and system for determining the driving situation,"a method and a system for ascertaining the driving situation of a motor vehicle by using data provided in the vehicle indicating the value of at least one state variable of the vehicle are provided. to relieve the burden on the driver a data record providing the history of the at least one state variable is supplied. in a neural network in the motor vehicle is supplied by a suitably programmed computer. the neural network has at least one input layer and one output layer, each of the layers having a plurality of perceptrons. the respective value of the at least one state variable of the respective point in time, preferably a normalized value, is supplied to a perceptron of the neural network. after the neural network has been trained, the current driving situation is output by the perceptrons of the output layer of the neural network.",2008-06-10,"The title of the patent is method and system for determining the driving situation and its abstract is a method and a system for ascertaining the driving situation of a motor vehicle by using data provided in the vehicle indicating the value of at least one state variable of the vehicle are provided. to relieve the burden on the driver a data record providing the history of the at least one state variable is supplied. in a neural network in the motor vehicle is supplied by a suitably programmed computer. the neural network has at least one input layer and one output layer, each of the layers having a plurality of perceptrons. the respective value of the at least one state variable of the respective point in time, preferably a normalized value, is supplied to a perceptron of the neural network. after the neural network has been trained, the current driving situation is output by the perceptrons of the output layer of the neural network. dated 2008-06-10"
7386448,biometric voice authentication,"a system and method enrolls a speaker with an enrollment utterance and authenticates a user with a biometric analysis of an authentication utterance, without the need for a pin (personal identification number). during authentication, the system uses the same authentication utterance to identify who a speaker claims to be with speaker recognition, and verify whether is the speaker is actually the claimed person. thus, it is not necessary for the speaker to identify biometric data using a pin. the biometric analysis includes a neural tree network to determine unique aspects of the authentication utterances for comparison to the enrollment authentication. the biometric analysis leverages a statistical analysis using hidden markov models to before authorizing the speaker.",2008-06-10,"The title of the patent is biometric voice authentication and its abstract is a system and method enrolls a speaker with an enrollment utterance and authenticates a user with a biometric analysis of an authentication utterance, without the need for a pin (personal identification number). during authentication, the system uses the same authentication utterance to identify who a speaker claims to be with speaker recognition, and verify whether is the speaker is actually the claimed person. thus, it is not necessary for the speaker to identify biometric data using a pin. the biometric analysis includes a neural tree network to determine unique aspects of the authentication utterances for comparison to the enrollment authentication. the biometric analysis leverages a statistical analysis using hidden markov models to before authorizing the speaker. dated 2008-06-10"
7389278,detection of pump cavitation/blockage and seal failure via current signature analysis,"a system and method is provided for monitoring the operating condition of a pump by evaluating fault data encoded in the instantaneous current of the motor driving the pump. the data is converted to a frequency spectrum which is analyzed to create a fault signature having fault attributes relating to various fault conditions associated with the pump. the fault signature is then input to a neural network that operates in conjunction with a preprocessing and post processing module to perform decisions and output those decisions to a user interface. a stand alone module is also provided that includes an adaptive preprocessing module, a one-shot unsupervised neural network and a fuzzy based expert system to provide a decision making module that operates with limited human supervision.",2008-06-17,"The title of the patent is detection of pump cavitation/blockage and seal failure via current signature analysis and its abstract is a system and method is provided for monitoring the operating condition of a pump by evaluating fault data encoded in the instantaneous current of the motor driving the pump. the data is converted to a frequency spectrum which is analyzed to create a fault signature having fault attributes relating to various fault conditions associated with the pump. the fault signature is then input to a neural network that operates in conjunction with a preprocessing and post processing module to perform decisions and output those decisions to a user interface. a stand alone module is also provided that includes an adaptive preprocessing module, a one-shot unsupervised neural network and a fuzzy based expert system to provide a decision making module that operates with limited human supervision. dated 2008-06-17"
7389701,method and sensor arrangement for load measurement on rolling element bearing based on model deformation,"method and sensor arrangement for determining a load vector acting on a rolling element bearing (1) in operation. a plurality of n sensors (8) are provided which measure displacement and/or strain for determining displacement and/or strain in one of the elements (5, 6, 7) of the rolling element bearing (1). furthermore, a mode shape coefficients calculator (11) is provided, connected to the plurality of n sensors (8), for determining a deformation of the element (5, 6, 7) by calculating amplitude and phase of n/2 fourier terms representing at least one radial mode shape of the ring shape element (5, 6, 7). also, a bearing neural network (12) is present, connected to the mode shape coefficients calculator (11), the bearing neural network (12) being trained to provide the load vector on the rolling element bearing (1) from the n/2 fourier terms.",2008-06-24,"The title of the patent is method and sensor arrangement for load measurement on rolling element bearing based on model deformation and its abstract is method and sensor arrangement for determining a load vector acting on a rolling element bearing (1) in operation. a plurality of n sensors (8) are provided which measure displacement and/or strain for determining displacement and/or strain in one of the elements (5, 6, 7) of the rolling element bearing (1). furthermore, a mode shape coefficients calculator (11) is provided, connected to the plurality of n sensors (8), for determining a deformation of the element (5, 6, 7) by calculating amplitude and phase of n/2 fourier terms representing at least one radial mode shape of the ring shape element (5, 6, 7). also, a bearing neural network (12) is present, connected to the mode shape coefficients calculator (11), the bearing neural network (12) being trained to provide the load vector on the rolling element bearing (1) from the n/2 fourier terms. dated 2008-06-24"
7390284,vehicle transmission shift quality,"a vehicle powertrain modelling system includes a powertrain model (20), a vehicle model (30), a seat model (55), a driver model (60) and a correlation element (50) comprising a neural network. as a result vehicle development is enhanced, using the neural network (50) to correlate modelled shift aspects with previous subjectively obtained ratings.",2008-06-24,"The title of the patent is vehicle transmission shift quality and its abstract is a vehicle powertrain modelling system includes a powertrain model (20), a vehicle model (30), a seat model (55), a driver model (60) and a correlation element (50) comprising a neural network. as a result vehicle development is enhanced, using the neural network (50) to correlate modelled shift aspects with previous subjectively obtained ratings. dated 2008-06-24"
7392161,identifying a state of a system using an artificial neural network generated model,"the state or condition of a system may be evaluated by comparing a set of selected parameter values, converted into a trial vector, with a number of model or exemplar vectors, each of which was represents a particular state or condition of a sample system. examples of such conditions may include “good”, “marginal”, “unacceptable”, “worn”, “defective”, or other general or specific conditions. sets of parameter values from the system are converted into input vectors. unprocessed vectors are then processed against the input vectors in an artificial neural network to generate the exemplar vectors. the exemplar vectors are stored in a memory of an operational system. during operation of the system, the trial vector is compared with the exemplar vectors. the exemplar vector which is closest to the trial vector represents a state which most closely represents the current state of the system. thus, a high similarity between the trial vector and an exemplar vector which represent a “good” system is likely to have come from a “good” system.",2008-06-24,"The title of the patent is identifying a state of a system using an artificial neural network generated model and its abstract is the state or condition of a system may be evaluated by comparing a set of selected parameter values, converted into a trial vector, with a number of model or exemplar vectors, each of which was represents a particular state or condition of a sample system. examples of such conditions may include “good”, “marginal”, “unacceptable”, “worn”, “defective”, or other general or specific conditions. sets of parameter values from the system are converted into input vectors. unprocessed vectors are then processed against the input vectors in an artificial neural network to generate the exemplar vectors. the exemplar vectors are stored in a memory of an operational system. during operation of the system, the trial vector is compared with the exemplar vectors. the exemplar vector which is closest to the trial vector represents a state which most closely represents the current state of the system. thus, a high similarity between the trial vector and an exemplar vector which represent a “good” system is likely to have come from a “good” system. dated 2008-06-24"
7392230,physical neural network liquid state machine utilizing nanotechnology,"a physical neural network is disclosed, which comprises a liquid state machine. the physical neural network is configured from molecular connections located within a dielectric solvent between pre-synaptic and post-synaptic electrodes thereof, such that the molecular connections are strengthened or weakened according to an application of an electric field or a frequency thereof to provide physical neural network connections thereof. a supervised learning mechanism is associated with the liquid state machine, whereby connections strengths of the molecular connections are determined by pre-synaptic and post-synaptic activity respectively associated with the pre-synaptic and post-synaptic electrodes, wherein the liquid state machine comprises a dynamic fading memory mechanism.",2008-06-24,"The title of the patent is physical neural network liquid state machine utilizing nanotechnology and its abstract is a physical neural network is disclosed, which comprises a liquid state machine. the physical neural network is configured from molecular connections located within a dielectric solvent between pre-synaptic and post-synaptic electrodes thereof, such that the molecular connections are strengthened or weakened according to an application of an electric field or a frequency thereof to provide physical neural network connections thereof. a supervised learning mechanism is associated with the liquid state machine, whereby connections strengths of the molecular connections are determined by pre-synaptic and post-synaptic activity respectively associated with the pre-synaptic and post-synaptic electrodes, wherein the liquid state machine comprises a dynamic fading memory mechanism. dated 2008-06-24"
7395251,neural networks for prediction and control,"neural networks for optimal estimation (including prediction) and/or control involve an execution step and a learning step, and are characterized by the learning step being performed by neural computations. the set of learning rules cause the circuit's connection strengths to learn to approximate the optimal estimation and/or control function that minimizes estimation error and/or a measure of control cost. the classical kalman filter and the classical kalman optimal controller are important examples of such an optimal estimation and/or control function. the circuit uses only a stream of noisy measurements to infer relevant properties of the external dynamical system, learn the optimal estimation and/or control function, and apply its learning of this optimal function to input data streams in an online manner. in this way, the circuit simultaneously learns and generates estimates and/or control output signals that are optimal, given the network's current state of learning.",2008-07-01,"The title of the patent is neural networks for prediction and control and its abstract is neural networks for optimal estimation (including prediction) and/or control involve an execution step and a learning step, and are characterized by the learning step being performed by neural computations. the set of learning rules cause the circuit's connection strengths to learn to approximate the optimal estimation and/or control function that minimizes estimation error and/or a measure of control cost. the classical kalman filter and the classical kalman optimal controller are important examples of such an optimal estimation and/or control function. the circuit uses only a stream of noisy measurements to infer relevant properties of the external dynamical system, learn the optimal estimation and/or control function, and apply its learning of this optimal function to input data streams in an online manner. in this way, the circuit simultaneously learns and generates estimates and/or control output signals that are optimal, given the network's current state of learning. dated 2008-07-01"
7398259,training of a physical neural network,"physical neural network systems and methods are disclosed. a physical neural network can be configured utilizing molecular technology, wherein said physical neural network comprises a plurality of molecular conductors, which form neural network connections thereof. a training mechanism can be provided for training said physical neural network to accomplish a particular neural network task based on a neural network training rule. the neural network connections are formed between pre-synaptic and post-synaptic components of said physical neural network. the neural network generally includes dynamic and modifiable connections for adaptive signal processing. the neural network training mechanism can be based, for example, on the anti-hebbian and hebbian (ahah) rule and/or other plasticity rules.",2008-07-08,"The title of the patent is training of a physical neural network and its abstract is physical neural network systems and methods are disclosed. a physical neural network can be configured utilizing molecular technology, wherein said physical neural network comprises a plurality of molecular conductors, which form neural network connections thereof. a training mechanism can be provided for training said physical neural network to accomplish a particular neural network task based on a neural network training rule. the neural network connections are formed between pre-synaptic and post-synaptic components of said physical neural network. the neural network generally includes dynamic and modifiable connections for adaptive signal processing. the neural network training mechanism can be based, for example, on the anti-hebbian and hebbian (ahah) rule and/or other plasticity rules. dated 2008-07-08"
7400291,local positioning system which operates based on reflected wireless signals,"a local positioning system is proposed for wirelessly locating an object using existing features within a static environment, such as walls, as the references for determining the position of the system. an antenna 16 attached to the object transmits rf signals which are reflected by the surroundings. during a training mode, the reflected signals are used to train a neural network 22, 43 to map the position of the object to the characteristics of the reflected signals. during a working mode, the trained neural network is to identify the position of the object based on reflected signals in working mode. optionally, the reflected signals may be subject to a clustering process before input to the neural network.",2008-07-15,"The title of the patent is local positioning system which operates based on reflected wireless signals and its abstract is a local positioning system is proposed for wirelessly locating an object using existing features within a static environment, such as walls, as the references for determining the position of the system. an antenna 16 attached to the object transmits rf signals which are reflected by the surroundings. during a training mode, the reflected signals are used to train a neural network 22, 43 to map the position of the object to the characteristics of the reflected signals. during a working mode, the trained neural network is to identify the position of the object based on reflected signals in working mode. optionally, the reflected signals may be subject to a clustering process before input to the neural network. dated 2008-07-15"
7400588,dynamic rate adaptation using neural networks for transmitting video data,"an adaptative source rate control method and apparatus utilizing a neural network are presented for controlling a data transmission of a media object over a communication network. a back propagation method transmitting control parameters related to the operation of a communications network is used for to dynamically adjust the performance of the network, in view of network parameters being sourced to a point of transmission. the bit rate and quantization level of the data stream are dynamically adjusted and shaped by the neural network in response to control parameters.",2008-07-15,"The title of the patent is dynamic rate adaptation using neural networks for transmitting video data and its abstract is an adaptative source rate control method and apparatus utilizing a neural network are presented for controlling a data transmission of a media object over a communication network. a back propagation method transmitting control parameters related to the operation of a communications network is used for to dynamically adjust the performance of the network, in view of network parameters being sourced to a point of transmission. the bit rate and quantization level of the data stream are dynamically adjusted and shaped by the neural network in response to control parameters. dated 2008-07-15"
7400943,methods and systems for analyzing engine unbalance conditions,"methods and systems for analyzing engine unbalance conditions are disclosed. in one embodiment, a method includes receiving vibrational data from a plurality of locations distributed over an engine and a surrounding engine support structure, and inputting the vibrational data into a neural network inverse model. the neural network inverse model establishes a relationship between the vibrational data and an unbalance condition of the engine, and outputs diagnostic information indicating the unbalance condition of the engine. in a further embodiment, a method further includes subjecting the vibrational data to a fast fourier transformation to extract a desired once per revolution vibrational data prior to input to the neural network inverse model.",2008-07-15,"The title of the patent is methods and systems for analyzing engine unbalance conditions and its abstract is methods and systems for analyzing engine unbalance conditions are disclosed. in one embodiment, a method includes receiving vibrational data from a plurality of locations distributed over an engine and a surrounding engine support structure, and inputting the vibrational data into a neural network inverse model. the neural network inverse model establishes a relationship between the vibrational data and an unbalance condition of the engine, and outputs diagnostic information indicating the unbalance condition of the engine. in a further embodiment, a method further includes subjecting the vibrational data to a fast fourier transformation to extract a desired once per revolution vibrational data prior to input to the neural network inverse model. dated 2008-07-15"
7401056,method and apparatus for multivariable analysis of biological measurements,"in a method and apparatus for analyzing multivariable data sets, a general computerized platform is provided for evaluating the relationship between large number of measurements of sets of variables characterizing components of complex states of a system under induced stimulation or controlled conditions. the linked responses of variables and their temporal relations tell about the network of interactions and their hierarchy. processing of data sets by a simple neural network gives a matrix of weight parameters, that allow to identify fingerprints of complex states characterized by patterns of measured variable and estimate the interactions between the components characterized by the measured variables. the results are provided numerically and by color-coded presentation indicating dominating relations between variables and strongly responding variables. when applied to dynamic responses of a system, the analysis can construct a schematic hierarchical architecture of the network of interaction between the components of the studied system. applications in biology include analysis of measurements characterizing responses of molecular components in cells under changes induced by stimuli (e.g. drugs, growth factors, hormones, mutations or forced expression of a proteins), and identification of complex cellular states (e.g. proliferation, differentiation, transformation, starvation, necrosis, apoptosis, and the time dependencies of the above effects).",2008-07-15,"The title of the patent is method and apparatus for multivariable analysis of biological measurements and its abstract is in a method and apparatus for analyzing multivariable data sets, a general computerized platform is provided for evaluating the relationship between large number of measurements of sets of variables characterizing components of complex states of a system under induced stimulation or controlled conditions. the linked responses of variables and their temporal relations tell about the network of interactions and their hierarchy. processing of data sets by a simple neural network gives a matrix of weight parameters, that allow to identify fingerprints of complex states characterized by patterns of measured variable and estimate the interactions between the components characterized by the measured variables. the results are provided numerically and by color-coded presentation indicating dominating relations between variables and strongly responding variables. when applied to dynamic responses of a system, the analysis can construct a schematic hierarchical architecture of the network of interaction between the components of the studied system. applications in biology include analysis of measurements characterizing responses of molecular components in cells under changes induced by stimuli (e.g. drugs, growth factors, hormones, mutations or forced expression of a proteins), and identification of complex cellular states (e.g. proliferation, differentiation, transformation, starvation, necrosis, apoptosis, and the time dependencies of the above effects). dated 2008-07-15"
7403928,identify data sources for neural network,"a system, method, and device for identifying data sources for a neural network are disclosed. the exemplary system may have a module for determining load curves for each selected data set. the system may also have a module for determining a global difference measure and a global similarity measure for each load curve of each selected data set. the system may have a module for determining a set of data sets with lowest value global difference measure. the system may also have a module for determining a set of data sets with largest value global similarity measure. the system may also have a module for determining a union of the sets of lowest value difference measure and the sets of largest value similarity measure. the system may also have a module for determining for each set in the union one of a local similarity measure and a local difference measure and a module for selecting a set of reduced data sets based on one of the local similarity measure and the local difference measure.",2008-07-22,"The title of the patent is identify data sources for neural network and its abstract is a system, method, and device for identifying data sources for a neural network are disclosed. the exemplary system may have a module for determining load curves for each selected data set. the system may also have a module for determining a global difference measure and a global similarity measure for each load curve of each selected data set. the system may have a module for determining a set of data sets with lowest value global difference measure. the system may also have a module for determining a set of data sets with largest value global similarity measure. the system may also have a module for determining a union of the sets of lowest value difference measure and the sets of largest value similarity measure. the system may also have a module for determining for each set in the union one of a local similarity measure and a local difference measure and a module for selecting a set of reduced data sets based on one of the local similarity measure and the local difference measure. dated 2008-07-22"
7403929,apparatus and methods for evaluating hyperdocuments using a trained artificial neural network,"an embodiment of a computer implemented method for determining the disposition of a hyperdocument includes retrieving a hyperdocument from an information source, providing information about content of the hyperdocument to a trained artificial neural network (ann), the ann being capable of evaluating the information and providing results reflecting the evaluation, and determining the disposition of the hyperdocument based upon results of the ann.",2008-07-22,"The title of the patent is apparatus and methods for evaluating hyperdocuments using a trained artificial neural network and its abstract is an embodiment of a computer implemented method for determining the disposition of a hyperdocument includes retrieving a hyperdocument from an information source, providing information about content of the hyperdocument to a trained artificial neural network (ann), the ann being capable of evaluating the information and providing results reflecting the evaluation, and determining the disposition of the hyperdocument based upon results of the ann. dated 2008-07-22"
7403930,neural network for determining the endpoint in a process,"there is provided a system and method for pattern recognition of an endpoint curve for a dry etch process. the system trains a neural network with a group of training curves corresponding to the dry etch process, wherein the training curves contain normal and abnormal features. the system receives an endpoint curve at the neural network representing a dry etch process and detects an abnormal feature in the endpoint curve.",2008-07-22,"The title of the patent is neural network for determining the endpoint in a process and its abstract is there is provided a system and method for pattern recognition of an endpoint curve for a dry etch process. the system trains a neural network with a group of training curves corresponding to the dry etch process, wherein the training curves contain normal and abnormal features. the system receives an endpoint curve at the neural network representing a dry etch process and detects an abnormal feature in the endpoint curve. dated 2008-07-22"
7403931,artificial intelligence trending system,"a data processing system program to develop, train, and implement a neural network for identifying customers who represent a bad debt risk is disclosed. a feature vector is applied to a neural network to generate outputs that approximate the relative likelihood that customers who are the subjects of the records used to generate the feature vector will be a bad debt risk. statistical values relating categorical attributes of the customers to the likelihood of their becoming a bad debt risk are substituted for the categorical attributes, and the attributes are normalized before the feature vector is applied to the network. in one embodiment the customers are customers of a long distance service provider.",2008-07-22,"The title of the patent is artificial intelligence trending system and its abstract is a data processing system program to develop, train, and implement a neural network for identifying customers who represent a bad debt risk is disclosed. a feature vector is applied to a neural network to generate outputs that approximate the relative likelihood that customers who are the subjects of the records used to generate the feature vector will be a bad debt risk. statistical values relating categorical attributes of the customers to the likelihood of their becoming a bad debt risk are substituted for the categorical attributes, and the attributes are normalized before the feature vector is applied to the network. in one embodiment the customers are customers of a long distance service provider. dated 2008-07-22"
7406386,system and method for sensing and interpreting dynamic forces,"the present invention relates to a sensing system which is capable of discriminating types of causes of changing loads on a surface, such as the type of motion of a human subject. the system has wide ranging applications including sports performance (e.g. golf club swing analysis). the system comprises a deformable load bearing surface (2), a plurality of mutually spaced sensors (6), a processor (8) and an output (10). the sensors (6) are coupled through the deformation response of the surface (2) to an applied load (4) to receive local sensory data from the surface (2). the processor (8) is operatively coupled to the sensors (6) and is arranged to transform the sensory data into information data relating to a load (4) applied to the surface (2), e.g. by means of a neural network algorithm. in an alternative embodiment, a housing including the deformable load bearing surface (2) contains a flowable material (e.g. liquid) which flows in response to the deformation of the surface.",2008-07-29,"The title of the patent is system and method for sensing and interpreting dynamic forces and its abstract is the present invention relates to a sensing system which is capable of discriminating types of causes of changing loads on a surface, such as the type of motion of a human subject. the system has wide ranging applications including sports performance (e.g. golf club swing analysis). the system comprises a deformable load bearing surface (2), a plurality of mutually spaced sensors (6), a processor (8) and an output (10). the sensors (6) are coupled through the deformation response of the surface (2) to an applied load (4) to receive local sensory data from the surface (2). the processor (8) is operatively coupled to the sensors (6) and is arranged to transform the sensory data into information data relating to a load (4) applied to the surface (2), e.g. by means of a neural network algorithm. in an alternative embodiment, a housing including the deformable load bearing surface (2) contains a flowable material (e.g. liquid) which flows in response to the deformation of the surface. dated 2008-07-29"
7406417,method for conditioning a database for automatic speech processing,"a neural network can be trained for synthesizing or recognizing speech with the aid of a database produced by automatically matching graphemes and phonemes. first, graphemes and phonemes are matched for words which have the same number of graphemes and phonemes. next, graphemes and phonemes are matched for words that have more graphemes than phonemes in a series of steps that combine graphemes with preceding phonemes. then, graphemes and phonemes are matched for words that have fewer graphemes than phonemes. after each step, infrequent and unsuccessful matches made in the preceding step are are erased. after this process is completed, the database can be used to train the neural network and graphemes, or letters of a text can be converted into the corresponding phonemes with the aid of the trained neural network.",2008-07-29,"The title of the patent is method for conditioning a database for automatic speech processing and its abstract is a neural network can be trained for synthesizing or recognizing speech with the aid of a database produced by automatically matching graphemes and phonemes. first, graphemes and phonemes are matched for words which have the same number of graphemes and phonemes. next, graphemes and phonemes are matched for words that have more graphemes than phonemes in a series of steps that combine graphemes with preceding phonemes. then, graphemes and phonemes are matched for words that have fewer graphemes than phonemes. after each step, infrequent and unsuccessful matches made in the preceding step are are erased. after this process is completed, the database can be used to train the neural network and graphemes, or letters of a text can be converted into the corresponding phonemes with the aid of the trained neural network. dated 2008-07-29"
7409340,method and device for determining prosodic markers by neural autoassociators,a neural network is used to obtain more robust performance in determining prosodic markers on the basis of linguistic categories.,2008-08-05,The title of the patent is method and device for determining prosodic markers by neural autoassociators and its abstract is a neural network is used to obtain more robust performance in determining prosodic markers on the basis of linguistic categories. dated 2008-08-05
7409372,neural network trained with spatial errors,"a neural network is trained with input data. the neural network is used to rescale the input data. errors for the rescaled values are determined, and neighborhoods of the errors are used adjust connection weights of the neural network.",2008-08-05,"The title of the patent is neural network trained with spatial errors and its abstract is a neural network is trained with input data. the neural network is used to rescale the input data. errors for the rescaled values are determined, and neighborhoods of the errors are used adjust connection weights of the neural network. dated 2008-08-05"
7409373,pattern analysis system and method,"method and arrangement for providing a computerized system having an interface arrangement for interfacing a data source. the data source delivering data related to motion of a person, a memory arrangement for storing said data, a processor for processing the data, an artificial neural network (ann) using the processor, means for collecting a second set of data from the person, means for calculating one or several parameters distinctive of various features of said person, and means for feeding the parameter values to the ann trained to recognize the various features.",2008-08-05,"The title of the patent is pattern analysis system and method and its abstract is method and arrangement for providing a computerized system having an interface arrangement for interfacing a data source. the data source delivering data related to motion of a person, a memory arrangement for storing said data, a processor for processing the data, an artificial neural network (ann) using the processor, means for collecting a second set of data from the person, means for calculating one or several parameters distinctive of various features of said person, and means for feeding the parameter values to the ann trained to recognize the various features. dated 2008-08-05"
7412426,neural networks with learning and expression capability,a neural network comprising a plurality of neurons in which any one of the plurality of neurons is able to associate with itself or another neuron in the plurality of neurons via active connections to a further neuron in the plurality of neurons.,2008-08-12,The title of the patent is neural networks with learning and expression capability and its abstract is a neural network comprising a plurality of neurons in which any one of the plurality of neurons is able to associate with itself or another neuron in the plurality of neurons via active connections to a further neuron in the plurality of neurons. dated 2008-08-12
7412428,application of hebbian and anti-hebbian learning to nanotechnology-based physical neural networks,"methods and systems are disclosed herein in which a physical neural network can be configured utilizing nanotechnology. such a physical neural network can comprise a plurality of molecular conductors (e.g., nanoconductors) which form neural connections between pre-synaptic and post-synaptic components of the physical neural network. additionally, a learning mechanism can be applied for implementing hebbian learning via the physical neural network. such a learning mechanism can utilize a voltage gradient or voltage gradient dependencies to implement hebbian and/or anti-hebbian plasticity within the physical neural network. the learning mechanism can also utilize pre-synaptic and post-synaptic frequencies to provide hebbian and/or anti-hebbian learning within the physical neural network.",2008-08-12,"The title of the patent is application of hebbian and anti-hebbian learning to nanotechnology-based physical neural networks and its abstract is methods and systems are disclosed herein in which a physical neural network can be configured utilizing nanotechnology. such a physical neural network can comprise a plurality of molecular conductors (e.g., nanoconductors) which form neural connections between pre-synaptic and post-synaptic components of the physical neural network. additionally, a learning mechanism can be applied for implementing hebbian learning via the physical neural network. such a learning mechanism can utilize a voltage gradient or voltage gradient dependencies to implement hebbian and/or anti-hebbian plasticity within the physical neural network. the learning mechanism can also utilize pre-synaptic and post-synaptic frequencies to provide hebbian and/or anti-hebbian learning within the physical neural network. dated 2008-08-12"
7415311,system and method for adaptive control of uncertain nonlinear processes,"a computer system for controlling a nonlinear physical process. the computer system comprises a linear controller and a neural network. the linear controller receives a command signal for control of the nonlinear physical process and a measured output signal from the output of the nonlinear physical process. the linear controller generates a control signal based on the command signal, a measured output signal, and a fixed linear model for the process. the neural network receives the control signal from the linear controller and the measured output signal from the output of the nonlinear physical process. the neural network uses the measured output signal to modify the connection weights of the neural network. the neural network also generates a modified control signal supplied to the linear controller to iterate a fixed point solution for the modified control signal used to control the nonlinear physical process.",2008-08-19,"The title of the patent is system and method for adaptive control of uncertain nonlinear processes and its abstract is a computer system for controlling a nonlinear physical process. the computer system comprises a linear controller and a neural network. the linear controller receives a command signal for control of the nonlinear physical process and a measured output signal from the output of the nonlinear physical process. the linear controller generates a control signal based on the command signal, a measured output signal, and a fixed linear model for the process. the neural network receives the control signal from the linear controller and the measured output signal from the output of the nonlinear physical process. the neural network uses the measured output signal to modify the connection weights of the neural network. the neural network also generates a modified control signal supplied to the linear controller to iterate a fixed point solution for the modified control signal used to control the nonlinear physical process. dated 2008-08-19"
7418432,adaptive control system having direct output feedback and related apparatuses and methods,"an adaptive control system (acs) uses direct output feedback to control a plant. the acs uses direct adaptive output feedback control developed for highly uncertain nonlinear systems, that does not rely on state estimation. the approach is also applicable to systems of unknown, but bounded dimension, whose output has known, but otherwise arbitrary relative degree. this includes systems with both parameter uncertainty and unmodeled dynamics. the result is achieved by extending the universal function approximation property of linearly parameterized neural networks to model unknown system dynamics from input/output data. the network weight adaptation rule is derived from lyapunov stability analysis, and guarantees that the adapted weight errors and the tracking error are bounded.",2008-08-26,"The title of the patent is adaptive control system having direct output feedback and related apparatuses and methods and its abstract is an adaptive control system (acs) uses direct output feedback to control a plant. the acs uses direct adaptive output feedback control developed for highly uncertain nonlinear systems, that does not rely on state estimation. the approach is also applicable to systems of unknown, but bounded dimension, whose output has known, but otherwise arbitrary relative degree. this includes systems with both parameter uncertainty and unmodeled dynamics. the result is achieved by extending the universal function approximation property of linearly parameterized neural networks to model unknown system dynamics from input/output data. the network weight adaptation rule is derived from lyapunov stability analysis, and guarantees that the adapted weight errors and the tracking error are bounded. dated 2008-08-26"
7421104,method of automatically assessing skeletal age of hand radiographs,"a method of automatically assessing skeletal age of hand radiographs, comprising: providing a radiographic image of both two hands; cropping a first image of the left hand or the right hand; rotating the first image to make the fingertip of the medius point upwards the vertical; cropping a second image of the medius; segmenting the phalanges of the medius to acquire a third image; extracting a plurality of physiological features of the third image to acquire a first data; extracting a plurality of morphological features of the third image to obtain a second data; delivering the first data and the second data to a neural network for training; and outputting an assessment of the skeletal age.",2008-09-02,"The title of the patent is method of automatically assessing skeletal age of hand radiographs and its abstract is a method of automatically assessing skeletal age of hand radiographs, comprising: providing a radiographic image of both two hands; cropping a first image of the left hand or the right hand; rotating the first image to make the fingertip of the medius point upwards the vertical; cropping a second image of the medius; segmenting the phalanges of the medius to acquire a third image; extracting a plurality of physiological features of the third image to acquire a first data; extracting a plurality of morphological features of the third image to obtain a second data; delivering the first data and the second data to a neural network for training; and outputting an assessment of the skeletal age. dated 2008-09-02"
7424370,computational method for identifying adhesin and adhesin-like proteins of therapeutic potential,"a computational method for identifying adhesin and adhesin-like proteins, said method comprising steps of computing the sequence-based attributes of a neural network software wherein the attributes are (i) amino acid frequencies, (ii) multiplet frequency, (iii) dipeptide frequencies, (iv),charge composition, and (v) hydrophobic composition, training the artificial neural network (ann) for each of the computed five attributes, and identifying the adhesin and adhesin-like proteins having probability of being an adhesin (pad) as ≧0.51; a computer system for performing the method; and genes and proteins encoding adhesin and adhesin-like proteins.",2008-09-09,"The title of the patent is computational method for identifying adhesin and adhesin-like proteins of therapeutic potential and its abstract is a computational method for identifying adhesin and adhesin-like proteins, said method comprising steps of computing the sequence-based attributes of a neural network software wherein the attributes are (i) amino acid frequencies, (ii) multiplet frequency, (iii) dipeptide frequencies, (iv),charge composition, and (v) hydrophobic composition, training the artificial neural network (ann) for each of the computed five attributes, and identifying the adhesin and adhesin-like proteins having probability of being an adhesin (pad) as ≧0.51; a computer system for performing the method; and genes and proteins encoding adhesin and adhesin-like proteins. dated 2008-09-09"
7426501,nanotechnology neural network methods and systems,"a physical neural network is disclosed, which includes a connection network comprising a plurality of molecular conducting connections suspended within a connection gap formed between one or more input electrodes and one or more output electrodes. one or more molecular connections of the molecular conducting connections can be strengthened or weakened according to an application of an electric field across said connection gap. thus, a plurality of physical neurons can be formed from said molecular conducting connections of said connection network. additionally, a gate can be located adjacent said connection gap and which comes into contact with said connection network. the gate can be connected to logic circuitry which can activate or deactivate individual physical neurons among said plurality of physical neurons.",2008-09-16,"The title of the patent is nanotechnology neural network methods and systems and its abstract is a physical neural network is disclosed, which includes a connection network comprising a plurality of molecular conducting connections suspended within a connection gap formed between one or more input electrodes and one or more output electrodes. one or more molecular connections of the molecular conducting connections can be strengthened or weakened according to an application of an electric field across said connection gap. thus, a plurality of physical neurons can be formed from said molecular conducting connections of said connection network. additionally, a gate can be located adjacent said connection gap and which comes into contact with said connection network. the gate can be connected to logic circuitry which can activate or deactivate individual physical neurons among said plurality of physical neurons. dated 2008-09-16"
7428516,handwriting recognition using neural networks,"new neural networks for handwriting recognition may be build from existing neural networks. an existing neural network pre-trained for a starting language is chosen based on a desired target language. the neural network is modified so that it may be used to recognize characters of the target language, and the modified neural network is used in a handwriting recognizer for the target language. modification includes copying one or more of the primary outputs of the existing neural network. an appropriate starting language may be chosen based on the desired target language. in addition, a “super network” may be provided that is a relatively large neural network configured to recognize characters from a number of different languages. one may customize a handwriting recognizer using such a super network by programming a mask to block outputs from the super network that are not necessary for the language desired to be recognized.",2008-09-23,"The title of the patent is handwriting recognition using neural networks and its abstract is new neural networks for handwriting recognition may be build from existing neural networks. an existing neural network pre-trained for a starting language is chosen based on a desired target language. the neural network is modified so that it may be used to recognize characters of the target language, and the modified neural network is used in a handwriting recognizer for the target language. modification includes copying one or more of the primary outputs of the existing neural network. an appropriate starting language may be chosen based on the desired target language. in addition, a “super network” may be provided that is a relatively large neural network configured to recognize characters from a number of different languages. one may customize a handwriting recognizer using such a super network by programming a mask to block outputs from the super network that are not necessary for the language desired to be recognized. dated 2008-09-23"
7430308,computer aided diagnosis of mammographic microcalcification clusters,"computer aided diagnosis techniques in medical imaging are developed for the automated differentiation between benign and malignant lesions and go beyond computer aided detection by providing cancer likelihood for a detected lesion given image and/or patient characteristics. a computer aided detection and diagnosis algorithm for mammographic calcification clusters is developed and evaluated. the emphasis is on the diagnostic component although the algorithm includes automated detection, segmentation, and classification steps based on wavelet filters and artificial neural networks. classification features are selected primarily from descriptors of the morphology of the individual calcifications and the distribution of the cluster as well as patient's demographics as input to the network. te selected features are robust morphological and distributional descriptors, relatively insensitive to segmentation and detection errors such as false positive signals and variations among imaging sources or imaging equipment.",2008-09-30,"The title of the patent is computer aided diagnosis of mammographic microcalcification clusters and its abstract is computer aided diagnosis techniques in medical imaging are developed for the automated differentiation between benign and malignant lesions and go beyond computer aided detection by providing cancer likelihood for a detected lesion given image and/or patient characteristics. a computer aided detection and diagnosis algorithm for mammographic calcification clusters is developed and evaluated. the emphasis is on the diagnostic component although the algorithm includes automated detection, segmentation, and classification steps based on wavelet filters and artificial neural networks. classification features are selected primarily from descriptors of the morphology of the individual calcifications and the distribution of the cluster as well as patient's demographics as input to the network. te selected features are robust morphological and distributional descriptors, relatively insensitive to segmentation and detection errors such as false positive signals and variations among imaging sources or imaging equipment. dated 2008-09-30"
7430546,applications of an algorithm that mimics cortical processing,"an information processing system having neuron-like signal processors that are interconnected by synapse-like processing junctions that simulates and extends capabilities of biological neural networks. the information processing systems uses integrate-and-fire neurons and temporally asymmetric hebbian learning (spike timing-dependent learning) to adapt the synaptic strengths. the synaptic strengths of each neuron are guaranteed to become optimal during the course of learning either for estimating the parameters of a dynamic system (system identification) or for computing the first principal component. this neural network is well-suited for hardware implementations, since the learning rule for the synaptic strengths only requires computing either spike-time differences or correlations. such hardware implementation may be used for predicting and recognizing audiovisual information or for improving cortical processing by a prosthetic device.",2008-09-30,"The title of the patent is applications of an algorithm that mimics cortical processing and its abstract is an information processing system having neuron-like signal processors that are interconnected by synapse-like processing junctions that simulates and extends capabilities of biological neural networks. the information processing systems uses integrate-and-fire neurons and temporally asymmetric hebbian learning (spike timing-dependent learning) to adapt the synaptic strengths. the synaptic strengths of each neuron are guaranteed to become optimal during the course of learning either for estimating the parameters of a dynamic system (system identification) or for computing the first principal component. this neural network is well-suited for hardware implementations, since the learning rule for the synaptic strengths only requires computing either spike-time differences or correlations. such hardware implementation may be used for predicting and recognizing audiovisual information or for improving cortical processing by a prosthetic device. dated 2008-09-30"
7433851,system and method for inferring geological classes,a system for inferring geological classes from oilfield well input data is described using a neural network for inferring class probabilities and class sequencing knowledge and optimising the class probabilities according to the sequencing knowledge.,2008-10-07,The title of the patent is system and method for inferring geological classes and its abstract is a system for inferring geological classes from oilfield well input data is described using a neural network for inferring class probabilities and class sequencing knowledge and optimising the class probabilities according to the sequencing knowledge. dated 2008-10-07
7436994,system of using neural network to distinguish text and picture in images and method thereof,"this specification discloses a system of using a neural network to distinguish text and pictures in an image and the method thereof. using the knowledge of text recognition learned by the neural network in advance, images data of color brightness and gray levels in an image block are processed to generate a greatest text faith value. the system determines the text status of the image block by comparing a text threshold with the greatest text faith value. if the greatest text faith value is larger than the text threshold, then the image block is determined to contain text pixels; otherwise, the image block contains purely picture pixels. this achieves the goal of separating text and pictures in an image.",2008-10-14,"The title of the patent is system of using neural network to distinguish text and picture in images and method thereof and its abstract is this specification discloses a system of using a neural network to distinguish text and pictures in an image and the method thereof. using the knowledge of text recognition learned by the neural network in advance, images data of color brightness and gray levels in an image block are processed to generate a greatest text faith value. the system determines the text status of the image block by comparing a text threshold with the greatest text faith value. if the greatest text faith value is larger than the text threshold, then the image block is determined to contain text pixels; otherwise, the image block contains purely picture pixels. this achieves the goal of separating text and pictures in an image. dated 2008-10-14"
7437339,"pulse signal circuit, parallel processing circuit, pattern recognition system, and image input system","a synaptic connection element for connecting neuron elements inputs a plurality of pulsed signals from different neuron elements n1 through n4, effects a common modulation (time window integration or pulse phase/width modulation) on a plurality of predetermined signals among the plurality of pulse signals, and outputs the modulated pulse signals to different signal lines to a neuron element m1. a neural network for representing and processing pattern information by the pulse modulation is thereby downsized in scale.",2008-10-14,"The title of the patent is pulse signal circuit, parallel processing circuit, pattern recognition system, and image input system and its abstract is a synaptic connection element for connecting neuron elements inputs a plurality of pulsed signals from different neuron elements n1 through n4, effects a common modulation (time window integration or pulse phase/width modulation) on a plurality of predetermined signals among the plurality of pulse signals, and outputs the modulated pulse signals to different signal lines to a neuron element m1. a neural network for representing and processing pattern information by the pulse modulation is thereby downsized in scale. dated 2008-10-14"
7440857,method and a system for detecting and locating an adjustment error or a defect of a rotorcraft rotor,"the invention relates to a method of detecting and identifying a defect or an adjustment error of a rotorcraft rotor using an artificial neural network (ann), the rotor having a plurality of blades and a plurality of adjustment members associated with each blade; the network (ann) is a supervised competitive learning network (sson, scln, ssom) having an input to which vibration spectral data measured on the rotorcraft is applied, the network outputting data representative of which rotor blade presents a defect or an adjustment error or data representative of no defect, and where appropriate data representative of the type of defect that has been detected.",2008-10-21,"The title of the patent is method and a system for detecting and locating an adjustment error or a defect of a rotorcraft rotor and its abstract is the invention relates to a method of detecting and identifying a defect or an adjustment error of a rotorcraft rotor using an artificial neural network (ann), the rotor having a plurality of blades and a plurality of adjustment members associated with each blade; the network (ann) is a supervised competitive learning network (sson, scln, ssom) having an input to which vibration spectral data measured on the rotorcraft is applied, the network outputting data representative of which rotor blade presents a defect or an adjustment error or data representative of no defect, and where appropriate data representative of the type of defect that has been detected. dated 2008-10-21"
7444282,method of setting optimum-partitioned classified neural network and method and apparatus for automatic labeling using optimum-partitioned classified neural network,"a method of automatic labeling using an optimum-partitioned classified neural network includes searching for neural networks having minimum errors with respect to a number of l phoneme combinations from a number of k neural network combinations generated at an initial stage or updated, updating weights during learning of the k neural networks by k phoneme combination groups searched with the same neural networks, and composing an optimum-partitioned classified neural network combination using the k neural networks of which a total error sum has converged; and tuning a phoneme boundary of a first label file by using the phoneme combination group classification result and the optimum-partitioned classified neural network combination, and generating a final label file reflecting the tuning result.",2008-10-28,"The title of the patent is method of setting optimum-partitioned classified neural network and method and apparatus for automatic labeling using optimum-partitioned classified neural network and its abstract is a method of automatic labeling using an optimum-partitioned classified neural network includes searching for neural networks having minimum errors with respect to a number of l phoneme combinations from a number of k neural network combinations generated at an initial stage or updated, updating weights during learning of the k neural networks by k phoneme combination groups searched with the same neural networks, and composing an optimum-partitioned classified neural network combination using the k neural networks of which a total error sum has converged; and tuning a phoneme boundary of a first label file by using the phoneme combination group classification result and the optimum-partitioned classified neural network combination, and generating a final label file reflecting the tuning result. dated 2008-10-28"
7444311,system and method for real-time recognition of driving patterns,"system and method for real-time, automatic, recognition of large time-scale driving patterns employs a statistical pattern recognition framework, implemented by means of feed-forward neural network utilizing models developed for recognizing, for example, four classes of driving environments, namely highway, main road, suburban traffic and city traffic, from vehicle performance data. a vehicle control application effects changes in vehicle performance aspects based on the recognized driving environment.",2008-10-28,"The title of the patent is system and method for real-time recognition of driving patterns and its abstract is system and method for real-time, automatic, recognition of large time-scale driving patterns employs a statistical pattern recognition framework, implemented by means of feed-forward neural network utilizing models developed for recognizing, for example, four classes of driving environments, namely highway, main road, suburban traffic and city traffic, from vehicle performance data. a vehicle control application effects changes in vehicle performance aspects based on the recognized driving environment. dated 2008-10-28"
7447547,neural prosthesis based on photomechanical deflectors and tactile sensory cells,"an interface for selective excitation of a biological neural network is provided. the interface includes a microelectromechanical (mems) device having a deformable membrane, and a tactile-sensitive neural cell disposed on the deformable membrane. the cell on the deformable membrane senses motion or deformation of the membrane and provides a signal, responsive to membrane motion or deformation, to the biological neural network. preferably, the deformable membrane and cell have about equal areas, to provide selective excitation. an interface array including at least two such interfaces is also provided. a retinal prosthesis interface array having, in each element of the array, a photodiode within the mems device for electrostatically actuating the deformable membrane is also provided. for this alternative, the cells and deformable membranes are preferably transparent.",2008-11-04,"The title of the patent is neural prosthesis based on photomechanical deflectors and tactile sensory cells and its abstract is an interface for selective excitation of a biological neural network is provided. the interface includes a microelectromechanical (mems) device having a deformable membrane, and a tactile-sensitive neural cell disposed on the deformable membrane. the cell on the deformable membrane senses motion or deformation of the membrane and provides a signal, responsive to membrane motion or deformation, to the biological neural network. preferably, the deformable membrane and cell have about equal areas, to provide selective excitation. an interface array including at least two such interfaces is also provided. a retinal prosthesis interface array having, in each element of the array, a photodiode within the mems device for electrostatically actuating the deformable membrane is also provided. for this alternative, the cells and deformable membranes are preferably transparent. dated 2008-11-04"
7447614,methods and systems for modeling material behavior,"a method for modeling material behavior includes using empirical three dimensional non-uniform stress and strain data to train a self-organizing computational model such as a neural network. a laboratory device for measuring non-uniform stress and strain data from material includes an enclosure with an inclusion in it. as the enclosure is compressed, the inclusion induces a non-uniform state of stress and strain. a field testing device includes a body having a moveable section. when the body is inserted in a material and the moveable section moved, a non-uniform state of stress and strain can be characterized.",2008-11-04,"The title of the patent is methods and systems for modeling material behavior and its abstract is a method for modeling material behavior includes using empirical three dimensional non-uniform stress and strain data to train a self-organizing computational model such as a neural network. a laboratory device for measuring non-uniform stress and strain data from material includes an enclosure with an inclusion in it. as the enclosure is compressed, the inclusion induces a non-uniform state of stress and strain. a field testing device includes a body having a moveable section. when the body is inserted in a material and the moveable section moved, a non-uniform state of stress and strain can be characterized. dated 2008-11-04"
7447664,neural network predictive control cost function designer,"a method, a computer-readable medium, and a system for tuning a cost function to control an operational plant are provided. a plurality of cost function parameters is selected. predicted future states generated by the neural network model are selectively incorporated into the cost function, and an input weight is applied to a control input signal. a series of known signals are iteratively applied as control input inputs, and the cost output is calculated. a phase is taken of the control and plant outputs in response to each of the known signals and combined, thereby allowing effective combinations of the cost function parameters, the input weight, and the predicted future states to be identified.",2008-11-04,"The title of the patent is neural network predictive control cost function designer and its abstract is a method, a computer-readable medium, and a system for tuning a cost function to control an operational plant are provided. a plurality of cost function parameters is selected. predicted future states generated by the neural network model are selectively incorporated into the cost function, and an input weight is applied to a control input signal. a series of known signals are iteratively applied as control input inputs, and the cost output is calculated. a phase is taken of the control and plant outputs in response to each of the known signals and combined, thereby allowing effective combinations of the cost function parameters, the input weight, and the predicted future states to be identified. dated 2008-11-04"
7450052,object detection method and apparatus,"method and apparatus for detecting objects. in one embodiment, a person entering a secured zone is illuminated with low-power polarized radio waves. differently polarized waves which are reflected back from the person are collected. concealed weapons are detected by measuring various parameters of the reflected signals and then calculating various selected differences between them. these differences create patterns when plotted as a function of time. preferably a trained neural network pattern recognition program is then used to evaluate these patterns and autonomously render a decision on the presence of a weapon. an interrupted continuous wave system may be employed. multiple units may be used to detect various azimuthal angles and to improve accuracy.",2008-11-11,"The title of the patent is object detection method and apparatus and its abstract is method and apparatus for detecting objects. in one embodiment, a person entering a secured zone is illuminated with low-power polarized radio waves. differently polarized waves which are reflected back from the person are collected. concealed weapons are detected by measuring various parameters of the reflected signals and then calculating various selected differences between them. these differences create patterns when plotted as a function of time. preferably a trained neural network pattern recognition program is then used to evaluate these patterns and autonomously render a decision on the presence of a weapon. an interrupted continuous wave system may be employed. multiple units may be used to detect various azimuthal angles and to improve accuracy. dated 2008-11-11"
7450986,non-invasive method and apparatus for determining onset of physiological conditions,"the invention relates to the modelling and design of early warning systems for detecting medical conditions using physiological responses. the device comprises sensors for monitoring physiological parameters such as skin impedance, heart rate, and qt interval of a patient, means for establishing when those parameters change, the rate of change of the parameters, and a neural network processor for processing the information obtained by the sensors. the neural network processor is programmed with a fast learning algorithm. when the neural network establishes that a physiological condition is present in the patient an alarm signal will be generated. the invention extends to a method of non-invasive monitoring of a person using a neural network programmed with a fast learning algorithm. a non-invasive hypoglycaemia monitor is specifically described.",2008-11-11,"The title of the patent is non-invasive method and apparatus for determining onset of physiological conditions and its abstract is the invention relates to the modelling and design of early warning systems for detecting medical conditions using physiological responses. the device comprises sensors for monitoring physiological parameters such as skin impedance, heart rate, and qt interval of a patient, means for establishing when those parameters change, the rate of change of the parameters, and a neural network processor for processing the information obtained by the sensors. the neural network processor is programmed with a fast learning algorithm. when the neural network establishes that a physiological condition is present in the patient an alarm signal will be generated. the invention extends to a method of non-invasive monitoring of a person using a neural network programmed with a fast learning algorithm. a non-invasive hypoglycaemia monitor is specifically described. dated 2008-11-11"
7451122,empirical design of experiments using neural network models,methods and apparatus are provided pertaining to a design of experiments. the method comprises generating a data set from historical data; identifying and removing any fault data points in the data set so as to create a revised data set; supplying the data points from the revised data set into a nonlinear neural network model; and deriving a simulator model characterizing a relationship between the input variables and the output variables. the apparatus comprises means for generating a data set from historical data; means for identifying and removing any fault data points in the data set so as to create a revised data set; means for supplying the data points from the revised data set into a nonlinear neural network model; and means for deriving a simulator model characterizing a relationship between the input variables and the output variables.,2008-11-11,The title of the patent is empirical design of experiments using neural network models and its abstract is methods and apparatus are provided pertaining to a design of experiments. the method comprises generating a data set from historical data; identifying and removing any fault data points in the data set so as to create a revised data set; supplying the data points from the revised data set into a nonlinear neural network model; and deriving a simulator model characterizing a relationship between the input variables and the output variables. the apparatus comprises means for generating a data set from historical data; means for identifying and removing any fault data points in the data set so as to create a revised data set; means for supplying the data points from the revised data set into a nonlinear neural network model; and means for deriving a simulator model characterizing a relationship between the input variables and the output variables. dated 2008-11-11
7454388,device for the autonomous bootstrapping of useful information,"a discovery system employing a neural network, training within this system, that is stimulated to generate novel output patterns through various forms of perturbation applied to it, a critic neural network likewise capable of training in situ within this system, that learns to associate such novel patterns with their utility or value while triggering reinforcement learning of the more useful or valuable of these patterns within the former net. the device is capable of bootstrapping itself to progressively higher levels of adaptive or creative competence, starting from no learning whatsoever, through cumulative cycles of experimentation and learning. optional feedback mechanisms between the latter and former self-learning artificial neural networks are used to accelerate the convergence of this system toward useful concepts or plans of action.",2008-11-18,"The title of the patent is device for the autonomous bootstrapping of useful information and its abstract is a discovery system employing a neural network, training within this system, that is stimulated to generate novel output patterns through various forms of perturbation applied to it, a critic neural network likewise capable of training in situ within this system, that learns to associate such novel patterns with their utility or value while triggering reinforcement learning of the more useful or valuable of these patterns within the former net. the device is capable of bootstrapping itself to progressively higher levels of adaptive or creative competence, starting from no learning whatsoever, through cumulative cycles of experimentation and learning. optional feedback mechanisms between the latter and former self-learning artificial neural networks are used to accelerate the convergence of this system toward useful concepts or plans of action. dated 2008-11-18"
7457787,neural network component,"a neural network component includes a plurality of inputs, at least one processing element, at least one output, and a digital memory storing values at addresses respectively corresponding to the at least one processing element, wherein the at least one processing element is arranged to receive a value from the digital memory in response to an input signal, and is instructed to execute one of a plurality of operations by the value that is received from the digital memory.",2008-11-25,"The title of the patent is neural network component and its abstract is a neural network component includes a plurality of inputs, at least one processing element, at least one output, and a digital memory storing values at addresses respectively corresponding to the at least one processing element, wherein the at least one processing element is arranged to receive a value from the digital memory in response to an input signal, and is instructed to execute one of a plurality of operations by the value that is received from the digital memory. dated 2008-11-25"
7457788,reducing number of computations in a neural network modeling several data sets,"an approach that enables reducing the number of computations while modeling data sets using a neural network. to model a first system characterized by a data set, a determination is made as to whether the data elements of the data set follow a similar pattern as data elements of another data set. if such an another data set exists, the weights determined with a system associated with the another data set, are used as initial weights while modeling the first system. due to such a feature, number of computations in a neural network can be reduced while modeling several data sets.",2008-11-25,"The title of the patent is reducing number of computations in a neural network modeling several data sets and its abstract is an approach that enables reducing the number of computations while modeling data sets using a neural network. to model a first system characterized by a data set, a determination is made as to whether the data elements of the data set follow a similar pattern as data elements of another data set. if such an another data set exists, the weights determined with a system associated with the another data set, are used as initial weights while modeling the first system. due to such a feature, number of computations in a neural network can be reduced while modeling several data sets. dated 2008-11-25"
7458342,method and system for sootblowing optimization,"a controller determines and adjusts system parameters, including cleanliness levels or sootblower operating settings, that are useful for maintaining the cleanliness of a fossil fuel boiler at an efficient level. some embodiments use a direct controller to determine cleanliness levels and/or sootblower operating settings. some embodiments use an indirect controller, with a system model, to determine cleanliness levels and/or sootblower settings. the controller may use a model that is, for example, a neural network, or a mass energy balance, or a genetically programmed model. the controller uses input about the actual performance or state of the boiler for adaptation. the controller may operate in conjunction with a sootblower optimization system that controls the actual settings of the sootblowers. the controller may coordinate cleanliness settings for multiple sootblowers and/or across a plurality of heat zones in the boiler.",2008-12-02,"The title of the patent is method and system for sootblowing optimization and its abstract is a controller determines and adjusts system parameters, including cleanliness levels or sootblower operating settings, that are useful for maintaining the cleanliness of a fossil fuel boiler at an efficient level. some embodiments use a direct controller to determine cleanliness levels and/or sootblower operating settings. some embodiments use an indirect controller, with a system model, to determine cleanliness levels and/or sootblower settings. the controller may use a model that is, for example, a neural network, or a mass energy balance, or a genetically programmed model. the controller uses input about the actual performance or state of the boiler for adaptation. the controller may operate in conjunction with a sootblower optimization system that controls the actual settings of the sootblowers. the controller may coordinate cleanliness settings for multiple sootblowers and/or across a plurality of heat zones in the boiler. dated 2008-12-02"
7461036,method for controlling risk in a computer security artificial neural network expert system,"a computer implemented method for monitoring system events and providing real-time response to security threats. system data is collected by monitors in the computing system. the expert system of the present invention compares the data against information in a knowledge base to identify a security threat to a system resource in a form of a system event and an action for mitigating effects of the system event. a determination is made as to whether a threat risk value of the system event is greater than an action risk value of the action for mitigating the system event. if the threat risk value is greater, a determination is made as to whether a trust value set by a user is greater than the action risk value. if the trust value is greater, the expert system executes the action against the security threat.",2008-12-02,"The title of the patent is method for controlling risk in a computer security artificial neural network expert system and its abstract is a computer implemented method for monitoring system events and providing real-time response to security threats. system data is collected by monitors in the computing system. the expert system of the present invention compares the data against information in a knowledge base to identify a security threat to a system resource in a form of a system event and an action for mitigating effects of the system event. a determination is made as to whether a threat risk value of the system event is greater than an action risk value of the action for mitigating the system event. if the threat risk value is greater, a determination is made as to whether a trust value set by a user is greater than the action risk value. if the trust value is greater, the expert system executes the action against the security threat. dated 2008-12-02"
7467117,artificial intelligence and global normalization methods for genotyping,"described herein are systems and methods for normalizing data without the use of external controls. also described herein are systems and methods for analyzing cluster data, such as genotyping data, using an artificial neural network.",2008-12-16,"The title of the patent is artificial intelligence and global normalization methods for genotyping and its abstract is described herein are systems and methods for normalizing data without the use of external controls. also described herein are systems and methods for analyzing cluster data, such as genotyping data, using an artificial neural network. dated 2008-12-16"
7469203,wireless network hybrid simulation,"a simulation method and system partitions network traffic into background traffic and explicit traffic, wherein explicit traffic is processed in detail, and background traffic is processed at a more abstract level. the packets of explicit traffic are modeled in complete detail, so that precise timing and behavior characteristics can be determined, whereas large volumes of traffic are modeled more abstractly as background flows, and only certain aspects, such as routing through the network, are simulated. tracer packets are used to model the background traffic and carry a number of characteristics of interest for generating simulation results. in this manner, the effect of the background traffic on the explicit traffic can be modeled at each network element. the abstract processing of background traffic is facilitated by techniques that include multi-variate table look-up, neural networks, and the like.",2008-12-23,"The title of the patent is wireless network hybrid simulation and its abstract is a simulation method and system partitions network traffic into background traffic and explicit traffic, wherein explicit traffic is processed in detail, and background traffic is processed at a more abstract level. the packets of explicit traffic are modeled in complete detail, so that precise timing and behavior characteristics can be determined, whereas large volumes of traffic are modeled more abstractly as background flows, and only certain aspects, such as routing through the network, are simulated. tracer packets are used to model the background traffic and carry a number of characteristics of interest for generating simulation results. in this manner, the effect of the background traffic on the explicit traffic can be modeled at each network element. the abstract processing of background traffic is facilitated by techniques that include multi-variate table look-up, neural networks, and the like. dated 2008-12-23"
7469237,method and apparatus for fractal computation,"fractal computers are neural network architectures that exploit the characteristics of fractal attractors to perform general computation. this disclosure explains neural network implementations for each of the critical components of computation: composition, minimalization, and recursion. it then describes the creation of fractal attractors within these implementations by means of selective amplification or inhibition of input signals, and it describes how to estimate critical parameters for each implementation by using results from studies of fractal percolation. these implementation provide standardizable implicit alternatives to traditional neural network designs. consequently, fractal computers permit the exploitation of alternative technologies for computation based on dynamic systems with underlying fractal attractors.",2008-12-23,"The title of the patent is method and apparatus for fractal computation and its abstract is fractal computers are neural network architectures that exploit the characteristics of fractal attractors to perform general computation. this disclosure explains neural network implementations for each of the critical components of computation: composition, minimalization, and recursion. it then describes the creation of fractal attractors within these implementations by means of selective amplification or inhibition of input signals, and it describes how to estimate critical parameters for each implementation by using results from studies of fractal percolation. these implementation provide standardizable implicit alternatives to traditional neural network designs. consequently, fractal computers permit the exploitation of alternative technologies for computation based on dynamic systems with underlying fractal attractors. dated 2008-12-23"
7471997,landing-control device and landing-control method for aircraft,"a landing-control device is provided having a new structure for carrying out landing control of an aircraft. a detecting unit 10 detects at least a relative altitude from a landing surface to an aircraft. a parameter-generating unit 20 is constructed by a neural network having a feedback loop which receives a detection value detected by the detecting unit 10 and outputs a landing-control parameter of the aircraft, an output of a first node among plural nodes constituting the neural network being input to a second node different from the first node. a controlling unit 30 controls the aircraft based on the control parameter output from the parameter-generating unit 20.",2008-12-30,"The title of the patent is landing-control device and landing-control method for aircraft and its abstract is a landing-control device is provided having a new structure for carrying out landing control of an aircraft. a detecting unit 10 detects at least a relative altitude from a landing surface to an aircraft. a parameter-generating unit 20 is constructed by a neural network having a feedback loop which receives a detection value detected by the detecting unit 10 and outputs a landing-control parameter of the aircraft, an output of a first node among plural nodes constituting the neural network being input to a second node different from the first node. a controlling unit 30 controls the aircraft based on the control parameter output from the parameter-generating unit 20. dated 2008-12-30"
7472007,method of classifying vehicle occupants,"a method of classifying vehicle occupants utilizes a neural network engine having a state machine for determining if an occupant has changed preferably between an adult and adult, an adult and child, and a child and child from a pair of images. if no change has occurred, the method utilizes the prior occupant type and then decides if the occupant has changed in position. if no, the occupant is deemed static and the prior type is valid as a classification or output to preferably a vehicle restraint system. if the occupant has changed in position, a dynamic classification process is initiated by either an adult or a child dynamic classifier as dictated by the state machine. valid dynamic classifier outputs or classifications can be sent to the restraint system and invalid dynamic classifier outputs are sent to a static classifier for update of the occupant type.",2008-12-30,"The title of the patent is method of classifying vehicle occupants and its abstract is a method of classifying vehicle occupants utilizes a neural network engine having a state machine for determining if an occupant has changed preferably between an adult and adult, an adult and child, and a child and child from a pair of images. if no change has occurred, the method utilizes the prior occupant type and then decides if the occupant has changed in position. if no, the occupant is deemed static and the prior type is valid as a classification or output to preferably a vehicle restraint system. if the occupant has changed in position, a dynamic classification process is initiated by either an adult or a child dynamic classifier as dictated by the state machine. valid dynamic classifier outputs or classifications can be sent to the restraint system and invalid dynamic classifier outputs are sent to a static classifier for update of the occupant type. dated 2008-12-30"
7472097,employee selection via multiple neural networks,"a plurality of neural networks or other models can be used in employee selection technologies. a hiring recommendation can be based at least on processing performed by a plurality of neural networks. for example, parallel or series processing by neural networks can be performed. a neural network can be coupled to one or more other neural networks. a binary or other n-ary output can be generated by one or more of the neural networks. in a series arrangement, candidates can be processed sequentially in multiple stages, and those surviving the stages are recommended for hire.",2008-12-30,"The title of the patent is employee selection via multiple neural networks and its abstract is a plurality of neural networks or other models can be used in employee selection technologies. a hiring recommendation can be based at least on processing performed by a plurality of neural networks. for example, parallel or series processing by neural networks can be performed. a neural network can be coupled to one or more other neural networks. a binary or other n-ary output can be generated by one or more of the neural networks. in a series arrangement, candidates can be processed sequentially in multiple stages, and those surviving the stages are recommended for hire. dated 2008-12-30"
7478013,method and system of monitoring and prognostics,a neural network learns the operating modes of a system being monitored under normal operating conditions. anomalies can be automatically detected and learned. a control command can be issued or an alert can be issued in response thereto.,2009-01-13,The title of the patent is method and system of monitoring and prognostics and its abstract is a neural network learns the operating modes of a system being monitored under normal operating conditions. anomalies can be automatically detected and learned. a control command can be issued or an alert can be issued in response thereto. dated 2009-01-13
7483868,automatic neural-net model generation and maintenance,"method of incrementally forming and adaptively updating a neural net model are provided. a function approximation node is incrementally added to the neural net model. function parameters for the function approximation node are determined and function parameters of other nodes in the neural network model are updated, by using the function parameters of the other nodes prior to addition of the function approximation node to the neural network model.",2009-01-27,"The title of the patent is automatic neural-net model generation and maintenance and its abstract is method of incrementally forming and adaptively updating a neural net model are provided. a function approximation node is incrementally added to the neural net model. function parameters for the function approximation node are determined and function parameters of other nodes in the neural network model are updated, by using the function parameters of the other nodes prior to addition of the function approximation node to the neural network model. dated 2009-01-27"
7487066,classifying a work machine operation,"a method for analyzing the use of a work machine is disclosed. in one embodiment, the method may include providing a computer with a neural network on the work machine. further, the method may include inputting data to the computer, at least a portion of the data associated with a load experienced by one of the components of the work machine. the neural network, when executed by the computer may then classify a current operation of the work machine into one of a plurality of types of operations.",2009-02-03,"The title of the patent is classifying a work machine operation and its abstract is a method for analyzing the use of a work machine is disclosed. in one embodiment, the method may include providing a computer with a neural network on the work machine. further, the method may include inputting data to the computer, at least a portion of the data associated with a load experienced by one of the components of the work machine. the neural network, when executed by the computer may then classify a current operation of the work machine into one of a plurality of types of operations. dated 2009-02-03"
7487132,method for filtering content using neural networks,provided is a method for filtering communications received from over a network for a person-to-person communication program. a communication is received for the person-to person communication program. the communication is processed to determine predefined language statements. information on the determined language statements is inputted into a neural network to produce an output value. a determination is made as to whether the output value indicates that the communication is unacceptable. the communication is forwarded to the person-to-person communication program unchanged if the output value indicates that the communication is acceptable. an action is performed with respect to the communication upon determining that the communication is unacceptable that differs from the forwarding of the communication that occurs if the output value indicates that the communication is acceptable.,2009-02-03,The title of the patent is method for filtering content using neural networks and its abstract is provided is a method for filtering communications received from over a network for a person-to-person communication program. a communication is received for the person-to person communication program. the communication is processed to determine predefined language statements. information on the determined language statements is inputted into a neural network to produce an output value. a determination is made as to whether the output value indicates that the communication is unacceptable. the communication is forwarded to the person-to-person communication program unchanged if the output value indicates that the communication is acceptable. an action is performed with respect to the communication upon determining that the communication is unacceptable that differs from the forwarding of the communication that occurs if the output value indicates that the communication is acceptable. dated 2009-02-03
7490030,power modelling of a circuit,"stimulation signals (22) are applied to a first circuit model (20) and the power behaviour of the circuit being modelled is determined from the behaviour of the first circuit model (20). in parallel, the same stimulation signals (22) are applied to a second circuit model (26) and the state variable changes within that second circuit model are calculated. the calculated power behaviour and the calculated state variable changes are then applied as training data inputs to a self learning power model, such as a neural network (28), which learns the relationship between state variable changes between the second model (26) and power behaviour of the circuit being simulated. in this way, a detailed first circuit model (20) may be used to calculate power behaviour and to train a separate power model (28, 30) which once trained can be publicly released without having to release sensitive information within the first circuit model (20).",2009-02-10,"The title of the patent is power modelling of a circuit and its abstract is stimulation signals (22) are applied to a first circuit model (20) and the power behaviour of the circuit being modelled is determined from the behaviour of the first circuit model (20). in parallel, the same stimulation signals (22) are applied to a second circuit model (26) and the state variable changes within that second circuit model are calculated. the calculated power behaviour and the calculated state variable changes are then applied as training data inputs to a self learning power model, such as a neural network (28), which learns the relationship between state variable changes between the second model (26) and power behaviour of the circuit being simulated. in this way, a detailed first circuit model (20) may be used to calculate power behaviour and to train a separate power model (28, 30) which once trained can be publicly released without having to release sensitive information within the first circuit model (20). dated 2009-02-10"
7496376,outer loop power control method and apparatus for wireless communications systems,"outer loop power control (olpc) method and apparatus for mobile communications systems which allow rapid adjustment of the target desired signal to interference ratio (sirtarget) satisfying a target block error rate (blertarget). specifically, the outer loop power control method proposed herein is termed “outage-based olpc” and establishes that the target desired signal to interference ratio (sirtarget) is given as the sum of two components: the first component (siroutage-tgt) is calculated by means of a dynamic adjusting function, for example, a neural network which makes a quality criterion based on outage probabilities correspond with one based on the target block error rate (blertarget), taking as input the fading margins associated with the different outage probabilities considered; the other component (sirbler-tgt) is that which acts to correct the possible deviations in the target block error rate (blertarget) due to the non-ideal behavior of the previous component (siroutage-tgt).",2009-02-24,"The title of the patent is outer loop power control method and apparatus for wireless communications systems and its abstract is outer loop power control (olpc) method and apparatus for mobile communications systems which allow rapid adjustment of the target desired signal to interference ratio (sirtarget) satisfying a target block error rate (blertarget). specifically, the outer loop power control method proposed herein is termed “outage-based olpc” and establishes that the target desired signal to interference ratio (sirtarget) is given as the sum of two components: the first component (siroutage-tgt) is calculated by means of a dynamic adjusting function, for example, a neural network which makes a quality criterion based on outage probabilities correspond with one based on the target block error rate (blertarget), taking as input the fading margins associated with the different outage probabilities considered; the other component (sirbler-tgt) is that which acts to correct the possible deviations in the target block error rate (blertarget) due to the non-ideal behavior of the previous component (siroutage-tgt). dated 2009-02-24"
7496546,"interconnecting neural network system, interconnecting neural network structure construction method, self-organizing neural network structure construction method, and construction programs therefor","this invention provides an interconnecting neural network system capable of freely taking a network form for inputting a plurality of input vectors, and facilitating additionally training an artificial neural network structure. the artificial neural network structure is constructed by interconnecting rbf elements relating to each other among all rbf elements via a weight. each rbf element outputs an excitation strength according to a similarity between each input vector and a centroid vector based on a radius base function when the rbf element is excited by the input vector applied from an outside, and outputs a pseudo excitation strength obtained based on the excitation strength output from the other rbf element when the rbf element is excited in a chain reaction to excitation of the other neuron connected to the neuron.",2009-02-24,"The title of the patent is interconnecting neural network system, interconnecting neural network structure construction method, self-organizing neural network structure construction method, and construction programs therefor and its abstract is this invention provides an interconnecting neural network system capable of freely taking a network form for inputting a plurality of input vectors, and facilitating additionally training an artificial neural network structure. the artificial neural network structure is constructed by interconnecting rbf elements relating to each other among all rbf elements via a weight. each rbf element outputs an excitation strength according to a similarity between each input vector and a centroid vector based on a radius base function when the rbf element is excited by the input vector applied from an outside, and outputs a pseudo excitation strength obtained based on the excitation strength output from the other rbf element when the rbf element is excited in a chain reaction to excitation of the other neuron connected to the neuron. dated 2009-02-24"
7496547,handwriting recognition using a comparative neural network,"handwriting recognition techniques employing a personalized handwriting recognition engine. the recognition techniques use examples of an individual's previous writing style to help recognize new pen input from that individual. the techniques also employ a shape trainer to select samples of an individual's handwriting that accurately represent the individual's writing style, for use as prototypes to recognize subsequent handwriting from the individual. the techniques also alternately or additionally employ an intelligent combiner to combine the recognition results from the personalized recognition engine and the conventional recognition engine (or engines). the combiner may use a comparative neural network to combine the recognition results from multiple recognition engines. the combiner alternately may use a rule-based system based on prior knowledge of different recognition engines.",2009-02-24,"The title of the patent is handwriting recognition using a comparative neural network and its abstract is handwriting recognition techniques employing a personalized handwriting recognition engine. the recognition techniques use examples of an individual's previous writing style to help recognize new pen input from that individual. the techniques also employ a shape trainer to select samples of an individual's handwriting that accurately represent the individual's writing style, for use as prototypes to recognize subsequent handwriting from the individual. the techniques also alternately or additionally employ an intelligent combiner to combine the recognition results from the personalized recognition engine and the conventional recognition engine (or engines). the combiner may use a comparative neural network to combine the recognition results from multiple recognition engines. the combiner alternately may use a rule-based system based on prior knowledge of different recognition engines. dated 2009-02-24"
7496548,neural network for electronic search applications,"a system, method and computer program product for information searching includes (a) a first layer with a first plurality of neurons, each of the first plurality of neurons being associated with a word and with a set of connections to at least some neurons of the first layer; (b) a second layer with a second plurality of neurons, each of the second plurality of neurons being associated with an object and with a set of connections to at least some neurons of the second layer, and with a set of connections to some neurons of the first layer; (c) a third layer with a third plurality of neurons, each of the third plurality of neurons being associated with a sentence and with a set of connections to at least some neurons of the third layer, and with a set of connections to at least some neurons of the first layer and to at least some neurons of the second layer; and (d) a fourth layer with a fourth plurality of neurons, each of the fourth plurality of neurons being associated with a document and with a set of connections to at least some neurons of the fourth layer, and with a set of connections to at least some neurons of other layers. a query to the first layer identifies to a user, through the fourth layer, a set of documents that are contextually relevant to the query. each connection has a corresponding weight and optional flags.",2009-02-24,"The title of the patent is neural network for electronic search applications and its abstract is a system, method and computer program product for information searching includes (a) a first layer with a first plurality of neurons, each of the first plurality of neurons being associated with a word and with a set of connections to at least some neurons of the first layer; (b) a second layer with a second plurality of neurons, each of the second plurality of neurons being associated with an object and with a set of connections to at least some neurons of the second layer, and with a set of connections to some neurons of the first layer; (c) a third layer with a third plurality of neurons, each of the third plurality of neurons being associated with a sentence and with a set of connections to at least some neurons of the third layer, and with a set of connections to at least some neurons of the first layer and to at least some neurons of the second layer; and (d) a fourth layer with a fourth plurality of neurons, each of the fourth plurality of neurons being associated with a document and with a set of connections to at least some neurons of the fourth layer, and with a set of connections to at least some neurons of other layers. a query to the first layer identifies to a user, through the fourth layer, a set of documents that are contextually relevant to the query. each connection has a corresponding weight and optional flags. dated 2009-02-24"
7496561,method and system of ranking and clustering for document indexing and retrieval,"a relevancy ranking and clustering method and system that determines the relevance of a document relative to a user's query using a similarity comparison process. input queries are parsed into one or more query predicate structures using an ontological parser. the ontological parser parses a set of known documents to generate one or more document predicate structures. a comparison of each query predicate structure with each document predicate structure is performed to determine a matching degree, represented by a real number. a multilevel modifier strategy is implemented to assign different relevance values to the different parts of each predicate structure match to calculate the predicate structure's matching degree. the relevance of a document to a user's query is determined by calculating a similarity coefficient, based on the structures of each pair of query predicates and document predicates. documents are autonomously clustered using a self-organizing neural network that provides a coordinate system that makes judgments in a non-subjective fashion.",2009-02-24,"The title of the patent is method and system of ranking and clustering for document indexing and retrieval and its abstract is a relevancy ranking and clustering method and system that determines the relevance of a document relative to a user's query using a similarity comparison process. input queries are parsed into one or more query predicate structures using an ontological parser. the ontological parser parses a set of known documents to generate one or more document predicate structures. a comparison of each query predicate structure with each document predicate structure is performed to determine a matching degree, represented by a real number. a multilevel modifier strategy is implemented to assign different relevance values to the different parts of each predicate structure match to calculate the predicate structure's matching degree. the relevance of a document to a user's query is determined by calculating a similarity coefficient, based on the structures of each pair of query predicates and document predicates. documents are autonomously clustered using a self-organizing neural network that provides a coordinate system that makes judgments in a non-subjective fashion. dated 2009-02-24"
7499588,low resolution ocr for camera acquired documents,"a global optimization framework for optical character recognition (ocr) of low-resolution photographed documents that combines a binarization-type process, segmentation, and recognition into a single process. the framework includes a machine learning approach trained on a large amount of data. a convolutional neural network can be employed to compute a classification function at multiple positions and take grey-level input which eliminates binarization. the framework utilizes preprocessing, layout analysis, character recognition, and word recognition to output high recognition rates. the framework also employs dynamic programming and language models to arrive at the desired output.",2009-03-03,"The title of the patent is low resolution ocr for camera acquired documents and its abstract is a global optimization framework for optical character recognition (ocr) of low-resolution photographed documents that combines a binarization-type process, segmentation, and recognition into a single process. the framework includes a machine learning approach trained on a large amount of data. a convolutional neural network can be employed to compute a classification function at multiple positions and take grey-level input which eliminates binarization. the framework utilizes preprocessing, layout analysis, character recognition, and word recognition to output high recognition rates. the framework also employs dynamic programming and language models to arrive at the desired output. dated 2009-03-03"
7499894,cerebral programming,"""a method of training a biological neural network using a controller, comprising:    """,2009-03-03,"The title of the patent is cerebral programming and its abstract is ""a method of training a biological neural network using a controller, comprising:    "" dated 2009-03-03"
7502763,artificial neural network design and evaluation tool,"disclosed herein is a programming tool stored on a computer-readable medium and adapted for implementation by a computer for designing an artificial neural network. the programming tool includes a network configuration module to provide a first display interface to support configuration of the artificial neural network, and a pattern data module to provide a second display interface to support establishment and modification of first and second pattern data sets for training and testing the artificial neural network, respectively.",2009-03-10,"The title of the patent is artificial neural network design and evaluation tool and its abstract is disclosed herein is a programming tool stored on a computer-readable medium and adapted for implementation by a computer for designing an artificial neural network. the programming tool includes a network configuration module to provide a first display interface to support configuration of the artificial neural network, and a pattern data module to provide a second display interface to support establishment and modification of first and second pattern data sets for training and testing the artificial neural network, respectively. dated 2009-03-10"
7502766,neural networks decoder,"a method of training a neural network to perform decoding of a time-varying signal comprising a sequence of input symbols, which is coded by a coder such that each coded output symbol depends on more than one input symbol, characterized by repetitively: providing a plurality of successive input symbols to the neural network and to the coder, comparing the network outputs with the input signals; and adapting the network parameters to reduce the differences therebetween.",2009-03-10,"The title of the patent is neural networks decoder and its abstract is a method of training a neural network to perform decoding of a time-varying signal comprising a sequence of input symbols, which is coded by a coder such that each coded output symbol depends on more than one input symbol, characterized by repetitively: providing a plurality of successive input symbols to the neural network and to the coder, comparing the network outputs with the input signals; and adapting the network parameters to reduce the differences therebetween. dated 2009-03-10"
7502768,system and method for predicting building thermal loads,"a system for forecasting predicted thermal loads for a building comprises a thermal condition forecaster for forecasting weather conditions to be compensated by a building environmental control system and a thermal load predictor for modeling building environmental management system components to generate a predicted thermal load for a building for maintaining a set of environmental conditions. the thermal load predictor of the present invention is a neural network and, preferably, the neural network is a recurrent neural network that generates the predicted thermal load from short-term data. the recurrent neural network is trained by inputting building thermal mass data and building occupancy data for actual weather conditions and comparing the predicted thermal load generated by the recurrent neural network to the actual thermal load measured at the building. training error is attributed to weights of the neurons processing the building thermal mass data and building occupancy data. iteratively adjusting these weights to minimize the error optimizes the design of the recurrent neural network for these non-weather inputs.",2009-03-10,"The title of the patent is system and method for predicting building thermal loads and its abstract is a system for forecasting predicted thermal loads for a building comprises a thermal condition forecaster for forecasting weather conditions to be compensated by a building environmental control system and a thermal load predictor for modeling building environmental management system components to generate a predicted thermal load for a building for maintaining a set of environmental conditions. the thermal load predictor of the present invention is a neural network and, preferably, the neural network is a recurrent neural network that generates the predicted thermal load from short-term data. the recurrent neural network is trained by inputting building thermal mass data and building occupancy data for actual weather conditions and comparing the predicted thermal load generated by the recurrent neural network to the actual thermal load measured at the building. training error is attributed to weights of the neurons processing the building thermal mass data and building occupancy data. iteratively adjusting these weights to minimize the error optimizes the design of the recurrent neural network for these non-weather inputs. dated 2009-03-10"
7511819,light source for a downhole spectrometer,"the present invention provides an apparatus and method for high resolution spectroscopy using a narrow light beam source such as a superluminescent diode (sld) and a tunable optical filter (tof) for analyzing a formation fluid sample downhole and at the surface to determine formation fluid parameters. the sld and tof have a matching etendue. the analysis comprises determination of gas oil ratio, api gravity and various other fluid parameters which can be estimated after developing correlations to a training set of samples using a neural network or a chemometric equation.",2009-03-31,"The title of the patent is light source for a downhole spectrometer and its abstract is the present invention provides an apparatus and method for high resolution spectroscopy using a narrow light beam source such as a superluminescent diode (sld) and a tunable optical filter (tof) for analyzing a formation fluid sample downhole and at the surface to determine formation fluid parameters. the sld and tof have a matching etendue. the analysis comprises determination of gas oil ratio, api gravity and various other fluid parameters which can be estimated after developing correlations to a training set of samples using a neural network or a chemometric equation. dated 2009-03-31"
7512573,optical processor for an artificial neural network,"an optical processor adapted to emulate an artificial neural network (ann) having a plurality of interconnected layers, each layer having one or more artificial neurons, the processor having a spatial light modulator (slm) optically coupled, via an optical mask, to a photodetector array. in one embodiment, the slm has a plurality of pixels, each pixel being configurable to emulate an output portion of a corresponding artificial neuron in a signal-sending ann layer. the optical mask has a hologram that encodes the weights corresponding to interlayer connections in the ann and spatially modulates the light transmitted by the slm. the photodetectors of the array spatially resolve the interference pattern produced by the spatially modulated light, with each photodetector being configurable to emulate an input portion of a corresponding artificial neuron in a signal-receiving ann layer.",2009-03-31,"The title of the patent is optical processor for an artificial neural network and its abstract is an optical processor adapted to emulate an artificial neural network (ann) having a plurality of interconnected layers, each layer having one or more artificial neurons, the processor having a spatial light modulator (slm) optically coupled, via an optical mask, to a photodetector array. in one embodiment, the slm has a plurality of pixels, each pixel being configurable to emulate an output portion of a corresponding artificial neuron in a signal-sending ann layer. the optical mask has a hologram that encodes the weights corresponding to interlayer connections in the ann and spatially modulates the light transmitted by the slm. the photodetectors of the array spatially resolve the interference pattern produced by the spatially modulated light, with each photodetector being configurable to emulate an input portion of a corresponding artificial neuron in a signal-receiving ann layer. dated 2009-03-31"
7516022,method and system for assessing quality of spot welds,"a system and method for assessing the quality of spot weld joints between pieces of metal includes an ultrasound transducer probing a spot weld joint. the ultrasound transducer transmits ultrasonic radiation into the spot weld joint, receives corresponding echoes, and transforms the echoes into electrical signals. an image reconstructor connected to the ultrasound transducer transforms the electrical signals into numerical data representing an ultrasound image. a neural network connected to the image reconstructor analyzes the numerical data and an output system presents information representing the quality of the spot weld joint. the system is trained to assess the quality of spot weld joints by scanning a spot weld joint with an ultrasound transducer to produce the data set representing the joint; then physically deconstructing the joint to assess the joint quality.",2009-04-07,"The title of the patent is method and system for assessing quality of spot welds and its abstract is a system and method for assessing the quality of spot weld joints between pieces of metal includes an ultrasound transducer probing a spot weld joint. the ultrasound transducer transmits ultrasonic radiation into the spot weld joint, receives corresponding echoes, and transforms the echoes into electrical signals. an image reconstructor connected to the ultrasound transducer transforms the electrical signals into numerical data representing an ultrasound image. a neural network connected to the image reconstructor analyzes the numerical data and an output system presents information representing the quality of the spot weld joint. the system is trained to assess the quality of spot weld joints by scanning a spot weld joint with an ultrasound transducer to produce the data set representing the joint; then physically deconstructing the joint to assess the joint quality. dated 2009-04-07"
7519485,method and apparatus for determining energy savings by using a baseline energy use model that incorporates a neural network algorithm,"a computer-based system, computer-implemented method and computer program product facilitate determining energy cost savings in an energy-consuming facility, such as a commercial building, using a neural network model that projects or estimates the amount of energy that would have been consumed by the facility but for the implementation of energy efficiency or conservation measures. energy savings are represented by the difference between the estimate of energy that would have been consumed but for the measures and the actual amount of energy consumed by the facility under actual conditions during a time interval after the measures have been implemented.",2009-04-14,"The title of the patent is method and apparatus for determining energy savings by using a baseline energy use model that incorporates a neural network algorithm and its abstract is a computer-based system, computer-implemented method and computer program product facilitate determining energy cost savings in an energy-consuming facility, such as a commercial building, using a neural network model that projects or estimates the amount of energy that would have been consumed by the facility but for the implementation of energy efficiency or conservation measures. energy savings are represented by the difference between the estimate of energy that would have been consumed but for the measures and the actual amount of energy consumed by the facility under actual conditions during a time interval after the measures have been implemented. dated 2009-04-14"
7519488,signal processing method and system for noise removal and signal extraction,"a signal processing method and system combining smooth level wavelet pre-processing together with artificial neural networks all in the wavelet domain for signal denoising and extraction. upon receiving a signal corrupted with noise, an n-level decomposition of the signal is performed using a discrete wavelet transform to produce a smooth component and a rough component for each decomposition level. the nth level smooth component is then inputted into a corresponding neural network pre-trained to filter out noise in that component by pattern recognition in the wavelet domain. additional rough components, beginning at the highest level, may also be retained and inputted into corresponding neural networks pre-trained to filter out noise in those components also by pattern recognition in the wavelet domain. in any case, an inverse discrete wavelet transform is performed on the combined output from all the neural networks to recover a clean signal back in the time domain.",2009-04-14,"The title of the patent is signal processing method and system for noise removal and signal extraction and its abstract is a signal processing method and system combining smooth level wavelet pre-processing together with artificial neural networks all in the wavelet domain for signal denoising and extraction. upon receiving a signal corrupted with noise, an n-level decomposition of the signal is performed using a discrete wavelet transform to produce a smooth component and a rough component for each decomposition level. the nth level smooth component is then inputted into a corresponding neural network pre-trained to filter out noise in that component by pattern recognition in the wavelet domain. additional rough components, beginning at the highest level, may also be retained and inputted into corresponding neural networks pre-trained to filter out noise in those components also by pattern recognition in the wavelet domain. in any case, an inverse discrete wavelet transform is performed on the combined output from all the neural networks to recover a clean signal back in the time domain. dated 2009-04-14"
7526463,neural network using spatially dependent data for controlling a web-based process,"system and method for controlling a process with spatially dependent conditions for producing a product with spatially dependent properties, e.g., a web/sheet-based process for producing a web/sheet-based product. input data comprising a plurality of input data sets are provided to a neural network (analog or computer-based), each data set comprising values for one or more input parameters, each comprising a respective process condition or product property. the input data preserve spatial relationships of the input data. the neural network generates output data in accordance with the input data, the output data comprising a plurality of output data sets, each comprising values for one or more output parameters, each comprising a predicted process condition or product property. the output data preserve spatial relationships of the output data, which correspond to the spatial relationships of the input data. the output data are useable by a controller or operator to control the process.",2009-04-28,"The title of the patent is neural network using spatially dependent data for controlling a web-based process and its abstract is system and method for controlling a process with spatially dependent conditions for producing a product with spatially dependent properties, e.g., a web/sheet-based process for producing a web/sheet-based product. input data comprising a plurality of input data sets are provided to a neural network (analog or computer-based), each data set comprising values for one or more input parameters, each comprising a respective process condition or product property. the input data preserve spatial relationships of the input data. the neural network generates output data in accordance with the input data, the output data comprising a plurality of output data sets, each comprising values for one or more output parameters, each comprising a predicted process condition or product property. the output data preserve spatial relationships of the output data, which correspond to the spatial relationships of the input data. the output data are useable by a controller or operator to control the process. dated 2009-04-28"
7529703,method and system for artificial neural networks to predict price movements in the financial markets,"the present invention relates to methods and systems for devising and implementing automated artificial neural networks to predict market performance and direction movements of the u.s. treasury market, mortgage option-adjusted spreads (oas), interest rate swap spreads, and u.s. dollar/mexican peso exchange rate. the methods and systems of the present invention employ techniques used in actual neural networks naturally occurring in biological organisms to develop artificial neural network models for predicting movements in the financial market that are capable of extracting in a very consistent fashion non-linear relationships among input variables of the models that are readily apparent to the human traders.",2009-05-05,"The title of the patent is method and system for artificial neural networks to predict price movements in the financial markets and its abstract is the present invention relates to methods and systems for devising and implementing automated artificial neural networks to predict market performance and direction movements of the u.s. treasury market, mortgage option-adjusted spreads (oas), interest rate swap spreads, and u.s. dollar/mexican peso exchange rate. the methods and systems of the present invention employ techniques used in actual neural networks naturally occurring in biological organisms to develop artificial neural network models for predicting movements in the financial market that are capable of extracting in a very consistent fashion non-linear relationships among input variables of the models that are readily apparent to the human traders. dated 2009-05-05"
7529722,automatic creation of neuro-fuzzy expert system from online anlytical processing (olap) tools,"a method for automatic generation of a neuro-fuzzy expert system (fuzzy logic expert system implemented as a neural network) from data. the method comprising a data interface allowing description of location, type, and structure of the data. the interface also allows designation of input attributes and output attributes in the data structure; automatic neuro-fuzzy expert system generation driven by the data; training of the expert system's neural network on the data and the presentation of results which include new knowledge embedded in the parameters and structure of the trained neuro-fuzzy expert system to a user.",2009-05-05,"The title of the patent is automatic creation of neuro-fuzzy expert system from online anlytical processing (olap) tools and its abstract is a method for automatic generation of a neuro-fuzzy expert system (fuzzy logic expert system implemented as a neural network) from data. the method comprising a data interface allowing description of location, type, and structure of the data. the interface also allows designation of input attributes and output attributes in the data structure; automatic neuro-fuzzy expert system generation driven by the data; training of the expert system's neural network on the data and the presentation of results which include new knowledge embedded in the parameters and structure of the trained neuro-fuzzy expert system to a user. dated 2009-05-05"
7529743,gui for subject matter navigation using maps and search terms,"a system, method and computer program product for navigating categorized information, including (a) a two-dimensional map displayed to a user on a screen, the map showing search terms relating to a subject matter, where the display of the search terms corresponds to relationship between the terms, and wherein a manner of display of the terms corresponds to their relative importance to the subject matter; and (b) a neural network underlying the map, wherein the manner of display and a selection of the search terms is derived from the neural network. the manner of display includes font color, font size, font transparency, distance between search terms and positioning of the search terms within the map. positioning of a cursor over one of the search terms rearranges the search terms on the map to correspond to an increased relevance of the one of the search terms, while the cursor is over the one of the search terms. clicking on the one of the search terms corresponds to navigating into a sub-subject matter of the one of the search terms.",2009-05-05,"The title of the patent is gui for subject matter navigation using maps and search terms and its abstract is a system, method and computer program product for navigating categorized information, including (a) a two-dimensional map displayed to a user on a screen, the map showing search terms relating to a subject matter, where the display of the search terms corresponds to relationship between the terms, and wherein a manner of display of the terms corresponds to their relative importance to the subject matter; and (b) a neural network underlying the map, wherein the manner of display and a selection of the search terms is derived from the neural network. the manner of display includes font color, font size, font transparency, distance between search terms and positioning of the search terms within the map. positioning of a cursor over one of the search terms rearranges the search terms on the map to correspond to an increased relevance of the one of the search terms, while the cursor is over the one of the search terms. clicking on the one of the search terms corresponds to navigating into a sub-subject matter of the one of the search terms. dated 2009-05-05"
7536231,method for determining acceptability of proposed color solution using an artificial intelligence model,"a method for determining if a proposed color solution, such as paint, pigments, or dye formulations, is acceptable, is provided. the inputs to the system are the actual color values of an item, differential color values, a proposed color solution, and second color values associated with the proposed color solution. the system includes an artificial intelligence model to analyze the inputs and produce an output for communicating whether the proposed color solution is acceptable. the artificial intelligence model may be embodied in a neural network.",2009-05-19,"The title of the patent is method for determining acceptability of proposed color solution using an artificial intelligence model and its abstract is a method for determining if a proposed color solution, such as paint, pigments, or dye formulations, is acceptable, is provided. the inputs to the system are the actual color values of an item, differential color values, a proposed color solution, and second color values associated with the proposed color solution. the system includes an artificial intelligence model to analyze the inputs and produce an output for communicating whether the proposed color solution is acceptable. the artificial intelligence model may be embodied in a neural network. dated 2009-05-19"
7536232,model predictive control of air pollution control processes,"a controller for directing operation of an air pollution control system performing a process to control emissions of a pollutant has multiple process parameters (mpps). one or more of the mpps is a controllable process parameter (ctpp) and one of the mpps is an amount of the pollutant (aop) emitted by the system. a defined aop value (aopv) represents an objective or limit on an actual value (av) of the emitted aop. the controller includes either a neural network process model or a non-neural network process model representing a relationship between each ctpp and the emitted aop. a control processor has the logic to predict, based on the model, how changes to the current value of each ctpp will affect a future av of emitted aop, to select one of the changes in one ctpp based on the predicted affect of that change and on the aopv, and to direct control of the one ctpp in accordance with the selected change for that ctpp.",2009-05-19,"The title of the patent is model predictive control of air pollution control processes and its abstract is a controller for directing operation of an air pollution control system performing a process to control emissions of a pollutant has multiple process parameters (mpps). one or more of the mpps is a controllable process parameter (ctpp) and one of the mpps is an amount of the pollutant (aop) emitted by the system. a defined aop value (aopv) represents an objective or limit on an actual value (av) of the emitted aop. the controller includes either a neural network process model or a non-neural network process model representing a relationship between each ctpp and the emitted aop. a control processor has the logic to predict, based on the model, how changes to the current value of each ctpp will affect a future av of emitted aop, to select one of the changes in one ctpp based on the predicted affect of that change and on the aopv, and to direct control of the one ctpp in accordance with the selected change for that ctpp. dated 2009-05-19"
7536348,enhancing delinquent debt collection using statistical models of debt historical information and account events,"a predictive model, for example, a neural network, evaluates individual debt holder accounts and predicts the amount that will be collected on each account based on learned relationships among known variables. the predictive model is generated using historical data of delinquent debt accounts, the collection methods used to collect the debts in the accounts, and the success of the collection methods. in one embodiment, the predictive model is generated using profiles of delinquent debt accounts summarizing patterns of events in the accounts, and the success of the collection effort in each account. in another embodiment, the predictive model includes a mathematical representation of the collector's notes created during the collection period for each account.",2009-05-19,"The title of the patent is enhancing delinquent debt collection using statistical models of debt historical information and account events and its abstract is a predictive model, for example, a neural network, evaluates individual debt holder accounts and predicts the amount that will be collected on each account based on learned relationships among known variables. the predictive model is generated using historical data of delinquent debt accounts, the collection methods used to collect the debts in the accounts, and the success of the collection methods. in one embodiment, the predictive model is generated using profiles of delinquent debt accounts summarizing patterns of events in the accounts, and the success of the collection effort in each account. in another embodiment, the predictive model includes a mathematical representation of the collector's notes created during the collection period for each account. dated 2009-05-19"
7536370,inferential diagnosing engines for grid-based computing systems,"disclosed herein is the creation and utilization of automated diagnostic agents that are used by service engineers to diagnose faults, errors and other events or conditions within a grid-based computing system, and provide a derived list of suspect root causes for the events. related computerized processes and network architectures and systems supporting such agents are also disclosed. the automated diagnostic agents utilize software driven rules engines that operate on facts or data, such as telemetry and event information and data in particular, according to a set of rules. the rules engine utilize a neural network analysis environment to predict in accordance with the rules, facts and data found in the grid-based system to make probabilistic determinations about the grid. particular memory allocations, diagnostic process and subprocess interactions, and rule constructs are disclosed.",2009-05-19,"The title of the patent is inferential diagnosing engines for grid-based computing systems and its abstract is disclosed herein is the creation and utilization of automated diagnostic agents that are used by service engineers to diagnose faults, errors and other events or conditions within a grid-based computing system, and provide a derived list of suspect root causes for the events. related computerized processes and network architectures and systems supporting such agents are also disclosed. the automated diagnostic agents utilize software driven rules engines that operate on facts or data, such as telemetry and event information and data in particular, according to a set of rules. the rules engine utilize a neural network analysis environment to predict in accordance with the rules, facts and data found in the grid-based system to make probabilistic determinations about the grid. particular memory allocations, diagnostic process and subprocess interactions, and rule constructs are disclosed. dated 2009-05-19"
7539549,"motorized system integrated control and diagnostics using vibration, pressure, temperature, speed, and/or current analysis","systems and methods are disclosed for controlling and diagnosing the health of a motorized system. the systems may comprise a diagnostics system and a controller, wherein the diagnostics system employs a neural network, an expert system, and/or a data fusion component in order to assess the health of the motorized system according to one or more attributes associated therewith. the controller may operate the motorized system in accordance with a setpoint and/or a diagnostics signal from the diagnostics system. also disclosed are methodologies for controlling and diagnosing the health of a motorized system, comprising operating a motor in the motorized system in a controlled fashion, and diagnosing the health of the motorized system according to a measured attribute associated with the motorized system, wherein the motor may be operated according to a setpoint and/or the diagnostics signal.",2009-05-26,"The title of the patent is motorized system integrated control and diagnostics using vibration, pressure, temperature, speed, and/or current analysis and its abstract is systems and methods are disclosed for controlling and diagnosing the health of a motorized system. the systems may comprise a diagnostics system and a controller, wherein the diagnostics system employs a neural network, an expert system, and/or a data fusion component in order to assess the health of the motorized system according to one or more attributes associated therewith. the controller may operate the motorized system in accordance with a setpoint and/or a diagnostics signal from the diagnostics system. also disclosed are methodologies for controlling and diagnosing the health of a motorized system, comprising operating a motor in the motorized system in a controlled fashion, and diagnosing the health of the motorized system according to a measured attribute associated with the motorized system, wherein the motor may be operated according to a setpoint and/or the diagnostics signal. dated 2009-05-26"
7542950,method and apparatus for producing three dimensional shapes,a method and system for automatically producing data representative of a modified head shape from data representative of a deformed head is provided. the method includes a step of processing captured data for the deformed head utilizing principal component analysis (pca) to generate pca data representative of the deformed head. the method also includes the steps of providing the pca data as input to a neural network; and utilizing the neural network to process the pca data to provide data representative of a corresponding modified head shape.,2009-06-02,The title of the patent is method and apparatus for producing three dimensional shapes and its abstract is a method and system for automatically producing data representative of a modified head shape from data representative of a deformed head is provided. the method includes a step of processing captured data for the deformed head utilizing principal component analysis (pca) to generate pca data representative of the deformed head. the method also includes the steps of providing the pca data as input to a neural network; and utilizing the neural network to process the pca data to provide data representative of a corresponding modified head shape. dated 2009-06-02
7545975,three-dimensional object recognizing system,"a three-dimensional object recognizing system comprises a distance image generating portion for generating a distance image by using image pairs picked up by a stereoscopic camera, a grouping processing portion for grouping the distance data indicating the same three-dimensional object on the distance image, an input value setting portion for setting an area containing distance data group of grouped three-dimensional object on the distance image and also setting input values having typical distance data as elements every small area that is obtained by dividing the area by a set number of partition, a computing portion for computing output values having a pattern that responds to a previously set three-dimensional object by using a neural network that has at least the input values xin as inputs to an input layer, and a discriminating portion for discriminating the type of the three-dimensional object based on the pattern of the output values.",2009-06-09,"The title of the patent is three-dimensional object recognizing system and its abstract is a three-dimensional object recognizing system comprises a distance image generating portion for generating a distance image by using image pairs picked up by a stereoscopic camera, a grouping processing portion for grouping the distance data indicating the same three-dimensional object on the distance image, an input value setting portion for setting an area containing distance data group of grouped three-dimensional object on the distance image and also setting input values having typical distance data as elements every small area that is obtained by dividing the area by a set number of partition, a computing portion for computing output values having a pattern that responds to a previously set three-dimensional object by using a neural network that has at least the input values xin as inputs to an input layer, and a discriminating portion for discriminating the type of the three-dimensional object based on the pattern of the output values. dated 2009-06-09"
7546280,use of neural networks for keyword generation,"a system for identifying keywords in search results includes a plurality of neurons connected as a neural network, the neurons being associated with words and documents. an activity regulator regulates a minimum and/or maximum number of neurons of the neural network that are excited at any given time. means for displaying the neurons to a user and identifying the neurons that correspond to keywords can be provided. means for changing positions of the neurons relative to each other based on input from the user can be provided. the change in position of one neuron changes the keywords. the input from the user can be dragging a neuron on a display device, or changing a relevance of two neurons relative to each other. the neural network can be excited by a query that comprises words selected by a user. the neural network can be a bidirectional network. the user can inhibit neurons of the neural network by indicating irrelevance of a document. the neural network can be excited by a query that identifies a document considered relevant by a user. the neural network can also include neurons that represent groups of words. the neural network can be excited by a query that identifies a plurality of documents considered relevant by a user, and can output keywords associated with the plurality of documents.",2009-06-09,"The title of the patent is use of neural networks for keyword generation and its abstract is a system for identifying keywords in search results includes a plurality of neurons connected as a neural network, the neurons being associated with words and documents. an activity regulator regulates a minimum and/or maximum number of neurons of the neural network that are excited at any given time. means for displaying the neurons to a user and identifying the neurons that correspond to keywords can be provided. means for changing positions of the neurons relative to each other based on input from the user can be provided. the change in position of one neuron changes the keywords. the input from the user can be dragging a neuron on a display device, or changing a relevance of two neurons relative to each other. the neural network can be excited by a query that comprises words selected by a user. the neural network can be a bidirectional network. the user can inhibit neurons of the neural network by indicating irrelevance of a document. the neural network can be excited by a query that identifies a document considered relevant by a user. the neural network can also include neurons that represent groups of words. the neural network can be excited by a query that identifies a plurality of documents considered relevant by a user, and can output keywords associated with the plurality of documents. dated 2009-06-09"
7548894,artificial neural network,"an artificial neural network that can act like the real neural network according to the input history of signals input. the network includes a learning circuit that stores an input history of an input signal, an output circuit that is connected to the learning circuit, and a reset circuit that resets the input history stored in the learning circuit. the learning circuit changes a potential-change characteristic of an internal node included in the output circuit, according to the input history. the output circuit starts an output operation of data when a potential at the internal node exceeds a threshold value. the artificial neural network of this invention can operate almost in the same way as the real neural network, because it performs an output operation, such as an oscillating operation, in response to the history of the input signal.",2009-06-16,"The title of the patent is artificial neural network and its abstract is an artificial neural network that can act like the real neural network according to the input history of signals input. the network includes a learning circuit that stores an input history of an input signal, an output circuit that is connected to the learning circuit, and a reset circuit that resets the input history stored in the learning circuit. the learning circuit changes a potential-change characteristic of an internal node included in the output circuit, according to the input history. the output circuit starts an output operation of data when a potential at the internal node exceeds a threshold value. the artificial neural network of this invention can operate almost in the same way as the real neural network, because it performs an output operation, such as an oscillating operation, in response to the history of the input signal. dated 2009-06-16"
7550968,"measurement of wall thicknesses particularly of a blade, by eddy currents","a method for evaluating the wall thickness of a hollow part, of the turbomachine blade type, at least at a point having a determined radius of curvature at this point, within determined ranges of radii of curvature and thicknesses, including the determination of impedance values of an electrical circuit formed by an eddy current detector applied to the wall, and the insertion of these values into a digital processing unit with a neural network, wherein the network parameters have been defined in advance by learning on spacers having a determined radius of curvature and thickness in the ranges.",2009-06-23,"The title of the patent is measurement of wall thicknesses particularly of a blade, by eddy currents and its abstract is a method for evaluating the wall thickness of a hollow part, of the turbomachine blade type, at least at a point having a determined radius of curvature at this point, within determined ranges of radii of curvature and thicknesses, including the determination of impedance values of an electrical circuit formed by an eddy current detector applied to the wall, and the insertion of these values into a digital processing unit with a neural network, wherein the network parameters have been defined in advance by learning on spacers having a determined radius of curvature and thickness in the ranges. dated 2009-06-23"
7554296,method and apparatus for detecting charged state of secondary battery based on neural network calculation,"a neural network type of apparatus is provided to detect an internal state of a secondary battery implemented in a battery system. the apparatus comprises a detecting unit, producing unit and estimating unit. the detecting unit detects electric signals indicating an operating state of the battery. the producing unit produces, using the electric signals, an input parameter required for estimating the internal state of the battery. the input parameter reflects calibration of a present charged state of the battery which is attributable to at least one of a present degraded state of the battery and a difference in types of the battery. the estimating unit estimates an output parameter indicating the charged state of the battery by applying the input parameter to neural network calculation.",2009-06-30,"The title of the patent is method and apparatus for detecting charged state of secondary battery based on neural network calculation and its abstract is a neural network type of apparatus is provided to detect an internal state of a secondary battery implemented in a battery system. the apparatus comprises a detecting unit, producing unit and estimating unit. the detecting unit detects electric signals indicating an operating state of the battery. the producing unit produces, using the electric signals, an input parameter required for estimating the internal state of the battery. the input parameter reflects calibration of a present charged state of the battery which is attributable to at least one of a present degraded state of the battery and a difference in types of the battery. the estimating unit estimates an output parameter indicating the charged state of the battery by applying the input parameter to neural network calculation. dated 2009-06-30"
7555468,neural network-based node mobility and network connectivty predictions for mobile ad hoc radio networks,"a self managed ad hoc communications network nodes and node mobility management. nodes include an artificial neural network (ann) that determines connection to other network nodes. the ann may use free space propagation link life estimation, inverse modeling for partition prediction, stochastic approximation, and/or coarse estimation. the node includes storage storing network tables and matrices indicating network connectivity and connection to other nodes. also, a wireless communications unit provides for wireless communicating with other nodes.",2009-06-30,"The title of the patent is neural network-based node mobility and network connectivty predictions for mobile ad hoc radio networks and its abstract is a self managed ad hoc communications network nodes and node mobility management. nodes include an artificial neural network (ann) that determines connection to other network nodes. the ann may use free space propagation link life estimation, inverse modeling for partition prediction, stochastic approximation, and/or coarse estimation. the node includes storage storing network tables and matrices indicating network connectivity and connection to other nodes. also, a wireless communications unit provides for wireless communicating with other nodes. dated 2009-06-30"
7555469,reconfigurable neural network systems and methods utilizing fpgas having packet routers,"systems and methods are disclosed for forming reconfigurable neural networks with interconnected fpgas each having a packet router. neural network nodes are formed within the fpgas and connections between nodes within an fpga and connections to nodes external to the fpga are made using packet routers that are configured within each fpga. the fpgas can be connected to each other using high-speed interconnects, such as high-speed serial digital interconnects. the fpga arrays with packet routing allow for dynamic and reconfigurable neural networks to be formed thereby greatly improving the performance and intelligence of the neural network.",2009-06-30,"The title of the patent is reconfigurable neural network systems and methods utilizing fpgas having packet routers and its abstract is systems and methods are disclosed for forming reconfigurable neural networks with interconnected fpgas each having a packet router. neural network nodes are formed within the fpgas and connections between nodes within an fpga and connections to nodes external to the fpga are made using packet routers that are configured within each fpga. the fpgas can be connected to each other using high-speed interconnects, such as high-speed serial digital interconnects. the fpga arrays with packet routing allow for dynamic and reconfigurable neural networks to be formed thereby greatly improving the performance and intelligence of the neural network. dated 2009-06-30"
7565231,crash prediction network with graded warning for vehicle,"a method for facilitating the avoidance of a vehicle collision with an object includes the following steps: a) providing a neural network, b) evolving a good driver, c) evolving a crash predictor, and d) outputting a graded warning signal.",2009-07-21,"The title of the patent is crash prediction network with graded warning for vehicle and its abstract is a method for facilitating the avoidance of a vehicle collision with an object includes the following steps: a) providing a neural network, b) evolving a good driver, c) evolving a crash predictor, and d) outputting a graded warning signal. dated 2009-07-21"
7574410,fast 3d inversion of electromagnetic survey data using a trained neural network in the forward modeling branch,"a method for interpreting electromagnetic survey data includes acquiring electromagnetic survey data near a top of a portion of the earth's subsurface. an initial model of the portion of the earth's subsurface is generated. the model includes at least spatial distribution of formation resistivity within the portion. the initial model is applied to an artificial neural network trained to generate expected electromagnetic survey instrument response to the initial model. the acquired electromagnetic survey data are compared to an output of the artificial neural network. the initial model is adjusted, and the applying the model to the artificial neural network and the comparing are repeated until differences between the output of the network and the acquired survey data fall below a selected threshold.",2009-08-11,"The title of the patent is fast 3d inversion of electromagnetic survey data using a trained neural network in the forward modeling branch and its abstract is a method for interpreting electromagnetic survey data includes acquiring electromagnetic survey data near a top of a portion of the earth's subsurface. an initial model of the portion of the earth's subsurface is generated. the model includes at least spatial distribution of formation resistivity within the portion. the initial model is applied to an artificial neural network trained to generate expected electromagnetic survey instrument response to the initial model. the acquired electromagnetic survey data are compared to an output of the artificial neural network. the initial model is adjusted, and the applying the model to the artificial neural network and the comparing are repeated until differences between the output of the network and the acquired survey data fall below a selected threshold. dated 2009-08-11"
7576278,song search system and song search method,"a characteristic-data-extraction unit 13 extracts characteristic data containing changing information from song data, then an impression-data-conversion unit 14 uses a pre-learned hierarchical neural network to convert the characteristic data extracted by the characteristic-data-extraction unit 13 to impression data and stores it together with song data into a song database 15. a song search unit 18 searches the song database 15 based on impression data input from a pc-control unit 19, and outputs the search results to a search-results-output unit 21.",2009-08-18,"The title of the patent is song search system and song search method and its abstract is a characteristic-data-extraction unit 13 extracts characteristic data containing changing information from song data, then an impression-data-conversion unit 14 uses a pre-learned hierarchical neural network to convert the characteristic data extracted by the characteristic-data-extraction unit 13 to impression data and stores it together with song data into a song database 15. a song search unit 18 searches the song database 15 based on impression data input from a pc-control unit 19, and outputs the search results to a search-results-output unit 21. dated 2009-08-18"
7576797,automatic white balancing via illuminant scoring autoexposure by neural network mapping,"automatic white balancing and/or autoexposure as useful in a digital camera extracts color channel gains from comparisons of image colors with reference colors under various color temperature illuminants and/or extracts exposure settings from illuminance mean, illuminance variance, illuminance minimum, and illuminance maximum in areas of an image with a trained neural network.",2009-08-18,"The title of the patent is automatic white balancing via illuminant scoring autoexposure by neural network mapping and its abstract is automatic white balancing and/or autoexposure as useful in a digital camera extracts color channel gains from comparisons of image colors with reference colors under various color temperature illuminants and/or extracts exposure settings from illuminance mean, illuminance variance, illuminance minimum, and illuminance maximum in areas of an image with a trained neural network. dated 2009-08-18"
7577290,"image processing method, image processing apparatus and image processing program","the image for the face candidate region is obtained, extraction conditions are determined for extracting the eye and the mouth pixels from the face candidate region based on the size of the face candidate region. the eye pixels and the mouth pixels are then extracted from the face candidate region based on the determined extraction conditions, and then the area ratio of the eyes and the area ratio of mouth are calculated for the face candidate region based on the eye pixels and the mouth pixels. next the area ratio for the eyes and mouth is input into the neural network and a determinations is made as to whether the face candidate region is the face of a person using the neural network.",2009-08-18,"The title of the patent is image processing method, image processing apparatus and image processing program and its abstract is the image for the face candidate region is obtained, extraction conditions are determined for extracting the eye and the mouth pixels from the face candidate region based on the size of the face candidate region. the eye pixels and the mouth pixels are then extracted from the face candidate region based on the determined extraction conditions, and then the area ratio of the eyes and the area ratio of mouth are calculated for the face candidate region based on the eye pixels and the mouth pixels. next the area ratio for the eyes and mouth is input into the neural network and a determinations is made as to whether the face candidate region is the face of a person using the neural network. dated 2009-08-18"
7577623,method for controlling risk in a computer security artificial neural network expert system,"a computer implemented method, data processing system, and computer program product for monitoring system events and providing real-time response to security threats. system data is collected by monitors in the computing system. the expert system of the present invention compares the data against information in a knowledge base to identify a security threat to a system resource in a form of a system event and an action for mitigating effects of the system event. a determination is made as to whether a threat risk value of the system event is greater than an action risk value of the action for mitigating the system event. if the threat risk value is greater, a determination is made as to whether a trust value set by a user is greater than the action risk value. if the trust value is greater, the expert system executes the action against the security threat.",2009-08-18,"The title of the patent is method for controlling risk in a computer security artificial neural network expert system and its abstract is a computer implemented method, data processing system, and computer program product for monitoring system events and providing real-time response to security threats. system data is collected by monitors in the computing system. the expert system of the present invention compares the data against information in a knowledge base to identify a security threat to a system resource in a form of a system event and an action for mitigating effects of the system event. a determination is made as to whether a threat risk value of the system event is greater than an action risk value of the action for mitigating the system event. if the threat risk value is greater, a determination is made as to whether a trust value set by a user is greater than the action risk value. if the trust value is greater, the expert system executes the action against the security threat. dated 2009-08-18"
7577624,convergent construction of traditional scorecards,a neural model for simulating a scorecard comprises a neural network for transforming one or more inputs into an output. each input of the neural model has a squashing function applied thereto for simulating a bin of the simulated scorecard. the squashing function includes a control variable for controlling the steepness of the response to the squashing function's input so that during training of the neural model the steepness can be controlled. the output of the neural model represents the score of the simulated scorecard. the neural network is trained to behave like a scorecard by providing plurality of example values to the inputs of the neural network. each output score produced is compared to an expected score to produce an error value. each error value is back-propagated to adjust the neural network transformation to reduce the error value. the steepness of each squashing function is controlled using the respective control variable to affect the response of each squashing function.,2009-08-18,The title of the patent is convergent construction of traditional scorecards and its abstract is a neural model for simulating a scorecard comprises a neural network for transforming one or more inputs into an output. each input of the neural model has a squashing function applied thereto for simulating a bin of the simulated scorecard. the squashing function includes a control variable for controlling the steepness of the response to the squashing function's input so that during training of the neural model the steepness can be controlled. the output of the neural model represents the score of the simulated scorecard. the neural network is trained to behave like a scorecard by providing plurality of example values to the inputs of the neural network. each output score produced is compared to an expected score to produce an error value. each error value is back-propagated to adjust the neural network transformation to reduce the error value. the steepness of each squashing function is controlled using the respective control variable to affect the response of each squashing function. dated 2009-08-18
7577626,multi-scale radial basis function neural network,"a network architecture of radial basis function neural network system utilizes a blocking layer (4) to exclude successfully mapped neighborhoods from later node influence. a signal is inserted into the system at input nodes (i1, i2, . . . in), which then promulgates to a non-linear layer (2). the non-linear layer (2) comprises a number of non-linear activation function nodes (10). after passing through the non-linear layer (2), the signal passes through the blocking layer (4) that is comprised of either binary signal blocking nodes, or inverted symmetrical sigmoidal signal blocking nodes (12) that act in a binary fashion. finally, the signal is weighted by a weighting function (6a, 6b, 6c, 6n), summed at a summer (8) and outputted at (o).",2009-08-18,"The title of the patent is multi-scale radial basis function neural network and its abstract is a network architecture of radial basis function neural network system utilizes a blocking layer (4) to exclude successfully mapped neighborhoods from later node influence. a signal is inserted into the system at input nodes (i1, i2, . . . in), which then promulgates to a non-linear layer (2). the non-linear layer (2) comprises a number of non-linear activation function nodes (10). after passing through the non-linear layer (2), the signal passes through the blocking layer (4) that is comprised of either binary signal blocking nodes, or inverted symmetrical sigmoidal signal blocking nodes (12) that act in a binary fashion. finally, the signal is weighted by a weighting function (6a, 6b, 6c, 6n), summed at a summer (8) and outputted at (o). dated 2009-08-18"
7583059,apparatus and method for estimating state of charge of battery using neural network,"disclosed are an apparatus and a method for estimating a state of charge of a battery representing a non-linear characteristic by using a neural network. the apparatus includes a sensing section for detecting current, voltage and a temperature from a battery cell, a neural network performing a neural network algorithm and a learning algorithm based on data of the current, voltage and temperature transmitted thereto from the sensing section and present time data, thereby outputting the soc of the battery estimated through a final learning algorithm, and a comparator for comparing an output value of the neural network with a predetermined target value and making the neural network iteratively perform the learning algorithm if a difference between the output value of the neural network and the predetermined target value is out of a predetermined limit, thereby update the learning algorithm to generate the final learning algorithm. the state of charge of the battery is precisely estimated through the neural network algorithm.",2009-09-01,"The title of the patent is apparatus and method for estimating state of charge of battery using neural network and its abstract is disclosed are an apparatus and a method for estimating a state of charge of a battery representing a non-linear characteristic by using a neural network. the apparatus includes a sensing section for detecting current, voltage and a temperature from a battery cell, a neural network performing a neural network algorithm and a learning algorithm based on data of the current, voltage and temperature transmitted thereto from the sensing section and present time data, thereby outputting the soc of the battery estimated through a final learning algorithm, and a comparator for comparing an output value of the neural network with a predetermined target value and making the neural network iteratively perform the learning algorithm if a difference between the output value of the neural network and the predetermined target value is out of a predetermined limit, thereby update the learning algorithm to generate the final learning algorithm. the state of charge of the battery is precisely estimated through the neural network algorithm. dated 2009-09-01"
7587280,genomic data mining using clustering logic and filtering criteria,"a method for automatic analysis of genomic information in order to determine relationships among genes allows one to determine complex relationships among genes. first a clustering algorithm is chosen and is applied to the table, obtaining sub-tables of data relative to groups of genes that satisfy the chosen clustering criterion. therefore, all possible combinations of pair of sub-tables are generated and characteristic parameters are calculated for genes contained in these sub-tables. finally, for each combination a characteristic value is calculated with a decision algorithm defined in function of these parameters, by considering the genes of the combination as constituting a “gene network” if this characteristic value exceeds a pre-defined threshold. the method is preferably is implemented by a relative system of identification of groups of co-expressed and co-regulated genes comprising an intelligent fuzzy sub-system trained off-line identified by a neural network.",2009-09-08,"The title of the patent is genomic data mining using clustering logic and filtering criteria and its abstract is a method for automatic analysis of genomic information in order to determine relationships among genes allows one to determine complex relationships among genes. first a clustering algorithm is chosen and is applied to the table, obtaining sub-tables of data relative to groups of genes that satisfy the chosen clustering criterion. therefore, all possible combinations of pair of sub-tables are generated and characteristic parameters are calculated for genes contained in these sub-tables. finally, for each combination a characteristic value is calculated with a decision algorithm defined in function of these parameters, by considering the genes of the combination as constituting a “gene network” if this characteristic value exceeds a pre-defined threshold. the method is preferably is implemented by a relative system of identification of groups of co-expressed and co-regulated genes comprising an intelligent fuzzy sub-system trained off-line identified by a neural network. dated 2009-09-08"
7587373,neural network based well log synthesis with reduced usage of radioisotopic sources,logging systems and methods are disclosed to reduce usage of radioisotopic sources. some embodiments comprise collecting at least one output log of a training well bore from measurements with a radioisotopic source; collecting at least one input log of the training well bore from measurements by a non-radioisotopic logging tool; training a neural network to predict the output log from the at least one input log; collecting at least one input log of a development well bore from measurements by the non-radioisotopic logging tool; and processing the at least one input log of the development well bore to synthesize at least one output log of the development well bore. the output logs may include formation density and neutron porosity logs.,2009-09-08,The title of the patent is neural network based well log synthesis with reduced usage of radioisotopic sources and its abstract is logging systems and methods are disclosed to reduce usage of radioisotopic sources. some embodiments comprise collecting at least one output log of a training well bore from measurements with a radioisotopic source; collecting at least one input log of the training well bore from measurements by a non-radioisotopic logging tool; training a neural network to predict the output log from the at least one input log; collecting at least one input log of a development well bore from measurements by the non-radioisotopic logging tool; and processing the at least one input log of the development well bore to synthesize at least one output log of the development well bore. the output logs may include formation density and neutron porosity logs. dated 2009-09-08
7593535,neural network filtering techniques for compensating linear and non-linear distortion of an audio transducer,"neural networks provide efficient, robust and precise filtering techniques for compensating linear and non-linear distortion of an audio transducer such as a speaker, amplified broadcast antenna or perhaps a microphone. these techniques include both a method of characterizing the audio transducer to compute the inverse transfer functions and a method of implementing those inverse transfer functions for reproduction. the inverse transfer functions are preferably extracted using time domain calculations such as provided by linear and non-linear neural networks, which more accurately represent the properties of audio signals and the audio transducer than conventional frequency domain or modeling based approaches. although the preferred approach is to compensate for both linear and non-linear distortion, the neural network filtering techniques may be applied independently.",2009-09-22,"The title of the patent is neural network filtering techniques for compensating linear and non-linear distortion of an audio transducer and its abstract is neural networks provide efficient, robust and precise filtering techniques for compensating linear and non-linear distortion of an audio transducer such as a speaker, amplified broadcast antenna or perhaps a microphone. these techniques include both a method of characterizing the audio transducer to compute the inverse transfer functions and a method of implementing those inverse transfer functions for reproduction. the inverse transfer functions are preferably extracted using time domain calculations such as provided by linear and non-linear neural networks, which more accurately represent the properties of audio signals and the audio transducer than conventional frequency domain or modeling based approaches. although the preferred approach is to compensate for both linear and non-linear distortion, the neural network filtering techniques may be applied independently. dated 2009-09-22"
7593796,torque estimator for internal combustion engine,"an apparatus for the determination of engine torque comprises a neural network receiving engine operational data, such as crankshaft rotation data, and providing an output corresponding to engine torque. the neural network may be, for example, a recurrent neural network (rnn) that is configured using training data obtained using a training process. by comparing a determined engine torque with an intended engine torque, for example determined from engine control input values such as throttle position, a useful engine diagnostic is obtained.",2009-09-22,"The title of the patent is torque estimator for internal combustion engine and its abstract is an apparatus for the determination of engine torque comprises a neural network receiving engine operational data, such as crankshaft rotation data, and providing an output corresponding to engine torque. the neural network may be, for example, a recurrent neural network (rnn) that is configured using training data obtained using a training process. by comparing a determined engine torque with an intended engine torque, for example determined from engine control input values such as throttle position, a useful engine diagnostic is obtained. dated 2009-09-22"
7596535,apparatus for the classification of physiological events,"an apparatus according to the invention for the classification of physiological events has a signal input for the input of a physiological signal representing or constituting a physiological event and a classification unit 1 for classifying the physiological signal on the basis of its signal shape. the classification unit 1 includes a transformation unit 3 which is designed to carry out transformation of the physiological signal in such a way that as the output signal it outputs a number of values representing the physiological signal and based on the transformation; and a probabilistic neural network which is connected to the transformation unit 3 to receive the values and which contains a number of event classes which represent physiological events and which in turn are each represented by a set of comparative values, which probabilistic neural network is adapted on the basis of the comparison of the values with the comparative values to effect an association of the physiological signal represented by the values with one of the event classes.",2009-09-29,"The title of the patent is apparatus for the classification of physiological events and its abstract is an apparatus according to the invention for the classification of physiological events has a signal input for the input of a physiological signal representing or constituting a physiological event and a classification unit 1 for classifying the physiological signal on the basis of its signal shape. the classification unit 1 includes a transformation unit 3 which is designed to carry out transformation of the physiological signal in such a way that as the output signal it outputs a number of values representing the physiological signal and based on the transformation; and a probabilistic neural network which is connected to the transformation unit 3 to receive the values and which contains a number of event classes which represent physiological events and which in turn are each represented by a set of comparative values, which probabilistic neural network is adapted on the basis of the comparison of the values with the comparative values to effect an association of the physiological signal represented by the values with one of the event classes. dated 2009-09-29"
7603004,neural network demodulation for an optical sensor,methods and systems of neural network demodulation for an optical sensor. an optical sensor may be coupled to a structure and be capable of reflecting a reflected optical signal. a wavelength of the reflected optical signal may be spread based on a strain being applied to the structure. a replication device may receive the reflected optical signal from the optical sensor and produce a plurality of optical signals. a filter may be coupled to the replication device to receive an optical signal from the plurality of optical signals and filter the received optical signal. a detector may receive the filtered optical signal and provide a voltage output proportional to an amount of the filtered optical signal received. a neural network may receive the voltage output and determine the strain on the structure.,2009-10-13,The title of the patent is neural network demodulation for an optical sensor and its abstract is methods and systems of neural network demodulation for an optical sensor. an optical sensor may be coupled to a structure and be capable of reflecting a reflected optical signal. a wavelength of the reflected optical signal may be spread based on a strain being applied to the structure. a replication device may receive the reflected optical signal from the optical sensor and produce a plurality of optical signals. a filter may be coupled to the replication device to receive an optical signal from the plurality of optical signals and filter the received optical signal. a detector may receive the filtered optical signal and provide a voltage output proportional to an amount of the filtered optical signal received. a neural network may receive the voltage output and determine the strain on the structure. dated 2009-10-13
7603330,meta learning for question classification,"a system and a method are disclosed for automatic question classification and answering. a multipart artificial neural network (ann) comprising a main ann and an auxiliary ann classifies a received question according to one of a plurality of defined categories. unlabeled data is received from a source, such as a plurality of human volunteers. the unlabeled data comprises additional questions that might be asked of an autonomous machine such as a humanoid robot, and is used to train the auxiliary ann in an unsupervised mode. the unsupervised training can comprise multiple auxiliary tasks that generate labeled data from the unlabeled data, thereby learning an underlying structure. once the auxiliary ann has trained, the weights are frozen and transferred to the main ann. the main ann can then be trained using labeled questions. the original question to be answered is applied to the trained main ann, which assigns one of the defined categories. the assigned category is used to map the original question to a database that most likely contains the appropriate answer. an object and/or a property within the original question can be identified and used to formulate a query, using, for example, system query language (sql), to search for the answer within the chosen database. the invention makes efficient use of available information, and improves training time and error rate relative to use of single part anns.",2009-10-13,"The title of the patent is meta learning for question classification and its abstract is a system and a method are disclosed for automatic question classification and answering. a multipart artificial neural network (ann) comprising a main ann and an auxiliary ann classifies a received question according to one of a plurality of defined categories. unlabeled data is received from a source, such as a plurality of human volunteers. the unlabeled data comprises additional questions that might be asked of an autonomous machine such as a humanoid robot, and is used to train the auxiliary ann in an unsupervised mode. the unsupervised training can comprise multiple auxiliary tasks that generate labeled data from the unlabeled data, thereby learning an underlying structure. once the auxiliary ann has trained, the weights are frozen and transferred to the main ann. the main ann can then be trained using labeled questions. the original question to be answered is applied to the trained main ann, which assigns one of the defined categories. the assigned category is used to map the original question to a database that most likely contains the appropriate answer. an object and/or a property within the original question can be identified and used to formulate a query, using, for example, system query language (sql), to search for the answer within the chosen database. the invention makes efficient use of available information, and improves training time and error rate relative to use of single part anns. dated 2009-10-13"
7609877,tactical image parameter adjustment method for stereo pair correlation,a method for the timely processing of tactical images which allows a military pilot adequate time to track and engage a target. the method uses a correlation process and back propagated neural network to home in on a correct position parameters for the tactical image including the heading and the range for the tactical image.,2009-10-27,The title of the patent is tactical image parameter adjustment method for stereo pair correlation and its abstract is a method for the timely processing of tactical images which allows a military pilot adequate time to track and engage a target. the method uses a correlation process and back propagated neural network to home in on a correct position parameters for the tactical image including the heading and the range for the tactical image. dated 2009-10-27
7610185,gui for subject matter navigation using maps and search terms,"a system, method and computer program product for navigating categorized information, including (a) a two-dimensional map displayed to a user on a screen, the map showing search terms relating to a subject matter, where the display of the search terms corresponds to relationship between the terms, and wherein a manner of display of the terms corresponds to their relative importance to the subject matter; and (b) a neural network underlying the map, wherein the manner of display and a selection of the search terms is derived from the neural network. the manner of display includes font color, font size, font transparency, distance between search terms and positioning of the search terms within the map. positioning of a cursor over one of the search terms rearranges the search terms on the map to correspond to an increased relevance of the one of the search terms, while the cursor is over the one of the search terms. clicking on the one of the search terms corresponds to navigating into a sub-subject matter of the one of the search terms.",2009-10-27,"The title of the patent is gui for subject matter navigation using maps and search terms and its abstract is a system, method and computer program product for navigating categorized information, including (a) a two-dimensional map displayed to a user on a screen, the map showing search terms relating to a subject matter, where the display of the search terms corresponds to relationship between the terms, and wherein a manner of display of the terms corresponds to their relative importance to the subject matter; and (b) a neural network underlying the map, wherein the manner of display and a selection of the search terms is derived from the neural network. the manner of display includes font color, font size, font transparency, distance between search terms and positioning of the search terms within the map. positioning of a cursor over one of the search terms rearranges the search terms on the map to correspond to an increased relevance of the one of the search terms, while the cursor is over the one of the search terms. clicking on the one of the search terms corresponds to navigating into a sub-subject matter of the one of the search terms. dated 2009-10-27"
7613663,intelligent control with hierarchal stacked neural networks,"an intelligent control system based on an explicit model of cognitive development (table 1) performs high-level functions. it comprises up to o hierarchically stacked neural networks, nm, . . . , nm+(o−1), where m denotes the stage/order tasks performed in the first neural network, nm, and o denotes the highest stage/order tasks performed in the highest-level neural network. the type of processing actions performed in a network, nm, corresponds to the complexity for stage/order m. thus n1 performs tasks at the level corresponding to stage/order 1. n5 processes information at the level corresponding to stage/order 5. stacked neural networks begin and end at any stage/order, but information must be processed by each stage in ascending order sequence. stages/orders cannot be skipped. each neural network in a stack may use different architectures, interconnections, algorithms, and training methods, depending on the stage/order of the neural network and the type of intelligent control system implemented.",2009-11-03,"The title of the patent is intelligent control with hierarchal stacked neural networks and its abstract is an intelligent control system based on an explicit model of cognitive development (table 1) performs high-level functions. it comprises up to o hierarchically stacked neural networks, nm, . . . , nm+(o−1), where m denotes the stage/order tasks performed in the first neural network, nm, and o denotes the highest stage/order tasks performed in the highest-level neural network. the type of processing actions performed in a network, nm, corresponds to the complexity for stage/order m. thus n1 performs tasks at the level corresponding to stage/order 1. n5 processes information at the level corresponding to stage/order 5. stacked neural networks begin and end at any stage/order, but information must be processed by each stage in ascending order sequence. stages/orders cannot be skipped. each neural network in a stack may use different architectures, interconnections, algorithms, and training methods, depending on the stage/order of the neural network and the type of intelligent control system implemented. dated 2009-11-03"
7613665,ensembles of neural networks with different input sets,"methods of creating and using robust neural network ensembles are disclosed. some embodiments take the form of computer-based methods that comprise receiving a set of available inputs; receiving training data; training at least one neural network for each of at least two different subsets of the set of available inputs; and providing at least two trained neural networks having different subsets of the available inputs as components of a neural network ensemble configured to transform the available inputs into at least one output. the neural network ensemble may be applied as a log synthesis method that comprises: receiving a set of downhole logs; applying a first subset of downhole logs to a first neural network to obtain an estimated log; applying a second, different subset of the downhole logs to a second neural network to obtain an estimated log; and combining the estimated logs to obtain a synthetic log.",2009-11-03,"The title of the patent is ensembles of neural networks with different input sets and its abstract is methods of creating and using robust neural network ensembles are disclosed. some embodiments take the form of computer-based methods that comprise receiving a set of available inputs; receiving training data; training at least one neural network for each of at least two different subsets of the set of available inputs; and providing at least two trained neural networks having different subsets of the available inputs as components of a neural network ensemble configured to transform the available inputs into at least one output. the neural network ensemble may be applied as a log synthesis method that comprises: receiving a set of downhole logs; applying a first subset of downhole logs to a first neural network to obtain an estimated log; applying a second, different subset of the downhole logs to a second neural network to obtain an estimated log; and combining the estimated logs to obtain a synthetic log. dated 2009-11-03"
7616791,confidence determination,"the present invention relates to a method, a system, a neural network and a computer program product for determining the quality of an analytical process, preferably the confidence value. the analytical process is performed in a microchannel structure of a microfluidic device, from which data information of the analytic process is acquired by scanning at least one search area of the microchannel structure for signal data. the search area comprises the result of the analytical process and the acquired data information is stored as an image, one image for each scanned search area.",2009-11-10,"The title of the patent is confidence determination and its abstract is the present invention relates to a method, a system, a neural network and a computer program product for determining the quality of an analytical process, preferably the confidence value. the analytical process is performed in a microchannel structure of a microfluidic device, from which data information of the analytic process is acquired by scanning at least one search area of the microchannel structure for signal data. the search area comprises the result of the analytical process and the acquired data information is stored as an image, one image for each scanned search area. dated 2009-11-10"
7617101,method and system for utterance verification,"a method and system for utterance verification is disclosed. it first extracts a sequence of feature vectors from speech signal. at least one candidate string is obtained after speech recognition. then, speech signal is segmented into speech segments according to the verification-unit-specified structure of candidate string for making each speech segment corresponding to a verification unit. after calculating the verification feature vectors of speech segments, these verification feature vectors are sequentially used to generate verification scores of speech segments in verification process. this invention uses neural networks for calculating verification scores, where each neural network is a multi-layer perceptron (mlp) developed for each verification unit. verification score is obtained through using feed-forward process of mlp. finally, utterance verification score is obtained by combining all verification scores of speech segments and is used to compare with a pre-defined threshold for the decision of acceptance or rejection of the candidate string.",2009-11-10,"The title of the patent is method and system for utterance verification and its abstract is a method and system for utterance verification is disclosed. it first extracts a sequence of feature vectors from speech signal. at least one candidate string is obtained after speech recognition. then, speech signal is segmented into speech segments according to the verification-unit-specified structure of candidate string for making each speech segment corresponding to a verification unit. after calculating the verification feature vectors of speech segments, these verification feature vectors are sequentially used to generate verification scores of speech segments in verification process. this invention uses neural networks for calculating verification scores, where each neural network is a multi-layer perceptron (mlp) developed for each verification unit. verification score is obtained through using feed-forward process of mlp. finally, utterance verification score is obtained by combining all verification scores of speech segments and is used to compare with a pre-defined threshold for the decision of acceptance or rejection of the candidate string. dated 2009-11-10"
7617166,neural network for aeroelastic analysis,"a system and method of performing aeroelastic analysis using a neural network. input parameters, such as mass and location, contributing to aeroelastic characterization are determined and constrained. a model of a structure to be analyzed can be constructed. the model can include a number of locations where the input parameters can be varied. the aeroelastic characteristic of the structure can be analyzed using a finite element model to determine a number of output characteristics, each of which can correspond to at least one of a plurality of input samples. a neural network can be generated for determining the aeroelastic characteristic based on input parameters. the input sample/output characteristic pairs can be used to train the neural network. the weights and bias values from the trained neural network can be used to generate a non-linear transfer function that generates the aeroelastic characteristic in response to input parameters.",2009-11-10,"The title of the patent is neural network for aeroelastic analysis and its abstract is a system and method of performing aeroelastic analysis using a neural network. input parameters, such as mass and location, contributing to aeroelastic characterization are determined and constrained. a model of a structure to be analyzed can be constructed. the model can include a number of locations where the input parameters can be varied. the aeroelastic characteristic of the structure can be analyzed using a finite element model to determine a number of output characteristics, each of which can correspond to at least one of a plurality of input samples. a neural network can be generated for determining the aeroelastic characteristic based on input parameters. the input sample/output characteristic pairs can be used to train the neural network. the weights and bias values from the trained neural network can be used to generate a non-linear transfer function that generates the aeroelastic characteristic in response to input parameters. dated 2009-11-10"
7617415,code coverage quality estimator,"a method for estimating a quality of code coverage of a test is described. the method includes training a neural network, using the neural network to generate a risk factor for each code element, and determining a coverage quality based on risk factors of executed code elements and risk factors of unexecuted code elements. the neural network is trained by inputting suggestive data as input and error severity data as output. suggestive data may be data that correlates to a likelihood that a code element contains an error, and the error severity data is an evaluation of a severity of any error that was present in the code element. a coverage quality can be determined based on the risk factors of the code elements tested during the test and the risk factors of the code elements not tested during the test.",2009-11-10,"The title of the patent is code coverage quality estimator and its abstract is a method for estimating a quality of code coverage of a test is described. the method includes training a neural network, using the neural network to generate a risk factor for each code element, and determining a coverage quality based on risk factors of executed code elements and risk factors of unexecuted code elements. the neural network is trained by inputting suggestive data as input and error severity data as output. suggestive data may be data that correlates to a likelihood that a code element contains an error, and the error severity data is an evaluation of a severity of any error that was present in the code element. a coverage quality can be determined based on the risk factors of the code elements tested during the test and the risk factors of the code elements not tested during the test. dated 2009-11-10"
7620546,isolating speech signals utilizing neural networks,"a speech signal isolation system configured to isolate and reconstruct a speech signal transmitted in an environment in which frequency components of the speech signal are masked by background noise. the speech signal isolation system obtains a noisy speech signal from an audio source. the noisy speech signal may then be fed through a neural network that has been trained to isolate and reconstruct a clean speech signal from against background noise. once the noisy speech signal has been fed through the neural network, the speech signal isolation system generates an estimated speech signal with substantially reduced noise.",2009-11-17,"The title of the patent is isolating speech signals utilizing neural networks and its abstract is a speech signal isolation system configured to isolate and reconstruct a speech signal transmitted in an environment in which frequency components of the speech signal are masked by background noise. the speech signal isolation system obtains a noisy speech signal from an audio source. the noisy speech signal may then be fed through a neural network that has been trained to isolate and reconstruct a clean speech signal from against background noise. once the noisy speech signal has been fed through the neural network, the speech signal isolation system generates an estimated speech signal with substantially reduced noise. dated 2009-11-17"
7620607,system and method for using a bidirectional neural network to identify sentences for use as document annotations,"a system for generating annotations of a document, including a plurality of neurons connected as a neural network, the neurons being associated with words, sentences and documents. an activity regulator regulates a minimum and/or maximum number of neurons of the neural network that are excited at any given time. the neurons are displayed to a user and identify the neurons that correspond to sentences containing a predetermined percentage of document meaning. the annotations can be also based on a context of the user's search query. the query can include keywords, documents considered relevant by the user, or both. positions of the neurons relative to each other can be changed on a display device, based on input from the user, with the change in position of one neuron changing the resulting annotations. the input from the user can also include changing a relevance of neurons relative to each other, or indicating relevance or irrelevance of a document or sentence.",2009-11-17,"The title of the patent is system and method for using a bidirectional neural network to identify sentences for use as document annotations and its abstract is a system for generating annotations of a document, including a plurality of neurons connected as a neural network, the neurons being associated with words, sentences and documents. an activity regulator regulates a minimum and/or maximum number of neurons of the neural network that are excited at any given time. the neurons are displayed to a user and identify the neurons that correspond to sentences containing a predetermined percentage of document meaning. the annotations can be also based on a context of the user's search query. the query can include keywords, documents considered relevant by the user, or both. positions of the neurons relative to each other can be changed on a display device, based on input from the user, with the change in position of one neuron changing the resulting annotations. the input from the user can also include changing a relevance of neurons relative to each other, or indicating relevance or irrelevance of a document or sentence. dated 2009-11-17"
7620819,system and method for classifying regions of keystroke density with a neural network,"we develop a system consisting of a neural architecture resulting in classifying regions corresponding to users' keystroke patterns. we extend the adaptation properties to classification phase resulting in learning of changes over time. classification results on login attempts of 43 users (216 valid, 657 impersonation samples) show considerable improvements over existing methods.",2009-11-17,"The title of the patent is system and method for classifying regions of keystroke density with a neural network and its abstract is we develop a system consisting of a neural architecture resulting in classifying regions corresponding to users' keystroke patterns. we extend the adaptation properties to classification phase resulting in learning of changes over time. classification results on login attempts of 43 users (216 valid, 657 impersonation samples) show considerable improvements over existing methods. dated 2009-11-17"
7622929,pulse-discharge battery testing methods and apparatus,"a method for evaluating the conditions a battery comprises applying a discharge pulse to the battery and monitoring a response of the battery to the discharge pulse. in some embodiments a measure of battery condition is based at least in part on at least one of first and second parameters. the first parameter is related to the decrease in battery voltage after the onset of the discharge pulse. the second parameter is related to the recovery of the battery voltage after the discharge pulse. the first and/or second parameters may be supplied as inputs to an evaluation system such as a neural network, a fuzzy logic inference engine or the like.",2009-11-24,"The title of the patent is pulse-discharge battery testing methods and apparatus and its abstract is a method for evaluating the conditions a battery comprises applying a discharge pulse to the battery and monitoring a response of the battery to the discharge pulse. in some embodiments a measure of battery condition is based at least in part on at least one of first and second parameters. the first parameter is related to the decrease in battery voltage after the onset of the discharge pulse. the second parameter is related to the recovery of the battery voltage after the discharge pulse. the first and/or second parameters may be supplied as inputs to an evaluation system such as a neural network, a fuzzy logic inference engine or the like. dated 2009-11-24"
7630521,biometric identification apparatus and method using bio signals and artificial neural network,"a biometric identification apparatus and method using bio signals and an artificial neural network, are provided. the biometric identification apparatus includes: a periodic signal extraction unit which extracts one or more periodic signals from an input bio signal; a template calculation unit which calculates a template value using the extracted periodic signals; a template storage unit which stores a plurality of template values corresponding to a plurality of living bodies; and a reading unit which reads the template value that is most approximate to the template value calculated by the template calculation unit from the template storage unit. accordingly, it is possible to identify a living body by taking into consideration all of the characteristics of bio signals detected from the living body.",2009-12-08,"The title of the patent is biometric identification apparatus and method using bio signals and artificial neural network and its abstract is a biometric identification apparatus and method using bio signals and an artificial neural network, are provided. the biometric identification apparatus includes: a periodic signal extraction unit which extracts one or more periodic signals from an input bio signal; a template calculation unit which calculates a template value using the extracted periodic signals; a template storage unit which stores a plurality of template values corresponding to a plurality of living bodies; and a reading unit which reads the template value that is most approximate to the template value calculated by the template calculation unit from the template storage unit. accordingly, it is possible to identify a living body by taking into consideration all of the characteristics of bio signals detected from the living body. dated 2009-12-08"
7630943,method and system for indoor geolocation using an impulse response fingerprinting technique,"a system and method for predicting the location of a transmitter in an indoor zone of interest, including fixed receiver for receiving a signal from the transmitter, the receiver deriving a fingerprint from the received signal, and a trained neural network. the trained neural network predicts the transmitter location from the fingerprint. the method includes receiving a signal transmitted from the transmitter at a fixed-location receiver, deriving a fingerprint from the received signal, supplying the fingerprint to a trained neural network, and reading the predicted location from the neural network. the artificial neural network may further be trained and include a plurality of weights and biases is also shown. the method may include collecting a training data set of fingerprints and corresponding locations, inputting the training data set to the neural network, and adjusting the weights and biases by minimizing a sum of squares error function.",2009-12-08,"The title of the patent is method and system for indoor geolocation using an impulse response fingerprinting technique and its abstract is a system and method for predicting the location of a transmitter in an indoor zone of interest, including fixed receiver for receiving a signal from the transmitter, the receiver deriving a fingerprint from the received signal, and a trained neural network. the trained neural network predicts the transmitter location from the fingerprint. the method includes receiving a signal transmitted from the transmitter at a fixed-location receiver, deriving a fingerprint from the received signal, supplying the fingerprint to a trained neural network, and reading the predicted location from the neural network. the artificial neural network may further be trained and include a plurality of weights and biases is also shown. the method may include collecting a training data set of fingerprints and corresponding locations, inputting the training data set to the neural network, and adjusting the weights and biases by minimizing a sum of squares error function. dated 2009-12-08"
7639752,process for selecting a transmission channel and receiver of signals with antenna diversity,"a process for selecting a transmission channel from several transmission channels of a receiver of orthogonal frequency division multiplexing (ofdm) radio signals with antenna diversity, with a view to favoring the transmission channel delivering a signal with the lowest binary error rate, consists in estimating the binary error rate for each transmission channel by feeding a neural network with data radio frequency channel (rfc) representative of the frequency response of the transmission channel.",2009-12-29,"The title of the patent is process for selecting a transmission channel and receiver of signals with antenna diversity and its abstract is a process for selecting a transmission channel from several transmission channels of a receiver of orthogonal frequency division multiplexing (ofdm) radio signals with antenna diversity, with a view to favoring the transmission channel delivering a signal with the lowest binary error rate, consists in estimating the binary error rate for each transmission channel by feeding a neural network with data radio frequency channel (rfc) representative of the frequency response of the transmission channel. dated 2009-12-29"
7640067,estimated parameter based control of a process for controlling emission of a pollutant into the air,"a controller directs a process primarily performed to control emission of a particular pollutant into the air. the process has multiple process parameters (mpps), including a parameter representing an amount of the particular pollutant. the controller includes either a neural network process model or a non-neural network process model. in either case, the model represents a relationship between a first of the mpps and one or more of the other mpps. the one or more other mpps include a second of the mpps which is other than the parameter representing the amount of the emitted particular pollutant. also included is a processor configured with logic to estimate a value of the second mpp, and to direct control of the first mpp based on the estimated value of the second mpp and the model.",2009-12-29,"The title of the patent is estimated parameter based control of a process for controlling emission of a pollutant into the air and its abstract is a controller directs a process primarily performed to control emission of a particular pollutant into the air. the process has multiple process parameters (mpps), including a parameter representing an amount of the particular pollutant. the controller includes either a neural network process model or a non-neural network process model. in either case, the model represents a relationship between a first of the mpps and one or more of the other mpps. the one or more other mpps include a second of the mpps which is other than the parameter representing the amount of the emitted particular pollutant. also included is a processor configured with logic to estimate a value of the second mpp, and to direct control of the first mpp based on the estimated value of the second mpp and the model. dated 2009-12-29"
7643354,neural network model for instruments that store and retrieve sequential information,"a method and design is provided for distributing and storing sets of temporally ordered information in a systematic and sequential fashion. this method is based on a model of how the brain functions in the distribution and storage of temporally ordered memories, but it can also be applied to the design of new biological, electronic or optical devices. these devices may be used in the testing and development of new therapeutic drugs, in the detection of toxic agents or impaired performance, or in the development of new industrial and consumer devices in which the orderly storage of sequential information is important.",2010-01-05,"The title of the patent is neural network model for instruments that store and retrieve sequential information and its abstract is a method and design is provided for distributing and storing sets of temporally ordered information in a systematic and sequential fashion. this method is based on a model of how the brain functions in the distribution and storage of temporally ordered memories, but it can also be applied to the design of new biological, electronic or optical devices. these devices may be used in the testing and development of new therapeutic drugs, in the detection of toxic agents or impaired performance, or in the development of new industrial and consumer devices in which the orderly storage of sequential information is important. dated 2010-01-05"
7643925,automatic transmission clutch timing optimization apparatus and method,"a method for determining when in the course of a shift event an on-coming clutch gains torque capacity is provided. the method includes closed-loop controlling an off-going clutch to maintain a predetermined slip threshold by generating an off-going clutch pressure command, causing the on-coming clutch to engage during the closed loop control of the off-going clutch, generating a first derivative with respect to time of the off-going clutch pressure command, and using the first derivative to determine when the on-coming clutch gained torque capacity. a neural network method is preferably employed in analyzing the first derivative to locate a transition in the rate of commanded pressure indicative of off-going clutch release. a corresponding apparatus is also provided.",2010-01-05,"The title of the patent is automatic transmission clutch timing optimization apparatus and method and its abstract is a method for determining when in the course of a shift event an on-coming clutch gains torque capacity is provided. the method includes closed-loop controlling an off-going clutch to maintain a predetermined slip threshold by generating an off-going clutch pressure command, causing the on-coming clutch to engage during the closed loop control of the off-going clutch, generating a first derivative with respect to time of the off-going clutch pressure command, and using the first derivative to determine when the on-coming clutch gained torque capacity. a neural network method is preferably employed in analyzing the first derivative to locate a transition in the rate of commanded pressure indicative of off-going clutch release. a corresponding apparatus is also provided. dated 2010-01-05"
7646334,radar signal processing system and method for analyzing downhole oil and gas well environments,"a radar signal processing system utilizing a specialized neural network generative topographic mapping method which employs an alternative method for characterizing multiple radar signatures wherein specialized phased-codes and wavelet waveform modulations are used. the invention comprises three layers each including a plurality of nodes: input layer, mapping layer, and output presentation layer. the input layer takes the collected signals and congregates them into groups of linear data points. the mapping layer provides a method for projecting the data into points of lower dimensions. the data is assumed to arise by first probabilistically picking a point in a low-dimensional space, mapping the point to be observed in high-dimensional input space, then adding noise. the result is a system in which radar data with multiple reflections in close proximity is manipulated into multiple signatures in systematic ways for use in downhole structure signatures.",2010-01-12,"The title of the patent is radar signal processing system and method for analyzing downhole oil and gas well environments and its abstract is a radar signal processing system utilizing a specialized neural network generative topographic mapping method which employs an alternative method for characterizing multiple radar signatures wherein specialized phased-codes and wavelet waveform modulations are used. the invention comprises three layers each including a plurality of nodes: input layer, mapping layer, and output presentation layer. the input layer takes the collected signals and congregates them into groups of linear data points. the mapping layer provides a method for projecting the data into points of lower dimensions. the data is assumed to arise by first probabilistically picking a point in a low-dimensional space, mapping the point to be observed in high-dimensional input space, then adding noise. the result is a system in which radar data with multiple reflections in close proximity is manipulated into multiple signatures in systematic ways for use in downhole structure signatures. dated 2010-01-12"
7646940,robust indexing and retrieval of electronic ink,"a unique system and method that facilitates indexing and retrieving electronic ink objects with improved efficiency and accuracy is provided. handwritten words or characters are mapped to a low dimension through a process of segmentation, stroke classification using a neural network, and projection along directions found using opca, for example. the employment of opca makes these low dimensional representations robust to handwriting variations or noise. each handwritten word or set of characters is stored along with neighborhood hyperrectangle that represents word variations. redundant bit vectors are used to index the hyperrectangles for efficient storage and retrieval. ink-based queries can be submitted in order to retrieve at least one ink object. to do so, the ink query is processed to determine its query point which is represented by a (query) hyperrectangle. a data store can be searched for any hyperrectangles that match the query hyperrectangle.",2010-01-12,"The title of the patent is robust indexing and retrieval of electronic ink and its abstract is a unique system and method that facilitates indexing and retrieving electronic ink objects with improved efficiency and accuracy is provided. handwritten words or characters are mapped to a low dimension through a process of segmentation, stroke classification using a neural network, and projection along directions found using opca, for example. the employment of opca makes these low dimensional representations robust to handwriting variations or noise. each handwritten word or set of characters is stored along with neighborhood hyperrectangle that represents word variations. redundant bit vectors are used to index the hyperrectangles for efficient storage and retrieval. ink-based queries can be submitted in order to retrieve at least one ink object. to do so, the ink query is processed to determine its query point which is represented by a (query) hyperrectangle. a data store can be searched for any hyperrectangles that match the query hyperrectangle. dated 2010-01-12"
7647284,fixed-weight recurrent neural network controller with fixed long-term and adaptive short-term memory,"a controller for a plant having a fixed-weight recurrent neural network with at least one external input signal representative of a desired condition of the plant and actual condition of the plant, and an output connected as a control signal to the plant. the fixed recurrent neural network includes a set of nodes with fixed weight interconnections between the nodes and at least one feedback input interconnecting an output from at least one of the nodes to an input of at least one node. these nodes collectively determine the value of the output from the neural network as a function of the input signal and the feedback input. the controller also includes an adaptive neural network having a plurality of nodes with variable weight interconnections between the nodes. a cost input from the plant is connected to the adaptive neural network while an output from the adaptive neural network is coupled as a processed feedback signal to nodes of the fixed-weight recurrent neural network.",2010-01-12,"The title of the patent is fixed-weight recurrent neural network controller with fixed long-term and adaptive short-term memory and its abstract is a controller for a plant having a fixed-weight recurrent neural network with at least one external input signal representative of a desired condition of the plant and actual condition of the plant, and an output connected as a control signal to the plant. the fixed recurrent neural network includes a set of nodes with fixed weight interconnections between the nodes and at least one feedback input interconnecting an output from at least one of the nodes to an input of at least one node. these nodes collectively determine the value of the output from the neural network as a function of the input signal and the feedback input. the controller also includes an adaptive neural network having a plurality of nodes with variable weight interconnections between the nodes. a cost input from the plant is connected to the adaptive neural network while an output from the adaptive neural network is coupled as a processed feedback signal to nodes of the fixed-weight recurrent neural network. dated 2010-01-12"
7657313,adaptive cardiac resynchronization therapy system,"a system including a learning module and an algorithmic module for learning a physiological aspect of a patient body and regulating the delivery of a physiological agent to the body. an embodiment of the invention is an adaptive crt device performing biventricular pacing in which the av delay and vv interval parameters are changed dynamically according to the information supplied by the iegm, hemodynamic sensor and online processed data, in order to achieve optimal hemodynamic performance.a learning module, preferably using artificial neural network, performs the adaptive part of the algorithm supervised by an algorithmic deterministic module, internally or externally from the implanted pacemaker or defibrillator.",2010-02-02,"The title of the patent is adaptive cardiac resynchronization therapy system and its abstract is a system including a learning module and an algorithmic module for learning a physiological aspect of a patient body and regulating the delivery of a physiological agent to the body. an embodiment of the invention is an adaptive crt device performing biventricular pacing in which the av delay and vv interval parameters are changed dynamically according to the information supplied by the iegm, hemodynamic sensor and online processed data, in order to achieve optimal hemodynamic performance.a learning module, preferably using artificial neural network, performs the adaptive part of the algorithm supervised by an algorithmic deterministic module, internally or externally from the implanted pacemaker or defibrillator. dated 2010-02-02"
7660437,neural network systems for vehicles,"method for obtaining information about an occupying item in a space in a vehicle in accordance with the invention includes obtaining images of an area above a seat in the vehicle in which the occupying item is situated and classifying the occupying item by inputting signals derived from the images into a trained neural network form which is trained to output an indication of the class of occupying item from one of a predetermined number of possible classes. the method is applicable for various vehicles including automobiles, trucks, buses, airplanes and boats. the images may be pre-processed to remove background portions of the images and then converted into signals for input into the neural network form.",2010-02-09,"The title of the patent is neural network systems for vehicles and its abstract is method for obtaining information about an occupying item in a space in a vehicle in accordance with the invention includes obtaining images of an area above a seat in the vehicle in which the occupying item is situated and classifying the occupying item by inputting signals derived from the images into a trained neural network form which is trained to output an indication of the class of occupying item from one of a predetermined number of possible classes. the method is applicable for various vehicles including automobiles, trucks, buses, airplanes and boats. the images may be pre-processed to remove background portions of the images and then converted into signals for input into the neural network form. dated 2010-02-09"
7660713,systems and methods that detect a desired signal via a linear discriminative classifier that utilizes an estimated posterior signal-to-noise ratio (snr),"the present invention provides systems and methods for signal detection and enhancement. the systems and methods utilize one or more discriminative classifiers (e.g., a logistic regression model and a convolutional neural network) to estimate a posterior probability that indicates whether a desired signal is present in a received signal. the discriminative estimators generate the estimated probability based on one or more signal-to-noise ratio (snrs) (e.g., a normalized logarithmic posterior snr (nlpsnr) and a mel-transformed nlpsnr (mel-nlpsnr)) and an estimated noise model. depending on the resolution desired, the estimated snr can be generated at a frame level or at an atom level, wherein the atom level estimates are utilized to generate the frame level estimate. the novel systems and methods can be utilized to facilitate speech detection, speech recognition, speech coding, noise adaptation, speech enhancement, microphone arrays and echo-cancellation.",2010-02-09,"The title of the patent is systems and methods that detect a desired signal via a linear discriminative classifier that utilizes an estimated posterior signal-to-noise ratio (snr) and its abstract is the present invention provides systems and methods for signal detection and enhancement. the systems and methods utilize one or more discriminative classifiers (e.g., a logistic regression model and a convolutional neural network) to estimate a posterior probability that indicates whether a desired signal is present in a received signal. the discriminative estimators generate the estimated probability based on one or more signal-to-noise ratio (snrs) (e.g., a normalized logarithmic posterior snr (nlpsnr) and a mel-transformed nlpsnr (mel-nlpsnr)) and an estimated noise model. depending on the resolution desired, the estimated snr can be generated at a frame level or at an atom level, wherein the atom level estimates are utilized to generate the frame level estimate. the novel systems and methods can be utilized to facilitate speech detection, speech recognition, speech coding, noise adaptation, speech enhancement, microphone arrays and echo-cancellation. dated 2010-02-09"
7660774,nonlinear neural network fault detection system and method,a system and method for fault detection is provided. the fault detection system provides the ability to detect symptoms of fault in turbine engines and other mechanical systems that have nonlinear relationships. the fault detection system uses a neural network to perform a data representation and feature extraction where the extracted features are analogous to principal components derived in a principal component analysis. this neural network data representation analysis can then be used to determine the likelihood of a fault in the system.,2010-02-09,The title of the patent is nonlinear neural network fault detection system and method and its abstract is a system and method for fault detection is provided. the fault detection system provides the ability to detect symptoms of fault in turbine engines and other mechanical systems that have nonlinear relationships. the fault detection system uses a neural network to perform a data representation and feature extraction where the extracted features are analogous to principal components derived in a principal component analysis. this neural network data representation analysis can then be used to determine the likelihood of a fault in the system. dated 2010-02-09
7662706,nanostructures formed of branched nanowhiskers and methods of producing the same,"a method of forming a nanostructure having the form of a tree, comprises a first stage and a second stage. the first stage includes providing one or more catalytic particles on a substrate surface, and growing a first nanowhisker via each catalytic particle. the second stage includes providing, on the periphery of each first nanowhisker, one or more second catalytic particles, and growing, from each second catalytic particle, a second nanowhisker extending transversely from the periphery of the respective first nanowhisker. further stages may be included to grow one or more further nanowhiskers extending from the nanowhisker(s) of the preceding stage. heterostructures may be created within the nanowhiskers. such nanostructures may form the components of a solar cell array or a light emitting flat panel, where the nanowhiskers are formed of a photosensitive material. a neural network may be formed by positioning the first nanowhiskers close together so that adjacent trees contact one another through nanowhiskers grown in a subsequent stage, and heterojunctions within the nanowhiskers create tunnel barriers to current flow.",2010-02-16,"The title of the patent is nanostructures formed of branched nanowhiskers and methods of producing the same and its abstract is a method of forming a nanostructure having the form of a tree, comprises a first stage and a second stage. the first stage includes providing one or more catalytic particles on a substrate surface, and growing a first nanowhisker via each catalytic particle. the second stage includes providing, on the periphery of each first nanowhisker, one or more second catalytic particles, and growing, from each second catalytic particle, a second nanowhisker extending transversely from the periphery of the respective first nanowhisker. further stages may be included to grow one or more further nanowhiskers extending from the nanowhisker(s) of the preceding stage. heterostructures may be created within the nanowhiskers. such nanostructures may form the components of a solar cell array or a light emitting flat panel, where the nanowhiskers are formed of a photosensitive material. a neural network may be formed by positioning the first nanowhiskers close together so that adjacent trees contact one another through nanowhiskers grown in a subsequent stage, and heterojunctions within the nanowhiskers create tunnel barriers to current flow. dated 2010-02-16"
7664548,distributed neuromodulation system for treatment of cardiovascular disease,a distributed system is described that employs electrical neural stimulation to modulate autonomic activity and which allows titration of the neural stimulation therapy in accordance with physiological measurements reflective of autonomic activity and/or physiological variables affected by the neural stimulation. such a system may include a plurality of implantable neuromodulation units that communicate with one another over a network.,2010-02-16,The title of the patent is distributed neuromodulation system for treatment of cardiovascular disease and its abstract is a distributed system is described that employs electrical neural stimulation to modulate autonomic activity and which allows titration of the neural stimulation therapy in accordance with physiological measurements reflective of autonomic activity and/or physiological variables affected by the neural stimulation. such a system may include a plurality of implantable neuromodulation units that communicate with one another over a network. dated 2010-02-16
7664593,method and system for estimating exhaust gas temperature and internal combustion engine equipped with such a system,"a method for estimating temperature of an internal combustion engine exhaust gases, a system for estimating exhaust gas temperature, and an engine equipped with such a system. the method uses an estimator with a neural network provided with a feedback loop returning directly or indirectly in the network input one or more quantities of the network output.",2010-02-16,"The title of the patent is method and system for estimating exhaust gas temperature and internal combustion engine equipped with such a system and its abstract is a method for estimating temperature of an internal combustion engine exhaust gases, a system for estimating exhaust gas temperature, and an engine equipped with such a system. the method uses an estimator with a neural network provided with a feedback loop returning directly or indirectly in the network input one or more quantities of the network output. dated 2010-02-16"
7664714,neural network element with reinforcement/attenuation learning,"a neural network element, outputting an output signal in response to a plurality of input signals, comprises a history memory for accumulating and storing the plurality of input signals in a temporal order as history values. it also includes an output module for outputting the output signal when an internal state exceeds a predetermined threshold value, the internal state being based on a sum of the product of a plurality of input signals and corresponding coupling coefficients. the history values depend on change of the internal state. the neural network element is configured to subtract a predetermined value from the internal state immediately after the output module fires and performs learning for reinforcing or attenuating the coupling coefficient according to the history values after the output module fires.",2010-02-16,"The title of the patent is neural network element with reinforcement/attenuation learning and its abstract is a neural network element, outputting an output signal in response to a plurality of input signals, comprises a history memory for accumulating and storing the plurality of input signals in a temporal order as history values. it also includes an output module for outputting the output signal when an internal state exceeds a predetermined threshold value, the internal state being based on a sum of the product of a plurality of input signals and corresponding coupling coefficients. the history values depend on change of the internal state. the neural network element is configured to subtract a predetermined value from the internal state immediately after the output module fires and performs learning for reinforcing or attenuating the coupling coefficient according to the history values after the output module fires. dated 2010-02-16"
7664715,"apparatus and method for compressing data, apparatus and method for analyzing data, and data management system","there are provided an apparatus and a method for compressing data, an apparatus and a method for analyzing data and a data management system, which are capable of compressing huge data and accurately reproducing the characteristics of the original data from the compressed data.the data compressing apparatus includes detection means for detecting a multiplicity of data sets, each including n parameter values that vary according to an operation of an object, where n is a natural number; and data compressing means for compressing the data sets by inputting the data sets into an n-dimensional space, arranging neurons smaller in number than the data sets in the n-dimensional space, carrying out unsupervised learning for a neural network on the neurons, and converting the data sets into a neuron model parameter characterizing a neuron model obtained by the unsupervised learning.",2010-02-16,"The title of the patent is apparatus and method for compressing data, apparatus and method for analyzing data, and data management system and its abstract is there are provided an apparatus and a method for compressing data, an apparatus and a method for analyzing data and a data management system, which are capable of compressing huge data and accurately reproducing the characteristics of the original data from the compressed data.the data compressing apparatus includes detection means for detecting a multiplicity of data sets, each including n parameter values that vary according to an operation of an object, where n is a natural number; and data compressing means for compressing the data sets by inputting the data sets into an n-dimensional space, arranging neurons smaller in number than the data sets in the n-dimensional space, carrying out unsupervised learning for a neural network on the neurons, and converting the data sets into a neuron model parameter characterizing a neuron model obtained by the unsupervised learning. dated 2010-02-16"
7664716,process and apparatus for realizing a digital neural network using electronic family,"provided are a method and system for providing member-specific information, which are capable of building a digital neural network by improving negative functions of the current internet, applying a paradigm of sound home- and lifestyle-oriented opened electronic home, and utilizing a closed groupware. the method for providing member-specific information in a digital neural network based on electronic home including my rooms, our homes, and towns includes the steps of: setting an initial value of a variable; extracting a priority while changing a weight of the variable; targeting members using the extracted priority information; grouping the targeted members; and pushing information to the member group or interoperating the information.",2010-02-16,"The title of the patent is process and apparatus for realizing a digital neural network using electronic family and its abstract is provided are a method and system for providing member-specific information, which are capable of building a digital neural network by improving negative functions of the current internet, applying a paradigm of sound home- and lifestyle-oriented opened electronic home, and utilizing a closed groupware. the method for providing member-specific information in a digital neural network based on electronic home including my rooms, our homes, and towns includes the steps of: setting an initial value of a variable; extracting a priority while changing a weight of the variable; targeting members using the extracted priority information; grouping the targeted members; and pushing information to the member group or interoperating the information. dated 2010-02-16"
7668397,apparatus and method for objective assessment of dct-coded video quality with or without an original video sequence,"a new approach to objective quality assessment of dct-coded video sequences, with or without a reference is proposed. the system is comprised of a proprietary segmentation algorithm, a feature extraction process and a nonlinear feed-forward-type neural network for feature analysis. the methods mimic function of the human visual system (hvs): a neural network training algorithm is used for determining the optimal network weights and biases for both system modes of operation. the proposed method allows for assessment of dct-coded video sequences without the original source being available (pseudo-reference mode). the pseudo-reference mode is also comprised of a proprietary dct-coded video (mpeg) noise reducer (mnr), co-pending patent application no. 60/592,143.",2010-02-23,"The title of the patent is apparatus and method for objective assessment of dct-coded video quality with or without an original video sequence and its abstract is a new approach to objective quality assessment of dct-coded video sequences, with or without a reference is proposed. the system is comprised of a proprietary segmentation algorithm, a feature extraction process and a nonlinear feed-forward-type neural network for feature analysis. the methods mimic function of the human visual system (hvs): a neural network training algorithm is used for determining the optimal network weights and biases for both system modes of operation. the proposed method allows for assessment of dct-coded video sequences without the original source being available (pseudo-reference mode). the pseudo-reference mode is also comprised of a proprietary dct-coded video (mpeg) noise reducer (mnr), co-pending patent application no. 60/592,143. dated 2010-02-23"
7669469,method and apparatus for a continuous data recorder for a downhole sample tank,"the present invention provides an apparatus and method for continuously monitoring the integrity of a pressurized well bore fluid sample collected downhole in an earth boring or well bore. the cdr continuous by measures the temperature and pressure for the down hole sample. near infrared, mid infrared and visible light analysis is also performed on the small amount of sample to provide an on site analysis of sample properties and contamination level. the onsite analysis comprises determination of gas oil ratio, api gravity and various other parameters which can be estimated by a trained neural network or chemometric equation a flexural mechanical resonator is also provided to measure fluid density and viscosity from which additional parameters can be estimated by a trained neural network or chemometric equation. the sample tank is overpressured or supercharged to obviate adverse pressure drop or other effects of diverting a small sample to the cdr.",2010-03-02,"The title of the patent is method and apparatus for a continuous data recorder for a downhole sample tank and its abstract is the present invention provides an apparatus and method for continuously monitoring the integrity of a pressurized well bore fluid sample collected downhole in an earth boring or well bore. the cdr continuous by measures the temperature and pressure for the down hole sample. near infrared, mid infrared and visible light analysis is also performed on the small amount of sample to provide an on site analysis of sample properties and contamination level. the onsite analysis comprises determination of gas oil ratio, api gravity and various other parameters which can be estimated by a trained neural network or chemometric equation a flexural mechanical resonator is also provided to measure fluid density and viscosity from which additional parameters can be estimated by a trained neural network or chemometric equation. the sample tank is overpressured or supercharged to obviate adverse pressure drop or other effects of diverting a small sample to the cdr. dated 2010-03-02"
7671983,method and apparatus for an advanced optical analyzer,"the present invention provides a sample tank having a window for introduction of electromagnetic energy into the sample tank for analyzing a formation fluid sample down hole or at the surface without disturbing the sample. near infrared, mid infrared and visible light analysis is performed on the sample to provide a downhole in situ or surface on site analysis of sample properties and contamination level. the onsite analysis comprises determination of gas oil ratio, api gravity and various other parameters which can be estimated by a trained neural network or chemometric equation. a flexural mechanical resonator is also provided to measure fluid density and viscosity from which additional parameters can be estimated by a trained neural network or chemometric equation. the sample tank is pressurized to obviate adverse pressure drop or other effects of diverting a small sample.",2010-03-02,"The title of the patent is method and apparatus for an advanced optical analyzer and its abstract is the present invention provides a sample tank having a window for introduction of electromagnetic energy into the sample tank for analyzing a formation fluid sample down hole or at the surface without disturbing the sample. near infrared, mid infrared and visible light analysis is performed on the sample to provide a downhole in situ or surface on site analysis of sample properties and contamination level. the onsite analysis comprises determination of gas oil ratio, api gravity and various other parameters which can be estimated by a trained neural network or chemometric equation. a flexural mechanical resonator is also provided to measure fluid density and viscosity from which additional parameters can be estimated by a trained neural network or chemometric equation. the sample tank is pressurized to obviate adverse pressure drop or other effects of diverting a small sample. dated 2010-03-02"
7676076,neural network based method for displaying an examination image with normalized grayscale values,"image processing method for a digital medical examination image, the pixels of which are assigned a gray-scale value in each instance, with a minimum and a maximum gray-scale value being defined as limit values for the purpose of displaying the examination image, with the pixels being subjected to an evaluation by means of a neural network, in order to determine such pixels and to disregard them when defining the gray-scale values which are located in a direct radiation region or in a projected collimator region.",2010-03-09,"The title of the patent is neural network based method for displaying an examination image with normalized grayscale values and its abstract is image processing method for a digital medical examination image, the pixels of which are assigned a gray-scale value in each instance, with a minimum and a maximum gray-scale value being defined as limit values for the purpose of displaying the examination image, with the pixels being subjected to an evaluation by means of a neural network, in order to determine such pixels and to disregard them when defining the gray-scale values which are located in a direct radiation region or in a projected collimator region. dated 2010-03-09"
7676441,"information processing apparatus, information processing method, pattern recognition apparatus, and pattern recognition method","in a hierarchical neural network having a module structure, learning necessary for detection of a new feature class is executed by a processing module which has not finished learning yet and includes a plurality of neurons which should learn an unlearned feature class and have an undetermined receptor field structure by presenting a predetermined pattern to a data input layer. thus, a feature class necessary for subject recognition can be learned automatically and efficiently.",2010-03-09,"The title of the patent is information processing apparatus, information processing method, pattern recognition apparatus, and pattern recognition method and its abstract is in a hierarchical neural network having a module structure, learning necessary for detection of a new feature class is executed by a processing module which has not finished learning yet and includes a plurality of neurons which should learn an unlearned feature class and have an undetermined receptor field structure by presenting a predetermined pattern to a data input layer. thus, a feature class necessary for subject recognition can be learned automatically and efficiently. dated 2010-03-09"
7680751,neural network based refrigerant charge detection algorithm for vapor compression systems,"methods and apparatus are provided for determining refrigerant charge in a vapor compressor system (vcs) of an aircraft. the methods and apparatus comprise the following steps of, and/or means for, generating a data set from historical data representative of a plurality of vcs operating conditions over time, identifying one or more steady-state data points in the generated data set, forming a revised data set that includes at least the steady-state data points, using principal components analysis (pca) to derive values for a plurality of minimally correlated input variables, supplying the derived values for the plurality of minimally correlated input variables and the corresponding values for the vcs refrigerant charge in the revised data set to a nonlinear neural network model, and deriving a simulator model characterizing a relationship between the plurality of minimally correlated input variables and the vcs refrigerant charge.",2010-03-16,"The title of the patent is neural network based refrigerant charge detection algorithm for vapor compression systems and its abstract is methods and apparatus are provided for determining refrigerant charge in a vapor compressor system (vcs) of an aircraft. the methods and apparatus comprise the following steps of, and/or means for, generating a data set from historical data representative of a plurality of vcs operating conditions over time, identifying one or more steady-state data points in the generated data set, forming a revised data set that includes at least the steady-state data points, using principal components analysis (pca) to derive values for a plurality of minimally correlated input variables, supplying the derived values for the plurality of minimally correlated input variables and the corresponding values for the vcs refrigerant charge in the revised data set to a nonlinear neural network model, and deriving a simulator model characterizing a relationship between the plurality of minimally correlated input variables and the vcs refrigerant charge. dated 2010-03-16"
7685081,bipedal walking simulation,"an artificial multiped is constructed (either in simulation or embodied) in such a way that its natural body dynamics allow the lower part of each leg to swing naturally under the influence of gravity. the upper part of each leg is actively actuated in the sagittal plane. the necessary input to drive the above-mentioned actuators is derived from a neural network controller. the latter is arranged as two bi-directionally coupled chains of neural oscillators, the number of which equals twice that of the legs to be actuated. parameter optimisation of the controllers is achieved by evolutionary computation in the form of a genetic algorithm.",2010-03-23,"The title of the patent is bipedal walking simulation and its abstract is an artificial multiped is constructed (either in simulation or embodied) in such a way that its natural body dynamics allow the lower part of each leg to swing naturally under the influence of gravity. the upper part of each leg is actively actuated in the sagittal plane. the necessary input to drive the above-mentioned actuators is derived from a neural network controller. the latter is arranged as two bi-directionally coupled chains of neural oscillators, the number of which equals twice that of the legs to be actuated. parameter optimisation of the controllers is achieved by evolutionary computation in the form of a genetic algorithm. dated 2010-03-23"
7689002,method of detecting bends on a road and system implementing same,"""the invention concerns a method of detecting from a vehicle a bend in a road comprising a surface and road edges, comprising the following operations:    the invention also concerns a system for implementing the method of the invention comprising a camera mounted in the vehicle, an image processing unit and a neural network.""",2010-03-30,"The title of the patent is method of detecting bends on a road and system implementing same and its abstract is ""the invention concerns a method of detecting from a vehicle a bend in a road comprising a surface and road edges, comprising the following operations:    the invention also concerns a system for implementing the method of the invention comprising a camera mounted in the vehicle, an image processing unit and a neural network."" dated 2010-03-30"
7693120,neural network-based mobility management for self-partition detection and identification of mobile ad hoc radio networks,"a method of managing an ad hoc communications network of wireless devices or nodes. the network is connected if all nodes can communicate with each other and otherwise partitioned. partitions are identified by recursively applying a connectivity function to a connectivity matrix representative of the network. the number of times the connectivity function is recursively applied is determined by the network diameter. if the result of the recursive application is a unity matrix, the network is connected; otherwise it is disconnected. also, if the network diameter exceeds a selected maximum length, the network may be voluntarily partitioned into connected sub-networks by recursively applying the connectivity function a lesser number of times to the connectivity matrix. the lesser number of times is determined by the selected maximum length or maximum allowable number of hops.",2010-04-06,"The title of the patent is neural network-based mobility management for self-partition detection and identification of mobile ad hoc radio networks and its abstract is a method of managing an ad hoc communications network of wireless devices or nodes. the network is connected if all nodes can communicate with each other and otherwise partitioned. partitions are identified by recursively applying a connectivity function to a connectivity matrix representative of the network. the number of times the connectivity function is recursively applied is determined by the network diameter. if the result of the recursive application is a unity matrix, the network is connected; otherwise it is disconnected. also, if the network diameter exceeds a selected maximum length, the network may be voluntarily partitioned into connected sub-networks by recursively applying the connectivity function a lesser number of times to the connectivity matrix. the lesser number of times is determined by the selected maximum length or maximum allowable number of hops. dated 2010-04-06"
7693177,navigational aid and carrier sense technique,"a navigational aid for use as an ais apparatus includes a memory for storing information about previous use of individual time slots synchronized with transmission schedules of other stations and a signal detector for judging whether an information signal exists in a time slot specified in accordance with a synchronization timing signal based on the information stored in the memory and a result of monitoring of the behavior of a baseband signal obtained from a received signal on an iq-plane. the navigational aid transmits information about own station based on a result of judgment by the signal detector. the monitoring of the behavior of the received baseband signal plotted on the iq-plane can be accomplished by performing pattern recognition operation, for which a carrier sense technique, such as a support vector machine, subspace method or neural network, can be used.",2010-04-06,"The title of the patent is navigational aid and carrier sense technique and its abstract is a navigational aid for use as an ais apparatus includes a memory for storing information about previous use of individual time slots synchronized with transmission schedules of other stations and a signal detector for judging whether an information signal exists in a time slot specified in accordance with a synchronization timing signal based on the information stored in the memory and a result of monitoring of the behavior of a baseband signal obtained from a received signal on an iq-plane. the navigational aid transmits information about own station based on a result of judgment by the signal detector. the monitoring of the behavior of the received baseband signal plotted on the iq-plane can be accomplished by performing pattern recognition operation, for which a carrier sense technique, such as a support vector machine, subspace method or neural network, can be used. dated 2010-04-06"
7698004,apc process control when process parameters are inaccurately measured,"a controller is provided for directing control of a process performed to control an amount of a pollutant emitted into the air. the process has multiple process parameters (mpps) the controller includes either a neural network process model or a non-neural network process model. whichever type model is included, it will represent a relationship between one of the mpps and other of the mpps. the controller also includes a control processor having the logic to determine the validity of a measured value of the one mpp based on the one model. the control processor directs control of the process in accordance with the measured value of the one mpp only if the measured value of the one mpp is determined to be valid. on the other hand, if the measured value is determined to be invalid, the control processor may direct control of the process in accordance with an estimated value of the one mpp.",2010-04-13,"The title of the patent is apc process control when process parameters are inaccurately measured and its abstract is a controller is provided for directing control of a process performed to control an amount of a pollutant emitted into the air. the process has multiple process parameters (mpps) the controller includes either a neural network process model or a non-neural network process model. whichever type model is included, it will represent a relationship between one of the mpps and other of the mpps. the controller also includes a control processor having the logic to determine the validity of a measured value of the one mpp based on the one model. the control processor directs control of the process in accordance with the measured value of the one mpp only if the measured value of the one mpp is determined to be valid. on the other hand, if the measured value is determined to be invalid, the control processor may direct control of the process in accordance with an estimated value of the one mpp. dated 2010-04-13"
7702145,adapting a neural network for individual style,"various technologies and techniques are disclosed for improving handwriting recognition using a neural network by allowing a user to provide samples. a recognition operation is performed on the user's handwritten input, and the user is not satisfied with the recognition result. the user selects an option to train the neural network on one or more characters to improve the recognition results. the user is prompted to specify samples for the certain character, word, or phrase, and the neural network is adjusted for the certain character, word, or phrase. handwritten input is later received from the user. a recognition operation is performed on the handwritten input using the neural network that was adjusted for the certain character or characters.",2010-04-20,"The title of the patent is adapting a neural network for individual style and its abstract is various technologies and techniques are disclosed for improving handwriting recognition using a neural network by allowing a user to provide samples. a recognition operation is performed on the user's handwritten input, and the user is not satisfied with the recognition result. the user selects an option to train the neural network on one or more characters to improve the recognition results. the user is prompted to specify samples for the certain character, word, or phrase, and the neural network is adjusted for the certain character, word, or phrase. handwritten input is later received from the user. a recognition operation is performed on the handwritten input using the neural network that was adjusted for the certain character or characters. dated 2010-04-20"
7702519,estimating an economic parameter related to a process for controlling emission of a pollutant into the air,"an economic parameter estimator is provided for a process that has multiple process parameters (mpps) and is performed to control emission of a pollutant into the air. the performance of the process is associated with one or more economic factors (efs). the estimator includes either a neural network process model or a non-neural network process model. in either case, the model represents a relationship between one or more of the mpps and an economic parameter. also included is a processor configured with logic, e.g. programmed software, to estimate a monetary value of the economic parameter based on a value of each of the one or more mpps, a value of each of at least one of the one or more efs, and the one model.",2010-04-20,"The title of the patent is estimating an economic parameter related to a process for controlling emission of a pollutant into the air and its abstract is an economic parameter estimator is provided for a process that has multiple process parameters (mpps) and is performed to control emission of a pollutant into the air. the performance of the process is associated with one or more economic factors (efs). the estimator includes either a neural network process model or a non-neural network process model. in either case, the model represents a relationship between one or more of the mpps and an economic parameter. also included is a processor configured with logic, e.g. programmed software, to estimate a monetary value of the economic parameter based on a value of each of the one or more mpps, a value of each of at least one of the one or more efs, and the one model. dated 2010-04-20"
7702598,methods and systems for predicting occurrence of an event,"embodiments of the present invention are directed to methods and systems for training a neural network having weighted connections for classification of data, as well as embodiments corresponding to the use of such a neural network for the classification of data, including, for example, prediction of an event (e.g., disease). the method may include inputting input training data into the neural network, processing, by the neural network, the input training data to produce an output, determining an error between the output and a desired output corresponding to the input training data, rating the performance neural network using an objective function, wherein the objective function comprises a function c substantially in accordance with an approximation of the concordance index and adapting the weighted connections of the neural network based upon results of the objective function.",2010-04-20,"The title of the patent is methods and systems for predicting occurrence of an event and its abstract is embodiments of the present invention are directed to methods and systems for training a neural network having weighted connections for classification of data, as well as embodiments corresponding to the use of such a neural network for the classification of data, including, for example, prediction of an event (e.g., disease). the method may include inputting input training data into the neural network, processing, by the neural network, the input training data to produce an output, determining an error between the output and a desired output corresponding to the input training data, rating the performance neural network using an objective function, wherein the objective function comprises a function c substantially in accordance with an approximation of the concordance index and adapting the weighted connections of the neural network based upon results of the objective function. dated 2010-04-20"
7702599,system and method for cognitive memory and auto-associative neural network based pattern recognition,"designs for cognitive memory systems storing input data, images, or patterns, and retrieving it without knowledge of where stored when cognitive memory is prompted by query pattern that is related to sought stored pattern. retrieval system of cognitive memory uses autoassociative neural networks and techniques for pre-processing query pattern to establish relationship between query pattern and sought stored pattern, to locate sought pattern, and to retrieve it and ancillary data. cognitive memory, when connected to computer or information appliance introduces computational architecture that applies to systems and methods for navigation, location and recognition of objects in images, character recognition, facial recognition, medical analysis and diagnosis, video image analysis, and to photographic search engines that when prompted with a query photograph containing faces and objects will retrieve related photographs stored in computer or other information appliance, and will identify url's of related photographs and documents stored on the world wide web.",2010-04-20,"The title of the patent is system and method for cognitive memory and auto-associative neural network based pattern recognition and its abstract is designs for cognitive memory systems storing input data, images, or patterns, and retrieving it without knowledge of where stored when cognitive memory is prompted by query pattern that is related to sought stored pattern. retrieval system of cognitive memory uses autoassociative neural networks and techniques for pre-processing query pattern to establish relationship between query pattern and sought stored pattern, to locate sought pattern, and to retrieve it and ancillary data. cognitive memory, when connected to computer or information appliance introduces computational architecture that applies to systems and methods for navigation, location and recognition of objects in images, character recognition, facial recognition, medical analysis and diagnosis, video image analysis, and to photographic search engines that when prompted with a query photograph containing faces and objects will retrieve related photographs stored in computer or other information appliance, and will identify url's of related photographs and documents stored on the world wide web. dated 2010-04-20"
7707130,"real-time predictive computer program, model, and method",a method for predicting a future occurrence of an event involves obtaining a history of prior occurrences of the event. a plurality of variables is created that are associated with the event. weights are assigned to each variable. an artificial neural network is accessed and trained with the history of past occurrences of the event by comparing an output of the artificial neural network to the past occurrence of the event. the weights are adjusted until the output corresponds to the past occurrence of the event.,2010-04-27,"The title of the patent is real-time predictive computer program, model, and method and its abstract is a method for predicting a future occurrence of an event involves obtaining a history of prior occurrences of the event. a plurality of variables is created that are associated with the event. weights are assigned to each variable. an artificial neural network is accessed and trained with the history of past occurrences of the event by comparing an output of the artificial neural network to the past occurrence of the event. the weights are adjusted until the output corresponds to the past occurrence of the event. dated 2010-04-27"
7711662,system and method for optimization of a database for the training and testing of prediction algorithms,"a system and method are provided for the training and testing of prediction algorithms. according to an exemplary embodiment of the invention the method generates optimum training, testing and/or validation data sets from a common general database by applying a genetic algorithm to populations of testing and training subsets used in connection with a given prediction algorithm. in exemplary embodiments the prediction algorithm operated upon is an artificial neural network. as well, in preferred exemplary embodiments, the most predictive independent variables of the records of the common database are automatically selected in a pre-processing phase. such preprocessing phase applies a genetic algorithm to populations of prediction algorithms which vary as to number and content of input variables, where the prediction algorithms representing the selections of input variables which have the best testing performances and the minimum input variables are promoted for the processing of the new generations according to a defined selection algorithm.",2010-05-04,"The title of the patent is system and method for optimization of a database for the training and testing of prediction algorithms and its abstract is a system and method are provided for the training and testing of prediction algorithms. according to an exemplary embodiment of the invention the method generates optimum training, testing and/or validation data sets from a common general database by applying a genetic algorithm to populations of testing and training subsets used in connection with a given prediction algorithm. in exemplary embodiments the prediction algorithm operated upon is an artificial neural network. as well, in preferred exemplary embodiments, the most predictive independent variables of the records of the common database are automatically selected in a pre-processing phase. such preprocessing phase applies a genetic algorithm to populations of prediction algorithms which vary as to number and content of input variables, where the prediction algorithms representing the selections of input variables which have the best testing performances and the minimum input variables are promoted for the processing of the new generations according to a defined selection algorithm. dated 2010-05-04"
7715907,"method and system for atrial fibrillation analysis, characterization, and mapping","a method and system for atrial fibrillation analysis, characterization, and mapping is disclosed. a finite element model (fem) representing a physical structure of a heart is generated. electrogram data can be sensed at various locations in the heart using an electrophysiology catheter, and the electrogram data is mapped to the elements of the fem. function parameters, which measure some characteristics of af arrhythmia, are then simultaneously calculated for all of the elements of the fem based on the electrogram data mapped to the elements of the fem. an artificial neural network (ann) can be used to calculate the function parameters.",2010-05-11,"The title of the patent is method and system for atrial fibrillation analysis, characterization, and mapping and its abstract is a method and system for atrial fibrillation analysis, characterization, and mapping is disclosed. a finite element model (fem) representing a physical structure of a heart is generated. electrogram data can be sensed at various locations in the heart using an electrophysiology catheter, and the electrogram data is mapped to the elements of the fem. function parameters, which measure some characteristics of af arrhythmia, are then simultaneously calculated for all of the elements of the fem based on the electrogram data mapped to the elements of the fem. an artificial neural network (ann) can be used to calculate the function parameters. dated 2010-05-11"
7716146,network management system utilizing a neural network,"preferred embodiments of the invention provide systems and methods to observe one or more network elements associated with a network, receive an indication of an event relating to one or more network element configurations associated with the network, observe a potential outcome associated with the network, store the potential outcome such that the potential outcome is associated with the event, determine a probable outcome based on the potential outcome and store the probable outcome such that the probable outcome is associated with the event.",2010-05-11,"The title of the patent is network management system utilizing a neural network and its abstract is preferred embodiments of the invention provide systems and methods to observe one or more network elements associated with a network, receive an indication of an event relating to one or more network element configurations associated with the network, observe a potential outcome associated with the network, store the potential outcome such that the potential outcome is associated with the event, determine a probable outcome based on the potential outcome and store the probable outcome such that the probable outcome is associated with the event. dated 2010-05-11"
7716147,"real-time predictive computer program, model, and method",a method for predicting a future occurrence of an event involves obtaining a history of prior occurrences of the event. a plurality of variables is created that are associated with the event. weights are assigned to each variable. an artificial neural network is accessed and trained with the history of past occurrences of the event by comparing an output of the artificial neural network to the past occurrence of the event. the weights are adjusted until the output corresponds to the past occurrence of the event.,2010-05-11,"The title of the patent is real-time predictive computer program, model, and method and its abstract is a method for predicting a future occurrence of an event involves obtaining a history of prior occurrences of the event. a plurality of variables is created that are associated with the event. weights are assigned to each variable. an artificial neural network is accessed and trained with the history of past occurrences of the event by comparing an output of the artificial neural network to the past occurrence of the event. the weights are adjusted until the output corresponds to the past occurrence of the event. dated 2010-05-11"
7720610,detection of psychological disorder activity patterns,"a method for detecting a psychological disorder in a person comprises collecting movement and, optionally, other data from the person by a device borne by the person; storing the data in a memory in contact with the device during the collection of data; transferring the stored data to a computer; calculating at least one set of parameter data distinctive of the movement data; feeding the least one set of parameter data to an artificial neural network trained to recognize in the data a feature specific for a psychological disorder or a group of such disorders. also is disclosed an assembly for carrying out the method.",2010-05-18,"The title of the patent is detection of psychological disorder activity patterns and its abstract is a method for detecting a psychological disorder in a person comprises collecting movement and, optionally, other data from the person by a device borne by the person; storing the data in a memory in contact with the device during the collection of data; transferring the stored data to a computer; calculating at least one set of parameter data distinctive of the movement data; feeding the least one set of parameter data to an artificial neural network trained to recognize in the data a feature specific for a psychological disorder or a group of such disorders. also is disclosed an assembly for carrying out the method. dated 2010-05-18"
7720615,system for detection and prediction of water quality events,"a method of evaluating a water sample for the presence or possible future presence of nitrification comprises obtaining data values of a number of parameters, processing the data values to determine correlation coefficients, to identify any linear dependencies, to standardize the scales, evaluating the data values over a plurality of proliferation time periods and neuron numbers, calculating mses and r2's from the evaluations, and estimating a valid likelihood of nitrification of the water sample. a method of evaluating a water sample for the presence or possible future presence of nitrification, comprises obtaining data values of a number of parameters, statistically pre-processing the data values and supplying the pre-processed data values to a neural network. apparatus, media and processors which are used in performing such methods.",2010-05-18,"The title of the patent is system for detection and prediction of water quality events and its abstract is a method of evaluating a water sample for the presence or possible future presence of nitrification comprises obtaining data values of a number of parameters, processing the data values to determine correlation coefficients, to identify any linear dependencies, to standardize the scales, evaluating the data values over a plurality of proliferation time periods and neuron numbers, calculating mses and r2's from the evaluations, and estimating a valid likelihood of nitrification of the water sample. a method of evaluating a water sample for the presence or possible future presence of nitrification, comprises obtaining data values of a number of parameters, statistically pre-processing the data values and supplying the pre-processed data values to a neural network. apparatus, media and processors which are used in performing such methods. dated 2010-05-18"
7721336,systems and methods for dynamic detection and prevention of electronic fraud,"the present invention provides systems and methods for dynamic detection and prevention of electronic fraud and network intrusion using an integrated set of intelligent technologies. the intelligent technologies include neural networks, multi-agents, data mining, case-based reasoning, rule-based reasoning, fuzzy logic, constraint programming, and genetic algorithms. the systems and methods of the present invention involve a fraud detection and prevention model that successfully detects and prevents electronic fraud and network intrusion in real-time. the model is not sensitive to known or unknown different types of fraud or network intrusion attacks, and can be used to detect and prevent fraud and network intrusion across multiple networks and industries.",2010-05-18,"The title of the patent is systems and methods for dynamic detection and prevention of electronic fraud and its abstract is the present invention provides systems and methods for dynamic detection and prevention of electronic fraud and network intrusion using an integrated set of intelligent technologies. the intelligent technologies include neural networks, multi-agents, data mining, case-based reasoning, rule-based reasoning, fuzzy logic, constraint programming, and genetic algorithms. the systems and methods of the present invention involve a fraud detection and prevention model that successfully detects and prevents electronic fraud and network intrusion in real-time. the model is not sensitive to known or unknown different types of fraud or network intrusion attacks, and can be used to detect and prevent fraud and network intrusion across multiple networks and industries. dated 2010-05-18"
7730086,data set request allocations to computers,"""a method of allocation a computer to service a request for a data set in a system having a plurality of computers. the method is implemented on a neural network having only an input layer having input nodes and an output layer having output nodes, where each output node is associated with a specific computer. connecting the input nodes to the output nodes are weights w(j,k). the method includes the steps of receiving a request for data set “i” and inputting to the input layer a vector r(i)    """,2010-06-01,"The title of the patent is data set request allocations to computers and its abstract is ""a method of allocation a computer to service a request for a data set in a system having a plurality of computers. the method is implemented on a neural network having only an input layer having input nodes and an output layer having output nodes, where each output node is associated with a specific computer. connecting the input nodes to the output nodes are weights w(j,k). the method includes the steps of receiving a request for data set “i” and inputting to the input layer a vector r(i)    "" dated 2010-06-01"
7734117,system for scaling images using neural networks,"an artificial neural network (ann) based system that is adapted to process an input pattern to generate an output pattern related thereto having a different number of components than the input pattern. the system (26) is comprised of an ann (27) and a memory (28), such as a dram memory, that are serially connected. the input pattern (23) is applied to a processor (22), where it can be processed or not (the most general case), before it is applied to the ann and stored therein as a prototype (if learned). a category is associated with each stored prototype. the processor computes the coefficients that allow the determination of the estimated values of the output pattern, these coefficients are the components of a so-called intermediate pattern (24). assuming the ann has already learned a number of input patterns, when a new input pattern is presented to the ann in the recognition phase, the category of the closest prototype is output therefrom and is used as a pointer to the memory. in turn, the memory outputs the corresponding intermediate pattern. the input pattern and the intermediate pattern are applied to the processor to construct the output pattern (25) using the coefficients. typically, the input pattern is a block of pixels in the field of scaling images.",2010-06-08,"The title of the patent is system for scaling images using neural networks and its abstract is an artificial neural network (ann) based system that is adapted to process an input pattern to generate an output pattern related thereto having a different number of components than the input pattern. the system (26) is comprised of an ann (27) and a memory (28), such as a dram memory, that are serially connected. the input pattern (23) is applied to a processor (22), where it can be processed or not (the most general case), before it is applied to the ann and stored therein as a prototype (if learned). a category is associated with each stored prototype. the processor computes the coefficients that allow the determination of the estimated values of the output pattern, these coefficients are the components of a so-called intermediate pattern (24). assuming the ann has already learned a number of input patterns, when a new input pattern is presented to the ann in the recognition phase, the category of the closest prototype is output therefrom and is used as a pointer to the memory. in turn, the memory outputs the corresponding intermediate pattern. the input pattern and the intermediate pattern are applied to the processor to construct the output pattern (25) using the coefficients. typically, the input pattern is a block of pixels in the field of scaling images. dated 2010-06-08"
7734555,separate learning system and method using two-layered neural network having target values for hidden nodes,"disclosed herein is a separate learning system and method using a two-layered neural network having target values for hidden nodes. the separate learning system of the present invention includes an input layer for receiving training data from a user, and including at least one input node. a hidden layer includes at least one hidden node. a first connection weight unit connects the input layer to the hidden layer, and changes a weight between the input node and the hidden node. an output layer outputs training data that has been completely learned. the second connection weight unit connects the hidden layer to the output layer, changing a weight between the output and the hidden node, and calculates a target value for the hidden node, based on a current error for the output node. a control unit stops learning, fixes the second connection weight unit, turns a learning direction to the first connection weight unit, and causes learning to be repeatedly performed between the input node and the hidden node if a learning speed decreases or a cost function increases due to local minima or plateaus when the first connection weight unit is fixed and learning is performed using only the second connection weight unit, thus allowing learning to be repeatedly performed until learning converges to the target value for the hidden node.",2010-06-08,"The title of the patent is separate learning system and method using two-layered neural network having target values for hidden nodes and its abstract is disclosed herein is a separate learning system and method using a two-layered neural network having target values for hidden nodes. the separate learning system of the present invention includes an input layer for receiving training data from a user, and including at least one input node. a hidden layer includes at least one hidden node. a first connection weight unit connects the input layer to the hidden layer, and changes a weight between the input node and the hidden node. an output layer outputs training data that has been completely learned. the second connection weight unit connects the hidden layer to the output layer, changing a weight between the output and the hidden node, and calculates a target value for the hidden node, based on a current error for the output node. a control unit stops learning, fixes the second connection weight unit, turns a learning direction to the first connection weight unit, and causes learning to be repeatedly performed between the input node and the hidden node if a learning speed decreases or a cost function increases due to local minima or plateaus when the first connection weight unit is fixed and learning is performed using only the second connection weight unit, thus allowing learning to be repeatedly performed until learning converges to the target value for the hidden node. dated 2010-06-08"
7742425,neural network-based mobility management for mobile ad hoc radio networks,a self managed ad hoc communications network and method of managing the network. the network includes wireless devices or nodes that include a neural network element and the ad hoc network operates as a neural network. one of the nodes is designated as a network management system (nms) that provides overall network management. clusters of nodes are organized around cluster leaders. each cluster leader manages a cluster of nodes and communications between node clusters. each cluster may also have other nodes identified as lower order cluster leaders.,2010-06-22,The title of the patent is neural network-based mobility management for mobile ad hoc radio networks and its abstract is a self managed ad hoc communications network and method of managing the network. the network includes wireless devices or nodes that include a neural network element and the ad hoc network operates as a neural network. one of the nodes is designated as a network management system (nms) that provides overall network management. clusters of nodes are organized around cluster leaders. each cluster leader manages a cluster of nodes and communications between node clusters. each cluster may also have other nodes identified as lower order cluster leaders. dated 2010-06-22
7742608,feedback elimination method and apparatus,"a method and apparatus for detecting a singing frequency in a signal processing system using two neural-networks is disclosed. the first one (a hit neural network) monitors the maximum spectral peak fft bin as it changes with time. the second one (change neural network) monitors the monotonic increasing behavior. the inputs to the neural-networks are the maximum spectral magnitude bin and its rate of change in time. the output is an indication whether howling is likely to occur and the corresponding singing frequency. once the singing frequency is identified, it can be suppressed using any one of many available techniques such as notch filters. several improvements of the base method or apparatus are also disclosed, where additional neural networks are used to detect more than one singing frequency.",2010-06-22,"The title of the patent is feedback elimination method and apparatus and its abstract is a method and apparatus for detecting a singing frequency in a signal processing system using two neural-networks is disclosed. the first one (a hit neural network) monitors the maximum spectral peak fft bin as it changes with time. the second one (change neural network) monitors the monotonic increasing behavior. the inputs to the neural-networks are the maximum spectral magnitude bin and its rate of change in time. the output is an indication whether howling is likely to occur and the corresponding singing frequency. once the singing frequency is identified, it can be suppressed using any one of many available techniques such as notch filters. several improvements of the base method or apparatus are also disclosed, where additional neural networks are used to detect more than one singing frequency. dated 2010-06-22"
7742612,method for training and operating a hearing aid,"the training of a hearing aid for individual situations is intended to be simpler and more comprehensive for the hearing aid wearer. the invention therefore provides for the hearing aid wearer just to have to associate a current acoustic situation with a predetermined hearing situation identification (3′). this association is learnt by a classifier, for example a neural network (5). after the training process, the neural network (5) can then reliably associate the corresponding hearing situation identification (3′) with an acoustic input signal (2). a current parameter set (4′) is varied or supplemented appropriately on the basis of this association.",2010-06-22,"The title of the patent is method for training and operating a hearing aid and its abstract is the training of a hearing aid for individual situations is intended to be simpler and more comprehensive for the hearing aid wearer. the invention therefore provides for the hearing aid wearer just to have to associate a current acoustic situation with a predetermined hearing situation identification (3′). this association is learnt by a classifier, for example a neural network (5). after the training process, the neural network (5) can then reliably associate the corresponding hearing situation identification (3′) with an acoustic input signal (2). a current parameter set (4′) is varied or supplemented appropriately on the basis of this association. dated 2010-06-22"
7747070,training convolutional neural networks on graphics processing units,"a convolutional neural network is implemented on a graphics processing unit. the network is then trained through a series of forward and backward passes, with convolutional kernels and bias matrices modified on each backward pass according to a gradient of an error function. the implementation takes advantage of parallel processing capabilities of pixel shader units on a gpu, and utilizes a set of start-to-finish formulas to program the computations on the pixel shaders. input and output to the program is done through textures, and a multi-pass summation process is used when sums are needed across pixel shader unit registers.",2010-06-29,"The title of the patent is training convolutional neural networks on graphics processing units and its abstract is a convolutional neural network is implemented on a graphics processing unit. the network is then trained through a series of forward and backward passes, with convolutional kernels and bias matrices modified on each backward pass according to a gradient of an error function. the implementation takes advantage of parallel processing capabilities of pixel shader units on a gpu, and utilizes a set of start-to-finish formulas to program the computations on the pixel shaders. input and output to the program is done through textures, and a multi-pass summation process is used when sums are needed across pixel shader unit registers. dated 2010-06-29"
7747419,prediction method of near field photolithography line fabrication using by the combination of taguchi method and neural network,"a method of building a set of experimental prediction model that requires fewer experimental frequency, shorter prediction time and higher prediction accuracy by using the advantages of combining the experimental data of taguchi method and neural network learning is disclosed. the error between the experimentally measured result of photolithography and the simulated result of the theoretical model of near field photolithography is set as an objective function of an inverse method for back calculating fiber probe aperture size, which is adopted in the following taguchi experiment. the analytical result of taguchi neural network model of the present invention proves that the taguchi neural network model can provide more accurate prediction result than the conventional taguchi network model, and at the same time, improve the demerit of requiring massive training examples of the conventional neural network.",2010-06-29,"The title of the patent is prediction method of near field photolithography line fabrication using by the combination of taguchi method and neural network and its abstract is a method of building a set of experimental prediction model that requires fewer experimental frequency, shorter prediction time and higher prediction accuracy by using the advantages of combining the experimental data of taguchi method and neural network learning is disclosed. the error between the experimentally measured result of photolithography and the simulated result of the theoretical model of near field photolithography is set as an objective function of an inverse method for back calculating fiber probe aperture size, which is adopted in the following taguchi experiment. the analytical result of taguchi neural network model of the present invention proves that the taguchi neural network model can provide more accurate prediction result than the conventional taguchi network model, and at the same time, improve the demerit of requiring massive training examples of the conventional neural network. dated 2010-06-29"
7747548,method of and system for evaluating tactile sensations of car seat covers using statistical recursive and artificial neural network models,"disclosed herein is a method of and system for evaluating tactile sensations of car seat covers using an artificial neural network, in which mechanical and thermal/physiological characteristics of the seat covers is measured with a measuring system, the physical quantity of the measured characteristics is acquired, and then the acquired physical quantity is converted into the amount of tactile qualities which a person feels using a statistical recursive model, thereby quantitatively evaluating tactile sensations of sticky, soft, elastic, coolness to the touch, and thermal and humid in case of car seat covers made of leather, and tactile sensations of soft, elastic, voluminous, smooth, coolness to the touch, and thermal and humid in case of car seat covers made of cloth, as well as quantitatively evaluating tactile sensations of sporty and high-class of the leather and cloths using an artificial neural network model.",2010-06-29,"The title of the patent is method of and system for evaluating tactile sensations of car seat covers using statistical recursive and artificial neural network models and its abstract is disclosed herein is a method of and system for evaluating tactile sensations of car seat covers using an artificial neural network, in which mechanical and thermal/physiological characteristics of the seat covers is measured with a measuring system, the physical quantity of the measured characteristics is acquired, and then the acquired physical quantity is converted into the amount of tactile qualities which a person feels using a statistical recursive model, thereby quantitatively evaluating tactile sensations of sticky, soft, elastic, coolness to the touch, and thermal and humid in case of car seat covers made of leather, and tactile sensations of soft, elastic, voluminous, smooth, coolness to the touch, and thermal and humid in case of car seat covers made of cloth, as well as quantitatively evaluating tactile sensations of sporty and high-class of the leather and cloths using an artificial neural network model. dated 2010-06-29"
7747549,long-term memory neural network modeling memory-chaining functions of the brain wherein a pointer holds information about mutually related neurons and neurons are classified hierarchically by degree of activation,"a stm network 11 for temporarily storing input pattern vectors is formed in phases 1 and 2, and then layered ltm networks 2 to l are formed successively by assigning output vectors provided by the stm network 11 as input vectors. in phase 4, a ltm network 1 for intuitive outputs to which input pattern vectors are applied directly is formed by taking the parameters of comparatively highly activated centroids among centroids in the ltm networks 2 to l. in phase 5, the parameters of the comparatively highly activated centroids among the centroids in the ltm networks 2 to l are fed back as the parameters of the centroids in the stm network. in phase 3, the ltm networks 2 to l are reconstructed at a particular time or in a fixed period by giving the centroid vectors of the ltm networks 2 to l again as input pattern vectors to the stm network 11.",2010-06-29,"The title of the patent is long-term memory neural network modeling memory-chaining functions of the brain wherein a pointer holds information about mutually related neurons and neurons are classified hierarchically by degree of activation and its abstract is a stm network 11 for temporarily storing input pattern vectors is formed in phases 1 and 2, and then layered ltm networks 2 to l are formed successively by assigning output vectors provided by the stm network 11 as input vectors. in phase 4, a ltm network 1 for intuitive outputs to which input pattern vectors are applied directly is formed by taking the parameters of comparatively highly activated centroids among centroids in the ltm networks 2 to l. in phase 5, the parameters of the comparatively highly activated centroids among the centroids in the ltm networks 2 to l are fed back as the parameters of the centroids in the stm network. in phase 3, the ltm networks 2 to l are reconstructed at a particular time or in a fixed period by giving the centroid vectors of the ltm networks 2 to l again as input pattern vectors to the stm network 11. dated 2010-06-29"
7752151,multilayer training in a physical neural network formed utilizing nanotechnology,"a method for and system for training a connection network located between neuron layers within a multi-layer physical neural network. a multi-layer physical neural network can be formed having a plurality of inputs and a plurality outputs thereof, wherein the multi-layer physical neural network comprises a plurality of layers, wherein each layer comprises one or more connection networks and associated neurons. thereafter, a training wave can be initiated across the connection networks associated with an initial layer of the multi-layer physical neural network which propagates thereafter through succeeding connection networks of succeeding layers of the neural network by successively closing and opening switches associated with each layer. one or more feedback signals thereof can be automatically provided to strengthen or weaken nanoconnections associated with each connection network.",2010-07-06,"The title of the patent is multilayer training in a physical neural network formed utilizing nanotechnology and its abstract is a method for and system for training a connection network located between neuron layers within a multi-layer physical neural network. a multi-layer physical neural network can be formed having a plurality of inputs and a plurality outputs thereof, wherein the multi-layer physical neural network comprises a plurality of layers, wherein each layer comprises one or more connection networks and associated neurons. thereafter, a training wave can be initiated across the connection networks associated with an initial layer of the multi-layer physical neural network which propagates thereafter through succeeding connection networks of succeeding layers of the neural network by successively closing and opening switches associated with each layer. one or more feedback signals thereof can be automatically provided to strengthen or weaken nanoconnections associated with each connection network. dated 2010-07-06"
7764622,"interplanetary communications network, interplanetary communications network backbone and method of managing interplanetary communications network","an interplanetary communications network, an interplanetary communications backbone network of artificial neural network (ann) nodes, an ann node and a method of managing interplanetary communications. the backbone network operates as a neural network with each node identifying optimum paths, e.g., end-to-end through the backbone network from a distant planet to an earth node. each node maintains a window matrix identifying reoccurring (e.g., periodically) communications windows between nodes and a propagation delay matrix identifying time varying propagation delays between nodes. each node determines whether and how long to store packets locally to minimize path delays. each node also maintains a link cost matrix indicating the cost of links to neighboring nodes and further determines whether and how long to store packets locally to minimize path delays at minimal link cost.",2010-07-27,"The title of the patent is interplanetary communications network, interplanetary communications network backbone and method of managing interplanetary communications network and its abstract is an interplanetary communications network, an interplanetary communications backbone network of artificial neural network (ann) nodes, an ann node and a method of managing interplanetary communications. the backbone network operates as a neural network with each node identifying optimum paths, e.g., end-to-end through the backbone network from a distant planet to an earth node. each node maintains a window matrix identifying reoccurring (e.g., periodically) communications windows between nodes and a propagation delay matrix identifying time varying propagation delays between nodes. each node determines whether and how long to store packets locally to minimize path delays. each node also maintains a link cost matrix indicating the cost of links to neighboring nodes and further determines whether and how long to store packets locally to minimize path delays at minimal link cost. dated 2010-07-27"
7765029,hybrid control device,"a brain-based device (bbd) for moving in a real-world environment has sensors that provide data about the environment, actuators to move the bbd, and a hybrid controller which includes a neural controller having a simulated nervous system being a model of selected areas of the human brain and a non-neural controller based on a computational algorithmic network. the neural controller and non-neural controller interact with one another to control movement of the bbd.",2010-07-27,"The title of the patent is hybrid control device and its abstract is a brain-based device (bbd) for moving in a real-world environment has sensors that provide data about the environment, actuators to move the bbd, and a hybrid controller which includes a neural controller having a simulated nervous system being a model of selected areas of the human brain and a non-neural controller based on a computational algorithmic network. the neural controller and non-neural controller interact with one another to control movement of the bbd. dated 2010-07-27"
7765174,linear associative memory-based hardware architecture for fault tolerant asic/fpga work-around,"a programmable logic unit (e.g., an asic or fpga) having a feedforward linear associative memory (lam) neural network checking circuit which classifies input vectors to a faulty hardware block as either good or not good and, when a new input vector is classified as not good, blocks a corresponding output vector of the faulty hardware block, enables a software work-around for the new input vector, and accepts the software work-around input as the output vector of the programmable logic circuit. the feedforward lam neural network checking circuit has a weight matrix whose elements are based on a set of known bad input vectors for said faulty hardware block. the feedforward lam neural network checking circuit may update the weight matrix online using one or more additional bad input vectors. a discrete hopfield algorithm is used to calculate the weight matrix w. the feedforward lam neural network checking circuit calculates an output vector a(m) by multiplying the weight matrix w by the new input vector b(m), that is, a(m)=wb(m), adjusts elements of the output vector a(m) by respective thresholds, and processes the elements using a plurality of non-linear units to provide an output of 1 when a given adjusted element is positive, and provide an output of 0 when a given adjusted element is not positive. if a vector constructed of the outputs of these non-linear units matches with an entry in a content-addressable memory (cam) storing the set of known bad vectors (a cam hit), then the new input vector is classified as not good.",2010-07-27,"The title of the patent is linear associative memory-based hardware architecture for fault tolerant asic/fpga work-around and its abstract is a programmable logic unit (e.g., an asic or fpga) having a feedforward linear associative memory (lam) neural network checking circuit which classifies input vectors to a faulty hardware block as either good or not good and, when a new input vector is classified as not good, blocks a corresponding output vector of the faulty hardware block, enables a software work-around for the new input vector, and accepts the software work-around input as the output vector of the programmable logic circuit. the feedforward lam neural network checking circuit has a weight matrix whose elements are based on a set of known bad input vectors for said faulty hardware block. the feedforward lam neural network checking circuit may update the weight matrix online using one or more additional bad input vectors. a discrete hopfield algorithm is used to calculate the weight matrix w. the feedforward lam neural network checking circuit calculates an output vector a(m) by multiplying the weight matrix w by the new input vector b(m), that is, a(m)=wb(m), adjusts elements of the output vector a(m) by respective thresholds, and processes the elements using a plurality of non-linear units to provide an output of 1 when a given adjusted element is positive, and provide an output of 0 when a given adjusted element is not positive. if a vector constructed of the outputs of these non-linear units matches with an entry in a content-addressable memory (cam) storing the set of known bad vectors (a cam hit), then the new input vector is classified as not good. dated 2010-07-27"
7765795,nox control using a neural network,a method of controlling engine nox production is provided. the method may include determining a desired amount of nox production for at least one engine cylinder at a first time and determining at least one engine operating parameter to produce the desired amount of nox using a feed-forward neural network.,2010-08-03,The title of the patent is nox control using a neural network and its abstract is a method of controlling engine nox production is provided. the method may include determining a desired amount of nox production for at least one engine cylinder at a first time and determining at least one engine operating parameter to produce the desired amount of nox using a feed-forward neural network. dated 2010-08-03
7769482,methods and systems for controlling a semiconductor fabrication process,"software for controlling processes in a heterogeneous semiconductor manufacturing environment may include a wafer-centric database, a real-time scheduler using a neural network, and a graphical user interface displaying simulated operation of the system. these features may be employed alone or in combination to offer improved usability and computational efficiency for real time control and monitoring of a semiconductor manufacturing process. more generally, these techniques may be usefully employed in a variety of real time control systems, particularly systems requiring complex scheduling decisions or heterogeneous systems constructed of hardware from numerous independent vendors.",2010-08-03,"The title of the patent is methods and systems for controlling a semiconductor fabrication process and its abstract is software for controlling processes in a heterogeneous semiconductor manufacturing environment may include a wafer-centric database, a real-time scheduler using a neural network, and a graphical user interface displaying simulated operation of the system. these features may be employed alone or in combination to offer improved usability and computational efficiency for real time control and monitoring of a semiconductor manufacturing process. more generally, these techniques may be usefully employed in a variety of real time control systems, particularly systems requiring complex scheduling decisions or heterogeneous systems constructed of hardware from numerous independent vendors. dated 2010-08-03"
7769513,image processing for vehicular applications applying edge detection technique,arrangement and method for obtaining information about objects in an environment in or around a vehicle includes one or more optical imagers for obtaining images of the environment and a processor coupled to the imager(s) for obtaining information about an object in one or more images obtained by the imager(s). the processor is arranged to process the obtained images to determine edges of objects in the images and input data about the edges into a trained pattern recognition algorithm which has been trained to provide information about the object as output. the pattern recognition algorithm may include a neural network or variation thereof. the information about the object may be used to control a vehicular component such as an airbag or light filter.,2010-08-03,The title of the patent is image processing for vehicular applications applying edge detection technique and its abstract is arrangement and method for obtaining information about objects in an environment in or around a vehicle includes one or more optical imagers for obtaining images of the environment and a processor coupled to the imager(s) for obtaining information about an object in one or more images obtained by the imager(s). the processor is arranged to process the obtained images to determine edges of objects in the images and input data about the edges into a trained pattern recognition algorithm which has been trained to provide information about the object as output. the pattern recognition algorithm may include a neural network or variation thereof. the information about the object may be used to control a vehicular component such as an airbag or light filter. dated 2010-08-03
7769580,method of optimising the execution of a neural network in a speech recognition system through conditionally skipping a variable number of frames,"a method of optimizing the execution of a neural network in a speech recognition system provides for conditionally skipping a variable number of frames, depending on a distance computed between output probabilities, or likelihoods, of a neural network. the distance is initially evaluated between two frames at times 1 and 1+k, where k is a predetermined maximum distance between frames, and if such distance is sufficiently small, the frames between times 1 and 1+k are calculated by interpolation, avoiding further executions of the neural network. if, on the contrary, such distance is not small enough, it means that the outputs of the network are changing quickly, and it is not possible to skip too many frames. in that case, the method attempts to skip remaining frames, calculating and evaluating a new distance.",2010-08-03,"The title of the patent is method of optimising the execution of a neural network in a speech recognition system through conditionally skipping a variable number of frames and its abstract is a method of optimizing the execution of a neural network in a speech recognition system provides for conditionally skipping a variable number of frames, depending on a distance computed between output probabilities, or likelihoods, of a neural network. the distance is initially evaluated between two frames at times 1 and 1+k, where k is a predetermined maximum distance between frames, and if such distance is sufficiently small, the frames between times 1 and 1+k are calculated by interpolation, avoiding further executions of the neural network. if, on the contrary, such distance is not small enough, it means that the outputs of the network are changing quickly, and it is not possible to skip too many frames. in that case, the method attempts to skip remaining frames, calculating and evaluating a new distance. dated 2010-08-03"
7769703,"system, apparatus and methods for augmenting a filter with an adaptive element for tracking targets","a system in accordance with the invention uses an adaptive element to augment a filter for tracking an observed system. the adaptive element only requires a single neural network and does not require an error observer. the adaptive element provides robustness to parameter uncertainty and unmodeled dynamics present in the observed system for improved tracking performance over the filter alone. the adaptive element can be implemented with a linearly parameterized neural network, whose weights are adapted online using error residuals generated from the filter. boundedness of the signals generated by the system can be proven using lyapunov's direct method and a backstepping argument. a related apparatus and method are also disclosed.",2010-08-03,"The title of the patent is system, apparatus and methods for augmenting a filter with an adaptive element for tracking targets and its abstract is a system in accordance with the invention uses an adaptive element to augment a filter for tracking an observed system. the adaptive element only requires a single neural network and does not require an error observer. the adaptive element provides robustness to parameter uncertainty and unmodeled dynamics present in the observed system for improved tracking performance over the filter alone. the adaptive element can be implemented with a linearly parameterized neural network, whose weights are adapted online using error residuals generated from the filter. boundedness of the signals generated by the system can be proven using lyapunov's direct method and a backstepping argument. a related apparatus and method are also disclosed. dated 2010-08-03"
7774149,water leakage-acoustic sensing method and apparatus in steam generator of sodium-cooled fast reactor using standard deviation by octave band analysis,"a water leakage-acoustic sensing method in a steam generator of a sodium-cooled fast reactor, the method including: calculating a standard deviation and an average of an octave band by octave band analysis of an input signal sound received from at least one predetermined acoustic sensor; comparing the calculated standard deviation and the calculated average of the octave band, and determining a size of the octave band based on a comparison result; calculating an average of standard deviations of the octave band recomposed by the determined size and normalizing the average of standard deviations; applying a predetermined weight, established by a predetermined neural network learning algorithm, to the normalized average of standard deviations; and generating leakage determination data based on the average of standard deviations to which the weight is applied.",2010-08-10,"The title of the patent is water leakage-acoustic sensing method and apparatus in steam generator of sodium-cooled fast reactor using standard deviation by octave band analysis and its abstract is a water leakage-acoustic sensing method in a steam generator of a sodium-cooled fast reactor, the method including: calculating a standard deviation and an average of an octave band by octave band analysis of an input signal sound received from at least one predetermined acoustic sensor; comparing the calculated standard deviation and the calculated average of the octave band, and determining a size of the octave band based on a comparison result; calculating an average of standard deviations of the octave band recomposed by the determined size and normalizing the average of standard deviations; applying a predetermined weight, established by a predetermined neural network learning algorithm, to the normalized average of standard deviations; and generating leakage determination data based on the average of standard deviations to which the weight is applied. dated 2010-08-10"
7778724,device for estimating machining dimension of machine tool,"a device for estimating machining dimensions of a machine tool which employs tool members each being rotatably driven by a driving unit includes: a vibration sensor; a characteristics extracting unit for extracting amounts of characteristics from an output of the vibration sensor; a neural network for classifying the amounts of characteristics into categories; and a conversion unit. amounts of characteristics of generated output by racing the tool member are used for training the neural network, and inputted again to the trained competitive learning neural network to excite neurons so that the relationships between euclidean distances and machining dimensions of workpieces are registered in the conversion unit. the euclidean distances are obtained between weight vectors of the excited neurons and respective corresponding training samples, and the machining dimensions are obtained when the workpieces are machined by the tool members at the same condition as the respective corresponding training samples are obtained.",2010-08-17,"The title of the patent is device for estimating machining dimension of machine tool and its abstract is a device for estimating machining dimensions of a machine tool which employs tool members each being rotatably driven by a driving unit includes: a vibration sensor; a characteristics extracting unit for extracting amounts of characteristics from an output of the vibration sensor; a neural network for classifying the amounts of characteristics into categories; and a conversion unit. amounts of characteristics of generated output by racing the tool member are used for training the neural network, and inputted again to the trained competitive learning neural network to excite neurons so that the relationships between euclidean distances and machining dimensions of workpieces are registered in the conversion unit. the euclidean distances are obtained between weight vectors of the excited neurons and respective corresponding training samples, and the machining dimensions are obtained when the workpieces are machined by the tool members at the same condition as the respective corresponding training samples are obtained. dated 2010-08-17"
7778946,neural networks with learning and expression capability,a neural network comprising a plurality of neurons in which any one of the plurality of neurons is able to associate with itself or another neuron in the plurality of neurons via active connections to a further neuron in the plurality of neurons.,2010-08-17,The title of the patent is neural networks with learning and expression capability and its abstract is a neural network comprising a plurality of neurons in which any one of the plurality of neurons is able to associate with itself or another neuron in the plurality of neurons via active connections to a further neuron in the plurality of neurons. dated 2010-08-17
7778947,anomaly monitoring device using two competitive neural networks,"an anomaly monitoring device includes two neural networks which are switchable between a training mode by using training samples and a checking mode for classifying, based on a training result, whether an amount of characteristics obtained by an operation of an apparatus indicates that the operation of an apparatus is normal and a mode switching unit controlling one of the neural networks to operate in training mode and the other neural network to operate in the checking mode. further, the anomaly monitoring device includes a switching determining unit computing a judgment evaluation value serving to evaluate reliability of a judgment result of the other neural network operating in the checking mode, and for instructing the mode switching unit to have the one of the neural networks operate in the checking mode and the other neural network operate in training mode when the judgment evaluation value does not meet evaluation criteria.",2010-08-17,"The title of the patent is anomaly monitoring device using two competitive neural networks and its abstract is an anomaly monitoring device includes two neural networks which are switchable between a training mode by using training samples and a checking mode for classifying, based on a training result, whether an amount of characteristics obtained by an operation of an apparatus indicates that the operation of an apparatus is normal and a mode switching unit controlling one of the neural networks to operate in training mode and the other neural network to operate in the checking mode. further, the anomaly monitoring device includes a switching determining unit computing a judgment evaluation value serving to evaluate reliability of a judgment result of the other neural network operating in the checking mode, and for instructing the mode switching unit to have the one of the neural networks operate in the checking mode and the other neural network operate in training mode when the judgment evaluation value does not meet evaluation criteria. dated 2010-08-17"
7782979,base-band digital pre-distortion-based method for improving efficiency of rf power amplifier,"the present invention relates to a bdpd-based method for improving efficiency of rf power amplifier, comprising: first, choose key neural network architecture and scale and input initial values of modeling data and network parameters necessary for establishing the neural network model for rf power amplifier; second, correct network parameters with back propagation method and output the neural network model for rf power amplifier when the error meets the criterion; next, solve the pre-distortion algorithm of the rf power amplifier with said model and then carry out pre-distortion processing for the input with the pre-distortion algorithm and feed the input to the rf power amplifier. the present invention can be used to establish a neural network model with adequate accuracy and easy to solve corresponding pre-distortion algorithm for rf power amplifier, in order to improve rf power amplifier efficiency, reduce costs, and suppress out-of-band spectrum leakage effectively through base-band digital pre-distortion technology.",2010-08-24,"The title of the patent is base-band digital pre-distortion-based method for improving efficiency of rf power amplifier and its abstract is the present invention relates to a bdpd-based method for improving efficiency of rf power amplifier, comprising: first, choose key neural network architecture and scale and input initial values of modeling data and network parameters necessary for establishing the neural network model for rf power amplifier; second, correct network parameters with back propagation method and output the neural network model for rf power amplifier when the error meets the criterion; next, solve the pre-distortion algorithm of the rf power amplifier with said model and then carry out pre-distortion processing for the input with the pre-distortion algorithm and feed the input to the rf power amplifier. the present invention can be used to establish a neural network model with adequate accuracy and easy to solve corresponding pre-distortion algorithm for rf power amplifier, in order to improve rf power amplifier efficiency, reduce costs, and suppress out-of-band spectrum leakage effectively through base-band digital pre-distortion technology. dated 2010-08-24"
7788194,method for controlling game character,"a method for controlling a game character is provided. the method includes analyzing a game situation in which a character appears; and controlling a behavior of the character depending on a result of the analyzing. accordingly, situation recognition and behavior control depending on the recognized situation are simultaneously performed using the same algorithm, so that the calculation amount can be reduced, and thus high artificial intelligence can be implemented with less computer resources. also, a game developer does not need to implement individual behavior rules of characters depending on game situations since a situation of the game may be recognized through learning of an artificial neural network using a game database of game situations, and thus the behavior of characters can be controlled depending on the recognized situation.",2010-08-31,"The title of the patent is method for controlling game character and its abstract is a method for controlling a game character is provided. the method includes analyzing a game situation in which a character appears; and controlling a behavior of the character depending on a result of the analyzing. accordingly, situation recognition and behavior control depending on the recognized situation are simultaneously performed using the same algorithm, so that the calculation amount can be reduced, and thus high artificial intelligence can be implemented with less computer resources. also, a game developer does not need to implement individual behavior rules of characters depending on game situations since a situation of the game may be recognized through learning of an artificial neural network using a game database of game situations, and thus the behavior of characters can be controlled depending on the recognized situation. dated 2010-08-31"
7788196,artificial neural network,an artificial neural network comprises at least one input layer with a predetermined number of input nodes and at least one output layer with a predetermined number of output nodes or also at least one intermediate hidden layer with a predetermined number of nodes between the input and the output layer. at least the nodes of the output layer and/or of the hidden layer and/or also of the input layer carry out a non linear transformation of a first non linear transformation of the input data for computing an output value to be fed as an input value to a following layer or the output data if the output layer is considered.,2010-08-31,The title of the patent is artificial neural network and its abstract is an artificial neural network comprises at least one input layer with a predetermined number of input nodes and at least one output layer with a predetermined number of output nodes or also at least one intermediate hidden layer with a predetermined number of nodes between the input and the output layer. at least the nodes of the output layer and/or of the hidden layer and/or also of the input layer carry out a non linear transformation of a first non linear transformation of the input data for computing an output value to be fed as an input value to a following layer or the output data if the output layer is considered. dated 2010-08-31
7792631,control system for internal combustion engine,a control system for an internal combustion engine having an exhaust gas recirculation device for recirculating a part of exhaust gases to an intake system of the engine is disclosed. an estimated exhaust gas recirculation amount is calculated using a neural network to which at least one engine operating parameter indicative of an operating condition of the engine is input. the neural network outputs an estimated value of an amount of exhaust gases recirculated by the exhaust gas recirculation device. at least one engine control parameter for controlling the engine is calculated based on the estimated exhaust gas recirculation amount.,2010-09-07,The title of the patent is control system for internal combustion engine and its abstract is a control system for an internal combustion engine having an exhaust gas recirculation device for recirculating a part of exhaust gases to an intake system of the engine is disclosed. an estimated exhaust gas recirculation amount is calculated using a neural network to which at least one engine operating parameter indicative of an operating condition of the engine is input. the neural network outputs an estimated value of an amount of exhaust gases recirculated by the exhaust gas recirculation device. at least one engine control parameter for controlling the engine is calculated based on the estimated exhaust gas recirculation amount. dated 2010-09-07
7792766,design of reconnaissance surveys using controlled source electromagnetic fields via probabilistic neural network,method for determining an expected value for a proposed reconnaissance electromagnetic (or any other type of geophysical) survey using a user-controlled source. the method requires only available geologic and economic information about the survey region. a series of calibration surveys are simulated with an assortment of resistive targets consistent with the known information. the calibration surveys are used to train pattern recognition software to assess the economic potential from anomalous resistivity maps. the calibrated classifier is then used on further simulated surveys of the area to generate probabilities that can be used in value of information theory to predict an expected value of a survey of the same design as the simulated surveys. the calibrated classifier technique can also be used to interpret actual csem survey results for economic potential.,2010-09-07,The title of the patent is design of reconnaissance surveys using controlled source electromagnetic fields via probabilistic neural network and its abstract is method for determining an expected value for a proposed reconnaissance electromagnetic (or any other type of geophysical) survey using a user-controlled source. the method requires only available geologic and economic information about the survey region. a series of calibration surveys are simulated with an assortment of resistive targets consistent with the known information. the calibration surveys are used to train pattern recognition software to assess the economic potential from anomalous resistivity maps. the calibrated classifier is then used on further simulated surveys of the area to generate probabilities that can be used in value of information theory to predict an expected value of a survey of the same design as the simulated surveys. the calibrated classifier technique can also be used to interpret actual csem survey results for economic potential. dated 2010-09-07
7792767,message routing using cyclical neural networks,"a system for routing business-to-business (“b2b”) messages includes a cyclical neural network. the cyclical neural network contains neurons for determining a needed destination of a message based on content type of the message, for example. neurons are monitored to establish a “state of understanding” of the network during processing, and tags may be applied to messages upon a determination of the needed destination.",2010-09-07,"The title of the patent is message routing using cyclical neural networks and its abstract is a system for routing business-to-business (“b2b”) messages includes a cyclical neural network. the cyclical neural network contains neurons for determining a needed destination of a message based on content type of the message, for example. neurons are monitored to establish a “state of understanding” of the network during processing, and tags may be applied to messages upon a determination of the needed destination. dated 2010-09-07"
7800052,method and system for stabilizing gain of a photomultipler used with a radiation detector,a method for controlling voltage applied to a photomultiplier used in a scintillation counter radiation detector includes determining numbers of voltage pulses having each of a plurality of predetermined amplitudes generated by the photomultiplier in response to radiation events being imparted to a scintillation detector. the numbers of voltage pulses at each of the predetermined amplitudes is conducted to a trained artificial neural network. the artificial neural network generates a signal corresponding to an amount of adjustment to the voltage applied to the photomultiplier.,2010-09-21,The title of the patent is method and system for stabilizing gain of a photomultipler used with a radiation detector and its abstract is a method for controlling voltage applied to a photomultiplier used in a scintillation counter radiation detector includes determining numbers of voltage pulses having each of a plurality of predetermined amplitudes generated by the photomultiplier in response to radiation events being imparted to a scintillation detector. the numbers of voltage pulses at each of the predetermined amplitudes is conducted to a trained artificial neural network. the artificial neural network generates a signal corresponding to an amount of adjustment to the voltage applied to the photomultiplier. dated 2010-09-21
7800490,electronic article surveillance system neural network minimizing false alarms and failures to deactivate,"a method, system and computer program product for managing false alarms in a security system. a detection zone is established. an alarm event is triggered based on the detection of a tag in the detection zone using an initial alarm trigger sensitivity. the initial alarm trigger sensitivity is based on an initial set of one or more detection criteria. the set of detection criteria is modified to adjust the alarm trigger sensitivity of the security system.",2010-09-21,"The title of the patent is electronic article surveillance system neural network minimizing false alarms and failures to deactivate and its abstract is a method, system and computer program product for managing false alarms in a security system. a detection zone is established. an alarm event is triggered based on the detection of a tag in the detection zone using an initial alarm trigger sensitivity. the initial alarm trigger sensitivity is based on an initial set of one or more detection criteria. the set of detection criteria is modified to adjust the alarm trigger sensitivity of the security system. dated 2010-09-21"
7809432,event detection—apparatus and method for measuring the activity of neural networks,"apparatus for measuring neural network activity with a textured semiconductor substrate. sensor elements have a respective detection electrode on the substrate surface for detecting neural network signals, and the detected neural signals are a basis for outputting electrical sensor output signals via respective sensor element outputs. each amplifier element has an input and an output. each of the sensor elements has associated therewith one of the amplifier elements whose input is connected to the sensor output of the respective sensor element. the amplified sensor output signal is output the amplifier output as an amplifier output signal. an activity evaluator has an input, which is connected to at least one of the amplifier outputs, and an output. the activity evaluation device produces an activity signal, which is a measure of activity of the neural network, based on the amplifier output signal, and outputs the amplifier output signal via the evaluation output.",2010-10-05,"The title of the patent is event detection—apparatus and method for measuring the activity of neural networks and its abstract is apparatus for measuring neural network activity with a textured semiconductor substrate. sensor elements have a respective detection electrode on the substrate surface for detecting neural network signals, and the detected neural signals are a basis for outputting electrical sensor output signals via respective sensor element outputs. each amplifier element has an input and an output. each of the sensor elements has associated therewith one of the amplifier elements whose input is connected to the sensor output of the respective sensor element. the amplified sensor output signal is output the amplifier output as an amplifier output signal. an activity evaluator has an input, which is connected to at least one of the amplifier outputs, and an output. the activity evaluation device produces an activity signal, which is a measure of activity of the neural network, based on the amplifier output signal, and outputs the amplifier output signal via the evaluation output. dated 2010-10-05"
7813543,computer modeling of physical scenes,"the present invention relates to automatic modeling of a physical scene. at least two images (i1, i2) of the scene are received, which are taken from different angles and/or positions. a matching module (130) matches image objects in the first image (i1) against image objects in the second image (i2), by first loading pixel values for at least one first portion of the first image (i1) into an artificial neural network (133). then, the artificial neural network (133) scans the second image (i2) in search of pixels representing a respective second portion corresponding to each of the at least one first portion; determines a position of the respective second portion upon fulfillment of a match criterion; and produces a representative matching result (m12). based on the matching result (m12), a first calculation module (140) calculates a fundamental matrix (f12), which defines a relationship between the first and second images (i1, i2). based on the fundamental matrix (f12), in turn, a second calculation module (150) calculates a depth map (d12), which describes distance differences between a set of image points in the first image (i1) and a corresponding set of image points in the second image (i2). finally, the depth map (d12) constitutes a basis for a synthetic model of the scene.",2010-10-12,"The title of the patent is computer modeling of physical scenes and its abstract is the present invention relates to automatic modeling of a physical scene. at least two images (i1, i2) of the scene are received, which are taken from different angles and/or positions. a matching module (130) matches image objects in the first image (i1) against image objects in the second image (i2), by first loading pixel values for at least one first portion of the first image (i1) into an artificial neural network (133). then, the artificial neural network (133) scans the second image (i2) in search of pixels representing a respective second portion corresponding to each of the at least one first portion; determines a position of the respective second portion upon fulfillment of a match criterion; and produces a representative matching result (m12). based on the matching result (m12), a first calculation module (140) calculates a fundamental matrix (f12), which defines a relationship between the first and second images (i1, i2). based on the fundamental matrix (f12), in turn, a second calculation module (150) calculates a depth map (d12), which describes distance differences between a set of image points in the first image (i1) and a corresponding set of image points in the second image (i2). finally, the depth map (d12) constitutes a basis for a synthetic model of the scene. dated 2010-10-12"
7814036,processing well logging data with neural network,"an artificial neural network, ann, and method of training the ann for inversion of logging tool signals into well logs of formation parameters is disclosed. properly selected synthetic models of earth formations are used to train the ann. the models include oklahoma and chirp type of formations. in each model parameter contrasts of from 10 to 1 to about 100 to 1 are included. models including maximum and minimum parameter values spanning the operating range of the selected logging tool are included. parameter contrasts at interfaces are limited to realistic values found in earth formations. the selected models are used to generate synthetic tool signals, which are then used as inputs to the ann for training. when the ann coefficients are properly adjusted to produce an output matching the original models, the ann can be used for inversion of any real signals from the selected logging tool.",2010-10-12,"The title of the patent is processing well logging data with neural network and its abstract is an artificial neural network, ann, and method of training the ann for inversion of logging tool signals into well logs of formation parameters is disclosed. properly selected synthetic models of earth formations are used to train the ann. the models include oklahoma and chirp type of formations. in each model parameter contrasts of from 10 to 1 to about 100 to 1 are included. models including maximum and minimum parameter values spanning the operating range of the selected logging tool are included. parameter contrasts at interfaces are limited to realistic values found in earth formations. the selected models are used to generate synthetic tool signals, which are then used as inputs to the ann for training. when the ann coefficients are properly adjusted to produce an output matching the original models, the ann can be used for inversion of any real signals from the selected logging tool. dated 2010-10-12"
7814038,feedback-tolerant method and device producing weight-adjustment factors for pre-synaptic neurons in artificial neural networks,"in an artificial neural network a method and neuron device that produce weight-adjustment factors, also called error values (116), for pre-synaptic neurons (302a . . . 302c) that are used to adjust the values of connection weights (106 . . . 106n) in neurons (100) used in artificial neural networks (anns). the amount of influence a pre-synaptic neuron has had over a post-synaptic neuron is calculated during signal propagation in the post-synaptic neuron (422a . . . 422n) and accumulated for the pre-synaptic neuron (426) for each post-synaptic neuron to which the pre-synaptic neuron's output is connected (428). influence values calculated for use by pre-synaptic neurons may further be modified by the post-synaptic neuron's output value (102) (option 424), and its error value (116) (option 1110).",2010-10-12,"The title of the patent is feedback-tolerant method and device producing weight-adjustment factors for pre-synaptic neurons in artificial neural networks and its abstract is in an artificial neural network a method and neuron device that produce weight-adjustment factors, also called error values (116), for pre-synaptic neurons (302a . . . 302c) that are used to adjust the values of connection weights (106 . . . 106n) in neurons (100) used in artificial neural networks (anns). the amount of influence a pre-synaptic neuron has had over a post-synaptic neuron is calculated during signal propagation in the post-synaptic neuron (422a . . . 422n) and accumulated for the pre-synaptic neuron (426) for each post-synaptic neuron to which the pre-synaptic neuron's output is connected (428). influence values calculated for use by pre-synaptic neurons may further be modified by the post-synaptic neuron's output value (102) (option 424), and its error value (116) (option 1110). dated 2010-10-12"
7817848,"apparatus, method, and computer product for discriminating object","an apparatus discriminates a potential obstacle in the path of a vehicle from among various objects in an image shot by a monocular camera. first, an object detecting unit detects an object in the image by applying a saliency calculation to the image. second, an object discriminating unit discriminates an object from among the objects detected by the object detecting unit as a potential obstacle by applying a neural network method to the objects.",2010-10-19,"The title of the patent is apparatus, method, and computer product for discriminating object and its abstract is an apparatus discriminates a potential obstacle in the path of a vehicle from among various objects in an image shot by a monocular camera. first, an object detecting unit detects an object in the image by applying a saliency calculation to the image. second, an object discriminating unit discriminates an object from among the objects detected by the object detecting unit as a potential obstacle by applying a neural network method to the objects. dated 2010-10-19"
7817857,combiner for improving handwriting recognition,"various technologies and techniques are disclosed that improve handwriting recognition operations. handwritten input is received in training mode and run through several base recognizers to generate several alternate lists. the alternate lists are unioned together into a combined alternate list. if the correct result is in the combined list, each correct/incorrect alternate pair is used to generate training patterns. the weights associated with the alternate pairs are stored. at runtime, the combined alternate list is generated just as training time. the trained comparator-net can be used to compare any two alternates in the combined list. a template matching base recognizer is used with one or more neural network base recognizers to improve recognition operations. the system provides comparator-net and reorder-net processes trained on print and cursive data, and ones that have been trained on cursive-only data. the respective comparator-net and reorder-net processes are used accordingly.",2010-10-19,"The title of the patent is combiner for improving handwriting recognition and its abstract is various technologies and techniques are disclosed that improve handwriting recognition operations. handwritten input is received in training mode and run through several base recognizers to generate several alternate lists. the alternate lists are unioned together into a combined alternate list. if the correct result is in the combined list, each correct/incorrect alternate pair is used to generate training patterns. the weights associated with the alternate pairs are stored. at runtime, the combined alternate list is generated just as training time. the trained comparator-net can be used to compare any two alternates in the combined list. a template matching base recognizer is used with one or more neural network base recognizers to improve recognition operations. the system provides comparator-net and reorder-net processes trained on print and cursive data, and ones that have been trained on cursive-only data. the respective comparator-net and reorder-net processes are used accordingly. dated 2010-10-19"
7821673,method and apparatus for removing visible artefacts in video images,a method and apparatus are provided for removing regularly occurring visible artifacts in decompressed video images. firstly a decompressed video signal is received. this is filtered frame-by-frame to extract data related to the artifacts. the thus extracted data is then processed in a neural network processor which has been trained to identify the artifacts in order to produce data identifying their locations. the video signal is then corrected to reduce the effect of the thus identified artifacts.,2010-10-26,The title of the patent is method and apparatus for removing visible artefacts in video images and its abstract is a method and apparatus are provided for removing regularly occurring visible artifacts in decompressed video images. firstly a decompressed video signal is received. this is filtered frame-by-frame to extract data related to the artifacts. the thus extracted data is then processed in a neural network processor which has been trained to identify the artifacts in order to produce data identifying their locations. the video signal is then corrected to reduce the effect of the thus identified artifacts. dated 2010-10-26
7822698,spike domain and pulse domain non-linear processors,"a neural network has an array of interconnected processors, each processor operating either the pulse domain or spike domain. each processor has (i) first inputs selectively coupled to other processors in the array of processors, each first input having an associated 1 bit dac coupled to a summing node, (ii) second inputs selectively coupled to inputs of the neural network, the second inputs having current generators associated therewith coupled to said summing node, (iii) a filter/integrator for generating an analog signal corresponding to current arriving at the summing node, (iv) an optional nonlinear element coupled to the filter/integrator, and (v) an analog-to-pulse converter, if the processors operate in the pulse domain, or an analog-to-spike convertor, if the processors operate in the spike domain, for converting an analog signal output by the optional nonlinear element or by the filter/integrator to either the pulse domain or spike domain, and providing the converted analog signal as an unquantized pulse or spike domain signal at an output of the processor. the array of processors are selectively interconnected with either unquantized pulse domain or spike domain signals.",2010-10-26,"The title of the patent is spike domain and pulse domain non-linear processors and its abstract is a neural network has an array of interconnected processors, each processor operating either the pulse domain or spike domain. each processor has (i) first inputs selectively coupled to other processors in the array of processors, each first input having an associated 1 bit dac coupled to a summing node, (ii) second inputs selectively coupled to inputs of the neural network, the second inputs having current generators associated therewith coupled to said summing node, (iii) a filter/integrator for generating an analog signal corresponding to current arriving at the summing node, (iv) an optional nonlinear element coupled to the filter/integrator, and (v) an analog-to-pulse converter, if the processors operate in the pulse domain, or an analog-to-spike convertor, if the processors operate in the spike domain, for converting an analog signal output by the optional nonlinear element or by the filter/integrator to either the pulse domain or spike domain, and providing the converted analog signal as an unquantized pulse or spike domain signal at an output of the processor. the array of processors are selectively interconnected with either unquantized pulse domain or spike domain signals. dated 2010-10-26"
7826642,electro-optical method and apparatus for evaluating protrusions of fibers from a fabric surface,an electro-optical method and apparatus for evaluating the dimensions of any protrusion from the threshold of the fabric surface is achieved by bending any length of fabric over a rotating roller so that the contoured area of the protrusion body above the surface can be visualized. the image of the silhouette as seen by a digital camera is processed by image processing algorithms then processed statistically and then by a neural network to yield an integrated picture of the fabric protrusions. the grading results of pilling are well correlated to the human visual method of pilling evaluation.,2010-11-02,The title of the patent is electro-optical method and apparatus for evaluating protrusions of fibers from a fabric surface and its abstract is an electro-optical method and apparatus for evaluating the dimensions of any protrusion from the threshold of the fabric surface is achieved by bending any length of fabric over a rotating roller so that the contoured area of the protrusion body above the surface can be visualized. the image of the silhouette as seen by a digital camera is processed by image processing algorithms then processed statistically and then by a neural network to yield an integrated picture of the fabric protrusions. the grading results of pilling are well correlated to the human visual method of pilling evaluation. dated 2010-11-02
7827031,method for accelerating the execution of speech recognition neural networks and the related speech recognition device,"a neural network in a speech-recognition system has computing units organized in levels including at least one hidden level and one output level. the computing units of the hidden level are connected to the computing units of the output level via weighted connections, and the computing units of the output level correspond to acoustic-phonetic units of the general vocabulary. this network executes the following steps:determining a subset of acoustic-phonetic units necessary for recognizing all the words contained in the general vocabulary subset;eliminating from the neural network all the weighted connections afferent to computing units of the output level that correspond to acoustic-phonetic units not contained in the previously determined subset of acoustic-phonetic units, thus obtaining a compacted neural network optimized for recognition of the words contained in the general vocabulary subset; andexecuting, at each moment in time, only the compacted neural network.",2010-11-02,"The title of the patent is method for accelerating the execution of speech recognition neural networks and the related speech recognition device and its abstract is a neural network in a speech-recognition system has computing units organized in levels including at least one hidden level and one output level. the computing units of the hidden level are connected to the computing units of the output level via weighted connections, and the computing units of the output level correspond to acoustic-phonetic units of the general vocabulary. this network executes the following steps:determining a subset of acoustic-phonetic units necessary for recognizing all the words contained in the general vocabulary subset;eliminating from the neural network all the weighted connections afferent to computing units of the output level that correspond to acoustic-phonetic units not contained in the previously determined subset of acoustic-phonetic units, thus obtaining a compacted neural network optimized for recognition of the words contained in the general vocabulary subset; andexecuting, at each moment in time, only the compacted neural network. dated 2010-11-02"
7827129,crystal lookup table generation using neural network-based algorithm,"a crystal lookup table used to define a matching relationship between a signal position of a detected event in a pet scanner and a corresponding detector pixel location is generated using a neural network-based algorithm, and is implemented by a fpga.",2010-11-02,"The title of the patent is crystal lookup table generation using neural network-based algorithm and its abstract is a crystal lookup table used to define a matching relationship between a signal position of a detected event in a pet scanner and a corresponding detector pixel location is generated using a neural network-based algorithm, and is implemented by a fpga. dated 2010-11-02"
7827131,high density synapse chip using nanoparticles,"a physical neural network synapse chip and a method for forming such a synapse chip. the synapse chip can be configured to include an input layer comprising a plurality of input electrodes and an output layer comprising a plurality of output electrodes, such that the output electrodes are located perpendicular to the input electrodes. a gap is generally formed between the input layer and the output layer. a solution can then be provided which is prepared from a plurality of nanoconductors and a dielectric solvent. the solution is located within the gap, such that an electric field is applied across the gap from the input layer to the output layer to form nanoconnections of a physical neural network implemented by the synapse chip. such a gap can thus be configured as an electrode gap. the input electrodes can be configured as an array of input electrodes, while the output electrodes can be configured as an array of output electrodes.",2010-11-02,"The title of the patent is high density synapse chip using nanoparticles and its abstract is a physical neural network synapse chip and a method for forming such a synapse chip. the synapse chip can be configured to include an input layer comprising a plurality of input electrodes and an output layer comprising a plurality of output electrodes, such that the output electrodes are located perpendicular to the input electrodes. a gap is generally formed between the input layer and the output layer. a solution can then be provided which is prepared from a plurality of nanoconductors and a dielectric solvent. the solution is located within the gap, such that an electric field is applied across the gap from the input layer to the output layer to form nanoconnections of a physical neural network implemented by the synapse chip. such a gap can thus be configured as an electrode gap. the input electrodes can be configured as an array of input electrodes, while the output electrodes can be configured as an array of output electrodes. dated 2010-11-02"
7831530,optimizing method of learning data set for signal discrimination apparatus and signal discrimination apparatus capable of optimizing learning data set by using a neural network,"a method of the present invention is processed by a selector. the selector selects each member constituting a learning data set from a data set source. each member of the source is feature data extracted through a transducer and assigned to any one of categories in advance. the selector calculates each member's divergence degree of the source to obtain an average divergence degree. if an output neuron of the output layer of a neural network is related to different categories of all the categories represented by the output layer, the selector includes every member of the source corresponding to the category of the minimum average divergence degree in the selection from the source to the learning data set. the selector also excludes, from the selection, every member of the source corresponding to every remaining category of the different categories.",2010-11-09,"The title of the patent is optimizing method of learning data set for signal discrimination apparatus and signal discrimination apparatus capable of optimizing learning data set by using a neural network and its abstract is a method of the present invention is processed by a selector. the selector selects each member constituting a learning data set from a data set source. each member of the source is feature data extracted through a transducer and assigned to any one of categories in advance. the selector calculates each member's divergence degree of the source to obtain an average divergence degree. if an output neuron of the output layer of a neural network is related to different categories of all the categories represented by the output layer, the selector includes every member of the source corresponding to the category of the minimum average divergence degree in the selection from the source to the learning data set. the selector also excludes, from the selection, every member of the source corresponding to every remaining category of the different categories. dated 2010-11-09"
7835576,apparatus and method for editing optimized color preference,"an apparatus and method for editing an optimized color preference are provided. the apparatus includes a color information controlling unit which extracts data about a preference by comparing color information of a transformed image generated by transforming color information of an original image and the original image according to a user preference; a learning unit which teaches a neural network about the preference, based on the extracted data, and predicts color information variation by the neural network; and an image correcting unit which corrects color information of an input image according to the predicted color information variation. the method includes extracting data about a preference; teaching a neural network about the preference, based on the extracted data; predicting color information variation by the neural network; and correcting color information of an input image according to the predicted color information variation.",2010-11-16,"The title of the patent is apparatus and method for editing optimized color preference and its abstract is an apparatus and method for editing an optimized color preference are provided. the apparatus includes a color information controlling unit which extracts data about a preference by comparing color information of a transformed image generated by transforming color information of an original image and the original image according to a user preference; a learning unit which teaches a neural network about the preference, based on the extracted data, and predicts color information variation by the neural network; and an image correcting unit which corrects color information of an input image according to the predicted color information variation. the method includes extracting data about a preference; teaching a neural network about the preference, based on the extracted data; predicting color information variation by the neural network; and correcting color information of an input image according to the predicted color information variation. dated 2010-11-16"
7835999,"recognizing input gestures using a multi-touch input device, calculated graphs, and a neural network with link weights","the present invention extends to methods, systems, and computer program products for recognizing input gestures. a neural network is trained using example inputs and backpropagation to recognize specified input patterns. input gesture data is representative of movements in contact on a multi-touch input display surface relative to one or more axes over time. example inputs used for training the neural network to recognize a specified input pattern can be created from sampling input gesture data for example input gestures known to represent the specified input pattern. trained neural networks can subsequently be used to recognize input gestures that are similar to known input gestures as the specified input pattern corresponding to the known input gestures.",2010-11-16,"The title of the patent is recognizing input gestures using a multi-touch input device, calculated graphs, and a neural network with link weights and its abstract is the present invention extends to methods, systems, and computer program products for recognizing input gestures. a neural network is trained using example inputs and backpropagation to recognize specified input patterns. input gesture data is representative of movements in contact on a multi-touch input display surface relative to one or more axes over time. example inputs used for training the neural network to recognize a specified input pattern can be created from sampling input gesture data for example input gestures known to represent the specified input pattern. trained neural networks can subsequently be used to recognize input gestures that are similar to known input gestures as the specified input pattern corresponding to the known input gestures. dated 2010-11-16"
7840569,enterprise relevancy ranking using a neural network,a neural network is used to process a set of ranking features in order to determine the relevancy ranking for a set of documents or other items. the neural network calculates a predicted relevancy score for each document and the documents can then be ordered by that score. alternate embodiments apply a set of data transformations to the ranking features before they are input to the neural network. training can be used to adapt both the neural network and certain of the data transformations to target environments.,2010-11-23,The title of the patent is enterprise relevancy ranking using a neural network and its abstract is a neural network is used to process a set of ranking features in order to determine the relevancy ranking for a set of documents or other items. the neural network calculates a predicted relevancy score for each document and the documents can then be ordered by that score. alternate embodiments apply a set of data transformations to the ranking features before they are input to the neural network. training can be used to adapt both the neural network and certain of the data transformations to target environments. dated 2010-11-23
7844423,component-based modeling of wireless mac protocols for efficient simulations,"channel access delays and reception uncertainty are modeled as protocol-independent generic processes that are optimized for improved simulation performance. the generic process components are designed such that each different protocol can be modeled using an arrangement of these components that is specific to the protocol. in this way, speed and/or accuracy improvements to the generic process components are reflected in each of such protocol models. if an accurate analytic model is not available for the generic process component, a prediction engine, such as a neural network, is preferably used. the prediction engine is trained using the existing detailed models of network devices. once trained, the prediction engine is used to model the generic process, and the protocol model that includes the generic component is used in lieu of the detailed models, thereby saving substantial processing time.",2010-11-30,"The title of the patent is component-based modeling of wireless mac protocols for efficient simulations and its abstract is channel access delays and reception uncertainty are modeled as protocol-independent generic processes that are optimized for improved simulation performance. the generic process components are designed such that each different protocol can be modeled using an arrangement of these components that is specific to the protocol. in this way, speed and/or accuracy improvements to the generic process components are reflected in each of such protocol models. if an accurate analytic model is not available for the generic process component, a prediction engine, such as a neural network, is preferably used. the prediction engine is trained using the existing detailed models of network devices. once trained, the prediction engine is used to model the generic process, and the protocol model that includes the generic component is used in lieu of the detailed models, thereby saving substantial processing time. dated 2010-11-30"
7847225,optical neural network,an input layer outputs light having a relatively narrow emission angle distribution to a middle layer as an output signal if the signal level of input signal is relatively high and outputs light having a relatively broad emission angle distribution to the middle layer as the output signal if the signal level of input signal is relatively low. the middle layer outputs light having a relatively narrow emission angle distribution as an output signal to an output layer if the signal level of the output signal from input layer is relatively high and outputs light having a relatively broad emission angle distribution to the output layer as an output signal if the signal level of the output signal from the input layer is relatively low.,2010-12-07,The title of the patent is optical neural network and its abstract is an input layer outputs light having a relatively narrow emission angle distribution to a middle layer as an output signal if the signal level of input signal is relatively high and outputs light having a relatively broad emission angle distribution to the middle layer as the output signal if the signal level of input signal is relatively low. the middle layer outputs light having a relatively narrow emission angle distribution as an output signal to an output layer if the signal level of the output signal from input layer is relatively high and outputs light having a relatively broad emission angle distribution to the output layer as an output signal if the signal level of the output signal from the input layer is relatively low. dated 2010-12-07
7848262,neural network-based mobility management for healing mobile ad hoc radio networks,"a self healing ad hoc communications network and method of training for and healing the network. the network includes wireless devices or nodes that include a neural network element and the ad hoc network operates as a neural network. some of the nodes are designated as healing nodes that are identified during network training and are strategically located in the network coverage area. whenever one group of nodes loses connection with another a healing node may reposition itself to reconnect the two groups. thus, the network can maintain connectivity without constraining node movement.",2010-12-07,"The title of the patent is neural network-based mobility management for healing mobile ad hoc radio networks and its abstract is a self healing ad hoc communications network and method of training for and healing the network. the network includes wireless devices or nodes that include a neural network element and the ad hoc network operates as a neural network. some of the nodes are designated as healing nodes that are identified during network training and are strategically located in the network coverage area. whenever one group of nodes loses connection with another a healing node may reposition itself to reconnect the two groups. thus, the network can maintain connectivity without constraining node movement. dated 2010-12-07"
7849029,comprehensive identity protection system,"a system and method for protecting identity fraud are disclosed. a system includes a detection subsystem to identify applications and/or accounts at risk of identity fraud, and a disposition subsystem to process data provided by the detection system and to determine whether identity fraud exists in the applications and/or accounts. according to an implementation, one or more neural network models are defined, each neural network model being configured to handle a class of cases related to the subject and a specific data configuration describing a case of the class. the one or more neural network models are run to generate data requests about the subject's identity, and the data requests are passed to a detection system that monitor transactions associated with the subject. additional data associated with the transactions is requested until a threshold certainty is achieved or until available data or models are exhausted.",2010-12-07,"The title of the patent is comprehensive identity protection system and its abstract is a system and method for protecting identity fraud are disclosed. a system includes a detection subsystem to identify applications and/or accounts at risk of identity fraud, and a disposition subsystem to process data provided by the detection system and to determine whether identity fraud exists in the applications and/or accounts. according to an implementation, one or more neural network models are defined, each neural network model being configured to handle a class of cases related to the subject and a specific data configuration describing a case of the class. the one or more neural network models are run to generate data requests about the subject's identity, and the data requests are passed to a detection system that monitor transactions associated with the subject. additional data associated with the transactions is requested until a threshold certainty is achieved or until available data or models are exhausted. dated 2010-12-07"
7849032,intelligent sampling for neural network data mining models,"a method, system, and computer program product provides automated determination of the size of the sample that is to be used in training a neural network data mining model that is large enough to properly train the neural network data mining model, yet is no larger than is necessary. a method of performing training of a neural network data mining model comprises the steps of: a) providing a training dataset for training an untrained neural network data mining model, the first training dataset comprising a plurality of rows of data, b) selecting a row of data from the training dataset for performing training processing on the neural network data mining model, c) computing an estimate of a gradient or cost function of the neural network data mining model, d) determining whether the gradient or cost function of the neural network data mining model has converged, based on the computed estimate of the gradient or cost function of the neural network data mining model, e) repeating steps b)-d), if the gradient or cost function of the neural network data mining model has not converged, and f) updating weights of the neural network data mining model, if the gradient or cost function of the neural network data mining model has converged.",2010-12-07,"The title of the patent is intelligent sampling for neural network data mining models and its abstract is a method, system, and computer program product provides automated determination of the size of the sample that is to be used in training a neural network data mining model that is large enough to properly train the neural network data mining model, yet is no larger than is necessary. a method of performing training of a neural network data mining model comprises the steps of: a) providing a training dataset for training an untrained neural network data mining model, the first training dataset comprising a plurality of rows of data, b) selecting a row of data from the training dataset for performing training processing on the neural network data mining model, c) computing an estimate of a gradient or cost function of the neural network data mining model, d) determining whether the gradient or cost function of the neural network data mining model has converged, based on the computed estimate of the gradient or cost function of the neural network data mining model, e) repeating steps b)-d), if the gradient or cost function of the neural network data mining model has not converged, and f) updating weights of the neural network data mining model, if the gradient or cost function of the neural network data mining model has converged. dated 2010-12-07"
7853323,selection of neurostimulator parameter configurations using neural networks,"in general, the invention is directed to a technique for selection of parameter configurations for a neurostimulator using neural networks. the technique may be employed by a programming device to allow a clinician to select parameter configurations, and then program an implantable neurostimulator to deliver therapy using the selected parameter configurations. the parameter configurations may include one or more of a variety of parameters, such as electrode configurations defining electrode combinations and polarities for an electrode set implanted in a patient. the electrode set may be carried by one or more implanted leads that are electrically coupled to the neurostimulator. in operation, the programming device executes a parameter configuration search algorithm to guide the clinician in the selection of parameter configurations. the search algorithm relies on a neural network that identifies potential optimum parameter configurations.",2010-12-14,"The title of the patent is selection of neurostimulator parameter configurations using neural networks and its abstract is in general, the invention is directed to a technique for selection of parameter configurations for a neurostimulator using neural networks. the technique may be employed by a programming device to allow a clinician to select parameter configurations, and then program an implantable neurostimulator to deliver therapy using the selected parameter configurations. the parameter configurations may include one or more of a variety of parameters, such as electrode configurations defining electrode combinations and polarities for an electrode set implanted in a patient. the electrode set may be carried by one or more implanted leads that are electrically coupled to the neurostimulator. in operation, the programming device executes a parameter configuration search algorithm to guide the clinician in the selection of parameter configurations. the search algorithm relies on a neural network that identifies potential optimum parameter configurations. dated 2010-12-14"
7860586,process parameter estimation in controlling emission of a non-particulate pollutant into the air,"a parameter value estimator is provided for a process performed primarily to control emission of a particular non-particulate pollutant, such as nox and so2, into the air. the process has multiple process parameters (mpps) including a parameter representing an amount of the particular non-particulate pollutant emitted. the parameter value estimator includes either a neural network process model or a non-neural network process model. in either case the model represents a relationship between one of the mpps, other than the parameter representing the amount of the emitted particular non-particulate pollutant, and one or more other of the mpps. also included is a processor configured with the logic, e.g. programmed software, to estimate a value of the one mpp based on a value of each of the one or more other mpps and the one model.",2010-12-28,"The title of the patent is process parameter estimation in controlling emission of a non-particulate pollutant into the air and its abstract is a parameter value estimator is provided for a process performed primarily to control emission of a particular non-particulate pollutant, such as nox and so2, into the air. the process has multiple process parameters (mpps) including a parameter representing an amount of the particular non-particulate pollutant emitted. the parameter value estimator includes either a neural network process model or a non-neural network process model. in either case the model represents a relationship between one of the mpps, other than the parameter representing the amount of the emitted particular non-particulate pollutant, and one or more other of the mpps. also included is a processor configured with the logic, e.g. programmed software, to estimate a value of the one mpp based on a value of each of the one or more other mpps and the one model. dated 2010-12-28"
7865018,personalized implicit and explicit character shape adaptation and recognition,"handwriting recognition techniques employing a personalized handwriting recognition engine. the recognition techniques use examples of an individual's previous writing style to help recognize new pen input from that individual. the techniques also employ a shape trainer to select samples of an individual's handwriting that accurately represent the individual's writing style, for use as prototypes to recognize subsequent handwriting from the individual. the techniques also alternately or additionally employ an intelligent combiner to combine the recognition results from the personalized recognition engine and the conventional recognition engine (or engines). the combiner may use a comparative neural network to combine the recognition results from multiple recognition engines. the combiner alternately may use a rule-based system based on prior knowledge of different recognition engines.",2011-01-04,"The title of the patent is personalized implicit and explicit character shape adaptation and recognition and its abstract is handwriting recognition techniques employing a personalized handwriting recognition engine. the recognition techniques use examples of an individual's previous writing style to help recognize new pen input from that individual. the techniques also employ a shape trainer to select samples of an individual's handwriting that accurately represent the individual's writing style, for use as prototypes to recognize subsequent handwriting from the individual. the techniques also alternately or additionally employ an intelligent combiner to combine the recognition results from the personalized recognition engine and the conventional recognition engine (or engines). the combiner may use a comparative neural network to combine the recognition results from multiple recognition engines. the combiner alternately may use a rule-based system based on prior knowledge of different recognition engines. dated 2011-01-04"
7865415,intelligent simulation analysis method and system,a method for calculating pricing information for a financial instrument consisting of a plurality of underlying financial instruments that includes the steps of: calculating a default time vector for a plurality of default scenarios wherein each default time vector includes a measure of a likelihood of default for each of the plurality of underlying financial instruments; calculating one or more cash flows for a subset of the default scenarios thereby forming a training set; training a neural network with the training set; and using the neural network to estimate one or more cash flows for a remaining number of the plurality of default scenarios.,2011-01-04,The title of the patent is intelligent simulation analysis method and system and its abstract is a method for calculating pricing information for a financial instrument consisting of a plurality of underlying financial instruments that includes the steps of: calculating a default time vector for a plurality of default scenarios wherein each default time vector includes a measure of a likelihood of default for each of the plurality of underlying financial instruments; calculating one or more cash flows for a subset of the default scenarios thereby forming a training set; training a neural network with the training set; and using the neural network to estimate one or more cash flows for a remaining number of the plurality of default scenarios. dated 2011-01-04
7865453,apparatus and methods for evaluating hyperdocuments using a trained artificial neural network,"an embodiment of a computer implemented method for determining the disposition of a hyperdocument includes retrieving a hyperdocument from an information source, providing information about content of the hyperdocument to a trained artificial neural network (ann), the ann being capable of evaluating the information and providing results reflecting the evaluation and determining the disposition of the hyperdocument based upon results of the ann.",2011-01-04,"The title of the patent is apparatus and methods for evaluating hyperdocuments using a trained artificial neural network and its abstract is an embodiment of a computer implemented method for determining the disposition of a hyperdocument includes retrieving a hyperdocument from an information source, providing information about content of the hyperdocument to a trained artificial neural network (ann), the ann being capable of evaluating the information and providing results reflecting the evaluation and determining the disposition of the hyperdocument based upon results of the ann. dated 2011-01-04"
7870118,search system,"a search engine and system for data, such as internet web pages, including a query analyser for processing a query to assign respective weights to terms of the query and to generate a query vector including the weights, and an index network responsive to the query vector to output at least one index to data in response to the query. the index network is a self-generating neural network built using training examples derived from a feature extractor. the feature extractor is used during both the search and training phase. a clusterer is used to group search results.",2011-01-11,"The title of the patent is search system and its abstract is a search engine and system for data, such as internet web pages, including a query analyser for processing a query to assign respective weights to terms of the query and to generate a query vector including the weights, and an index network responsive to the query vector to output at least one index to data in response to the query. the index network is a self-generating neural network built using training examples derived from a feature extractor. the feature extractor is used during both the search and training phase. a clusterer is used to group search results. dated 2011-01-11"
7873585,apparatus and methods for predicting a semiconductor parameter across an area of a wafer,"apparatus and methods are provided for predicting a plurality of unknown parameter values (e.g. overlay error or critical dimension) using a plurality of known parameter values. in one embodiment, the method involves training a neural network to predict the plurality of parameter values. in other embodiments, the prediction process does not depend on an optical property of a photolithography tool. such predictions may be used to determine wafer lot disposition.",2011-01-18,"The title of the patent is apparatus and methods for predicting a semiconductor parameter across an area of a wafer and its abstract is apparatus and methods are provided for predicting a plurality of unknown parameter values (e.g. overlay error or critical dimension) using a plurality of known parameter values. in one embodiment, the method involves training a neural network to predict the plurality of parameter values. in other embodiments, the prediction process does not depend on an optical property of a photolithography tool. such predictions may be used to determine wafer lot disposition. dated 2011-01-18"
7875536,nanostructures formed of branched nanowhiskers and methods of producing the same,"a method of forming a nanostructure having the form of a tree, comprises a first stage and a second stage. the first stage includes providing one or more catalytic particles on a substrate surface, and growing a first nanowhisker via each catalytic particle. the second stage includes providing, on the periphery of each first nanowhisker, one or more second catalytic particles, and growing, from each second catalytic particle, a second nanowhisker extending transversely from the periphery of the respective first nanowhisker. further stages may be included to grow one or more further nanowhiskers extending from the nanowhisker(s) of the preceding stage. heterostructures may be created within the nanowhiskers. such nanostructures may form the components of a solar cell array or a light emitting flat panel, where the nanowhiskers are formed of a photosensitive material. a neural network may be formed by positioning the first nanowhiskers close together so that adjacent trees contact one another through nanowhiskers grown in a subsequent stage, and heterojunctions within the nanowhiskers create tunnel barriers to current flow.",2011-01-25,"The title of the patent is nanostructures formed of branched nanowhiskers and methods of producing the same and its abstract is a method of forming a nanostructure having the form of a tree, comprises a first stage and a second stage. the first stage includes providing one or more catalytic particles on a substrate surface, and growing a first nanowhisker via each catalytic particle. the second stage includes providing, on the periphery of each first nanowhisker, one or more second catalytic particles, and growing, from each second catalytic particle, a second nanowhisker extending transversely from the periphery of the respective first nanowhisker. further stages may be included to grow one or more further nanowhiskers extending from the nanowhisker(s) of the preceding stage. heterostructures may be created within the nanowhiskers. such nanostructures may form the components of a solar cell array or a light emitting flat panel, where the nanowhiskers are formed of a photosensitive material. a neural network may be formed by positioning the first nanowhiskers close together so that adjacent trees contact one another through nanowhiskers grown in a subsequent stage, and heterojunctions within the nanowhiskers create tunnel barriers to current flow. dated 2011-01-25"
7877342,"neural network for processing arrays of data with existent topology, such as images and application of the network","a neural network for processing arrays of data with pertinent topology includes a n-dimensional array of cells (ki) corresponding to the knots of the neural network, each cell having connections to the directly adjacent cells (kj) forming the neighborhood of a cell (ki), each cell (ki) has inputs for each connection to directly adjacent cells; an output for the connection to one or more of the directly adjacent cells (kj), the connection between the cells being determined by weights (wij), and each cell being characterized by an internal value and being able to carry out signal processing for generating a cell output signal (ui), the output signal (ui) of a cell (ki) is a function of its internal value and of the input signals from the neighboring cells, each cell being associated univocally to a record of a n-dimensional database (pi) with pertinent topology and the value of each data record being the starting value of the corresponding cell. processing is carried out by considering the internal value or the output value (ui) of each cell (ki) after a certain number of iterative processing steps of the neural network as the new obtained value (ui) for the univocally associated data records (pi).",2011-01-25,"The title of the patent is neural network for processing arrays of data with existent topology, such as images and application of the network and its abstract is a neural network for processing arrays of data with pertinent topology includes a n-dimensional array of cells (ki) corresponding to the knots of the neural network, each cell having connections to the directly adjacent cells (kj) forming the neighborhood of a cell (ki), each cell (ki) has inputs for each connection to directly adjacent cells; an output for the connection to one or more of the directly adjacent cells (kj), the connection between the cells being determined by weights (wij), and each cell being characterized by an internal value and being able to carry out signal processing for generating a cell output signal (ui), the output signal (ui) of a cell (ki) is a function of its internal value and of the input signals from the neighboring cells, each cell being associated univocally to a record of a n-dimensional database (pi) with pertinent topology and the value of each data record being the starting value of the corresponding cell. processing is carried out by considering the internal value or the output value (ui) of each cell (ki) after a certain number of iterative processing steps of the neural network as the new obtained value (ui) for the univocally associated data records (pi). dated 2011-01-25"
7881889,method and apparatus for determining energy savings by using a baseline energy use model that incorporates an artificial intelligence algorithm,"a computer-based system, computer-implemented method and computer program product facilitate determining energy cost savings in an energy-consuming facility, such as a commercial building, using an artificial intelligence model, for example a neural network model, that projects or estimates the amount of energy that would have been consumed by the facility but for the implementation of energy efficiency or conservation measures. energy savings are represented by the difference between the estimate of energy that would have been consumed but for the measures and the actual amount of energy consumed by the facility under actual conditions during a time interval after the measures have been implemented.",2011-02-01,"The title of the patent is method and apparatus for determining energy savings by using a baseline energy use model that incorporates an artificial intelligence algorithm and its abstract is a computer-based system, computer-implemented method and computer program product facilitate determining energy cost savings in an energy-consuming facility, such as a commercial building, using an artificial intelligence model, for example a neural network model, that projects or estimates the amount of energy that would have been consumed by the facility but for the implementation of energy efficiency or conservation measures. energy savings are represented by the difference between the estimate of energy that would have been consumed but for the measures and the actual amount of energy consumed by the facility under actual conditions during a time interval after the measures have been implemented. dated 2011-02-01"
7882049,process control system using spatially dependent data for controlling a web-based process,"system and method for controlling a process with spatially dependent conditions for producing a product with spatially dependent properties, e.g., a web/sheet-based process for producing a web/sheet-based product. input data comprising a plurality of input data sets are provided to a neural network (analog or computer-based), each data set comprising values for one or more input parameters, each comprising a respective process condition or product property. the input data preserve spatial relationships of the input data. the neural network generates output data in accordance with the input data, the output data comprising a plurality of output data sets, each comprising values for one or more output parameters, each comprising a predicted process condition or product property. the output data preserve spatial relationships of the output data, which correspond to the spatial relationships of the input data. the output data are useable by a controller or operator to control the process.",2011-02-01,"The title of the patent is process control system using spatially dependent data for controlling a web-based process and its abstract is system and method for controlling a process with spatially dependent conditions for producing a product with spatially dependent properties, e.g., a web/sheet-based process for producing a web/sheet-based product. input data comprising a plurality of input data sets are provided to a neural network (analog or computer-based), each data set comprising values for one or more input parameters, each comprising a respective process condition or product property. the input data preserve spatial relationships of the input data. the neural network generates output data in accordance with the input data, the output data comprising a plurality of output data sets, each comprising values for one or more output parameters, each comprising a predicted process condition or product property. the output data preserve spatial relationships of the output data, which correspond to the spatial relationships of the input data. the output data are useable by a controller or operator to control the process. dated 2011-02-01"
7882052,evolutionary neural network and method of generating an evolutionary neural network,"an evolutionary neural network and a method of generating such a neural network is disclosed. the evolutionary neural network comprises an input set consisting of at least one input neuron, said input neurons being adapted for receiving an input signal form an external system, an output set consisting of at least one output neuron, said output neurons being adapted for producing an output signal for said external system, an internal network composed of a plurality of internal neurons, each internal neuron being adapted for processing a signal received from at least one of said input neurons or other internal neurons and producing a signal for at least one of said output neurons or other internal neurons, and a plurality of synapses constituting connections between said neurons, each of said synapses having a value of strength that can be adjusted by a learning process. each of said neurons is assigned to a neuron class, the parameter values of which are defined by the genotype of the neural network, and each of said synapses are assigned to a respective synapse class, the parameter values of which are also defined by said genotype of the neural network. at reproduction, the genotype of any new neural network is subject to genetic operations. the evolutionary neural network is associated with a neural space, said neural space comprising a plurality of neural layers. each neuron is associated with at least one neural layer and described by a set of topographical parameters with respect to said neural space. at least one of said topographical parameters of at least the internal neurons is encoded in the genotype of the evolutionary neural network in a statistical form.",2011-02-01,"The title of the patent is evolutionary neural network and method of generating an evolutionary neural network and its abstract is an evolutionary neural network and a method of generating such a neural network is disclosed. the evolutionary neural network comprises an input set consisting of at least one input neuron, said input neurons being adapted for receiving an input signal form an external system, an output set consisting of at least one output neuron, said output neurons being adapted for producing an output signal for said external system, an internal network composed of a plurality of internal neurons, each internal neuron being adapted for processing a signal received from at least one of said input neurons or other internal neurons and producing a signal for at least one of said output neurons or other internal neurons, and a plurality of synapses constituting connections between said neurons, each of said synapses having a value of strength that can be adjusted by a learning process. each of said neurons is assigned to a neuron class, the parameter values of which are defined by the genotype of the neural network, and each of said synapses are assigned to a respective synapse class, the parameter values of which are also defined by said genotype of the neural network. at reproduction, the genotype of any new neural network is subject to genetic operations. the evolutionary neural network is associated with a neural space, said neural space comprising a plurality of neural layers. each neuron is associated with at least one neural layer and described by a set of topographical parameters with respect to said neural space. at least one of said topographical parameters of at least the internal neurons is encoded in the genotype of the evolutionary neural network in a statistical form. dated 2011-02-01"
7885198,systems and methods for characterizing packet-switching networks,"a packet-network analyzer system for characterizing network conditions of a packet-network-under-test is provided. in this regard, one such system can be broadly summarized by a representative analyzer system that incorporates a data collection element to receive the raw digital data from a host analyzer, a data selection element to receive the raw digital data, a data processing element to process the selected data set to generate a normalized data set, a neural processing module to process the normalized data set to generate a set of rules and relationships, and a data mining module that uses the rules and relationships to generate a mined data set from the selected data set, the mined data set being used to characterize a packet-network-under test.",2011-02-08,"The title of the patent is systems and methods for characterizing packet-switching networks and its abstract is a packet-network analyzer system for characterizing network conditions of a packet-network-under-test is provided. in this regard, one such system can be broadly summarized by a representative analyzer system that incorporates a data collection element to receive the raw digital data from a host analyzer, a data selection element to receive the raw digital data, a data processing element to process the selected data set to generate a normalized data set, a neural processing module to process the normalized data set to generate a set of rules and relationships, and a data mining module that uses the rules and relationships to generate a mined data set from the selected data set, the mined data set being used to characterize a packet-network-under test. dated 2011-02-08"
7894072,laser-based gas differential spectral analysis,"a laser interferometer, such as a dual path michelson interferometer, is used to generate fringe patterns resulting from one of a plurality of optical paths passing through a sample gas. an artificial neural network, such as, for example, a kaser neural network, is used to recognize patterns in the fringe interference patterns corresponding to known target gases.",2011-02-22,"The title of the patent is laser-based gas differential spectral analysis and its abstract is a laser interferometer, such as a dual path michelson interferometer, is used to generate fringe patterns resulting from one of a plurality of optical paths passing through a sample gas. an artificial neural network, such as, for example, a kaser neural network, is used to recognize patterns in the fringe interference patterns corresponding to known target gases. dated 2011-02-22"
7894902,adaptive cardiac resyncronization therapy and vagal stimulation system,"an adaptive feed-back controlled system for regulating a physiological function of a heart in which a hemodynamic sensor continuously monitors the physiological performance of the heart. three implanted electrodes sense and pace the right atrial, right ventricle and left ventricle. a learning neural network module receives and processes information for the electrodes (18) and sensors (22), and is controlled by a deterministic module for limiting said learning module. a pulse generator (16), is also controlled by the deterministic module, and stimulates both the heart and the vagus (20).",2011-02-22,"The title of the patent is adaptive cardiac resyncronization therapy and vagal stimulation system and its abstract is an adaptive feed-back controlled system for regulating a physiological function of a heart in which a hemodynamic sensor continuously monitors the physiological performance of the heart. three implanted electrodes sense and pace the right atrial, right ventricle and left ventricle. a learning neural network module receives and processes information for the electrodes (18) and sensors (22), and is controlled by a deterministic module for limiting said learning module. a pulse generator (16), is also controlled by the deterministic module, and stimulates both the heart and the vagus (20). dated 2011-02-22"
7895140,"neural network learning device, method, and program","it is possible to acquire existing techniques in a neural network model currently studied and developed so as to generalize them as an element technique, and provide modeling of a basic unit of bottom-up approach using the neural network by adding new values to the existing techniques. a network learning device builds up a network of basic units in a network section, acquires an input from a sensor input section for evaluating it, changes a coupling weight coefficient by using a correlation operation so that the evaluation value satisfies a predetermined evaluation value, and inserts a new neural network according to need.",2011-02-22,"The title of the patent is neural network learning device, method, and program and its abstract is it is possible to acquire existing techniques in a neural network model currently studied and developed so as to generalize them as an element technique, and provide modeling of a basic unit of bottom-up approach using the neural network by adding new values to the existing techniques. a network learning device builds up a network of basic units in a network section, acquires an input from a sensor input section for evaluating it, changes a coupling weight coefficient by using a correlation operation so that the evaluation value satisfies a predetermined evaluation value, and inserts a new neural network according to need. dated 2011-02-22"
7896636,support apparatus of injection-molding machine,"a support apparatus of an injection-molding machine has a neural network that receives test molding data corresponding to molding conditions and a quality value obtained by measuring a non-defective molded article, and that determines a quality prediction function based on the received test molding data. a computer calculates a predicted value of the quality value using the quality prediction function. an input apparatus inputs into the neural network fixed values for the molding conditions except for a selected at least one of the molding conditions, and inputs a target value of the quality value. a graph generator generates a graphical relationship between the selected at least one molding condition and the predicted value. a graph correction unit corrects the graphical relationship generated by the graph generator on the basis of the target value. a display unit selectively displays the graphical relationship generated by the graph generator and the graphical relationship corrected by the graph correction unit.",2011-03-01,"The title of the patent is support apparatus of injection-molding machine and its abstract is a support apparatus of an injection-molding machine has a neural network that receives test molding data corresponding to molding conditions and a quality value obtained by measuring a non-defective molded article, and that determines a quality prediction function based on the received test molding data. a computer calculates a predicted value of the quality value using the quality prediction function. an input apparatus inputs into the neural network fixed values for the molding conditions except for a selected at least one of the molding conditions, and inputs a target value of the quality value. a graph generator generates a graphical relationship between the selected at least one molding condition and the predicted value. a graph correction unit corrects the graphical relationship generated by the graph generator on the basis of the target value. a display unit selectively displays the graphical relationship generated by the graph generator and the graphical relationship corrected by the graph correction unit. dated 2011-03-01"
7896812,system and method for diagnosis and treatment of a breathing pattern of a patient,"described is a system including a sensor and a processing arrangement. the sensor measures data corresponding to a patient's breathing patterns. the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of a rem sleep state. in another embodiment, the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of one of the following states: (i) a wake state and (ii) a sleep state. in another embodiment, a neural network analyzes the data to determine whether the breathing patterns are indicative of one of the following states: (i) a rem sleep state, (ii) a wake state and (iii) a sleep state. in another embodiment, the processing arrangement analyzes the data to determine whether the breathing pattern is indicative of an arousal.",2011-03-01,"The title of the patent is system and method for diagnosis and treatment of a breathing pattern of a patient and its abstract is described is a system including a sensor and a processing arrangement. the sensor measures data corresponding to a patient's breathing patterns. the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of a rem sleep state. in another embodiment, the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of one of the following states: (i) a wake state and (ii) a sleep state. in another embodiment, a neural network analyzes the data to determine whether the breathing patterns are indicative of one of the following states: (i) a rem sleep state, (ii) a wake state and (iii) a sleep state. in another embodiment, the processing arrangement analyzes the data to determine whether the breathing pattern is indicative of an arousal. dated 2011-03-01"
7899562,methods and systems for controlling a semiconductor fabrication process,"software for controlling processes in a heterogeneous semiconductor manufacturing environment may include a wafer-centric database, a real-time scheduler using a neural network, and a graphical user interface displaying simulated operation of the system. these features may be employed alone or in combination to offer improved usability and computational efficiency for real time control and monitoring of a semiconductor manufacturing process. more generally, these techniques may be usefully employed in a variety of real time control systems, particularly systems requiring complex scheduling decisions or heterogeneous systems constructed of hardware from numerous independent vendors.",2011-03-01,"The title of the patent is methods and systems for controlling a semiconductor fabrication process and its abstract is software for controlling processes in a heterogeneous semiconductor manufacturing environment may include a wafer-centric database, a real-time scheduler using a neural network, and a graphical user interface displaying simulated operation of the system. these features may be employed alone or in combination to offer improved usability and computational efficiency for real time control and monitoring of a semiconductor manufacturing process. more generally, these techniques may be usefully employed in a variety of real time control systems, particularly systems requiring complex scheduling decisions or heterogeneous systems constructed of hardware from numerous independent vendors. dated 2011-03-01"
7899765,"method of measuring taste using two phase radial basis function neural networks, a taste sensor, and a taste measuring apparatus","a method for measuring tastes, which can better simulate the human gustation than known methods, as well as a taste sensor, computer program and an apparatus for measuring tastes, is disclosed. in this method, data processing is carried out by a two-phase radial basis function neural network. that is, by sensors, each of which sensors can quantify at least one component representing, individually or cooperatively, the taste of saltiness, sourness, sweetness, umami or bitterness, to obtain a response value from each sensor, and each of the obtained response values is input to a first phase radial basis function neural network to calculate the concentration of each component from each response value. then, the concentration of each component is fed into a second phase radial basis function neural network, which correlates the concentration of each component with the intensities of saltiness, sourness, sweetness, umami and bitterness sensed by humans, to calculate the intensities of saltiness, sourness, sweetness, umami and bitterness sensed by humans.",2011-03-01,"The title of the patent is method of measuring taste using two phase radial basis function neural networks, a taste sensor, and a taste measuring apparatus and its abstract is a method for measuring tastes, which can better simulate the human gustation than known methods, as well as a taste sensor, computer program and an apparatus for measuring tastes, is disclosed. in this method, data processing is carried out by a two-phase radial basis function neural network. that is, by sensors, each of which sensors can quantify at least one component representing, individually or cooperatively, the taste of saltiness, sourness, sweetness, umami or bitterness, to obtain a response value from each sensor, and each of the obtained response values is input to a first phase radial basis function neural network to calculate the concentration of each component from each response value. then, the concentration of each component is fed into a second phase radial basis function neural network, which correlates the concentration of each component with the intensities of saltiness, sourness, sweetness, umami and bitterness sensed by humans, to calculate the intensities of saltiness, sourness, sweetness, umami and bitterness sensed by humans. dated 2011-03-01"
7904195,method for prognostic maintenance in semiconductor manufacturing equipments,"a method for prognostic maintenance in semiconductor manufacturing equipments is disclosed. the said method comprising: collecting a plurality of raw data from the default detection and classification system for equipments, preprocessing the raw data, using the neural network model (nn model) to find a plurality of health indices, generating health information by using the principal component analysis (pca) to identify the health indices, and using the partial least square discriminated analysis (pls-da) to find a health report. the health report provides the engineers with current risk levels of equipments. by the health report, the engineers can initiate prognostic maintenance and repair the equipments early.",2011-03-08,"The title of the patent is method for prognostic maintenance in semiconductor manufacturing equipments and its abstract is a method for prognostic maintenance in semiconductor manufacturing equipments is disclosed. the said method comprising: collecting a plurality of raw data from the default detection and classification system for equipments, preprocessing the raw data, using the neural network model (nn model) to find a plurality of health indices, generating health information by using the principal component analysis (pca) to identify the health indices, and using the partial least square discriminated analysis (pls-da) to find a health report. the health report provides the engineers with current risk levels of equipments. by the health report, the engineers can initiate prognostic maintenance and repair the equipments early. dated 2011-03-08"
7910873,biochip microsystem for bioinformatics recognition and analysis,"a system with applications in pattern recognition, or classification, of dna assay samples. because dna reference and sample material in wells of an assay may be caused to fluoresce depending upon dye added to the material, the resulting light may be imaged onto an embodiment comprising an array of photodetectors and an adaptive neural network, with applications to dna analysis. other embodiments are described and claimed.",2011-03-22,"The title of the patent is biochip microsystem for bioinformatics recognition and analysis and its abstract is a system with applications in pattern recognition, or classification, of dna assay samples. because dna reference and sample material in wells of an assay may be caused to fluoresce depending upon dye added to the material, the resulting light may be imaged onto an embodiment comprising an array of photodetectors and an adaptive neural network, with applications to dna analysis. other embodiments are described and claimed. dated 2011-03-22"
7912796,system and method for real-time recognition of driving patterns,"system and method for real-time, automatic, recognition of large time-scale driving patterns employs a statistical pattern recognition framework, implemented by means of feed-forward neural network utilizing models developed for recognizing, for example, four classes of driving environments, namely highway, main road, suburban traffic and city traffic, from vehicle performance data. a vehicle control application effects changes in vehicle performance aspects based on the recognized driving environment.",2011-03-22,"The title of the patent is system and method for real-time recognition of driving patterns and its abstract is system and method for real-time, automatic, recognition of large time-scale driving patterns employs a statistical pattern recognition framework, implemented by means of feed-forward neural network utilizing models developed for recognizing, for example, four classes of driving environments, namely highway, main road, suburban traffic and city traffic, from vehicle performance data. a vehicle control application effects changes in vehicle performance aspects based on the recognized driving environment. dated 2011-03-22"
7917335,method and system of monitoring prognostics,a neural network learns the operating modes of a system being monitored under normal operating conditions. anomalies can be automatically detected and learned. a control command can be issued or an alert can be issued in response thereto.,2011-03-29,The title of the patent is method and system of monitoring prognostics and its abstract is a neural network learns the operating modes of a system being monitored under normal operating conditions. anomalies can be automatically detected and learned. a control command can be issued or an alert can be issued in response thereto. dated 2011-03-29
7930259,apparatus for detecting vibrations of a test object using a competitive learning neural network in determining frequency characteristics generated,"a nondestructive inspection apparatus includes a sensor unit for detecting vibrations transmitted through a test object from a vibration generator and a signal input unit for extracting a target signal from an electric signal outputted from the sensor unit. an amount of characteristics extracting unit is also included for extracting multiple frequency components from the test signal as an amount of characteristics. further, a decision unit has a competitive learning neural network for determining whether the amount of the characteristics belongs to a category, wherein the competitive learning neural network has been trained by using training samples belong to the category representing an internal state of the test object, wherein distributions of membership degrees of the training samples are set in the decision unit.",2011-04-19,"The title of the patent is apparatus for detecting vibrations of a test object using a competitive learning neural network in determining frequency characteristics generated and its abstract is a nondestructive inspection apparatus includes a sensor unit for detecting vibrations transmitted through a test object from a vibration generator and a signal input unit for extracting a target signal from an electric signal outputted from the sensor unit. an amount of characteristics extracting unit is also included for extracting multiple frequency components from the test signal as an amount of characteristics. further, a decision unit has a competitive learning neural network for determining whether the amount of the characteristics belongs to a category, wherein the competitive learning neural network has been trained by using training samples belong to the category representing an internal state of the test object, wherein distributions of membership degrees of the training samples are set in the decision unit. dated 2011-04-19"
7933679,method for analyzing and optimizing a machining process,"a method for optimizing machining parameters for a cutting process performed on a work piece. finite element analysis of cutting tool and work material interaction is initially performed. mechanistic modeling of the cutting process, using results of the finite element analysis, is then performed to provide optimized machining parameters for improved rate of material removal and tool life. optionally, a two-stage artificial neural network may be supplementally employed, wherein a first stage of the network provides output parameters including peak tool temperature and cutting forces in x and y directions, for a combination of input reference parameters including tool rake angle, material cutting speed, and feed rate.",2011-04-26,"The title of the patent is method for analyzing and optimizing a machining process and its abstract is a method for optimizing machining parameters for a cutting process performed on a work piece. finite element analysis of cutting tool and work material interaction is initially performed. mechanistic modeling of the cutting process, using results of the finite element analysis, is then performed to provide optimized machining parameters for improved rate of material removal and tool life. optionally, a two-stage artificial neural network may be supplementally employed, wherein a first stage of the network provides output parameters including peak tool temperature and cutting forces in x and y directions, for a combination of input reference parameters including tool rake angle, material cutting speed, and feed rate. dated 2011-04-26"
7936916,system and method for video quality measurement based on packet metric and image metric,"a video quality measurement (vqm) system for a video stream includes a neural network vqm module having a constructed architecture to calculate a video quality metric based on a hybrid of image metric and packet metric of the video stream. an image metric measuring module receives the video stream and calculates the image metric of the video stream. a packet metric measuring module obtains information about packet-level characteristics of the video stream to calculate the packet metric. the image metric and the packet metric are inputted to the neural network vqm module to calculate the video quality metric. the vqm system further includes a vqm test-bed that determines and validates the architecture of the neural network vqm module. furthermore, a video quality measurement (vqm) method based on a hybrid of image metric and packet metric is also described.",2011-05-03,"The title of the patent is system and method for video quality measurement based on packet metric and image metric and its abstract is a video quality measurement (vqm) system for a video stream includes a neural network vqm module having a constructed architecture to calculate a video quality metric based on a hybrid of image metric and packet metric of the video stream. an image metric measuring module receives the video stream and calculates the image metric of the video stream. a packet metric measuring module obtains information about packet-level characteristics of the video stream to calculate the packet metric. the image metric and the packet metric are inputted to the neural network vqm module to calculate the video quality metric. the vqm system further includes a vqm test-bed that determines and validates the architecture of the neural network vqm module. furthermore, a video quality measurement (vqm) method based on a hybrid of image metric and packet metric is also described. dated 2011-05-03"
7937197,apparatus and methods for evaluating a dynamic system,"a method of evaluating whether a vehicle under test is operating as intended. parameters of the vehicle are sampled at a plurality of sample times to obtain a plurality of data samples. data samples from more than one of the sample times are included in a sample set. the sample set is input to an artificial neural network (ann). many time-varying parameters, e.g., response times in motor vehicle systems, can be detected and evaluated.",2011-05-03,"The title of the patent is apparatus and methods for evaluating a dynamic system and its abstract is a method of evaluating whether a vehicle under test is operating as intended. parameters of the vehicle are sampled at a plurality of sample times to obtain a plurality of data samples. data samples from more than one of the sample times are included in a sample set. the sample set is input to an artificial neural network (ann). many time-varying parameters, e.g., response times in motor vehicle systems, can be detected and evaluated. dated 2011-05-03"
7937343,method and apparatus for randomized verification of neural nets,described are techniques for using statistical analysis to reduce the number of samples required in accordance with statistical analysis confidence intervals to verify correctness of a component. these techniques may be used in verification of a neural network or other hardware or software component.,2011-05-03,The title of the patent is method and apparatus for randomized verification of neural nets and its abstract is described are techniques for using statistical analysis to reduce the number of samples required in accordance with statistical analysis confidence intervals to verify correctness of a component. these techniques may be used in verification of a neural network or other hardware or software component. dated 2011-05-03
7945348,methods and systems for controlling a semiconductor fabrication process,"software for controlling processes in a heterogeneous semiconductor manufacturing environment may include a wafer-centric database, a real-time scheduler using a neural network, and a graphical user interface displaying simulated operation of the system. these features may be employed alone or in combination to offer improved usability and computational efficiency for real time control and monitoring of a semiconductor manufacturing process. more generally, these techniques may be usefully employed in a variety of real time control systems, particularly systems requiring complex scheduling decisions or heterogeneous systems constructed of hardware from numerous independent vendors.",2011-05-17,"The title of the patent is methods and systems for controlling a semiconductor fabrication process and its abstract is software for controlling processes in a heterogeneous semiconductor manufacturing environment may include a wafer-centric database, a real-time scheduler using a neural network, and a graphical user interface displaying simulated operation of the system. these features may be employed alone or in combination to offer improved usability and computational efficiency for real time control and monitoring of a semiconductor manufacturing process. more generally, these techniques may be usefully employed in a variety of real time control systems, particularly systems requiring complex scheduling decisions or heterogeneous systems constructed of hardware from numerous independent vendors. dated 2011-05-17"
7947626,passaged neural stem cell-derived neuronal networks as sensing elements for detection of environmental threats,"this invention comprises a method for generating functional neural networks using neural progenitor cells on microelectrode arrays (meas). the method involves dissociating neural progenitor cells from an embryo, propagating the neural progenitor cells, passaging the neural progenitor cells and seeding the neural progenitor cells on meas to produce a functional neural network. the neural progenitor cells may be continuously passaged to propagate an endless supply of neural progenitor cells. the resultant passaged progenitor cell derived neural network mea may be used to detect and/or quantify various biological or chemical toxins.",2011-05-24,"The title of the patent is passaged neural stem cell-derived neuronal networks as sensing elements for detection of environmental threats and its abstract is this invention comprises a method for generating functional neural networks using neural progenitor cells on microelectrode arrays (meas). the method involves dissociating neural progenitor cells from an embryo, propagating the neural progenitor cells, passaging the neural progenitor cells and seeding the neural progenitor cells on meas to produce a functional neural network. the neural progenitor cells may be continuously passaged to propagate an endless supply of neural progenitor cells. the resultant passaged progenitor cell derived neural network mea may be used to detect and/or quantify various biological or chemical toxins. dated 2011-05-24"
7953279,combining online and offline recognizers in a handwriting recognition system,"described is a technology by which online recognition of handwritten input data is combined with offline recognition and processing to obtain a combined recognition result. in general, the combination improves overall recognition accuracy. in one aspect, online and offline recognition is separately performed to obtain online and offline character-level recognition scores for candidates (hypotheses). a statistical analysis-based combination algorithm, an adaboost algorithm, and/or a neural network-based combination may determine a combination function to combine the scores to produce a result set of one or more results. online and offline radical-level recognition may be performed. for example, a hmm recognizer may generate online radical scores used to build a radical graph, which is then rescored using the offline radical recognition scores. paths in the rescored graph are then searched to provide the combined recognition result, e.g., corresponding to the path with the highest score.",2011-05-31,"The title of the patent is combining online and offline recognizers in a handwriting recognition system and its abstract is described is a technology by which online recognition of handwritten input data is combined with offline recognition and processing to obtain a combined recognition result. in general, the combination improves overall recognition accuracy. in one aspect, online and offline recognition is separately performed to obtain online and offline character-level recognition scores for candidates (hypotheses). a statistical analysis-based combination algorithm, an adaboost algorithm, and/or a neural network-based combination may determine a combination function to combine the scores to produce a result set of one or more results. online and offline radical-level recognition may be performed. for example, a hmm recognizer may generate online radical scores used to build a radical graph, which is then rescored using the offline radical recognition scores. paths in the rescored graph are then searched to provide the combined recognition result, e.g., corresponding to the path with the highest score. dated 2011-05-31"
7953681,system and method of forecasting print job related demand,"systems and methods for forecasting print demand are disclosed. print demand data is collected and stored for each print job processed during a selected time interval, and processed with a computer implemented service manager to obtain a first demand series with multiple demand components and a second demand series with one demand component. each of the multiple demand components is less than, and the one demand component is greater than, a selected variability level. the service manager is adapted to (1) generate a first demand related forecast with a combination of the multiple demand components, and (2) use a neural network to generate a second demand related forecast with the one demand component. the neural network includes multiple neurons optimally weighted with respect to print-related demand data collected over selected time intervals. the number of neurons is optimized to improve forecasting accuracy and re-optimized after a selected time interval.",2011-05-31,"The title of the patent is system and method of forecasting print job related demand and its abstract is systems and methods for forecasting print demand are disclosed. print demand data is collected and stored for each print job processed during a selected time interval, and processed with a computer implemented service manager to obtain a first demand series with multiple demand components and a second demand series with one demand component. each of the multiple demand components is less than, and the one demand component is greater than, a selected variability level. the service manager is adapted to (1) generate a first demand related forecast with a combination of the multiple demand components, and (2) use a neural network to generate a second demand related forecast with the one demand component. the neural network includes multiple neurons optimally weighted with respect to print-related demand data collected over selected time intervals. the number of neurons is optimized to improve forecasting accuracy and re-optimized after a selected time interval. dated 2011-05-31"
7954579,adaptive control strategy and method for optimizing hybrid electric vehicles,"this invention relates a control strategy for a hybrid electric vehicle having an electric motor, a battery and an internal combustion engine. the control strategy improves fuel economy and reduces emissions while providing sufficient acceleration over a varying set of driving conditions through an adaptive control unit with an artificial neural network. the artificial neural network is trained on a pre-processed training set based on the highest fuel economies of multiple control strategies and multiple driving profiles. training the artificial neural network includes a training algorithm and a learning algorithm. the invention also includes a method of operating a hybrid electric vehicle with an adaptive control strategy using an artificial neural network.",2011-06-07,"The title of the patent is adaptive control strategy and method for optimizing hybrid electric vehicles and its abstract is this invention relates a control strategy for a hybrid electric vehicle having an electric motor, a battery and an internal combustion engine. the control strategy improves fuel economy and reduces emissions while providing sufficient acceleration over a varying set of driving conditions through an adaptive control unit with an artificial neural network. the artificial neural network is trained on a pre-processed training set based on the highest fuel economies of multiple control strategies and multiple driving profiles. training the artificial neural network includes a training algorithm and a learning algorithm. the invention also includes a method of operating a hybrid electric vehicle with an adaptive control strategy using an artificial neural network. dated 2011-06-07"
7957875,method and apparatus for predicting braking system friction,"a brake system control method determines vehicle operating conditions, compares the conditions to an allowable range, and uses a neural network to predict an expected coefficient of friction when the conditions are within the range. when the conditions fall outside of the range, the method determines an amount of required braking force using a constant coefficient of friction, and calculates the required braking force using the expected coefficient of friction when the conditions are within the range. the vehicle operating conditions include a vehicle speed, brake pressure, modeled brake rotor temperature, and apply state. the expected coefficient is multiplied by a constant or a calculated correction factor. a vehicle has an engine, transmission, and braking system, with a controller and an algorithm for predicting a coefficient of friction for two brake rotors, calculating a hydraulic brake pressure, and for applying the braking system using the hydraulic brake pressure.",2011-06-07,"The title of the patent is method and apparatus for predicting braking system friction and its abstract is a brake system control method determines vehicle operating conditions, compares the conditions to an allowable range, and uses a neural network to predict an expected coefficient of friction when the conditions are within the range. when the conditions fall outside of the range, the method determines an amount of required braking force using a constant coefficient of friction, and calculates the required braking force using the expected coefficient of friction when the conditions are within the range. the vehicle operating conditions include a vehicle speed, brake pressure, modeled brake rotor temperature, and apply state. the expected coefficient is multiplied by a constant or a calculated correction factor. a vehicle has an engine, transmission, and braking system, with a controller and an algorithm for predicting a coefficient of friction for two brake rotors, calculating a hydraulic brake pressure, and for applying the braking system using the hydraulic brake pressure. dated 2011-06-07"
7962429,neuromorphic device for proofreading connection adjustments in hardware artificial neural networks,"a hardware-implemented method for proofreading updates of connections in a hardware artificial neural network (hann) includes computing a draft weight change independently at a connection between neuroids and at a corresponding dedicated special purpose nousoid, determining whether the draft weight changes agree, and executing a weight change at the connection equal to the draft weight change upon determining that the draft weight changes agree.",2011-06-14,"The title of the patent is neuromorphic device for proofreading connection adjustments in hardware artificial neural networks and its abstract is a hardware-implemented method for proofreading updates of connections in a hardware artificial neural network (hann) includes computing a draft weight change independently at a connection between neuroids and at a corresponding dedicated special purpose nousoid, determining whether the draft weight changes agree, and executing a weight change at the connection equal to the draft weight change upon determining that the draft weight changes agree. dated 2011-06-14"
7966177,method and device for recognising a phonetic sound sequence or character sequence,"the invention relates to a method for recognizing a phonetic sound sequence or a character sequence, e.g. according to the ascii standards, comprising the following steps: a) the sequence is fed to a neural network, b) in said neural network, a sequence of characteristics is formed from the phonetic sequence or character sequence, by taking into consideration phonetic and/or lexical stored information and/or based on a character string sequence (blank characters), c) the character sequence thus formed is compared with a characteristic combination that has a defined statement content, said combination being formed from stored lexical and semantic information, based on the characteristic sequence, d) step c is repeated using new character combinations until, by the reduction of contradictions, a character combination is found that at least largely corresponds with the character sequence, e) the statement content of the character combination with the least number of contradictions is output as the result and/or an action assigned to the statement content is carried out.",2011-06-21,"The title of the patent is method and device for recognising a phonetic sound sequence or character sequence and its abstract is the invention relates to a method for recognizing a phonetic sound sequence or a character sequence, e.g. according to the ascii standards, comprising the following steps: a) the sequence is fed to a neural network, b) in said neural network, a sequence of characteristics is formed from the phonetic sequence or character sequence, by taking into consideration phonetic and/or lexical stored information and/or based on a character string sequence (blank characters), c) the character sequence thus formed is compared with a characteristic combination that has a defined statement content, said combination being formed from stored lexical and semantic information, based on the characteristic sequence, d) step c is repeated using new character combinations until, by the reduction of contradictions, a character combination is found that at least largely corresponds with the character sequence, e) the statement content of the character combination with the least number of contradictions is output as the result and/or an action assigned to the statement content is carried out. dated 2011-06-21"
7966273,predicting formation fluid property through downhole fluid analysis using artificial neural network,"apparatus and methods to perform downhole fluid analysis using an artificial neural network are disclosed. a disclosed example method involves obtaining a first formation fluid property value of a formation fluid sample from a downhole fluid analysis process. the first formation fluid property value is provided to an artificial neural network, and a second formation fluid property value of the formation fluid sample is generated by means of the artificial neural network.",2011-06-21,"The title of the patent is predicting formation fluid property through downhole fluid analysis using artificial neural network and its abstract is apparatus and methods to perform downhole fluid analysis using an artificial neural network are disclosed. a disclosed example method involves obtaining a first formation fluid property value of a formation fluid sample from a downhole fluid analysis process. the first formation fluid property value is provided to an artificial neural network, and a second formation fluid property value of the formation fluid sample is generated by means of the artificial neural network. dated 2011-06-21"
7970212,method for automatic detection and classification of objects and patterns in low resolution environments,"the invention is a method of using wavelet transformation and artificial neural network (ann) systems for automatic detecting and classifying objects. to train the system in object recognition different images, which usually contain desired objects alongside other objects are used. these objects may appear at different angles. different characteristics regarding the objects are extracted from the images and stored in a data bank. the system then determines the extent to which each inserted characteristic will be useful in future recognition and determines its relative weight. after the initial insertion of data, the operator tests the system with a set of new images, some of which contain the class objects and some of which contain similar and/or dissimilar objects of different classification. the system learns from the images containing similar objects of different classes as well as from the images containing the class objects, since each specific class characteristic needs to be set apart from other class characteristic. the system may be tested and trained again and again until the operator is satisfied with the system's success rate of object recognition and classification.",2011-06-28,"The title of the patent is method for automatic detection and classification of objects and patterns in low resolution environments and its abstract is the invention is a method of using wavelet transformation and artificial neural network (ann) systems for automatic detecting and classifying objects. to train the system in object recognition different images, which usually contain desired objects alongside other objects are used. these objects may appear at different angles. different characteristics regarding the objects are extracted from the images and stored in a data bank. the system then determines the extent to which each inserted characteristic will be useful in future recognition and determines its relative weight. after the initial insertion of data, the operator tests the system with a set of new images, some of which contain the class objects and some of which contain similar and/or dissimilar objects of different classification. the system learns from the images containing similar objects of different classes as well as from the images containing the class objects, since each specific class characteristic needs to be set apart from other class characteristic. the system may be tested and trained again and again until the operator is satisfied with the system's success rate of object recognition and classification. dated 2011-06-28"
7970764,gui for subject matter navigation using maps and search terms,"a system, method and computer program product for navigating categorized information, including (a) a two-dimensional map displayed to a user on a screen, the map showing search terms relating to a subject matter, where the display of the search terms corresponds to relationship between the terms, and wherein a manner of display of the terms corresponds to their relative importance to the subject matter; and (b) a neural network underlying the map, wherein the manner of display and a selection of the search terms is derived from the neural network. the manner of display includes font color, font size, font transparency, distance between search terms and positioning of the search terms within the map. positioning of a cursor over one of the search terms rearranges the search terms on the map to correspond to an increased relevance of the one of the search terms, while the cursor is over the one of the search terms. clicking on the one of the search terms corresponds to navigating into a sub-subject matter of the one of the search terms.",2011-06-28,"The title of the patent is gui for subject matter navigation using maps and search terms and its abstract is a system, method and computer program product for navigating categorized information, including (a) a two-dimensional map displayed to a user on a screen, the map showing search terms relating to a subject matter, where the display of the search terms corresponds to relationship between the terms, and wherein a manner of display of the terms corresponds to their relative importance to the subject matter; and (b) a neural network underlying the map, wherein the manner of display and a selection of the search terms is derived from the neural network. the manner of display includes font color, font size, font transparency, distance between search terms and positioning of the search terms within the map. positioning of a cursor over one of the search terms rearranges the search terms on the map to correspond to an increased relevance of the one of the search terms, while the cursor is over the one of the search terms. clicking on the one of the search terms corresponds to navigating into a sub-subject matter of the one of the search terms. dated 2011-06-28"
7970896,system and article of manufacturing for filtering content using neural networks,provided are a system and article of manufacture for filtering communications received from over a network for a person-to-person communication program. a communication is received for the person-to person communication program. the communication is processed to determine predefined language statements. information on the determined language statements is inputted into a neural network to produce an output value. a determination is made as to whether the output value indicates that the communication is unacceptable. the communication is forwarded to the person-to-person communication program unchanged if the output value indicates that the communication is acceptable. an action is performed with respect to the communication upon determining that the communication is unacceptable that differs from the forwarding of the communication that occurs if the output value indicates that the communication is acceptable.,2011-06-28,The title of the patent is system and article of manufacturing for filtering content using neural networks and its abstract is provided are a system and article of manufacture for filtering communications received from over a network for a person-to-person communication program. a communication is received for the person-to person communication program. the communication is processed to determine predefined language statements. information on the determined language statements is inputted into a neural network to produce an output value. a determination is made as to whether the output value indicates that the communication is unacceptable. the communication is forwarded to the person-to-person communication program unchanged if the output value indicates that the communication is acceptable. an action is performed with respect to the communication upon determining that the communication is unacceptable that differs from the forwarding of the communication that occurs if the output value indicates that the communication is acceptable. dated 2011-06-28
7971450,deep-freezer with neural network,"deep-freezing apparatus for foodstuffs comprising a deep-freezing compartment, signal display means, sensor means adapted to detect the temperature inside the foodstuffs stored in said deep-freezing compartment, processing means for the signals generated by said sensor means, wherein the apparatus further comprises a neural network adapted to receive the signals issued by a temperature sensor situated inside the foodstuff being deep-frozen and by an information on the time elapsed from the beginning of the deep-freezing process, and further adapted to provide a signal representative of the residual time needed to reach a pre-set (deep-freezing) temperature, as well as processing means adapted to receive the signal output by said neural network and provide in response an information representative of the predicted time needed for a pre-determined temperature to be reached on said first temperature sensor.",2011-07-05,"The title of the patent is deep-freezer with neural network and its abstract is deep-freezing apparatus for foodstuffs comprising a deep-freezing compartment, signal display means, sensor means adapted to detect the temperature inside the foodstuffs stored in said deep-freezing compartment, processing means for the signals generated by said sensor means, wherein the apparatus further comprises a neural network adapted to receive the signals issued by a temperature sensor situated inside the foodstuff being deep-frozen and by an information on the time elapsed from the beginning of the deep-freezing process, and further adapted to provide a signal representative of the residual time needed to reach a pre-set (deep-freezing) temperature, as well as processing means adapted to receive the signal output by said neural network and provide in response an information representative of the predicted time needed for a pre-determined temperature to be reached on said first temperature sensor. dated 2011-07-05"
7979370,neural network for electronic search applications,a system for information searching includes a first layer and a second layer. the first layer includes a first plurality of neurons each associated with a word and with a first set of dynamic connections to at least some of the first plurality of neurons. the second layer include a second plurality of neurons each associated with a document and with a second set of dynamic connections to at least some of the first plurality of neurons. the first set of dynamic connections and the second set of dynamic connections can be configured such that a query of at least one neuron of the first plurality of neurons excites at least one neuron of the second plurality of neurons. the excited at least one neuron of the second plurality of neurons can be contextually related to the queried at least one neuron of the first plurality of neurons.,2011-07-12,The title of the patent is neural network for electronic search applications and its abstract is a system for information searching includes a first layer and a second layer. the first layer includes a first plurality of neurons each associated with a word and with a first set of dynamic connections to at least some of the first plurality of neurons. the second layer include a second plurality of neurons each associated with a document and with a second set of dynamic connections to at least some of the first plurality of neurons. the first set of dynamic connections and the second set of dynamic connections can be configured such that a query of at least one neuron of the first plurality of neurons excites at least one neuron of the second plurality of neurons. the excited at least one neuron of the second plurality of neurons can be contextually related to the queried at least one neuron of the first plurality of neurons. dated 2011-07-12
7983744,neural network based learning engine to adapt therapies,a system for implementing a cardiac device having adaptive treatment therapies utilizing a neural network based learning engine is disclosed. the system includes an implantable cardiac device module and an external data processing system for specifying the operating characteristics of the cardiac device module. both the cardiac device module and the external processing system possess an artificial neural network to specify the operation of the cardiac device module as it provides adaptive treatment therapies. the external data processing system includes a complete neural network module that trains and validates the operation of the neural network to match the optimal treatment options with a received set of collected patient data. a runtime neural network module that provides real time operation of the neural network using collected patient data is located within the cardiac device module. the cardiac device module and the external processing module are connected via a communication link.,2011-07-19,The title of the patent is neural network based learning engine to adapt therapies and its abstract is a system for implementing a cardiac device having adaptive treatment therapies utilizing a neural network based learning engine is disclosed. the system includes an implantable cardiac device module and an external data processing system for specifying the operating characteristics of the cardiac device module. both the cardiac device module and the external processing system possess an artificial neural network to specify the operation of the cardiac device module as it provides adaptive treatment therapies. the external data processing system includes a complete neural network module that trains and validates the operation of the neural network to match the optimal treatment options with a received set of collected patient data. a runtime neural network module that provides real time operation of the neural network using collected patient data is located within the cardiac device module. the cardiac device module and the external processing module are connected via a communication link. dated 2011-07-19
7984001,neural network-based extension of global position timing,"a wireless communication system (20) includes a base station controller (22) that receives timing information from a data set (26) that is generated by a neural network (28). the data set (26) allows for generating timing information based upon previous time information received from a gps (24) and in one example, is capable of covering a time interval of up to two weeks during which effective communication with the gps may be interrupted. in one example, the data set is continuously updated so that the base station controller (24) continuously has up to two weeks of future time information available.",2011-07-19,"The title of the patent is neural network-based extension of global position timing and its abstract is a wireless communication system (20) includes a base station controller (22) that receives timing information from a data set (26) that is generated by a neural network (28). the data set (26) allows for generating timing information based upon previous time information received from a gps (24) and in one example, is capable of covering a time interval of up to two weeks during which effective communication with the gps may be interrupted. in one example, the data set is continuously updated so that the base station controller (24) continuously has up to two weeks of future time information available. dated 2011-07-19"
7991223,method for training of supervised prototype neural gas networks and their use in mass spectrometry,"a neural gas network used for pattern recognition, sequence and image processing is extended to a supervised classifier with labeled prototypes by extending a cost function of the neural gas network with additive terms, each of which increases with a difference between elements of the class labels of a prototype and a training data point and decreases with their distance. the extended cost function is then iteratively minimized by adapting weight vectors of the prototypes. the trained network can then be used to classify mass spectrometric data, especially mass spectrometric data derived from biological samples.",2011-08-02,"The title of the patent is method for training of supervised prototype neural gas networks and their use in mass spectrometry and its abstract is a neural gas network used for pattern recognition, sequence and image processing is extended to a supervised classifier with labeled prototypes by extending a cost function of the neural gas network with additive terms, each of which increases with a difference between elements of the class labels of a prototype and a training data point and decreases with their distance. the extended cost function is then iteratively minimized by adapting weight vectors of the prototypes. the trained network can then be used to classify mass spectrometric data, especially mass spectrometric data derived from biological samples. dated 2011-08-02"
7991714,cognitive memory and auto-associative neural network based search engine for computer and network located images and photographs,"designs for cognitive memory systems storing input data, images, or patterns, and retrieving it without knowledge of where stored when cognitive memory is prompted by query pattern that is related to sought stored pattern. retrieval system of cognitive memory uses autoassociative neural networks and techniques for pre-processing query pattern to establish relationship between query pattern and sought stored pattern, to locate sought pattern, and to retrieve it and ancillary data. cognitive memory, when connected to computer or information appliance introduces computational architecture that applies to systems and methods for navigation, location and recognition of objects in images, character recognition, facial recognition, medical analysis and diagnosis, video image analysis, and to photographic search engines that when prompted with a query photograph containing faces and objects will retrieve related photographs stored in computer or other information appliance, and will identify url's of related photographs and documents stored on the world wide web.",2011-08-02,"The title of the patent is cognitive memory and auto-associative neural network based search engine for computer and network located images and photographs and its abstract is designs for cognitive memory systems storing input data, images, or patterns, and retrieving it without knowledge of where stored when cognitive memory is prompted by query pattern that is related to sought stored pattern. retrieval system of cognitive memory uses autoassociative neural networks and techniques for pre-processing query pattern to establish relationship between query pattern and sought stored pattern, to locate sought pattern, and to retrieve it and ancillary data. cognitive memory, when connected to computer or information appliance introduces computational architecture that applies to systems and methods for navigation, location and recognition of objects in images, character recognition, facial recognition, medical analysis and diagnosis, video image analysis, and to photographic search engines that when prompted with a query photograph containing faces and objects will retrieve related photographs stored in computer or other information appliance, and will identify url's of related photographs and documents stored on the world wide web. dated 2011-08-02"
7991716,comprehensive identity protection system,"a system and method for protecting identity fraud are disclosed. a system includes a detection subsystem to identify applications and/or accounts at risk of identity fraud, and a disposition subsystem to process data provided by the detection system and to determine whether identity fraud exists in the applications and/or accounts. according to an implementation, one or more neural network models are defined, each neural network model being configured to handle a class of cases related to the subject and a specific data configuration describing a case of the class. the one or more neural network models are run to generate data requests about the subject's identity, and the data requests are passed to a detection system that monitor transactions associated with the subject. additional data associated with the transactions is requested until a threshold certainty is achieved or until available data or models are exhausted.",2011-08-02,"The title of the patent is comprehensive identity protection system and its abstract is a system and method for protecting identity fraud are disclosed. a system includes a detection subsystem to identify applications and/or accounts at risk of identity fraud, and a disposition subsystem to process data provided by the detection system and to determine whether identity fraud exists in the applications and/or accounts. according to an implementation, one or more neural network models are defined, each neural network model being configured to handle a class of cases related to the subject and a specific data configuration describing a case of the class. the one or more neural network models are run to generate data requests about the subject's identity, and the data requests are passed to a detection system that monitor transactions associated with the subject. additional data associated with the transactions is requested until a threshold certainty is achieved or until available data or models are exhausted. dated 2011-08-02"
7991719,"information processing method and apparatus, and image pickup device","an output value of neuron within an objective layer of a hierarchical neural network is computed. the data of the output value of neuron is stored in a memory only if the output value of neuron is greater than or equal to a predetermined value by referring to the computed output value of neuron within the objective layer. when the data of the output value of neuron on a former layer of objective layer is read from the memory, the data having a predetermined value is read, instead of the data of the output value of neuron not stored in the memory.",2011-08-02,"The title of the patent is information processing method and apparatus, and image pickup device and its abstract is an output value of neuron within an objective layer of a hierarchical neural network is computed. the data of the output value of neuron is stored in a memory only if the output value of neuron is greater than or equal to a predetermined value by referring to the computed output value of neuron within the objective layer. when the data of the output value of neuron on a former layer of objective layer is read from the memory, the data having a predetermined value is read, instead of the data of the output value of neuron not stored in the memory. dated 2011-08-02"
7996110,surgical robot and robotic controller,"the present invention was developed by a neurosurgeon and seeks to mimic the results of primate neurological research which is indicative of a human's actual neurological control structures and logic. specifically, the motor proprioceptive and tactile neurophysiology functioning of the surgeon's hands and internal hand control system from the muscular level through the intrafusal fiber system of the neural network is considered in creating the robot and method of operation of the present invention. therefore, the surgery is not slowed down as in the art, because the surgeon is in conscious and subconscious natural agreement and harmonization with the robotically actuated surgical instruments based on neurological mimicking of the surgeon's behavior with the functioning of the robot. therefore, the robot can enhance the surgeon's humanly limited senses while not introducing disruptive variables to the surgeon's naturally occurring operation of his neurophysiology. this is therefore also a new field, neurophysiological symbiotic robotics.",2011-08-09,"The title of the patent is surgical robot and robotic controller and its abstract is the present invention was developed by a neurosurgeon and seeks to mimic the results of primate neurological research which is indicative of a human's actual neurological control structures and logic. specifically, the motor proprioceptive and tactile neurophysiology functioning of the surgeon's hands and internal hand control system from the muscular level through the intrafusal fiber system of the neural network is considered in creating the robot and method of operation of the present invention. therefore, the surgery is not slowed down as in the art, because the surgeon is in conscious and subconscious natural agreement and harmonization with the robotically actuated surgical instruments based on neurological mimicking of the surgeon's behavior with the functioning of the robot. therefore, the robot can enhance the surgeon's humanly limited senses while not introducing disruptive variables to the surgeon's naturally occurring operation of his neurophysiology. this is therefore also a new field, neurophysiological symbiotic robotics. dated 2011-08-09"
8005908,system for detecting information leakage in outbound e-mails without using the content of the mail,"a system for detecting information leakage in e-mails using neural network and support vector machines is provided. this system does not use the content of the e-mail or the content of the attachments in the e-mail. instead, a set of non-sensitive variables or attributes is picked from the e-mails originating from a given establishment and also from the profiles of the users sending those mails. the said attributes are extracted for all outbound mails. this extraction process does not involve reading the main text of the mail and thus the sensitivity of the mail information is protected. these attributes are chosen using filters built into the detection hardware. neural networks and support vector machine built into the detection hardware are then used on these attributes to detect pattern violation and possible information leakage.",2011-08-23,"The title of the patent is system for detecting information leakage in outbound e-mails without using the content of the mail and its abstract is a system for detecting information leakage in e-mails using neural network and support vector machines is provided. this system does not use the content of the e-mail or the content of the attachments in the e-mail. instead, a set of non-sensitive variables or attributes is picked from the e-mails originating from a given establishment and also from the profiles of the users sending those mails. the said attributes are extracted for all outbound mails. this extraction process does not involve reading the main text of the mail and thus the sensitivity of the mail information is protected. these attributes are chosen using filters built into the detection hardware. neural networks and support vector machine built into the detection hardware are then used on these attributes to detect pattern violation and possible information leakage. dated 2011-08-23"
8010252,trailer oscillation detection and compensation method for a vehicle and trailer combination,"a system and method of controlling a vehicle with a trailer comprises determining the presence of a trailer, generating an oscillation signal indicative of trailer swaying relative to the vehicle, generating an initial weighted dynamic control signal for a vehicle dynamic control system in response to the oscillation signal, operating at least one vehicle dynamic system according to the dynamic control signal, and thereafter, iteratively generating a penalty function for the weighted dynamic control signal as a function of the oscillation signal response. a neural network with an associated trainer modifies the dynamic control signal as a function of trailer sway response.",2011-08-30,"The title of the patent is trailer oscillation detection and compensation method for a vehicle and trailer combination and its abstract is a system and method of controlling a vehicle with a trailer comprises determining the presence of a trailer, generating an oscillation signal indicative of trailer swaying relative to the vehicle, generating an initial weighted dynamic control signal for a vehicle dynamic control system in response to the oscillation signal, operating at least one vehicle dynamic system according to the dynamic control signal, and thereafter, iteratively generating a penalty function for the weighted dynamic control signal as a function of the oscillation signal response. a neural network with an associated trainer modifies the dynamic control signal as a function of trailer sway response. dated 2011-08-30"
8010468,method for wafer analysis with artificial neural network and system thereof,"a method for wafer analysis with artificial neural network and the system thereof are disclosed. the method of the system of the present invention has several steps, including: first of all, providing a test unit for wafer test and generating a plurality of test data; next, transmitting the test data to a processing unit for transferring to output data; then, comparing the output data with predictive value and modifying bias and making the output data close to the predictive value, and repeating the steps mentioned above to train this system; finally, analyzing wafers by the trained system. using this system to analyze wafers not only saves time, but also reduces manpower and the risk resulting from artificial analysis.",2011-08-30,"The title of the patent is method for wafer analysis with artificial neural network and system thereof and its abstract is a method for wafer analysis with artificial neural network and the system thereof are disclosed. the method of the system of the present invention has several steps, including: first of all, providing a test unit for wafer test and generating a plurality of test data; next, transmitting the test data to a processing unit for transferring to output data; then, comparing the output data with predictive value and modifying bias and making the output data close to the predictive value, and repeating the steps mentioned above to train this system; finally, analyzing wafers by the trained system. using this system to analyze wafers not only saves time, but also reduces manpower and the risk resulting from artificial analysis. dated 2011-08-30"
8015128,biometric security using neuroplastic fidelity,"a system, method and program product for providing biometric security using neuroplastic fidelity. a method is disclosed that includes: receiving biometric data; analyzing the biometric data with a probabilistic neural network and outputting a chromosome containing a binary string; mapping the binary string to a selected extractor and a selected matcher; apply the selected extractor to the biometric data to generate a template; using the selected matcher to compare the template to a set of stored templates to identify a match; and outputting a result.",2011-09-06,"The title of the patent is biometric security using neuroplastic fidelity and its abstract is a system, method and program product for providing biometric security using neuroplastic fidelity. a method is disclosed that includes: receiving biometric data; analyzing the biometric data with a probabilistic neural network and outputting a chromosome containing a binary string; mapping the binary string to a selected extractor and a selected matcher; apply the selected extractor to the biometric data to generate a template; using the selected matcher to compare the template to a set of stored templates to identify a match; and outputting a result. dated 2011-09-06"
8015130,"information processing apparatus, information processing method, pattern recognition apparatus, and pattern recognition method","in a hierarchical neural network having a module structure, learning necessary for detection of a new feature class is executed by a processing module which has not finished learning yet and includes a plurality of neurons which should learn an unlearned feature class and have an undetermined receptor field structure by presenting a predetermined pattern to a data input layer. thus, a feature class necessary for subject recognition can be learned automatically and efficiently.",2011-09-06,"The title of the patent is information processing apparatus, information processing method, pattern recognition apparatus, and pattern recognition method and its abstract is in a hierarchical neural network having a module structure, learning necessary for detection of a new feature class is executed by a processing module which has not finished learning yet and includes a plurality of neurons which should learn an unlearned feature class and have an undetermined receptor field structure by presenting a predetermined pattern to a data input layer. thus, a feature class necessary for subject recognition can be learned automatically and efficiently. dated 2011-09-06"
8019134,automatic image analysis and quantification for fluorescence in situ hybridization,"an analysis system automatically analyzes and counts fluorescence signals present in biopsy tissue marked using fluorescence in situ hybridization (fish). the user of the system specifies classes of a class network and process steps of a process hierarchy. then pixel values in image slices of biopsy tissue are acquired in three dimensions. a computer-implemented network structure is generated by linking pixel values to objects of a data network according to the class network and process hierarchy. objects associated with pixel values at different depths of the biopsy tissue are used to determine the number, volume and distance between cell components. in one application, fluorescence signals that mark her2/neural genes and centromeres of chromosome seventeen are counted to diagnose breast cancer. her2/neural genes that overlap one another or that are covered by centromeres can be accurately counted. signal artifacts that do not mark genes can be identified by their excessive volume.",2011-09-13,"The title of the patent is automatic image analysis and quantification for fluorescence in situ hybridization and its abstract is an analysis system automatically analyzes and counts fluorescence signals present in biopsy tissue marked using fluorescence in situ hybridization (fish). the user of the system specifies classes of a class network and process steps of a process hierarchy. then pixel values in image slices of biopsy tissue are acquired in three dimensions. a computer-implemented network structure is generated by linking pixel values to objects of a data network according to the class network and process hierarchy. objects associated with pixel values at different depths of the biopsy tissue are used to determine the number, volume and distance between cell components. in one application, fluorescence signals that mark her2/neural genes and centromeres of chromosome seventeen are counted to diagnose breast cancer. her2/neural genes that overlap one another or that are covered by centromeres can be accurately counted. signal artifacts that do not mark genes can be identified by their excessive volume. dated 2011-09-13"
8024123,subterranean formation properties prediction,"a method for predicting subterranean formation properties of a wellsite. the method includes obtaining seismic data for an area of interest, obtaining an initial seismic cube using the seismic data, and obtaining a shifted seismic cubes using the seismic data, where each of the shifted seismic cubes is shifted from the initial seismic cube obtaining a shifted seismic cubes using the seismic data, where each of the shifted seismic cubes is shifted from the initial seismic cube. the method further includes generating a neural network using the initial seismic cube, the shifted seismic cubes, and well log data and applying the neural network to the seismic data to obtain a model for the area of interest, where the model is used to adjust an operation of the wellsite.",2011-09-20,"The title of the patent is subterranean formation properties prediction and its abstract is a method for predicting subterranean formation properties of a wellsite. the method includes obtaining seismic data for an area of interest, obtaining an initial seismic cube using the seismic data, and obtaining a shifted seismic cubes using the seismic data, where each of the shifted seismic cubes is shifted from the initial seismic cube obtaining a shifted seismic cubes using the seismic data, where each of the shifted seismic cubes is shifted from the initial seismic cube. the method further includes generating a neural network using the initial seismic cube, the shifted seismic cubes, and well log data and applying the neural network to the seismic data to obtain a model for the area of interest, where the model is used to adjust an operation of the wellsite. dated 2011-09-20"
8027730,systems and methods for treating disorders of the central nervous system by modulation of brain networks,"the present invention involves methods and systems for treatment of brain disorders using neuromodulation of brain networks. treatment of one or more brain networks associated with a brain disorder is realized with a consideration of network dynamics and coupling effects such as indirect stimulation of non-target regions. a brain modulation system (bms) increases, decreases, or otherwise modulates network regional activity in a differential manner. therapy may aim to maintain electrical or chemical (relative) characteristics within a specified range. therapy is initiated/adjusted using network functional imaging data including the use of brain network modeling. linking rules may guide in the setting and subsequent adjusting of the therapy related to regions of brain network. novel techniques are described for deterring the emergence of neural adaptation and of unintentional/indirect modulation arising from connectivity between network structures.",2011-09-27,"The title of the patent is systems and methods for treating disorders of the central nervous system by modulation of brain networks and its abstract is the present invention involves methods and systems for treatment of brain disorders using neuromodulation of brain networks. treatment of one or more brain networks associated with a brain disorder is realized with a consideration of network dynamics and coupling effects such as indirect stimulation of non-target regions. a brain modulation system (bms) increases, decreases, or otherwise modulates network regional activity in a differential manner. therapy may aim to maintain electrical or chemical (relative) characteristics within a specified range. therapy is initiated/adjusted using network functional imaging data including the use of brain network modeling. linking rules may guide in the setting and subsequent adjusting of the therapy related to regions of brain network. novel techniques are described for deterring the emergence of neural adaptation and of unintentional/indirect modulation arising from connectivity between network structures. dated 2011-09-27"
8027942,method and circuits for associating a complex operator to each component of an input pattern presented to an artificial neural network,"the method and circuits of the present invention aim to associate a complex component operator (cc_op) to each component of an input pattern presented to an input space mapping algorithm based artificial neural network (ann) during the distance evaluation process. a complex operator consists in the description of a function and a set of parameters attached thereto. the function is a mathematical entity (either a logic operator e.g. match(ai,bi), abs(ai−bi), . . . or an arithmetic operator, e.g. >, <, . . . ) or a set of software instructions possibly with a condition. in a first embodiment, the ann is provided with a global memory, common for all the neurons of the ann, that stores all the cc_ops. in another embodiment, the set of cc_ops is stored in the prototype memory of the neuron, so that the global memory is no longer physically necessary. according to the present invention, a component of a stored prototype may now designate objects of different nature. in addition, either implementation significantly reduces the number of components that are required in the neurons, therefore saving room when the ann is integrated in a silicon chip.",2011-09-27,"The title of the patent is method and circuits for associating a complex operator to each component of an input pattern presented to an artificial neural network and its abstract is the method and circuits of the present invention aim to associate a complex component operator (cc_op) to each component of an input pattern presented to an input space mapping algorithm based artificial neural network (ann) during the distance evaluation process. a complex operator consists in the description of a function and a set of parameters attached thereto. the function is a mathematical entity (either a logic operator e.g. match(ai,bi), abs(ai−bi), . . . or an arithmetic operator, e.g. >, <, . . . ) or a set of software instructions possibly with a condition. in a first embodiment, the ann is provided with a global memory, common for all the neurons of the ann, that stores all the cc_ops. in another embodiment, the set of cc_ops is stored in the prototype memory of the neuron, so that the global memory is no longer physically necessary. according to the present invention, a component of a stored prototype may now designate objects of different nature. in addition, either implementation significantly reduces the number of components that are required in the neurons, therefore saving room when the ann is integrated in a silicon chip. dated 2011-09-27"
8032469,recommending similar content identified with a neural network,"methods, systems and computer-readable media for finding similarities between visual objects by evaluating user interactions with a collection of visual objects are provided. using a neural network, human interactions with a collection of visual objects are evaluated to ascertain relationships or connections between visual objects. the relationship between visual objects indicates that the visual objects are similar. once relationships between visual objects are identified, a user may select one or more visual objects and receive suggested visual objects that are similar to the one or more visual objects selected by the user.",2011-10-04,"The title of the patent is recommending similar content identified with a neural network and its abstract is methods, systems and computer-readable media for finding similarities between visual objects by evaluating user interactions with a collection of visual objects are provided. using a neural network, human interactions with a collection of visual objects are evaluated to ascertain relationships or connections between visual objects. the relationship between visual objects indicates that the visual objects are similar. once relationships between visual objects are identified, a user may select one or more visual objects and receive suggested visual objects that are similar to the one or more visual objects selected by the user. dated 2011-10-04"
8036425,neural network-controlled automatic tracking and recognizing system and method,"a neural network-controlled automatic tracking and recognizing system includes a fixed field of view collection module, a full functions variable field of view collection module, a video image recognition algorithm module, a neural network control module, a suspect object track-tracking module, a database comparison and alarm judgment module, a monitored characteristic recording and rule setting module, a light monitoring and control module, a backlight module, an alarm output/display/storage module, and security monitoring sensors. the invention relates also to the operation method of the system.",2011-10-11,"The title of the patent is neural network-controlled automatic tracking and recognizing system and method and its abstract is a neural network-controlled automatic tracking and recognizing system includes a fixed field of view collection module, a full functions variable field of view collection module, a video image recognition algorithm module, a neural network control module, a suspect object track-tracking module, a database comparison and alarm judgment module, a monitored characteristic recording and rule setting module, a light monitoring and control module, a backlight module, an alarm output/display/storage module, and security monitoring sensors. the invention relates also to the operation method of the system. dated 2011-10-11"
8046314,"apparatus, method and system for stochastic workflow in oilfield operations","the invention relates to a method for performing an oilfield operation. the method steps include obtaining oilfield data sets associated with oilfield entities, generating a stochastic database from the oilfield data sets based on an artificial neural network of the oilfield data sets, screening the oilfield data sets to identify candidates from the oilfield entities, wherein the screening is based on the stochastic database, performing a detail evaluation of each candidates, selecting an oilfield entity from the candidates based on the detail evaluation, and performing the oilfield operation for the selected oilfield entity.",2011-10-25,"The title of the patent is apparatus, method and system for stochastic workflow in oilfield operations and its abstract is the invention relates to a method for performing an oilfield operation. the method steps include obtaining oilfield data sets associated with oilfield entities, generating a stochastic database from the oilfield data sets based on an artificial neural network of the oilfield data sets, screening the oilfield data sets to identify candidates from the oilfield entities, wherein the screening is based on the stochastic database, performing a detail evaluation of each candidates, selecting an oilfield entity from the candidates based on the detail evaluation, and performing the oilfield operation for the selected oilfield entity. dated 2011-10-25"
8051019,neural network resource sizing apparatus for database applications,"a neural network resource sizing apparatus for database applications. through use of multiple database application metrics input into a neural network learning algorithm, recommended resource capacities are generated. input parameters such as the number of records, lookups, images, pdfs, fields, blobs and width of fields for example may be utilized to train a neural network to yield needed resource metrics such as the processing power, memory, disk and/or network capacities required to run the database application. training for the neural network may involve running tests over all desired cross interactions of input and output parameters beginning for example with a small repository and ending with the maximum complexity of data and schema test. the training data is input into the neural network for the given database application version and utilized to plan resource utilization. a portal or webservice may be utilized to provide an interface to the apparatus.",2011-11-01,"The title of the patent is neural network resource sizing apparatus for database applications and its abstract is a neural network resource sizing apparatus for database applications. through use of multiple database application metrics input into a neural network learning algorithm, recommended resource capacities are generated. input parameters such as the number of records, lookups, images, pdfs, fields, blobs and width of fields for example may be utilized to train a neural network to yield needed resource metrics such as the processing power, memory, disk and/or network capacities required to run the database application. training for the neural network may involve running tests over all desired cross interactions of input and output parameters beginning for example with a small repository and ending with the maximum complexity of data and schema test. the training data is input into the neural network for the given database application version and utilized to plan resource utilization. a portal or webservice may be utilized to provide an interface to the apparatus. dated 2011-11-01"
8051020,base oil properties expert system,"a method for predicting properties of lubricant base oil blends, comprising the steps of generating an nmr spectrum, hplc-uv spectrum, and fims spectrum of a sample of a blend of at least two lubricant base oils and determining at least one composite structural molecular parameter of the sample from said spectrums. simdist and hpo analyses of the sample are then generated in order to determine a composite boiling point distribution and molecular weight of the sample from such analysis. a composite structural molecular parameter is applied, and the composite boiling point distribution and the composite molecular weight to a trained neural network is trained to correlate with the composite structural molecular parameter composite boiling point distribution and the composite molecular weight so as to predict composite properties of the sample. the properties comprise kinematic viscosity at 40 c, kinematic viscosity at 100 c, viscosity index, cloud point, and oxidation performance.",2011-11-01,"The title of the patent is base oil properties expert system and its abstract is a method for predicting properties of lubricant base oil blends, comprising the steps of generating an nmr spectrum, hplc-uv spectrum, and fims spectrum of a sample of a blend of at least two lubricant base oils and determining at least one composite structural molecular parameter of the sample from said spectrums. simdist and hpo analyses of the sample are then generated in order to determine a composite boiling point distribution and molecular weight of the sample from such analysis. a composite structural molecular parameter is applied, and the composite boiling point distribution and the composite molecular weight to a trained neural network is trained to correlate with the composite structural molecular parameter composite boiling point distribution and the composite molecular weight so as to predict composite properties of the sample. the properties comprise kinematic viscosity at 40 c, kinematic viscosity at 100 c, viscosity index, cloud point, and oxidation performance. dated 2011-11-01"
8055018,object image detection method,"the present invention discloses an object image detection method, which uses a coarse-to-fine strategy to detect objects. the method of the present invention comprises steps: acquiring an image and pre-processing the image to achieve dimensional reduction and information fusion; using a trained filter to screen features; and sequentially using a coarse-level mlp verifier and a fine-level mlp verifier to perform a neural network image detection to determine whether the features of the image match the features of the image of a target object. the present invention simultaneously uses three mainstream image detection methods, including the statistic method, neural network method and adaboost method, to perform image detection. therefore, the present invention has the advantages of the rapidity of the adaboost method and the accuracy of the neural network method at the same time.",2011-11-08,"The title of the patent is object image detection method and its abstract is the present invention discloses an object image detection method, which uses a coarse-to-fine strategy to detect objects. the method of the present invention comprises steps: acquiring an image and pre-processing the image to achieve dimensional reduction and information fusion; using a trained filter to screen features; and sequentially using a coarse-level mlp verifier and a fine-level mlp verifier to perform a neural network image detection to determine whether the features of the image match the features of the image of a target object. the present invention simultaneously uses three mainstream image detection methods, including the statistic method, neural network method and adaboost method, to perform image detection. therefore, the present invention has the advantages of the rapidity of the adaboost method and the accuracy of the neural network method at the same time. dated 2011-11-08"
8059890,method for implementing n-dimensional object recognition using dynamic adaptive recognition layers,"in a method and a system for the implementation of multi-layered network object recognition in multi-dimensional space, the structure of a neural recognition network is dynamically generated and adapted to recognize objects. the layers of the network are capable of recognizing key features of the input data by using evaluation rules to establish a hierarchical structure that is independent of data position and orientation, and can adapt varying data densities, geometrical scaling, and faulty or missing data. adjacent layers of the hierarchy are mutually reinforcing to facilitate the convergence of a solution. information flow is both bottom-up and top-down during the recognition process providing feedback from higher hierarchical layers to lower layers to cascade the results of higher-level recognition decisions to elements in lower layers.",2011-11-15,"The title of the patent is method for implementing n-dimensional object recognition using dynamic adaptive recognition layers and its abstract is in a method and a system for the implementation of multi-layered network object recognition in multi-dimensional space, the structure of a neural recognition network is dynamically generated and adapted to recognize objects. the layers of the network are capable of recognizing key features of the input data by using evaluation rules to establish a hierarchical structure that is independent of data position and orientation, and can adapt varying data densities, geometrical scaling, and faulty or missing data. adjacent layers of the hierarchy are mutually reinforcing to facilitate the convergence of a solution. information flow is both bottom-up and top-down during the recognition process providing feedback from higher hierarchical layers to lower layers to cascade the results of higher-level recognition decisions to elements in lower layers. dated 2011-11-15"
8060240,injection molding control method,"a method for controlling injection molding using a neural network in a control device of an injection molding machine. a measurement monitor value is acquired in a measurement step during test injection molding and an injection monitor value is acquired in an injection step. the acquired measurement monitor value is designated as an input term and the injection monitor value is designated as an output term. a prediction function that incorporates the measurement monitor value is then determined using the neural network. a first value corresponding to the injection monitor value is predicted by substituting into the prediction function a measurement monitor value acquired at completion of a measurement step during mass-production injection molding. on the basis of the predicted first value, a second value corresponding to an injection condition is determined. injection control and pressure maintenance control are then implemented on the basis of the second value corresponding to the injection condition.",2011-11-15,"The title of the patent is injection molding control method and its abstract is a method for controlling injection molding using a neural network in a control device of an injection molding machine. a measurement monitor value is acquired in a measurement step during test injection molding and an injection monitor value is acquired in an injection step. the acquired measurement monitor value is designated as an input term and the injection monitor value is designated as an output term. a prediction function that incorporates the measurement monitor value is then determined using the neural network. a first value corresponding to the injection monitor value is predicted by substituting into the prediction function a measurement monitor value acquired at completion of a measurement step during mass-production injection molding. on the basis of the predicted first value, a second value corresponding to an injection condition is determined. injection control and pressure maintenance control are then implemented on the basis of the second value corresponding to the injection condition. dated 2011-11-15"
8065022,methods and systems for neural network modeling of turbine components,"embodiments of the invention can include methods and systems for controlling clearances in a turbine. in one embodiment, a method can include applying at least one operating parameter as an input to at least one neural network model, modeling via the neural network model a thermal expansion of at least one turbine component, and taking a control action based at least in part on the modeled thermal expansion of the one or more turbine components. an example system can include a controller operable to determine and apply the operating parameters as inputs to the neural network model, model thermal expansion via the neural network model, and generate a control action based at least in part on the modeled thermal expansion.",2011-11-22,"The title of the patent is methods and systems for neural network modeling of turbine components and its abstract is embodiments of the invention can include methods and systems for controlling clearances in a turbine. in one embodiment, a method can include applying at least one operating parameter as an input to at least one neural network model, modeling via the neural network model a thermal expansion of at least one turbine component, and taking a control action based at least in part on the modeled thermal expansion of the one or more turbine components. an example system can include a controller operable to determine and apply the operating parameters as inputs to the neural network model, model thermal expansion via the neural network model, and generate a control action based at least in part on the modeled thermal expansion. dated 2011-11-22"
8065244,neural-network based surrogate model construction methods and applications thereof,"various neural-network based surrogate model construction methods are disclosed herein, along with various applications of such models. designed for use when only a sparse amount of data is available (a “sparse data condition”), some embodiments of the disclosed systems and methods: create a pool of neural networks trained on a first portion of a sparse data set; generate for each of various multi-objective functions a set of neural network ensembles that minimize the multi-objective function; select a local ensemble from each set of ensembles based on data not included in said first portion of said sparse data set; and combine a subset of the local ensembles to form a global ensemble. this approach enables usage of larger candidate pools, multi-stage validation, and a comprehensive performance measure that provides more robust predictions in the voids of parameter space.",2011-11-22,"The title of the patent is neural-network based surrogate model construction methods and applications thereof and its abstract is various neural-network based surrogate model construction methods are disclosed herein, along with various applications of such models. designed for use when only a sparse amount of data is available (a “sparse data condition”), some embodiments of the disclosed systems and methods: create a pool of neural networks trained on a first portion of a sparse data set; generate for each of various multi-objective functions a set of neural network ensembles that minimize the multi-objective function; select a local ensemble from each set of ensembles based on data not included in said first portion of said sparse data set; and combine a subset of the local ensembles to form a global ensemble. this approach enables usage of larger candidate pools, multi-stage validation, and a comprehensive performance measure that provides more robust predictions in the voids of parameter space. dated 2011-11-22"
8068731,dynamic bandwidth allocation method of ethernet passive optical network,"a dynamic bandwidth allocation method of an ethernet passive optical network, comprises a predictor and a rule of qos-promoted dynamic bandwidth allocation (pq-dba); the predictor predicts a client behavior and numbers of various kinds of packets by using a pipeline scheduling predictor consisted of a pipelined recurrent neural network (prnn), and a learning rule of the extended recursive least squares (erls); the present invention establishes a better qos traffic management for the olt-allocated onu bandwidth and client packets sent by priority.",2011-11-29,"The title of the patent is dynamic bandwidth allocation method of ethernet passive optical network and its abstract is a dynamic bandwidth allocation method of an ethernet passive optical network, comprises a predictor and a rule of qos-promoted dynamic bandwidth allocation (pq-dba); the predictor predicts a client behavior and numbers of various kinds of packets by using a pipeline scheduling predictor consisted of a pipelined recurrent neural network (prnn), and a learning rule of the extended recursive least squares (erls); the present invention establishes a better qos traffic management for the olt-allocated onu bandwidth and client packets sent by priority. dated 2011-11-29"
8068958,method for monitoring the adjustment movement of a component driven by a drive device,"a method for monitoring the adjustment movement of a component, in particular a window pane or a sunroof in motor vehicles, which is driven by a drive device and can be adjusted in a translatory or rotary fashion. a plurality of input signals which can be derived from the drive device and which represent a deceleration of the adjustment movement of the drive device are input at input neurons of an input layer of a neural network with at least one hidden layer having hidden neurons. said network outputting, at at least one output neuron of an output layer, an output value which corresponds to the adjusting force or to a trapped state or nontrapped state.",2011-11-29,"The title of the patent is method for monitoring the adjustment movement of a component driven by a drive device and its abstract is a method for monitoring the adjustment movement of a component, in particular a window pane or a sunroof in motor vehicles, which is driven by a drive device and can be adjusted in a translatory or rotary fashion. a plurality of input signals which can be derived from the drive device and which represent a deceleration of the adjustment movement of the drive device are input at input neurons of an input layer of a neural network with at least one hidden layer having hidden neurons. said network outputting, at at least one output neuron of an output layer, an output value which corresponds to the adjusting force or to a trapped state or nontrapped state. dated 2011-11-29"
8069076,generating audience analytics,"the present invention is directed to generating audience analytics that includes providing a database containing a plurality of user input pattern profiles representing the group of users of terminal device, in which each user of the group is associated with one of the plurality of user input pattern profiles. a clickstream algorithm, tracking algorithm, neural network, bayes classifier algorithm, or affinity-day part algorithm can be used to generate the user input pattern profiles. a user input pattern is detected based upon use of the terminal device by the current user and the user input pattern of the current user is dynamically matched with one of the user input pattern profiles contained in the database. the current user is identified based upon dynamic matching of the user input pattern generated by the current user with one of the user input pattern profiles. the present invention processes each user input pattern profile to identify a demographic type. a plurality of biometric behavior models are employed to identify a unique demographic type. each user input pattern profile is compared against the plurality of biometric behavior models to match each user input pattern profile with one of the biometric behavior models such that each user input pattern profile is correlated with one demographic type. audience analytics are then based upon the identified demographic types.",2011-11-29,"The title of the patent is generating audience analytics and its abstract is the present invention is directed to generating audience analytics that includes providing a database containing a plurality of user input pattern profiles representing the group of users of terminal device, in which each user of the group is associated with one of the plurality of user input pattern profiles. a clickstream algorithm, tracking algorithm, neural network, bayes classifier algorithm, or affinity-day part algorithm can be used to generate the user input pattern profiles. a user input pattern is detected based upon use of the terminal device by the current user and the user input pattern of the current user is dynamically matched with one of the user input pattern profiles contained in the database. the current user is identified based upon dynamic matching of the user input pattern generated by the current user with one of the user input pattern profiles. the present invention processes each user input pattern profile to identify a demographic type. a plurality of biometric behavior models are employed to identify a unique demographic type. each user input pattern profile is compared against the plurality of biometric behavior models to match each user input pattern profile with one of the biometric behavior models such that each user input pattern profile is correlated with one demographic type. audience analytics are then based upon the identified demographic types. dated 2011-11-29"
8071926,stability multiplexed autopilot,"rolling airframe projectile guidance and stability systems are disclosed. flight control surfaces, such as canards and/or tail fins are attached to a projectile airframe that is designed to roll during flight. stepper motors are attached to the flight control surfaces and move the flight control surfaces in discrete increments. a control system generates signals that control the flight control surfaces. the control system may include a neural network that is trained to generate control signals in response to received inputs.",2011-12-06,"The title of the patent is stability multiplexed autopilot and its abstract is rolling airframe projectile guidance and stability systems are disclosed. flight control surfaces, such as canards and/or tail fins are attached to a projectile airframe that is designed to roll during flight. stepper motors are attached to the flight control surfaces and move the flight control surfaces in discrete increments. a control system generates signals that control the flight control surfaces. the control system may include a neural network that is trained to generate control signals in response to received inputs. dated 2011-12-06"
8078330,"automatic energy management and energy consumption reduction, especially in commercial and multi-building systems","automatic energy management is provided, in even the most complex multi-building system. the necessity of a human operator for managing energy in a complex, multi-building system is reduced and even eliminated. computer-based monitoring and computer-based recognition of adverse energy events (such as the approach of a new energy peak) is highly advantageous in energy management. immediate automatic querying of energy users within a system of buildings for energy curtailment possibilities is provided. such immediate, automatic querying may be answered by the energy users through artificial intelligence and/or neural network technology provided to or programmed into the energy users, and the queried energy users may respond in real-time. those real-time computerized responses with energy curtailment possibilities may be received automatically by a data processing facility, and processed in real-time. advantageously, the responses from queried energy users with energy curtailment possibilities may be automatically processed into a round-robin curtailment rotation which may be implemented by a computer-based control system. thus, impact on occupants is minimized, and energy use and energy cost may be beneficially reduced in an intelligent, real-time manner. the invention also provides for early-recognition of impending adverse energy events, optimal response to a particular energy situation, real-time analysis of energy-related data, etc.",2011-12-13,"The title of the patent is automatic energy management and energy consumption reduction, especially in commercial and multi-building systems and its abstract is automatic energy management is provided, in even the most complex multi-building system. the necessity of a human operator for managing energy in a complex, multi-building system is reduced and even eliminated. computer-based monitoring and computer-based recognition of adverse energy events (such as the approach of a new energy peak) is highly advantageous in energy management. immediate automatic querying of energy users within a system of buildings for energy curtailment possibilities is provided. such immediate, automatic querying may be answered by the energy users through artificial intelligence and/or neural network technology provided to or programmed into the energy users, and the queried energy users may respond in real-time. those real-time computerized responses with energy curtailment possibilities may be received automatically by a data processing facility, and processed in real-time. advantageously, the responses from queried energy users with energy curtailment possibilities may be automatically processed into a round-robin curtailment rotation which may be implemented by a computer-based control system. thus, impact on occupants is minimized, and energy use and energy cost may be beneficially reduced in an intelligent, real-time manner. the invention also provides for early-recognition of impending adverse energy events, optimal response to a particular energy situation, real-time analysis of energy-related data, etc. dated 2011-12-13"
8078557,use of neural networks for keyword generation,"a system for identifying keywords in search results includes a plurality of neurons connected as a neural network, the neurons being associated with words and documents. an activity regulator regulates a minimum and/or maximum number of neurons of the neural network that are excited at any given time. means for displaying the neurons to a user and identifying the neurons that correspond to keywords can be provided. means for changing positions of the neurons relative to each other based on input from the user can be provided. the change in position of one neuron changes the keywords. the input from the user can be dragging a neuron on a display device, or changing a relevance of two neurons relative to each other. the neural network can be excited by a query that comprises words selected by a user. the neural network can be a bidirectional network. the user can inhibit neurons of the neural network by indicating irrelevance of a document. the neural network can be excited by a query that identifies a document considered relevant by a user. the neural network can also include neurons that represent groups of words. the neural network can be excited by a query that identifies a plurality of documents considered relevant by a user, and can output keywords associated with the plurality of documents.",2011-12-13,"The title of the patent is use of neural networks for keyword generation and its abstract is a system for identifying keywords in search results includes a plurality of neurons connected as a neural network, the neurons being associated with words and documents. an activity regulator regulates a minimum and/or maximum number of neurons of the neural network that are excited at any given time. means for displaying the neurons to a user and identifying the neurons that correspond to keywords can be provided. means for changing positions of the neurons relative to each other based on input from the user can be provided. the change in position of one neuron changes the keywords. the input from the user can be dragging a neuron on a display device, or changing a relevance of two neurons relative to each other. the neural network can be excited by a query that comprises words selected by a user. the neural network can be a bidirectional network. the user can inhibit neurons of the neural network by indicating irrelevance of a document. the neural network can be excited by a query that identifies a document considered relevant by a user. the neural network can also include neurons that represent groups of words. the neural network can be excited by a query that identifies a plurality of documents considered relevant by a user, and can output keywords associated with the plurality of documents. dated 2011-12-13"
8080964,neural network and method for estimating regions of motor operation from information characterizing the motor,"a method for collecting operational parameters of a motor may include controlling the energization of a phase winding of the motor to establish an operating point, monitoring operational parameters of the motor that characterize a relationship between the energization control applied to the motor's phase winding and the motor's response to this control, and collecting information of the operational parameters for the operating point that characterizes the relationship between the applied energization control and the motor's response. the collected information characterizing the relationship between the applied energization control and the motor's response may be employed by a neural network to estimate the regions of operation of the motor. and a system for controlling the operation of motor may employ this information, the neural network, or both to regulate the energization of a motor's phase winding during a phase cycle.",2011-12-20,"The title of the patent is neural network and method for estimating regions of motor operation from information characterizing the motor and its abstract is a method for collecting operational parameters of a motor may include controlling the energization of a phase winding of the motor to establish an operating point, monitoring operational parameters of the motor that characterize a relationship between the energization control applied to the motor's phase winding and the motor's response to this control, and collecting information of the operational parameters for the operating point that characterizes the relationship between the applied energization control and the motor's response. the collected information characterizing the relationship between the applied energization control and the motor's response may be employed by a neural network to estimate the regions of operation of the motor. and a system for controlling the operation of motor may employ this information, the neural network, or both to regulate the energization of a motor's phase winding during a phase cycle. dated 2011-12-20"
8081816,apparatus and method for hardware implementation of object recognition from an image stream using artificial neural network,"the present invention is an apparatus and method for object recognition from at least an image stream from at least an image frame utilizing at least an artificial neural network. the present invention further comprises means for generating multiple components of an image pyramid simultaneously from a single image stream, means for providing the active pixel and interlayer neuron data to at least a subwindow processor, means for multiplying and accumulating the product of a pixel data or interlayer data and a synapse weight, and means for performing the activation of an accumulation. the present invention allows the artificial neural networks to be reconfigurable, thus embracing a broad range of object recognition applications in a flexible way. the subwindow processor in the present invention also further comprises means for performing neuron computations for at least a neuron. an exemplary embodiment of the present invention is used for object recognition, including face detection and gender recognition, in hardware. the apparatus comprises a digital circuitry system or ic that embodies the components of the present invention.",2011-12-20,"The title of the patent is apparatus and method for hardware implementation of object recognition from an image stream using artificial neural network and its abstract is the present invention is an apparatus and method for object recognition from at least an image stream from at least an image frame utilizing at least an artificial neural network. the present invention further comprises means for generating multiple components of an image pyramid simultaneously from a single image stream, means for providing the active pixel and interlayer neuron data to at least a subwindow processor, means for multiplying and accumulating the product of a pixel data or interlayer data and a synapse weight, and means for performing the activation of an accumulation. the present invention allows the artificial neural networks to be reconfigurable, thus embracing a broad range of object recognition applications in a flexible way. the subwindow processor in the present invention also further comprises means for performing neuron computations for at least a neuron. an exemplary embodiment of the present invention is used for object recognition, including face detection and gender recognition, in hardware. the apparatus comprises a digital circuitry system or ic that embodies the components of the present invention. dated 2011-12-20"
8082217,multiphase flow meter for electrical submersible pumps using artificial neural networks,"a multiphase flow meter used in conjunction with an electrical submersible pump system in a well bore includes sensors to determine and transmit well bore pressure measurements, including tubing and down hole pressure measurements. the multiphase flow meter also includes at least one artificial neural network device to be used for outputting flow characteristics of the well bore. the artificial neural network device is trained to output tubing and downhole flow characteristics responsive to multiphase-flow pressure gradient calculations and pump and reservoir models, combined with standard down-hole pressure, tubing surface pressure readings, and the frequency applied to the electrical submersible pump motor.",2011-12-20,"The title of the patent is multiphase flow meter for electrical submersible pumps using artificial neural networks and its abstract is a multiphase flow meter used in conjunction with an electrical submersible pump system in a well bore includes sensors to determine and transmit well bore pressure measurements, including tubing and down hole pressure measurements. the multiphase flow meter also includes at least one artificial neural network device to be used for outputting flow characteristics of the well bore. the artificial neural network device is trained to output tubing and downhole flow characteristics responsive to multiphase-flow pressure gradient calculations and pump and reservoir models, combined with standard down-hole pressure, tubing surface pressure readings, and the frequency applied to the electrical submersible pump motor. dated 2011-12-20"
8086052,hybrid video compression method,"the invention concerns a method for compressing a digitally coded video frame sequence. in the method, a given frame is divided into blocks, and the information content of selected blocks is modified, relying on information contained in a neighboring block or blocks (prediction), and the blocks are converted from spatial representation into frequency representation. the information content of the transformed blocks is encoded by arithmetic coding. the efficiency of the coding is improved by various methods, such as dynamically partitioning the blocks into sub-blocks, or performing a compressibility analysis is the blocks before carrying out further transformations. the entropy coding uses a neural network to determine the parameters of the arithmetic coding. the frames are dynamically re-scaled, depending on available bandwidth and quality of the coded image.",2011-12-27,"The title of the patent is hybrid video compression method and its abstract is the invention concerns a method for compressing a digitally coded video frame sequence. in the method, a given frame is divided into blocks, and the information content of selected blocks is modified, relying on information contained in a neighboring block or blocks (prediction), and the blocks are converted from spatial representation into frequency representation. the information content of the transformed blocks is encoded by arithmetic coding. the efficiency of the coding is improved by various methods, such as dynamically partitioning the blocks into sub-blocks, or performing a compressibility analysis is the blocks before carrying out further transformations. the entropy coding uses a neural network to determine the parameters of the arithmetic coding. the frames are dynamically re-scaled, depending on available bandwidth and quality of the coded image. dated 2011-12-27"
8090512,system and method for controlling a clutch fill event,"a method optimizes a fill event of an apply chamber of a fluid-actuated clutch, and includes determining input values describing the fill event, and then estimating a fill time using the input values. the method includes filling the apply chamber using the estimated fill time (eft) or within an allowable range of the eft. the input values can include a command line pressure, command fill stroke pressure, and an estimated viscosity of the fluid, although other values can be used. the input values are processed through a neural network having an input layer, an optional hidden layer, and an output layer. an assembly includes a fluid-actuated clutch having an apply chamber and a controller operable for estimating the fill time required for filling the apply chamber, and for controlling the fill of the apply chamber within the eft.",2012-01-03,"The title of the patent is system and method for controlling a clutch fill event and its abstract is a method optimizes a fill event of an apply chamber of a fluid-actuated clutch, and includes determining input values describing the fill event, and then estimating a fill time using the input values. the method includes filling the apply chamber using the estimated fill time (eft) or within an allowable range of the eft. the input values can include a command line pressure, command fill stroke pressure, and an estimated viscosity of the fluid, although other values can be used. the input values are processed through a neural network having an input layer, an optional hidden layer, and an output layer. an assembly includes a fluid-actuated clutch having an apply chamber and a controller operable for estimating the fill time required for filling the apply chamber, and for controlling the fill of the apply chamber within the eft. dated 2012-01-03"
8090672,detecting and evaluating operation-dependent processes in automated production utilizing fuzzy operators and neural network system,"a system (10) testing and rating operation-dependent processes and/or components (20) in automated production and test sequences comprises a robot (12) which by means of a minimum of one sensor (14, 16) detects test/measured values (m) of at least one operating and/or display element (22, 24) of the component (20) to be tested respectively rated and transmits to an analyzer (40) analyzing and rating the measured values (m) by means of defined quality functions (50), said quality functions by means of operators (52) imitating human rating schematics respectively rules and based on this processing result generating at least one rating.",2012-01-03,"The title of the patent is detecting and evaluating operation-dependent processes in automated production utilizing fuzzy operators and neural network system and its abstract is a system (10) testing and rating operation-dependent processes and/or components (20) in automated production and test sequences comprises a robot (12) which by means of a minimum of one sensor (14, 16) detects test/measured values (m) of at least one operating and/or display element (22, 24) of the component (20) to be tested respectively rated and transmits to an analyzer (40) analyzing and rating the measured values (m) by means of defined quality functions (50), said quality functions by means of operators (52) imitating human rating schematics respectively rules and based on this processing result generating at least one rating. dated 2012-01-03"
8095344,methods and systems for modeling material behavior,"a method for modeling material behavior includes using empirical three dimensional non-uniform stress and strain data to train a self-organizing computational model such as a neural network. a laboratory device for measuring non-uniform stress and strain data from material includes an enclosure with an inclusion in it. as the enclosure is compressed, the inclusion induces a non-uniform state of stress and strain. a field testing device includes a body having a moveable section. when the body is inserted in a material and the moveable section moved, a non-uniform state of stress and strain can be characterized.",2012-01-10,"The title of the patent is methods and systems for modeling material behavior and its abstract is a method for modeling material behavior includes using empirical three dimensional non-uniform stress and strain data to train a self-organizing computational model such as a neural network. a laboratory device for measuring non-uniform stress and strain data from material includes an enclosure with an inclusion in it. as the enclosure is compressed, the inclusion induces a non-uniform state of stress and strain. a field testing device includes a body having a moveable section. when the body is inserted in a material and the moveable section moved, a non-uniform state of stress and strain can be characterized. dated 2012-01-10"
8099311,system and method for routing tasks to a user in a workforce,"a routing system and method efficiently routes tasks to users who are members of a large and geographically diverse workforce. generally, limited information is known about each user's skills and behavioral factors. based on a profile containing the known information about a user, task is efficiently allocated and routed to a user by matching attributes of the task to the profile using a neural network and a stochastic model. feedback is collected by the routing system based on the user's handling of the task and on whether a solution provided by the user was accepted. over time, as more feedback is collected, the profile and/or the neural network are refined which allows for more efficient routing of future tasks.",2012-01-17,"The title of the patent is system and method for routing tasks to a user in a workforce and its abstract is a routing system and method efficiently routes tasks to users who are members of a large and geographically diverse workforce. generally, limited information is known about each user's skills and behavioral factors. based on a profile containing the known information about a user, task is efficiently allocated and routed to a user by matching attributes of the task to the profile using a neural network and a stochastic model. feedback is collected by the routing system based on the user's handling of the task and on whether a solution provided by the user was accepted. over time, as more feedback is collected, the profile and/or the neural network are refined which allows for more efficient routing of future tasks. dated 2012-01-17"
8103602,solving the distal reward problem through linkage of stdp and dopamine signaling,"in pavlovian and instrumental conditioning, rewards typically come seconds after reward-triggering actions, creating an explanatory conundrum known as the distal reward problem or the credit assignment problem. how does the brain know what firing patterns of what neurons are responsible for the reward if (1) the firing patterns are no longer there when the reward arrives and (2) most neurons and synapses are active during the waiting period to the reward? a model network and computer simulation of cortical spiking neurons with spike-timing-dependent plasticity (stdp) modulated by dopamine (da) is disclosed to answer this question. stdp is triggered by nearly-coincident firing patterns of a presynaptic neuron and a postsynaptic neuron on a millisecond time scale, with slow kinetics of subsequent synaptic plasticity being sensitive to changes in the extracellular dopamine da concentration during the critical period of a few seconds after the nearly-coincident firing patterns. random neuronal firings during the waiting period leading to the reward do not affect stdp, and hence make the neural network insensitive to this ongoing random firing activity. the importance of precise firing patterns in brain dynamics and the use of a global diffusive reinforcement signal in the form of extracellular dopamine da can selectively influence the right synapses at the right time.",2012-01-24,"The title of the patent is solving the distal reward problem through linkage of stdp and dopamine signaling and its abstract is in pavlovian and instrumental conditioning, rewards typically come seconds after reward-triggering actions, creating an explanatory conundrum known as the distal reward problem or the credit assignment problem. how does the brain know what firing patterns of what neurons are responsible for the reward if (1) the firing patterns are no longer there when the reward arrives and (2) most neurons and synapses are active during the waiting period to the reward? a model network and computer simulation of cortical spiking neurons with spike-timing-dependent plasticity (stdp) modulated by dopamine (da) is disclosed to answer this question. stdp is triggered by nearly-coincident firing patterns of a presynaptic neuron and a postsynaptic neuron on a millisecond time scale, with slow kinetics of subsequent synaptic plasticity being sensitive to changes in the extracellular dopamine da concentration during the critical period of a few seconds after the nearly-coincident firing patterns. random neuronal firings during the waiting period leading to the reward do not affect stdp, and hence make the neural network insensitive to this ongoing random firing activity. the importance of precise firing patterns in brain dynamics and the use of a global diffusive reinforcement signal in the form of extracellular dopamine da can selectively influence the right synapses at the right time. dated 2012-01-24"
8103606,"architecture, system and method for artificial neural network implementation","an architecture, systems and methods for a scalable artificial neural network, wherein the architecture includes: an input layer; at least one hidden layer; an output layer; and a parallelization subsystem configured to provide a variable degree of parallelization to the input layer, at least one hidden layer, and output layer. in a particular case, the architecture includes a back-propagation subsystem that is configured to adjust weights in the scalable artificial neural network in accordance with the variable degree of parallelization. systems and methods are also provided for selecting an appropriate degree of parallelization based on factors such as hardware resources and performance requirements.",2012-01-24,"The title of the patent is architecture, system and method for artificial neural network implementation and its abstract is an architecture, systems and methods for a scalable artificial neural network, wherein the architecture includes: an input layer; at least one hidden layer; an output layer; and a parallelization subsystem configured to provide a variable degree of parallelization to the input layer, at least one hidden layer, and output layer. in a particular case, the architecture includes a back-propagation subsystem that is configured to adjust weights in the scalable artificial neural network in accordance with the variable degree of parallelization. systems and methods are also provided for selecting an appropriate degree of parallelization based on factors such as hardware resources and performance requirements. dated 2012-01-24"
8108328,neural network based hermite interpolator for scatterometry parameter estimation,"generation of a meta-model for scatterometry analysis of a sample diffracting structure having unknown parameters. a training set comprising both a spectral signal evaluation and a derivative of the signal with respect to at least one parameter across a parameter space is rigorously computed. a neural network is trained with the training set to provide reference spectral information for a comparison to sample spectral information recorded from the sample diffracting structure. a neural network may be trained with derivative information using an algebraic method wherein a network bias vector is centered over both a primary sampling matrix and an auxiliary sampling matrix. the result of the algebraic method may be used for initializing neural network coefficients for training by optimization of the neural network weights, minimizing a difference between the actual signal and the modeled signal based on a objective function containing both function evaluations and derivatives.",2012-01-31,"The title of the patent is neural network based hermite interpolator for scatterometry parameter estimation and its abstract is generation of a meta-model for scatterometry analysis of a sample diffracting structure having unknown parameters. a training set comprising both a spectral signal evaluation and a derivative of the signal with respect to at least one parameter across a parameter space is rigorously computed. a neural network is trained with the training set to provide reference spectral information for a comparison to sample spectral information recorded from the sample diffracting structure. a neural network may be trained with derivative information using an algebraic method wherein a network bias vector is centered over both a primary sampling matrix and an auxiliary sampling matrix. the result of the algebraic method may be used for initializing neural network coefficients for training by optimization of the neural network weights, minimizing a difference between the actual signal and the modeled signal based on a objective function containing both function evaluations and derivatives. dated 2012-01-31"
8111174,acoustic signature recognition of running vehicles using spectro-temporal dynamic neural network,"a method and apparatus for identifying running vehicles in an area to be monitored using acoustic signature recognition. the apparatus includes an input sensor for capturing an acoustic waveform produced by a vehicle source, and a processing system. the waveform is digitized and divided into frames. each frame is filtered into a plurality of gammatone filtered signals. at least one spectral feature vector is computed for each frame. the vectors are integrated across a plurality of frames to create a spectro-temporal representation of the vehicle waveform. in a training mode, values from the spectro-temporal representation are used as inputs to a nonlinear hebbian learning function to extract acoustic signatures and synaptic weights. in an active mode, the synaptic weights and acoustic signatures are used as patterns in a supervised associative network to identify whether a vehicle is present in the area to be monitored. in response to a vehicle being present, the class of vehicle is identified. results may be provided to a central computer.",2012-02-07,"The title of the patent is acoustic signature recognition of running vehicles using spectro-temporal dynamic neural network and its abstract is a method and apparatus for identifying running vehicles in an area to be monitored using acoustic signature recognition. the apparatus includes an input sensor for capturing an acoustic waveform produced by a vehicle source, and a processing system. the waveform is digitized and divided into frames. each frame is filtered into a plurality of gammatone filtered signals. at least one spectral feature vector is computed for each frame. the vectors are integrated across a plurality of frames to create a spectro-temporal representation of the vehicle waveform. in a training mode, values from the spectro-temporal representation are used as inputs to a nonlinear hebbian learning function to extract acoustic signatures and synaptic weights. in an active mode, the synaptic weights and acoustic signatures are used as patterns in a supervised associative network to identify whether a vehicle is present in the area to be monitored. in response to a vehicle being present, the class of vehicle is identified. results may be provided to a central computer. dated 2012-02-07"
8116787,wireless network coverage based on quality of service,"coverage-based quality-of-service (qos) in wireless networks. a premium qos service is provided by the network to users who qualify to receive qos signals by moving handsets into a bounded premium service geographical coverage area. a mobile handset periodically transmits handset lat-long data to the wireless network via a control channel. the network includes data that defines the bounded geographical coverage area. the wireless network receives the handset lat-long data and maps the data to coverage area data. if the handset lat-long data maps into the coverage area data, the user handset is authorized to receive and use the premium qos signals over a 3g network; otherwise, the user falls back to default service on a 2g network. alternatively, the network utilizes a trained neural network to process the lat-long data to qualify the handset for premium services.",2012-02-14,"The title of the patent is wireless network coverage based on quality of service and its abstract is coverage-based quality-of-service (qos) in wireless networks. a premium qos service is provided by the network to users who qualify to receive qos signals by moving handsets into a bounded premium service geographical coverage area. a mobile handset periodically transmits handset lat-long data to the wireless network via a control channel. the network includes data that defines the bounded geographical coverage area. the wireless network receives the handset lat-long data and maps the data to coverage area data. if the handset lat-long data maps into the coverage area data, the user handset is authorized to receive and use the premium qos signals over a 3g network; otherwise, the user falls back to default service on a 2g network. alternatively, the network utilizes a trained neural network to process the lat-long data to qualify the handset for premium services. dated 2012-02-14"
8117143,using affinity measures with supervised classifiers,"a non-binary affinity measure between any two data points for a supervised classifier may be determined. for example, affinity measures may be determined for tree, kernel-based, nearest neighbor-based and neural network supervised classifiers. by providing non-binary affinity measures using supervised classifiers, more information may be provided for clustering, analyzing and, particularly, for visualizing the results of data mining.",2012-02-14,"The title of the patent is using affinity measures with supervised classifiers and its abstract is a non-binary affinity measure between any two data points for a supervised classifier may be determined. for example, affinity measures may be determined for tree, kernel-based, nearest neighbor-based and neural network supervised classifiers. by providing non-binary affinity measures using supervised classifiers, more information may be provided for clustering, analyzing and, particularly, for visualizing the results of data mining. dated 2012-02-14"
8121968,long-term memory in a video analysis system,"a long-term memory used to store and retrieve information learned while a video analysis system observes a stream of video frames is disclosed. the long-term memory provides a memory with a capacity that grows in size gracefully, as events are observed over time. additionally, the long-term memory may encode events, represented by sub-graphs of a neural network. further, rather than predefining a number of patterns recognized and manipulated by the long-term memory, embodiments of the invention provide a long-term memory where the size of a feature dimension (used to determine the similarity between different observed events) may grow dynamically as necessary, depending on the actual events observed in a sequence of video frames.",2012-02-21,"The title of the patent is long-term memory in a video analysis system and its abstract is a long-term memory used to store and retrieve information learned while a video analysis system observes a stream of video frames is disclosed. the long-term memory provides a memory with a capacity that grows in size gracefully, as events are observed over time. additionally, the long-term memory may encode events, represented by sub-graphs of a neural network. further, rather than predefining a number of patterns recognized and manipulated by the long-term memory, embodiments of the invention provide a long-term memory where the size of a feature dimension (used to determine the similarity between different observed events) may grow dynamically as necessary, depending on the actual events observed in a sequence of video frames. dated 2012-02-21"
8126710,conservative training method for adapting a neural network of an automatic speech recognition device,"a method of adapting a neural network of an automatic speech recognition device, includes the steps of: providing a neural network including an input stage, an intermediate stage and an output stage, the output stage outputting phoneme probabilities; providing a linear stage in the neural network; and training the linear stage by means of an adaptation set; wherein the step of providing the linear stage includes the step of providing the linear stage after the intermediate stage.",2012-02-28,"The title of the patent is conservative training method for adapting a neural network of an automatic speech recognition device and its abstract is a method of adapting a neural network of an automatic speech recognition device, includes the steps of: providing a neural network including an input stage, an intermediate stage and an output stage, the output stage outputting phoneme probabilities; providing a linear stage in the neural network; and training the linear stage by means of an adaptation set; wherein the step of providing the linear stage includes the step of providing the linear stage after the intermediate stage. dated 2012-02-28"
8131405,method and apparatuses for controlling high wing loaded parafoils,"an adaptive guidance system (ags) regulates the altitude and heading of a parasail to arrive at a target at a prescribed altitude. since the altitude profile depends both on unknown wing loading, and wind magnitude and direction, the ags estimates the glide slope and wind on the fly and provides a command to a stability augmentation system (sas) that results in the desired glide slope and heading by performing a sequence of maneuvers and sensing the response. according to one embodiment, the sas operates in the linear region and includes a pid controller that uses the difference between the actual heading and heading command to create an actuator output. the actuator output is limited by a position/rate limiter that imposes the physical limitations of the response time of the actuator/servos and position limits to prevent entering the nonlinear region. alternatively, an adaptive sas operates in both the linear and nonlinear region and includes a neural network (nn) that receives an error signal (difference between a reference model and actual heading) which is used to adapt the weights of the nn.",2012-03-06,"The title of the patent is method and apparatuses for controlling high wing loaded parafoils and its abstract is an adaptive guidance system (ags) regulates the altitude and heading of a parasail to arrive at a target at a prescribed altitude. since the altitude profile depends both on unknown wing loading, and wind magnitude and direction, the ags estimates the glide slope and wind on the fly and provides a command to a stability augmentation system (sas) that results in the desired glide slope and heading by performing a sequence of maneuvers and sensing the response. according to one embodiment, the sas operates in the linear region and includes a pid controller that uses the difference between the actual heading and heading command to create an actuator output. the actuator output is limited by a position/rate limiter that imposes the physical limitations of the response time of the actuator/servos and position limits to prevent entering the nonlinear region. alternatively, an adaptive sas operates in both the linear and nonlinear region and includes a neural network (nn) that receives an error signal (difference between a reference model and actual heading) which is used to adapt the weights of the nn. dated 2012-03-06"
8131655,spam filtering using feature relevance assignment in neural networks,"in some embodiments, a spam filtering method includes computing a pattern relevance for each of a set of message feature patterns, and using a neural network filter to classify incoming messages as spam or ham according to the pattern relevancies. each message feature pattern is characterized by the simultaneous presence within a message of a specific set of message features (e.g., the presence of certain keywords within the message body, various message header heuristics, various message layout features, etc.). each message feature may be spam- or ham-identifying, and may receive a tunable feature relevance weight from an external source (e.g. data file and/or human operator). the external feature relevance weights modulate the set of neuronal weights calculated through a training process of the neural network.",2012-03-06,"The title of the patent is spam filtering using feature relevance assignment in neural networks and its abstract is in some embodiments, a spam filtering method includes computing a pattern relevance for each of a set of message feature patterns, and using a neural network filter to classify incoming messages as spam or ham according to the pattern relevancies. each message feature pattern is characterized by the simultaneous presence within a message of a specific set of message features (e.g., the presence of certain keywords within the message body, various message header heuristics, various message layout features, etc.). each message feature may be spam- or ham-identifying, and may receive a tunable feature relevance weight from an external source (e.g. data file and/or human operator). the external feature relevance weights modulate the set of neuronal weights calculated through a training process of the neural network. dated 2012-03-06"
8131659,field-programmable gate array based accelerator system,"accelerator systems and methods are disclosed that utilize fpga technology to achieve better parallelism and processing speed. a field programmable gate array (fpga) is configured to have a hardware logic performing computations associated with a neural network training algorithm, especially a web relevance ranking algorithm such as lambarank. the training data is first processed and organized by a host computing device, and then streamed to the fpga for direct access by the fpga to perform high-bandwidth computation with increased training speed. thus, large data sets such as that related to web relevance ranking can be processed. the fpga may include a processing element performing computations of a hidden layer of the neural network training algorithm. parallel computing may be realized using a single instruction multiple data streams (simd) architecture with multiple arithmetic logic units in the fpga.",2012-03-06,"The title of the patent is field-programmable gate array based accelerator system and its abstract is accelerator systems and methods are disclosed that utilize fpga technology to achieve better parallelism and processing speed. a field programmable gate array (fpga) is configured to have a hardware logic performing computations associated with a neural network training algorithm, especially a web relevance ranking algorithm such as lambarank. the training data is first processed and organized by a host computing device, and then streamed to the fpga for direct access by the fpga to perform high-bandwidth computation with increased training speed. thus, large data sets such as that related to web relevance ranking can be processed. the fpga may include a processing element performing computations of a hidden layer of the neural network training algorithm. parallel computing may be realized using a single instruction multiple data streams (simd) architecture with multiple arithmetic logic units in the fpga. dated 2012-03-06"
8138766,flashover analysis tool,"a method to minimize human intervention during decision making processes while controlling an electrical power system by identifying an initiating element that cause a tripping of the transmission overhead lines and identifying potential future protection system failures that can initiate a cascading of tripping or total national blackout. a method of producing flashover analysis signal as a protection system analysis including processing a neutral current, three phase current profile, three phase voltage profile, and a plurality of digital signal of a transmission line using an artificial neural network to calculate pickup time, reset time, def confirmation time or total fault clearance time. a method of producing flashover analysis signal including as a flashover signature analysis to identify the cause of the flashover as a current transformer explosion, tree encroachment, crane, lightning strike or polluted insulator.",2012-03-20,"The title of the patent is flashover analysis tool and its abstract is a method to minimize human intervention during decision making processes while controlling an electrical power system by identifying an initiating element that cause a tripping of the transmission overhead lines and identifying potential future protection system failures that can initiate a cascading of tripping or total national blackout. a method of producing flashover analysis signal as a protection system analysis including processing a neutral current, three phase current profile, three phase voltage profile, and a plurality of digital signal of a transmission line using an artificial neural network to calculate pickup time, reset time, def confirmation time or total fault clearance time. a method of producing flashover analysis signal including as a flashover signature analysis to identify the cause of the flashover as a current transformer explosion, tree encroachment, crane, lightning strike or polluted insulator. dated 2012-03-20"
8146429,methods and systems for classifying the type and severity of defects in welds,a method for determining the type of a defect in a weld may include determining a defect location and a corresponding defect signal by analyzing ultrasonic response signals collected from a plurality of measurement locations along the weld. the defect signal and the plurality of defect proximity signals corresponding to ultrasonic response signals from measurement locations on each side of the defect location may then be input into a trained artificial neural network. the trained artificial neural network may be operable to identify the type of the defect located at the defect location based on the defect signal and the plurality of defect proximity signals and output the type of the defect located at the defect location. the trained artificial neural network may also be operable to determine a defect severity classification based on the defect signal and the plurality of defect proximity signals and output the severity classification.,2012-04-03,The title of the patent is methods and systems for classifying the type and severity of defects in welds and its abstract is a method for determining the type of a defect in a weld may include determining a defect location and a corresponding defect signal by analyzing ultrasonic response signals collected from a plurality of measurement locations along the weld. the defect signal and the plurality of defect proximity signals corresponding to ultrasonic response signals from measurement locations on each side of the defect location may then be input into a trained artificial neural network. the trained artificial neural network may be operable to identify the type of the defect located at the defect location based on the defect signal and the plurality of defect proximity signals and output the type of the defect located at the defect location. the trained artificial neural network may also be operable to determine a defect severity classification based on the defect signal and the plurality of defect proximity signals and output the severity classification. dated 2012-04-03
8156057,adaptive neural network utilizing nanotechnology-based components,"methods and systems for modifying at least one synapse of a physicallelectromechanical neural network. a physical/electromechanical neural network implemented as an adaptive neural network can be provided, which includes one or more neurons and one or more synapses thereof, wherein the neurons and synapses are formed from a plurality of nanoparticles disposed within a dielectric solution in association with one or more pre-synaptic electrodes and one or more post-synaptic electrodes and an applied electric field. at least one pulse can be generated from one or more of the neurons to one or more of the pre-synaptic electrodes of a succeeding neuron and one or more post-synaptic electrodes of one or more of the neurons of the physical/electromechanical neural network, thereby strengthening at least one nanoparticle of a plurality of nanoparticles disposed within the dielectric solution and at least one synapse thereof.",2012-04-10,"The title of the patent is adaptive neural network utilizing nanotechnology-based components and its abstract is methods and systems for modifying at least one synapse of a physicallelectromechanical neural network. a physical/electromechanical neural network implemented as an adaptive neural network can be provided, which includes one or more neurons and one or more synapses thereof, wherein the neurons and synapses are formed from a plurality of nanoparticles disposed within a dielectric solution in association with one or more pre-synaptic electrodes and one or more post-synaptic electrodes and an applied electric field. at least one pulse can be generated from one or more of the neurons to one or more of the pre-synaptic electrodes of a succeeding neuron and one or more post-synaptic electrodes of one or more of the neurons of the physical/electromechanical neural network, thereby strengthening at least one nanoparticle of a plurality of nanoparticles disposed within the dielectric solution and at least one synapse thereof. dated 2012-04-10"
8159976,neural network-based mobility management for healing mobile ad hoc radio networks,"a self healing ad hoc communications network and method of training for and healing the network. the network includes wireless devices or nodes that include a neural network element and the ad hoc network operates as a neural network. some of the nodes are designated as healing nodes that are identified during network training and are strategically located in the network coverage area. whenever one group of nodes loses connection with another a healing node may reposition itself to reconnect the two groups. thus, the network can maintain connectivity without constraining node movement.",2012-04-17,"The title of the patent is neural network-based mobility management for healing mobile ad hoc radio networks and its abstract is a self healing ad hoc communications network and method of training for and healing the network. the network includes wireless devices or nodes that include a neural network element and the ad hoc network operates as a neural network. some of the nodes are designated as healing nodes that are identified during network training and are strategically located in the network coverage area. whenever one group of nodes loses connection with another a healing node may reposition itself to reconnect the two groups. thus, the network can maintain connectivity without constraining node movement. dated 2012-04-17"
8160354,multi-stage image pattern recognizer,"an image-based pattern recognizer and a method and apparatus for making such a pattern recognizer are disclosed. by employing positional coding, the meaning of any feature present in an image can be defined implicitly in space. the pattern recognizer can be a neural network including a plurality of stages of observers. the observers are configured to cooperate to identify the presence of features in the input image and to recognize a pattern in the input image based on the features. each of the observers includes a plurality of neurons. the input image includes a plurality of units, and each of the observers is configured to generate a separate output set that includes zero or more coordinates of such units.",2012-04-17,"The title of the patent is multi-stage image pattern recognizer and its abstract is an image-based pattern recognizer and a method and apparatus for making such a pattern recognizer are disclosed. by employing positional coding, the meaning of any feature present in an image can be defined implicitly in space. the pattern recognizer can be a neural network including a plurality of stages of observers. the observers are configured to cooperate to identify the presence of features in the input image and to recognize a pattern in the input image based on the features. each of the observers includes a plurality of neurons. the input image includes a plurality of units, and each of the observers is configured to generate a separate output set that includes zero or more coordinates of such units. dated 2012-04-17"
8160362,combining online and offline recognizers in a handwriting recognition system,"described is a technology by which online recognition of handwritten input data is combined with offline recognition and processing to obtain a combined recognition result. in general, the combination improves overall recognition accuracy. in one aspect, online and offline recognition is separately performed to obtain online and offline character-level recognition scores for candidates (hypotheses). a statistical analysis-based combination algorithm, an adaboost algorithm, and/or a neural network-based combination may determine a combination function to combine the scores to produce a result set of one or more results. online and offline radical-level recognition may be performed. for example, a hmm recognizer may generate online radical scores used to build a radical graph, which is then rescored using the offline radical recognition scores. paths in the rescored graph are then searched to provide the combined recognition result, e.g., corresponding to the path with the highest score.",2012-04-17,"The title of the patent is combining online and offline recognizers in a handwriting recognition system and its abstract is described is a technology by which online recognition of handwritten input data is combined with offline recognition and processing to obtain a combined recognition result. in general, the combination improves overall recognition accuracy. in one aspect, online and offline recognition is separately performed to obtain online and offline character-level recognition scores for candidates (hypotheses). a statistical analysis-based combination algorithm, an adaboost algorithm, and/or a neural network-based combination may determine a combination function to combine the scores to produce a result set of one or more results. online and offline radical-level recognition may be performed. for example, a hmm recognizer may generate online radical scores used to build a radical graph, which is then rescored using the offline radical recognition scores. paths in the rescored graph are then searched to provide the combined recognition result, e.g., corresponding to the path with the highest score. dated 2012-04-17"
8160692,system and method for analyzing progress of labor and preterm labor,"systems and methods for monitoring uterus contraction activity and progress of labor. the system of the subject invention can comprises (1) a plurality of sensors; (2) an amplifying/filtering means; (3) a computing means; and (4) a graphical user interface. accurate clinical data, which can be extracted and provided to the user in real-time using the system of the invention, include without limitation, progress of labor, prediction and monitoring of preterm labor, and intrauterine pressure prediction. in a preferred embodiment, the system of the invention includes an intelligence means, such as a neural network system, to analyze and interpret clinical data for use in clinical diagnosis as well as delivery strategy.",2012-04-17,"The title of the patent is system and method for analyzing progress of labor and preterm labor and its abstract is systems and methods for monitoring uterus contraction activity and progress of labor. the system of the subject invention can comprises (1) a plurality of sensors; (2) an amplifying/filtering means; (3) a computing means; and (4) a graphical user interface. accurate clinical data, which can be extracted and provided to the user in real-time using the system of the invention, include without limitation, progress of labor, prediction and monitoring of preterm labor, and intrauterine pressure prediction. in a preferred embodiment, the system of the invention includes an intelligence means, such as a neural network system, to analyze and interpret clinical data for use in clinical diagnosis as well as delivery strategy. dated 2012-04-17"
8160847,hybrid multi-layer artificial immune system,"the hybrid artificial immune system consists of three main layers, including a solution application layer that interacts with the environment, a solution generation layer that solves combinatorial optimization problems and a modeling layer that analyzes problems and presents solution scenarios. the system solves evolutionary multi-objective optimization problems in network computing, robotics, artificial neural networks, protein network modeling, evolutionary systems and evolutionary hardware.",2012-04-17,"The title of the patent is hybrid multi-layer artificial immune system and its abstract is the hybrid artificial immune system consists of three main layers, including a solution application layer that interacts with the environment, a solution generation layer that solves combinatorial optimization problems and a modeling layer that analyzes problems and presents solution scenarios. the system solves evolutionary multi-objective optimization problems in network computing, robotics, artificial neural networks, protein network modeling, evolutionary systems and evolutionary hardware. dated 2012-04-17"
8160978,method for computer-aided control or regulation of a technical system,"a method for computer-aided control of any technical system is provided. the method includes two steps, the learning of the dynamic with historical data based on a recurrent neural network and a subsequent learning of an optimal regulation by coupling the recurrent neural network to a further neural network. the recurrent neural network has a hidden layer comprising a first and a second hidden state at a respective time point. the first hidden state is coupled to the second hidden state using a matrix to be learned. this allows a bottleneck structure to be created, in that the dimension of the first hidden state is smaller than the dimension of the second hidden state or vice versa. the autonomous dynamic is taken into account during the learning of the network, thereby improving the approximation capacity of the network. the technical system includes a gas turbine.",2012-04-17,"The title of the patent is method for computer-aided control or regulation of a technical system and its abstract is a method for computer-aided control of any technical system is provided. the method includes two steps, the learning of the dynamic with historical data based on a recurrent neural network and a subsequent learning of an optimal regulation by coupling the recurrent neural network to a further neural network. the recurrent neural network has a hidden layer comprising a first and a second hidden state at a respective time point. the first hidden state is coupled to the second hidden state using a matrix to be learned. this allows a bottleneck structure to be created, in that the dimension of the first hidden state is smaller than the dimension of the second hidden state or vice versa. the autonomous dynamic is taken into account during the learning of the network, thereby improving the approximation capacity of the network. the technical system includes a gas turbine. dated 2012-04-17"
8164742,photopolarimetric lidar dual-beam switching device and mueller matrix standoff detection system and method,"an optomechanical switching device, a control system, and a graphical user interface for a photopolarimetric lidar standoff detection that employs differential-absorption mueller matrix spectroscopy. an output train of alternate continuous-wave co2 laser beams [ . . . l1:l2 . . . ] is directed onto a suspect chemical-biological (cb) aerosol plume or the land mass it contaminates (s) vis-à-vis the osd, with l1 [l2] tuned on [detuned off] a resonant molecular absorption moiety of cb analyte. both incident beams and their backscattered radiances from s are polarization-modulated synchronously so as to produce gated temporal voltage waveforms (scattergrams) recorded on a focus at the receiver end of a sensor (lidar) system. all 16 elements of the mueller matrix (mij) of s are measured via digital or analog filtration of constituent frequency components in these running scattergram data streams (phase-sensitive detection). a collective set of normalized elements {δmi,j} (ratio to m11) susceptible to analyte, probed on-then-off its molecular absorption band, form a unique detection domain that is scrutinized; i.e., any mapping onto this domain by incoming lidar data—by means of a trained neural network pattern recognition system for instance—cues a standoff detection event.",2012-04-24,"The title of the patent is photopolarimetric lidar dual-beam switching device and mueller matrix standoff detection system and method and its abstract is an optomechanical switching device, a control system, and a graphical user interface for a photopolarimetric lidar standoff detection that employs differential-absorption mueller matrix spectroscopy. an output train of alternate continuous-wave co2 laser beams [ . . . l1:l2 . . . ] is directed onto a suspect chemical-biological (cb) aerosol plume or the land mass it contaminates (s) vis-à-vis the osd, with l1 [l2] tuned on [detuned off] a resonant molecular absorption moiety of cb analyte. both incident beams and their backscattered radiances from s are polarization-modulated synchronously so as to produce gated temporal voltage waveforms (scattergrams) recorded on a focus at the receiver end of a sensor (lidar) system. all 16 elements of the mueller matrix (mij) of s are measured via digital or analog filtration of constituent frequency components in these running scattergram data streams (phase-sensitive detection). a collective set of normalized elements {δmi,j} (ratio to m11) susceptible to analyte, probed on-then-off its molecular absorption band, form a unique detection domain that is scrutinized; i.e., any mapping onto this domain by incoming lidar data—by means of a trained neural network pattern recognition system for instance—cues a standoff detection event. dated 2012-04-24"
8165770,trailer oscillation detection and compensation method for a vehicle and trailer combination,"a system and method of controlling a vehicle with a trailer comprises determining the presence of a trailer, generating an oscillation signal indicative of trailer swaying relative to the vehicle, generating an initial weighted dynamic control signal for a vehicle dynamic control system in response to the oscillation signal, operating at least one vehicle dynamic system according to the dynamic control signal, and thereafter, iteratively generating a penalty function for the weighted dynamic control signal as a function of the oscillation signal response. a neural network with an associated trainer modifies the dynamic control signal as a function of trailer sway response.",2012-04-24,"The title of the patent is trailer oscillation detection and compensation method for a vehicle and trailer combination and its abstract is a system and method of controlling a vehicle with a trailer comprises determining the presence of a trailer, generating an oscillation signal indicative of trailer swaying relative to the vehicle, generating an initial weighted dynamic control signal for a vehicle dynamic control system in response to the oscillation signal, operating at least one vehicle dynamic system according to the dynamic control signal, and thereafter, iteratively generating a penalty function for the weighted dynamic control signal as a function of the oscillation signal response. a neural network with an associated trainer modifies the dynamic control signal as a function of trailer sway response. dated 2012-04-24"
8167803,system and method for bladder detection using harmonic imaging,"systems, methods, and ultrasound transceivers equipped and configured to execute harmonic analysis and extract harmonic information related to a targeted organ of a subject are described. the methods utilize neural network algorithms to establish improved segmentation accuracy of the targeted organ or structures within a region-of-interest. the neural network algorithms, refined for detection of the bladder and to ascertain the presence or absence of a uterus, is optimally applied to better segment and thus confer the capability to optimize measurement of bladder geometry, area, and volumes.",2012-05-01,"The title of the patent is system and method for bladder detection using harmonic imaging and its abstract is systems, methods, and ultrasound transceivers equipped and configured to execute harmonic analysis and extract harmonic information related to a targeted organ of a subject are described. the methods utilize neural network algorithms to establish improved segmentation accuracy of the targeted organ or structures within a region-of-interest. the neural network algorithms, refined for detection of the bladder and to ascertain the presence or absence of a uterus, is optimally applied to better segment and thus confer the capability to optimize measurement of bladder geometry, area, and volumes. dated 2012-05-01"
8180633,fast semantic extraction using a neural network architecture,"a system and method for semantic extraction using a neural network architecture includes indexing each word in an input sentence into a dictionary and using these indices to map each word to a d-dimensional vector (the features of which are learned). together with this, position information for a word of interest (the word to labeled) and a verb of interest (the verb that the semantic role is being predicted for) with respect to a given word are also used. these positions are integrated by employing a linear layer that is adapted to the input sentence. several linear transformations and squashing functions are then applied to output class probabilities for semantic role labels. all the weights for the whole architecture are trained by backpropagation.",2012-05-15,"The title of the patent is fast semantic extraction using a neural network architecture and its abstract is a system and method for semantic extraction using a neural network architecture includes indexing each word in an input sentence into a dictionary and using these indices to map each word to a d-dimensional vector (the features of which are learned). together with this, position information for a word of interest (the word to labeled) and a verb of interest (the verb that the semantic role is being predicted for) with respect to a given word are also used. these positions are integrated by employing a linear layer that is adapted to the input sentence. several linear transformations and squashing functions are then applied to output class probabilities for semantic role labels. all the weights for the whole architecture are trained by backpropagation. dated 2012-05-15"
8180754,semantic neural network for aggregating query searches,"a system, method and computer program product for implementation of a aggregate neural semantic network, which stores the relationships and semantic connections between the key search words for each user. the aggregate neural semantic network processes the search results produced by a standard search engine such as, for example, google or yahoo!. the set of hits produced by the standard search engine is processed by the aggregate neural semantic network, which selects the hits that are relevant to a particular user based on the previous search queries made by the user. the aggregate neural semantic network can also use the connections between the terms (i.e., key words) that are most frequently used by all of the previous aggregate neural semantic network users. the aggregate neural semantic network is constantly updating and self-teaching. the more user queries are processed by the aggregate neural semantic network, the more comprehensive processing of search engine outputs is provided by the aggregate neural semantic network to the subsequent user queries.",2012-05-15,"The title of the patent is semantic neural network for aggregating query searches and its abstract is a system, method and computer program product for implementation of a aggregate neural semantic network, which stores the relationships and semantic connections between the key search words for each user. the aggregate neural semantic network processes the search results produced by a standard search engine such as, for example, google or yahoo!. the set of hits produced by the standard search engine is processed by the aggregate neural semantic network, which selects the hits that are relevant to a particular user based on the previous search queries made by the user. the aggregate neural semantic network can also use the connections between the terms (i.e., key words) that are most frequently used by all of the previous aggregate neural semantic network users. the aggregate neural semantic network is constantly updating and self-teaching. the more user queries are processed by the aggregate neural semantic network, the more comprehensive processing of search engine outputs is provided by the aggregate neural semantic network to the subsequent user queries. dated 2012-05-15"
8181907,wing-drive mechanism and vehicle employing same,"a wing-drive mechanism is described that permits, with proper control, movement of a wing about multiple wing trajectories. the wing-drive is capable of independent movement about three rotational degrees of movement; movement about a flap axis is independent of movement about a yaw axis, and both are independent of changes in the pitch of the wing. methods of controlling the wing-drive mechanism to affect a desired wing trajectory include the use of a non-linear automated controller that generates input signals to the wing-drive mechanism by comparing actual and desired wing trajectories in real time. specification of wing trajectories is preferably also accomplished in real time using an automated trajectory specification system, which can include a fuzzy logic processor or a neural network.",2012-05-22,"The title of the patent is wing-drive mechanism and vehicle employing same and its abstract is a wing-drive mechanism is described that permits, with proper control, movement of a wing about multiple wing trajectories. the wing-drive is capable of independent movement about three rotational degrees of movement; movement about a flap axis is independent of movement about a yaw axis, and both are independent of changes in the pitch of the wing. methods of controlling the wing-drive mechanism to affect a desired wing trajectory include the use of a non-linear automated controller that generates input signals to the wing-drive mechanism by comparing actual and desired wing trajectories in real time. specification of wing trajectories is preferably also accomplished in real time using an automated trajectory specification system, which can include a fuzzy logic processor or a neural network. dated 2012-05-22"
8182424,diary-free calorimeter,"an indirect calorimeter estimates nutritional caloric intake by periodically monitoring weight and sensing physical exercise (i.e., physiological data and/or motion data related to physical exertion), which can then be used in a calorimetry model derived from regression analysis of a population (e.g., linear regression, feed-forward neural network, gaussian process, boosted regression tree, etc.). a strap-on user device for tracking exercise can detect one or more of heart rate, body temperature, skin resistance, motion/acceleration sensing (e.g., pedometer, accelerometer), velocity sensing (e.g., global positioning system (gps)), and an intelligent, integrated exercise machine (e.g., treadmill, exercise bike, etc.). to gain further fidelity, the user can fine-tune the estimate by undergoing a journal-based routine for a relatively short period of time or clinical calorimetry measurement (e.g., respiratory calorimeter), thereby providing a baseline for resting or exercising metabolic rate.",2012-05-22,"The title of the patent is diary-free calorimeter and its abstract is an indirect calorimeter estimates nutritional caloric intake by periodically monitoring weight and sensing physical exercise (i.e., physiological data and/or motion data related to physical exertion), which can then be used in a calorimetry model derived from regression analysis of a population (e.g., linear regression, feed-forward neural network, gaussian process, boosted regression tree, etc.). a strap-on user device for tracking exercise can detect one or more of heart rate, body temperature, skin resistance, motion/acceleration sensing (e.g., pedometer, accelerometer), velocity sensing (e.g., global positioning system (gps)), and an intelligent, integrated exercise machine (e.g., treadmill, exercise bike, etc.). to gain further fidelity, the user can fine-tune the estimate by undergoing a journal-based routine for a relatively short period of time or clinical calorimetry measurement (e.g., respiratory calorimeter), thereby providing a baseline for resting or exercising metabolic rate. dated 2012-05-22"
8183785,method of controlling a lighting system based on a target light distribution,"the invention relates to a method of controlling a lighting system with multiple controllable light sources 3a, 3b and a system therefor. according to a first aspect, influence data of the lighting system are obtained, which data represent the effect of one or more of the light sources 3a, 3b on the illumination of one or more sections of an illuminated environment. in an optimization method, sets of control commands are continuously determined, a predicted light distribution for these control commands is determined from the influence data, and a colorimetric difference between the predicted light distribution and a target light distribution is determined. a plurality of adjustment steps are performed to minimize the colorimetric difference. according to a second aspect, a neural network is trained with the influence data and a set of control commands for controlling the lighting system is determined with the use of the neural network.",2012-05-22,"The title of the patent is method of controlling a lighting system based on a target light distribution and its abstract is the invention relates to a method of controlling a lighting system with multiple controllable light sources 3a, 3b and a system therefor. according to a first aspect, influence data of the lighting system are obtained, which data represent the effect of one or more of the light sources 3a, 3b on the illumination of one or more sections of an illuminated environment. in an optimization method, sets of control commands are continuously determined, a predicted light distribution for these control commands is determined from the influence data, and a colorimetric difference between the predicted light distribution and a target light distribution is determined. a plurality of adjustment steps are performed to minimize the colorimetric difference. according to a second aspect, a neural network is trained with the influence data and a set of control commands for controlling the lighting system is determined with the use of the neural network. dated 2012-05-22"
8185909,predictive database resource utilization and load balancing using neural network model,"a preemptive neural network database load balancer configured to observe, learn and predict the resource utilization that given incoming tasks utilize. allows for efficient execution and use of system resources. preemptively assigns incoming tasks to particular servers based on predicted cpu, memory, disk and network utilization for the incoming tasks. direct write-based tasks to a master server and utilizes slave servers to handle read-based tasks. read-base tasks are analyzed with a neural network to learn and predict the amount of resources that tasks will utilize. tasks are assigned to a database server based on the predicted utilization of the incoming task and the predicted and observed resource utilization on each database server. the predicted resource utilization may be updated over time as the number of records, lookups, images, pdfs, fields, blobs and width of fields in the database change over time.",2012-05-22,"The title of the patent is predictive database resource utilization and load balancing using neural network model and its abstract is a preemptive neural network database load balancer configured to observe, learn and predict the resource utilization that given incoming tasks utilize. allows for efficient execution and use of system resources. preemptively assigns incoming tasks to particular servers based on predicted cpu, memory, disk and network utilization for the incoming tasks. direct write-based tasks to a master server and utilizes slave servers to handle read-based tasks. read-base tasks are analyzed with a neural network to learn and predict the amount of resources that tasks will utilize. tasks are assigned to a database server based on the predicted utilization of the incoming task and the predicted and observed resource utilization on each database server. the predicted resource utilization may be updated over time as the number of records, lookups, images, pdfs, fields, blobs and width of fields in the database change over time. dated 2012-05-22"
8190542,"neural network, a device for processing information, a method of operating a neural network, a program element and a computer-readable medium",a neural network includes neurons and wires adapted for connecting the neurons. some of the wires comprise input connections and exactly one output connection and/or a part of the wires comprise exactly one input connection and output connections. neurons are hierarchically arranged in groups. a lower group of neurons recognizes a pattern of information input to the neurons of this lower group. a higher group of neurons recognizes higher level patterns. a strength value is associated with a connection between different neurons. the strength value of a particular connection is indicative of a likelihood that information which is input to the neurons propagates via the particular connection. the strength value of each connection is modifiable based on an amount of traffic of information which is input to the neurons and which propagates via the particular connection and/or is modifiable based on a strength modification impulse.,2012-05-29,"The title of the patent is neural network, a device for processing information, a method of operating a neural network, a program element and a computer-readable medium and its abstract is a neural network includes neurons and wires adapted for connecting the neurons. some of the wires comprise input connections and exactly one output connection and/or a part of the wires comprise exactly one input connection and output connections. neurons are hierarchically arranged in groups. a lower group of neurons recognizes a pattern of information input to the neurons of this lower group. a higher group of neurons recognizes higher level patterns. a strength value is associated with a connection between different neurons. the strength value of a particular connection is indicative of a likelihood that information which is input to the neurons propagates via the particular connection. the strength value of each connection is modifiable based on an amount of traffic of information which is input to the neurons and which propagates via the particular connection and/or is modifiable based on a strength modification impulse. dated 2012-05-29"
8195598,method of and system for hierarchical human/crowd behavior detection,the present invention is directed to a computer automated method of selectively identifying a user specified behavior of a crowd. the method comprises receiving video data but can also include audio data and sensor data. the video data contains images a crowd. the video data is processed to extract hierarchical human and crowd features. the detected crowd features are processed to detect a selectable crowd behavior. the selected crowd behavior detected is specified by a configurable behavior rule. human detection is provided by a hybrid human detector algorithm which can include adaboost or convolutional neural network. crowd features are detected using textual analysis techniques. the configurable crowd behavior for detection can be defined by crowd behavioral language.,2012-06-05,The title of the patent is method of and system for hierarchical human/crowd behavior detection and its abstract is the present invention is directed to a computer automated method of selectively identifying a user specified behavior of a crowd. the method comprises receiving video data but can also include audio data and sensor data. the video data contains images a crowd. the video data is processed to extract hierarchical human and crowd features. the detected crowd features are processed to detect a selectable crowd behavior. the selected crowd behavior detected is specified by a configurable behavior rule. human detection is provided by a hybrid human detector algorithm which can include adaboost or convolutional neural network. crowd features are detected using textual analysis techniques. the configurable crowd behavior for detection can be defined by crowd behavioral language. dated 2012-06-05
8200593,method for efficiently simulating the information processing in cells and tissues of the nervous system with a temporal series compressed encoding neural network,"a neural network simulation represents components of neurons by finite state machines, called sectors, implemented using look-up tables. each sector has an internal state represented by a compressed history of data input to the sector and is factorized into distinct historical time intervals of the data input. the compressed history of data input to the sector may be computed by compressing the data input to the sector during a time interval, storing the compressed history of data input to the sector in memory, and computing from the stored compressed history of data input to the sector the data output from the sector.",2012-06-12,"The title of the patent is method for efficiently simulating the information processing in cells and tissues of the nervous system with a temporal series compressed encoding neural network and its abstract is a neural network simulation represents components of neurons by finite state machines, called sectors, implemented using look-up tables. each sector has an internal state represented by a compressed history of data input to the sector and is factorized into distinct historical time intervals of the data input. the compressed history of data input to the sector may be computed by compressing the data input to the sector during a time interval, storing the compressed history of data input to the sector in memory, and computing from the stored compressed history of data input to the sector the data output from the sector. dated 2012-06-12"
8200642,system and method for managing electronic documents in a litigation context,"a system and method for production of analyzing electronic documents includes document acquisition software; a database, comprising a document table; a document parser; a categorization schema; and a document processor operatively in communication with the database and the categorization schema. document acquisition software operatively resident in a first computer acquires an electronic document which is then parsed by a document parser operatively resident in a second computer to create a set of parsed data related to the acquired document. a predetermined set of data describing the parsed document, comprising the created parsed data, are stored into a document table of a database accessible to the second computer. a non-neural network process is used to process the created parsed data in a document processor operatively resident in a third computer according to a categorization schema to create an association between the acquired document and the categorization schema.",2012-06-12,"The title of the patent is system and method for managing electronic documents in a litigation context and its abstract is a system and method for production of analyzing electronic documents includes document acquisition software; a database, comprising a document table; a document parser; a categorization schema; and a document processor operatively in communication with the database and the categorization schema. document acquisition software operatively resident in a first computer acquires an electronic document which is then parsed by a document parser operatively resident in a second computer to create a set of parsed data related to the acquired document. a predetermined set of data describing the parsed document, comprising the created parsed data, are stored into a document table of a database accessible to the second computer. a non-neural network process is used to process the created parsed data in a document processor operatively resident in a third computer according to a categorization schema to create an association between the acquired document and the categorization schema. dated 2012-06-12"
8204292,feature based neural network regression for feature suppression,"a method of obtaining one or more components from an image may include normalizing and pre-processing the image to obtain a processed image. features may be extracted from the processed image. neural-network-based regression may then be performed on the set of extracted features to predict the one or more components. these techniques may be applied, for example, to the problem of extracting and removing bone components from radiographic images, which may be thoracic (lung) images.",2012-06-19,"The title of the patent is feature based neural network regression for feature suppression and its abstract is a method of obtaining one or more components from an image may include normalizing and pre-processing the image to obtain a processed image. features may be extracted from the processed image. neural-network-based regression may then be performed on the set of extracted features to predict the one or more components. these techniques may be applied, for example, to the problem of extracting and removing bone components from radiographic images, which may be thoracic (lung) images. dated 2012-06-19"
8208170,system and method for printing target colors with process colors utilizing parallel feedforward neural networks,a system and method for printing target colors includes a print-engine interface and a neural network component. the print-engine interface is in operative communication with a print engine of a printing system. the neural network component is calibrated to the print engine for printing a target color on a substrate. the neural network is in operative communication with the print-engine interface and communicates a parameter associated with printing the target color on the substrate utilizing the print engine.,2012-06-26,The title of the patent is system and method for printing target colors with process colors utilizing parallel feedforward neural networks and its abstract is a system and method for printing target colors includes a print-engine interface and a neural network component. the print-engine interface is in operative communication with a print engine of a printing system. the neural network component is calibrated to the print engine for printing a target color on a substrate. the neural network is in operative communication with the print-engine interface and communicates a parameter associated with printing the target color on the substrate utilizing the print engine. dated 2012-06-26
8209080,system for determining most probable cause of a problem in a plant,"a system for determining a most probable cause or causes of a problem in a plant is disclosed. the system includes a plant, the plant having a plurality of subsystems that contribute to the operation of the plant, the plurality of subsystems having operating functions that produce operational signals. a plurality of sensors that are operable to detect the operational signals from the plurality of subsystems and transmit data related to the signals is also provided. an advisory system is disclosed that receives an input, the input being in the form of data from the plurality of sensors, possible input root causes of the problem, possible input symptoms of the problem and/or combinations thereof. the advisory system has an autoencoder in the form of a recurrent neural network. the recurrent neural network has sparse connectivity in a plurality of nodes, and the autoencoder is also operable to receive the input and perform multiple iterations of computations at each of the plurality of nodes as a function of the input and provide an output. the output can be in the form of possible output causes of the problem, possible output symptoms of the problem and/or combinations thereof.",2012-06-26,"The title of the patent is system for determining most probable cause of a problem in a plant and its abstract is a system for determining a most probable cause or causes of a problem in a plant is disclosed. the system includes a plant, the plant having a plurality of subsystems that contribute to the operation of the plant, the plurality of subsystems having operating functions that produce operational signals. a plurality of sensors that are operable to detect the operational signals from the plurality of subsystems and transmit data related to the signals is also provided. an advisory system is disclosed that receives an input, the input being in the form of data from the plurality of sensors, possible input root causes of the problem, possible input symptoms of the problem and/or combinations thereof. the advisory system has an autoencoder in the form of a recurrent neural network. the recurrent neural network has sparse connectivity in a plurality of nodes, and the autoencoder is also operable to receive the input and perform multiple iterations of computations at each of the plurality of nodes as a function of the input and provide an output. the output can be in the form of possible output causes of the problem, possible output symptoms of the problem and/or combinations thereof. dated 2012-06-26"
8214312,system and method for calibrating radio frequency power of communication devices,"a radio frequency (rf) calibrating system and a method for calibrating rf power of communication devices are provided. the method collects rf signals transmitted from the communication devices, and generates a group of training samples by retrieving measurement data from the rf signals. the method further constructs a neural network according to the group of training samples, calibrate rf power of the communication devices using the neural network, and generate corresponding calibration results of the rf power. in addition, the method generates a frequency spectrum of the rf power according to the calibration results of the rf power, and displays the frequency spectrum on a display device of the rf calibrating system.",2012-07-03,"The title of the patent is system and method for calibrating radio frequency power of communication devices and its abstract is a radio frequency (rf) calibrating system and a method for calibrating rf power of communication devices are provided. the method collects rf signals transmitted from the communication devices, and generates a group of training samples by retrieving measurement data from the rf signals. the method further constructs a neural network according to the group of training samples, calibrate rf power of the communication devices using the neural network, and generate corresponding calibration results of the rf power. in addition, the method generates a frequency spectrum of the rf power according to the calibration results of the rf power, and displays the frequency spectrum on a display device of the rf calibrating system. dated 2012-07-03"
8214313,turn rate calculation,a data store includes item configuration data for a plurality of configurations of an item and inventory mix rate data. a computing device is configured to generate a matrix that combines the item configuration data with the inventory mix rate data such that cells in the matrix include an indication of an inventory mix rate related to an item feature; use the matrix to develop a plurality of neural network inventory turn rate models; and use the turn rate models to associate a value with each configuration in the plurality of configurations.,2012-07-03,The title of the patent is turn rate calculation and its abstract is a data store includes item configuration data for a plurality of configurations of an item and inventory mix rate data. a computing device is configured to generate a matrix that combines the item configuration data with the inventory mix rate data such that cells in the matrix include an indication of an inventory mix rate related to an item feature; use the matrix to develop a plurality of neural network inventory turn rate models; and use the turn rate models to associate a value with each configuration in the plurality of configurations. dated 2012-07-03
8218865,constructing a color transform using a neural network for colors outside the spectrum locus,"a color management module which provides color values in a destination color space by interpolation of a lut that maps from color values in a source color space to corresponding color values in the destination color space. the lut includes cells corresponding to color values within a spectrum locus and color values outside the spectrum locus. the lut is populated differently for cells within the spectrum locus and for those outside the spectrum locus. for cells within the spectrum locus, color values are calculated using a color transform constructed based on device profiles for the source device and for the destination device, and corresponding cells of the lut are populated based on the calculated values. for cells outside of the spectrum locus, an artificial neural network is trained using the calculated color values, and the corresponding cells are populated based on outputs of the trained neural network.",2012-07-10,"The title of the patent is constructing a color transform using a neural network for colors outside the spectrum locus and its abstract is a color management module which provides color values in a destination color space by interpolation of a lut that maps from color values in a source color space to corresponding color values in the destination color space. the lut includes cells corresponding to color values within a spectrum locus and color values outside the spectrum locus. the lut is populated differently for cells within the spectrum locus and for those outside the spectrum locus. for cells within the spectrum locus, color values are calculated using a color transform constructed based on device profiles for the source device and for the destination device, and corresponding cells of the lut are populated based on the calculated values. for cells outside of the spectrum locus, an artificial neural network is trained using the calculated color values, and the corresponding cells are populated based on outputs of the trained neural network. dated 2012-07-10"
8229209,neural network based pattern recognizer,"an image-based pattern recognizer and a method and apparatus for making such a pattern recognizer are disclosed. by employing positional coding, the meaning of any feature present in an image can be defined implicitly in space. the pattern recognizer can be a neural network including a plurality of stages of observers. the observers are configured to cooperate to identify the presence of features in the input image and to recognize a pattern in the input image based on the features. each of the observers includes a plurality of neurons. the input image includes a plurality of units, and each of the observers is configured to generate a separate output set that includes zero or more coordinates of such units.",2012-07-24,"The title of the patent is neural network based pattern recognizer and its abstract is an image-based pattern recognizer and a method and apparatus for making such a pattern recognizer are disclosed. by employing positional coding, the meaning of any feature present in an image can be defined implicitly in space. the pattern recognizer can be a neural network including a plurality of stages of observers. the observers are configured to cooperate to identify the presence of features in the input image and to recognize a pattern in the input image based on the features. each of the observers includes a plurality of neurons. the input image includes a plurality of units, and each of the observers is configured to generate a separate output set that includes zero or more coordinates of such units. dated 2012-07-24"
8229880,evaluation of acid fracturing treatments in an oilfield,"the invention relates to a method for performing acid fracturing operations of an oilfield. the method includes obtaining a plurality of historical data of acid fracturing treatments of the oilfield, generating a neural network based on the plurality of historical data, identifying a stimulation parameter, in the neural network, associated with optimal performance of the acid fracturing treatments, and establishing a best practice procedure for performing the acid fracturing operations based on the stimulation parameter.",2012-07-24,"The title of the patent is evaluation of acid fracturing treatments in an oilfield and its abstract is the invention relates to a method for performing acid fracturing operations of an oilfield. the method includes obtaining a plurality of historical data of acid fracturing treatments of the oilfield, generating a neural network based on the plurality of historical data, identifying a stimulation parameter, in the neural network, associated with optimal performance of the acid fracturing treatments, and establishing a best practice procedure for performing the acid fracturing operations based on the stimulation parameter. dated 2012-07-24"
8234227,timestamp neural network,"a timestamp neural network comprised of sensor elements, internal elements, and motor elements is responsive to timestamps. sensor elements transform a wide variety of signals into events that trigger the updating of timestamps. internal elements are responsive to timestamps. motor elements convert timestamps into useful output signals. a real time video pattern recognition system is implemented.",2012-07-31,"The title of the patent is timestamp neural network and its abstract is a timestamp neural network comprised of sensor elements, internal elements, and motor elements is responsive to timestamps. sensor elements transform a wide variety of signals into events that trigger the updating of timestamps. internal elements are responsive to timestamps. motor elements convert timestamps into useful output signals. a real time video pattern recognition system is implemented. dated 2012-07-31"
8239498,system and method for facilitating the implementation of changes to the configuration of resources in an enterprise,"system and method for facilitating the implementation of changes to the configuration of resources in an enterprise. embodiments of the present invention facilitate the use of historical information about an enterprise's it configuration to evaluate the risk and impact of proposed changes. risk can be evaluated using a success history for an organization within the enterprise that is responsible for the proposed change, or by applying a neural network to the historical data to detect recognizable patterns in the historical data. a risk evaluation can take into account sensitivity of the change to dates assigned on a change calendar, based on sensitivity dates gathered from the historical data. historical data can be maintained and provided by a configuration database.",2012-08-07,"The title of the patent is system and method for facilitating the implementation of changes to the configuration of resources in an enterprise and its abstract is system and method for facilitating the implementation of changes to the configuration of resources in an enterprise. embodiments of the present invention facilitate the use of historical information about an enterprise's it configuration to evaluate the risk and impact of proposed changes. risk can be evaluated using a success history for an organization within the enterprise that is responsible for the proposed change, or by applying a neural network to the historical data to detect recognizable patterns in the historical data. a risk evaluation can take into account sensitivity of the change to dates assigned on a change calendar, based on sensitivity dates gathered from the historical data. historical data can be maintained and provided by a configuration database. dated 2012-08-07"
8254670,self-learning object detection and classification systems and methods,"a method of object classification based upon fusion of a remote sensing system and a natural imaging system is provided. the method includes detecting an object using the remote sensing system. an angle of view of a video camera of the natural imaging system is varied. an image including the object is generated using the natural imaging system. the natural imaging system may zoom in on the object. the image represented in either pixel or transformed space is compared to a plurality of templates via a competition based neural network learning algorithm. each template has an associated label determined statistically. the template with a closest match to the image is determined. the image may be assigned the label associated with the relative location of the object, the relative speed of the object, and the label of the template determined statistically to be the closest match to the image.",2012-08-28,"The title of the patent is self-learning object detection and classification systems and methods and its abstract is a method of object classification based upon fusion of a remote sensing system and a natural imaging system is provided. the method includes detecting an object using the remote sensing system. an angle of view of a video camera of the natural imaging system is varied. an image including the object is generated using the natural imaging system. the natural imaging system may zoom in on the object. the image represented in either pixel or transformed space is compared to a plurality of templates via a competition based neural network learning algorithm. each template has an associated label determined statistically. the template with a closest match to the image is determined. the image may be assigned the label associated with the relative location of the object, the relative speed of the object, and the label of the template determined statistically to be the closest match to the image. dated 2012-08-28"
8255119,vehicle body slip angle-estimating device and method and engine control unit,"a vehicle body slip angle-estimating device which, in estimating a vehicle body slip angle with an algorithm using a nonlinear model, is capable of accurately estimating a vehicle body slip angle irrespective of whether the frequency of occurrence of a state during traveling of the vehicle. a basic value-calculating section calculates a basic value of a vehicle body slip angle with an algorithm using a neural network model. a turning state-determining section determines whether the vehicle is in a predetermined limit turning traveling state. a correction value-calculating section calculates a correction value with an algorithm using a predetermined linear model when the vehicle is in the predetermined state. in the other cases, the correction value is set to 0. a straight traveling-determining section sets the angle to the sum of the basic value and the correction value when the vehicle is in a turning traveling state.",2012-08-28,"The title of the patent is vehicle body slip angle-estimating device and method and engine control unit and its abstract is a vehicle body slip angle-estimating device which, in estimating a vehicle body slip angle with an algorithm using a nonlinear model, is capable of accurately estimating a vehicle body slip angle irrespective of whether the frequency of occurrence of a state during traveling of the vehicle. a basic value-calculating section calculates a basic value of a vehicle body slip angle with an algorithm using a neural network model. a turning state-determining section determines whether the vehicle is in a predetermined limit turning traveling state. a correction value-calculating section calculates a correction value with an algorithm using a predetermined linear model when the vehicle is in the predetermined state. in the other cases, the correction value is set to 0. a straight traveling-determining section sets the angle to the sum of the basic value and the correction value when the vehicle is in a turning traveling state. dated 2012-08-28"
8260428,method and system for training a visual prosthesis,a method for training a visual prosthesis includes presenting a non-visual reference stimulus corresponding to a reference image to a visual prosthesis patient. training data sets are generated by presenting a series of stimulation patterns to the patient through the visual prosthesis. each stimulation pattern in the series is determined at least in part on a received user perception input and a fitness function optimization algorithm. the presented stimulation patterns and the user perception inputs are stored and presented to a neural network off-line to determine a vision solution.,2012-09-04,The title of the patent is method and system for training a visual prosthesis and its abstract is a method for training a visual prosthesis includes presenting a non-visual reference stimulus corresponding to a reference image to a visual prosthesis patient. training data sets are generated by presenting a series of stimulation patterns to the patient through the visual prosthesis. each stimulation pattern in the series is determined at least in part on a received user perception input and a fitness function optimization algorithm. the presented stimulation patterns and the user perception inputs are stored and presented to a neural network off-line to determine a vision solution. dated 2012-09-04
8260732,method for identifying hammerstein models,"the computerized method for identifying hammerstein models is a method in which the linear dynamic part is modeled by a space-state model and the static nonlinear part is modeled using a radial basis function neural network (rbfnn). accurate identification of a hammerstein model requires that output error between the actual and estimated systems be minimized. thus, the problem of identification is an optimization problem. a hybrid algorithm, based on least mean square (lms) principles and the subspace identification method (sim) is developed for the identification of the hammerstein model. lms is a gradient-based optimization algorithm that searches for optimal solutions in the negative direction of the gradient of a cost index. in the method, lms is used for estimating the parameters of the rbfnn. for estimation of state-space matrices, the n4sid algorithm for subspace identification is used.",2012-09-04,"The title of the patent is method for identifying hammerstein models and its abstract is the computerized method for identifying hammerstein models is a method in which the linear dynamic part is modeled by a space-state model and the static nonlinear part is modeled using a radial basis function neural network (rbfnn). accurate identification of a hammerstein model requires that output error between the actual and estimated systems be minimized. thus, the problem of identification is an optimization problem. a hybrid algorithm, based on least mean square (lms) principles and the subspace identification method (sim) is developed for the identification of the hammerstein model. lms is a gradient-based optimization algorithm that searches for optimal solutions in the negative direction of the gradient of a cost index. in the method, lms is used for estimating the parameters of the rbfnn. for estimation of state-space matrices, the n4sid algorithm for subspace identification is used. dated 2012-09-04"
8260733,neural network system and method for controlling information output based on user feedback,"a system and method for controlling information output based on user feedback about the information is provided. at least one neural network module selects one or more of a plurality of objects to receive information from a plurality of information sources based at least in part on a plurality of inputs and a plurality of weight values during an epoch. the information sources may include electronic mail providers, chat participants, or page links. recipients of the objects provide feedback about the information during an epoch. at the conclusion of an epoch, the neural network takes the feedback and generates a rating value for each of the plurality of objects. based on the rating value and the selections made, the neural network redetermines the weight values within the network. the neural network then selects the objects to receive information during a subsequent epoch using the redetermined weight values and the inputs for that subsequent epoch.",2012-09-04,"The title of the patent is neural network system and method for controlling information output based on user feedback and its abstract is a system and method for controlling information output based on user feedback about the information is provided. at least one neural network module selects one or more of a plurality of objects to receive information from a plurality of information sources based at least in part on a plurality of inputs and a plurality of weight values during an epoch. the information sources may include electronic mail providers, chat participants, or page links. recipients of the objects provide feedback about the information during an epoch. at the conclusion of an epoch, the neural network takes the feedback and generates a rating value for each of the plurality of objects. based on the rating value and the selections made, the neural network redetermines the weight values within the network. the neural network then selects the objects to receive information during a subsequent epoch using the redetermined weight values and the inputs for that subsequent epoch. dated 2012-09-04"
8266083,large scale manifold transduction that predicts class labels with a neural network and uses a mean of the class labels,"a method for training a learning machine for use in discriminative classification and regression includes randomly selecting, in a first computer process, an unclassified datapoint associated with a phenomenon of interest; determining, in a second computer process, a set of datapoints associated with the phenomenon of interest that is likely to be in the same class as the selected unclassified datapoint; predicting, in a third computer process, a class label for the selected unclassified datapoint in a third computer process; predicting a class label for the set of datapoints in a fourth computer process; combining the predicted class labels in a fifth computer process, to predict a composite class label that describes the selected unclassified datapoint and the set of datapoints; and using the combined class label to adjust at least one parameter of the learning machine in a sixth computer process.",2012-09-11,"The title of the patent is large scale manifold transduction that predicts class labels with a neural network and uses a mean of the class labels and its abstract is a method for training a learning machine for use in discriminative classification and regression includes randomly selecting, in a first computer process, an unclassified datapoint associated with a phenomenon of interest; determining, in a second computer process, a set of datapoints associated with the phenomenon of interest that is likely to be in the same class as the selected unclassified datapoint; predicting, in a third computer process, a class label for the selected unclassified datapoint in a third computer process; predicting a class label for the set of datapoints in a fourth computer process; combining the predicted class labels in a fifth computer process, to predict a composite class label that describes the selected unclassified datapoint and the set of datapoints; and using the combined class label to adjust at least one parameter of the learning machine in a sixth computer process. dated 2012-09-11"
8275451,maternal-fetal monitoring system,"a maternal-fetal monitoring system for use during all stages of pregnancy, including antepartum and intrapartum stages. the maternal-fetal monitoring system of the subject invention comprises (1) a set of sensors; (2) an amplifying/filtering means; (3) a computing means; and (4) a graphical user interface. accurate clinical data, which can be extracted and provided to the user in real-time using the system of the invention, include without limitation, maternal electrocardiogram (ecg) signals, maternal uterine activity signals (ehg), maternal heart rate, fetal ecg signals, and fetal heart rate. in a preferred embodiment, the maternal-fetal monitoring system of the invention includes an intelligence means, such as a neural network system, to analyze and interpret clinical data for use in clinical diagnosis antepartum, intrapartum and postpartum, as well as delivery strategy.",2012-09-25,"The title of the patent is maternal-fetal monitoring system and its abstract is a maternal-fetal monitoring system for use during all stages of pregnancy, including antepartum and intrapartum stages. the maternal-fetal monitoring system of the subject invention comprises (1) a set of sensors; (2) an amplifying/filtering means; (3) a computing means; and (4) a graphical user interface. accurate clinical data, which can be extracted and provided to the user in real-time using the system of the invention, include without limitation, maternal electrocardiogram (ecg) signals, maternal uterine activity signals (ehg), maternal heart rate, fetal ecg signals, and fetal heart rate. in a preferred embodiment, the maternal-fetal monitoring system of the invention includes an intelligence means, such as a neural network system, to analyze and interpret clinical data for use in clinical diagnosis antepartum, intrapartum and postpartum, as well as delivery strategy. dated 2012-09-25"
8285531,neural net for use in drilling simulation,"a method of optimizing a drilling tool assembly including inputting well data into an optimization system, the optimization system having an experience data set and an artificial neural network. the method further including comparing the well data to the experience data set and developing an initial drilling tool assembly based on the comparing the well data to the experience data, wherein the drilling tool assembly is developed using the artificial neural network. additionally, the method including simulating the initial drilling tool assembly in the optimization system and creating result data in the optimization system based on the simulating.",2012-10-09,"The title of the patent is neural net for use in drilling simulation and its abstract is a method of optimizing a drilling tool assembly including inputting well data into an optimization system, the optimization system having an experience data set and an artificial neural network. the method further including comparing the well data to the experience data set and developing an initial drilling tool assembly based on the comparing the well data to the experience data, wherein the drilling tool assembly is developed using the artificial neural network. additionally, the method including simulating the initial drilling tool assembly in the optimization system and creating result data in the optimization system based on the simulating. dated 2012-10-09"
8285659,aircraft system modeling error and control error,"a method for modeling error-driven adaptive control of an aircraft. normal aircraft plant dynamics is modeled, using an original plant description in which a controller responds to a tracking error e(k) to drive the component to a normal reference value according to an asymptote curve. where the system senses that (1) at least one aircraft plant component is experiencing an excursion and (2) the return of this component value toward its reference value is not proceeding according to the expected controller characteristics, neural network (nn) modeling of aircraft plant operation may be changed. however, if (1) is satisfied but the error component is returning toward its reference value according to expected controller characteristics, the nn will continue to model operation of the aircraft plant according to an original description.",2012-10-09,"The title of the patent is aircraft system modeling error and control error and its abstract is a method for modeling error-driven adaptive control of an aircraft. normal aircraft plant dynamics is modeled, using an original plant description in which a controller responds to a tracking error e(k) to drive the component to a normal reference value according to an asymptote curve. where the system senses that (1) at least one aircraft plant component is experiencing an excursion and (2) the return of this component value toward its reference value is not proceeding according to the expected controller characteristics, neural network (nn) modeling of aircraft plant operation may be changed. however, if (1) is satisfied but the error component is returning toward its reference value according to expected controller characteristics, the nn will continue to model operation of the aircraft plant according to an original description. dated 2012-10-09"
8290250,method and apparatus for creating a pattern recognizer,"an image-based pattern recognizer and a method and apparatus for making such a pattern recognizer are disclosed. by employing positional coding, the meaning of any feature present in an image can be defined implicitly in space. the pattern recognizer can be a neural network including a plurality of stages of observers. the observers are configured to cooperate to identify the presence of features in the input image and to recognize a pattern in the input image based on the features. each of the observers includes a plurality of neurons. the input image includes a plurality of units, and each of the observers is configured to generate a separate output set that includes zero or more coordinates of such units.",2012-10-16,"The title of the patent is method and apparatus for creating a pattern recognizer and its abstract is an image-based pattern recognizer and a method and apparatus for making such a pattern recognizer are disclosed. by employing positional coding, the meaning of any feature present in an image can be defined implicitly in space. the pattern recognizer can be a neural network including a plurality of stages of observers. the observers are configured to cooperate to identify the presence of features in the input image and to recognize a pattern in the input image based on the features. each of the observers includes a plurality of neurons. the input image includes a plurality of units, and each of the observers is configured to generate a separate output set that includes zero or more coordinates of such units. dated 2012-10-16"
8296250,comprehensive identity protection system,"a system and method for protecting identity fraud are disclosed. a system includes a detection subsystem to identify applications and/or accounts at risk of identity fraud, and a disposition subsystem to process data provided by the detection system and to determine whether identity fraud exists in the applications and/or accounts. according to an implementation, one or more neural network models are defined, each neural network model being configured to handle a class of cases related to the subject and a specific data configuration describing a case of the class. the one or more neural network models are run to generate data requests about the subject's identity, and the data requests are passed to a detection system that monitor transactions associated with the subject. additional data associated with the transactions is requested until a threshold certainty is achieved or until available data or models are exhausted.",2012-10-23,"The title of the patent is comprehensive identity protection system and its abstract is a system and method for protecting identity fraud are disclosed. a system includes a detection subsystem to identify applications and/or accounts at risk of identity fraud, and a disposition subsystem to process data provided by the detection system and to determine whether identity fraud exists in the applications and/or accounts. according to an implementation, one or more neural network models are defined, each neural network model being configured to handle a class of cases related to the subject and a specific data configuration describing a case of the class. the one or more neural network models are run to generate data requests about the subject's identity, and the data requests are passed to a detection system that monitor transactions associated with the subject. additional data associated with the transactions is requested until a threshold certainty is achieved or until available data or models are exhausted. dated 2012-10-23"
8297047,exhaust gas purifying apparatus for internal combustion engine,"an exhaust gas purifying apparatus for an internal combustion engine having a lean nox catalyst in an exhaust system is provided. the lean nox catalyst traps nox in exhaust gases when the exhaust gases are in an oxidizing state, and discharges the trapped nox when the exhaust gases are in an reducing state. in this apparatus, an estimated trapped nox amount which is an estimated value of an amount of nox trapped in the lean nox catalyst, is calculated using a neural network. engine operating parameters indicative of an operating condition of the engine are input, and the neural network outputs at least one control parameter which is relevant to the lean nox catalyst. a reducing process of the nox trapped in the lean nox catalyst is performed according to the estimated trapped nox amount.",2012-10-30,"The title of the patent is exhaust gas purifying apparatus for internal combustion engine and its abstract is an exhaust gas purifying apparatus for an internal combustion engine having a lean nox catalyst in an exhaust system is provided. the lean nox catalyst traps nox in exhaust gases when the exhaust gases are in an oxidizing state, and discharges the trapped nox when the exhaust gases are in an reducing state. in this apparatus, an estimated trapped nox amount which is an estimated value of an amount of nox trapped in the lean nox catalyst, is calculated using a neural network. engine operating parameters indicative of an operating condition of the engine are input, and the neural network outputs at least one control parameter which is relevant to the lean nox catalyst. a reducing process of the nox trapped in the lean nox catalyst is performed according to the estimated trapped nox amount. dated 2012-10-30"
8301356,engine out nox virtual sensor using cylinder pressure sensor,"a method for estimating nox creation in a combustion process of a four-stroke internal combustion engine includes monitoring engine sensor inputs, modeling parameters descriptive of said combustion process based upon said engine sensor inputs, and estimating nox creation with an artificial neural network based upon said parameters.",2012-10-30,"The title of the patent is engine out nox virtual sensor using cylinder pressure sensor and its abstract is a method for estimating nox creation in a combustion process of a four-stroke internal combustion engine includes monitoring engine sensor inputs, modeling parameters descriptive of said combustion process based upon said engine sensor inputs, and estimating nox creation with an artificial neural network based upon said parameters. dated 2012-10-30"
8301576,weighted pattern learning for neural networks,"a method of training a neural net includes receiving a plurality of sets of data, each set representative of a plurality of inputs to the neural net and a resulting at least one output from the neural net and calculating a plurality of network weights for the neural network based on the received plurality of sets of data. calculating the plurality of network weights including attributing greater weight in the calculation to at least one set of the plurality of sets of data than at least one other set of the plurality of sets of data.",2012-10-30,"The title of the patent is weighted pattern learning for neural networks and its abstract is a method of training a neural net includes receiving a plurality of sets of data, each set representative of a plurality of inputs to the neural net and a resulting at least one output from the neural net and calculating a plurality of network weights for the neural network based on the received plurality of sets of data. calculating the plurality of network weights including attributing greater weight in the calculation to at least one set of the plurality of sets of data than at least one other set of the plurality of sets of data. dated 2012-10-30"
8306781,professional diagnosis method of battery performance analysis,"the present invention discloses a professional diagnosis method of battery performance analysis, through the overall evaluation of experiential data library, several parameters about the battery are input into the artificial neural network, outputting capacity prediction and service life prediction of each battery, etc. and giving useful advices for each battery. therefore the result is much more in conformity with the real condition of the battery. besides, it designs an adaptive learning function of the abovementioned artificial neural network. this invention effectively avoids the defect of evaluating the vrla battery performance at single moment, from single perspective and by single method, and it does the real-time monitoring and evaluating for the performance of the battery during vrla battery working period, which is easy to operate, and avoids checking discharge test to the battery so that it doesn't affect the cycle life of the vrla battery.",2012-11-06,"The title of the patent is professional diagnosis method of battery performance analysis and its abstract is the present invention discloses a professional diagnosis method of battery performance analysis, through the overall evaluation of experiential data library, several parameters about the battery are input into the artificial neural network, outputting capacity prediction and service life prediction of each battery, etc. and giving useful advices for each battery. therefore the result is much more in conformity with the real condition of the battery. besides, it designs an adaptive learning function of the abovementioned artificial neural network. this invention effectively avoids the defect of evaluating the vrla battery performance at single moment, from single perspective and by single method, and it does the real-time monitoring and evaluating for the performance of the battery during vrla battery working period, which is easy to operate, and avoids checking discharge test to the battery so that it doesn't affect the cycle life of the vrla battery. dated 2012-11-06"
8306931,"detecting, classifying, and tracking abnormal data in a data stream","the present invention extends to methods, systems, and computer program products for detecting, classifying, and tracking abnormal data in a data stream. embodiments include an integrated set of algorithms that enable an analyst to detect, characterize, and track abnormalities in real-time data streams based upon historical data labeled as predominantly normal or abnormal. embodiments of the invention can detect, identify relevant historical contextual similarity, and fuse unexpected and unknown abnormal signatures with other possibly related sensor and source information. the number, size, and connections of the neural networks all automatically adapted to the data. further, adaption appropriately and automatically integrates unknown and known abnormal signature training within one neural network architecture solution automatically. algorithms and neural networks architecture are data driven, resulting more affordable processing. expert knowledge can be incorporated to enhance the process, but sufficient performance is achievable without any system domain or neural networks expertise.",2012-11-06,"The title of the patent is detecting, classifying, and tracking abnormal data in a data stream and its abstract is the present invention extends to methods, systems, and computer program products for detecting, classifying, and tracking abnormal data in a data stream. embodiments include an integrated set of algorithms that enable an analyst to detect, characterize, and track abnormalities in real-time data streams based upon historical data labeled as predominantly normal or abnormal. embodiments of the invention can detect, identify relevant historical contextual similarity, and fuse unexpected and unknown abnormal signatures with other possibly related sensor and source information. the number, size, and connections of the neural networks all automatically adapted to the data. further, adaption appropriately and automatically integrates unknown and known abnormal signature training within one neural network architecture solution automatically. algorithms and neural networks architecture are data driven, resulting more affordable processing. expert knowledge can be incorporated to enhance the process, but sufficient performance is achievable without any system domain or neural networks expertise. dated 2012-11-06"
8306932,system and method for adaptive data masking,"a method for adaptive data masking of a database is provided. the method comprises extracting data from a first database and providing one or more predefined rules for masking the extracted data. the method further comprises masking a first portion of extracted data using a trained artificial neural network (ann), where the ann is trained for masking at least one database having properties similar to the first database. the masked and unmasked data is aggregated to arrive at an output structurally similar to the extracted data. the method furthermore comprises determining a deviation value between the arrived output and expected output of the extracted data, and adapting the trained ann automatically according to data masking requirements of the first database, if the deviation value is more than a predefined value.",2012-11-06,"The title of the patent is system and method for adaptive data masking and its abstract is a method for adaptive data masking of a database is provided. the method comprises extracting data from a first database and providing one or more predefined rules for masking the extracted data. the method further comprises masking a first portion of extracted data using a trained artificial neural network (ann), where the ann is trained for masking at least one database having properties similar to the first database. the masked and unmasked data is aggregated to arrive at an output structurally similar to the extracted data. the method furthermore comprises determining a deviation value between the arrived output and expected output of the extracted data, and adapting the trained ann automatically according to data masking requirements of the first database, if the deviation value is more than a predefined value. dated 2012-11-06"
8308646,trainable diagnostic system and method of use,"a trainable, adaptable system for analyzing functional or structural clinical data can be used to identify a given pathology based on functional data. the system includes a signal processor that receives functional data from a device monitoring a subject and normalizes the functional data over at least one cycle of functional data. the system also includes a neural network having a plurality of weights selected based on predetermined data and receiving and processing the normalized functional data based on the plurality of weights to generate at least one metric indicating a degree of relation between the normalized functional data to the predetermined data. a diagnostic interpretation module is included for receiving the at least one metric from the neural network and classifying the functional data as indicative of the given pathology or not indicative of the given pathology based on a comparison of the at least one metric to at least one probability distribution of a likelihood of the given pathology.",2012-11-13,"The title of the patent is trainable diagnostic system and method of use and its abstract is a trainable, adaptable system for analyzing functional or structural clinical data can be used to identify a given pathology based on functional data. the system includes a signal processor that receives functional data from a device monitoring a subject and normalizes the functional data over at least one cycle of functional data. the system also includes a neural network having a plurality of weights selected based on predetermined data and receiving and processing the normalized functional data based on the plurality of weights to generate at least one metric indicating a degree of relation between the normalized functional data to the predetermined data. a diagnostic interpretation module is included for receiving the at least one metric from the neural network and classifying the functional data as indicative of the given pathology or not indicative of the given pathology based on a comparison of the at least one metric to at least one probability distribution of a likelihood of the given pathology. dated 2012-11-13"
8311961,effort estimation using text analysis,"a system, method and program product for estimating effort of implementing a system based on a use case specification document. a system is provided that includes: a volumetrics processor that quantifies a structure of the document and evaluates a format of the document; a domain processor that identifies a domain of the system associated with the document; a complexity processor that defines a set of complexity variables associated with the document based on the structure of the document, a format of the document and a domain of the document; and a neural network that estimates an effort based on the set of complexity variables.",2012-11-13,"The title of the patent is effort estimation using text analysis and its abstract is a system, method and program product for estimating effort of implementing a system based on a use case specification document. a system is provided that includes: a volumetrics processor that quantifies a structure of the document and evaluates a format of the document; a domain processor that identifies a domain of the system associated with the document; a complexity processor that defines a set of complexity variables associated with the document based on the structure of the document, a format of the document and a domain of the document; and a neural network that estimates an effort based on the set of complexity variables. dated 2012-11-13"
8320629,multi-level neural network based characters identification method and system,"a system and method, which enable precise and automatic identification of characters, perform and calibrate data verification to ensure data reliability. the system can process these identified characters, such as override adverse conditions, adjusting and correcting unclear characters and their images.",2012-11-27,"The title of the patent is multi-level neural network based characters identification method and system and its abstract is a system and method, which enable precise and automatic identification of characters, perform and calibrate data verification to ensure data reliability. the system can process these identified characters, such as override adverse conditions, adjusting and correcting unclear characters and their images. dated 2012-11-27"
8326040,combiner for improving handwriting recognition,"various technologies and techniques are disclosed that improve handwriting recognition operations. handwritten input is received in training mode and run through several base recognizers to generate several alternate lists. the alternate lists are unioned together into a combined alternate list. if the correct result is in the combined list, each correct/incorrect alternate pair is used to generate training patterns. the weights associated with the alternate pairs are stored. at runtime, the combined alternate list is generated just as training time. the trained comparator-net can be used to compare any two alternates in the combined list. a template matching base recognizer is used with one or more neural network base recognizers to improve recognition operations. the system provides comparator-net and reorder-net processes trained on print and cursive data, and ones that have been trained on cursive-only data. the respective comparator-net and reorder-net processes are used accordingly.",2012-12-04,"The title of the patent is combiner for improving handwriting recognition and its abstract is various technologies and techniques are disclosed that improve handwriting recognition operations. handwritten input is received in training mode and run through several base recognizers to generate several alternate lists. the alternate lists are unioned together into a combined alternate list. if the correct result is in the combined list, each correct/incorrect alternate pair is used to generate training patterns. the weights associated with the alternate pairs are stored. at runtime, the combined alternate list is generated just as training time. the trained comparator-net can be used to compare any two alternates in the combined list. a template matching base recognizer is used with one or more neural network base recognizers to improve recognition operations. the system provides comparator-net and reorder-net processes trained on print and cursive data, and ones that have been trained on cursive-only data. the respective comparator-net and reorder-net processes are used accordingly. dated 2012-12-04"
8326047,image processing using neural network,"image processing method that includes the steps of considering each image point as a node of an artificial neural network, and of processing the image as function of values of the nodes and of connections of each image point undergoing processing with neighboring image points, the image points of the processed image being obtained by iterative evolution steps of parameters defining the appearance as evolution steps of the value of nodes or by iterative evolution steps of values of the set of connections or by a combination of the evolutions, wherein the processing occurs by evolution iterative steps that are functions of connections of neighboring image points with the image point under examination, each of the neighboring image points being further considered as neighboring one or more or all adjacent image points, the functions providing immediate feedback contributions for determining appearance values of all other image points.",2012-12-04,"The title of the patent is image processing using neural network and its abstract is image processing method that includes the steps of considering each image point as a node of an artificial neural network, and of processing the image as function of values of the nodes and of connections of each image point undergoing processing with neighboring image points, the image points of the processed image being obtained by iterative evolution steps of parameters defining the appearance as evolution steps of the value of nodes or by iterative evolution steps of values of the set of connections or by a combination of the evolutions, wherein the processing occurs by evolution iterative steps that are functions of connections of neighboring image points with the image point under examination, each of the neighboring image points being further considered as neighboring one or more or all adjacent image points, the functions providing immediate feedback contributions for determining appearance values of all other image points. dated 2012-12-04"
8326781,"method for the compressed transmission of data packet header fields in a packet-oriented data stream, method for compressing data packet header fields in a packet-oriented data stream, method for decompressing data packet header fields in a packet-oriented data stream, compression/decompression system, compression apparatus and decompression apparatus","in various embodiments, a method for compressed transmission of data packet header fields in a packet-oriented data stream may comprise: estimating in advance a data packet header field in a packet-oriented data stream from at least one preceding data packet header field; forming a piece of comparison information which indicates the difference between the data packet header field and the data packet header field estimated in advance using the transmitter neural network; transmitting the comparison information as a compressed data packet header field from the transmitter to a receiver; estimating in advance the data packet header field from at least one already transmitted data packet header field in the packet-oriented data stream using the receiver neural network; and generating the data packet header field from the data packet header field estimated in advance using the neural network of the receiver and from the transmitted piece of comparison information.",2012-12-04,"The title of the patent is method for the compressed transmission of data packet header fields in a packet-oriented data stream, method for compressing data packet header fields in a packet-oriented data stream, method for decompressing data packet header fields in a packet-oriented data stream, compression/decompression system, compression apparatus and decompression apparatus and its abstract is in various embodiments, a method for compressed transmission of data packet header fields in a packet-oriented data stream may comprise: estimating in advance a data packet header field in a packet-oriented data stream from at least one preceding data packet header field; forming a piece of comparison information which indicates the difference between the data packet header field and the data packet header field estimated in advance using the transmitter neural network; transmitting the comparison information as a compressed data packet header field from the transmitter to a receiver; estimating in advance the data packet header field from at least one already transmitted data packet header field in the packet-oriented data stream using the receiver neural network; and generating the data packet header field from the data packet header field estimated in advance using the neural network of the receiver and from the transmitted piece of comparison information. dated 2012-12-04"
8332337,condition-based monitoring system for machinery and associated methods,"real-time condition-based analysis is performed on a machine for providing diagnostic and prognostic outputs indicative of machine status includes a signal processor for receiving signals from sensors adapted for measuring machine performance parameters. the signal processor conditions and shapes at least some of the received signals into an input form for a neural network. a fuzzy adaptive resonance theory neural network receives at least some of the conditioned and shaped signals, and detects and classifies a state of the machine based upon the received conditioned and shaped signals, and upon a predetermined ontology of machine states, diagnostics, and prognostics. the neural network can also determine from the machine state a health status thereof, which can comprise an anomaly, and output a signal representative of the determined health status. a bayesian intelligence network receives the machine state from the neural network and determines a fault probability at a future time.",2012-12-11,"The title of the patent is condition-based monitoring system for machinery and associated methods and its abstract is real-time condition-based analysis is performed on a machine for providing diagnostic and prognostic outputs indicative of machine status includes a signal processor for receiving signals from sensors adapted for measuring machine performance parameters. the signal processor conditions and shapes at least some of the received signals into an input form for a neural network. a fuzzy adaptive resonance theory neural network receives at least some of the conditioned and shaped signals, and detects and classifies a state of the machine based upon the received conditioned and shaped signals, and upon a predetermined ontology of machine states, diagnostics, and prognostics. the neural network can also determine from the machine state a health status thereof, which can comprise an anomaly, and output a signal representative of the determined health status. a bayesian intelligence network receives the machine state from the neural network and determines a fault probability at a future time. dated 2012-12-11"
8335564,ventricle pacing during atrial fibrillation episodes,an adaptive cardiac resynchronization therapy system delivers biventricular stimulation to the heart with dynamic av delay and vv interval. the stimulation is modified continuously in correlation with the hemodynamic performance of the heart. the system uses a spiking neural network comprising spike controller (42) that learns to associate the va interval based on hemodynamic sensor temporal patterns. the associated va interval replaces the sensed atrial event signal during atrial fibrillation episodes.,2012-12-18,The title of the patent is ventricle pacing during atrial fibrillation episodes and its abstract is an adaptive cardiac resynchronization therapy system delivers biventricular stimulation to the heart with dynamic av delay and vv interval. the stimulation is modified continuously in correlation with the hemodynamic performance of the heart. the system uses a spiking neural network comprising spike controller (42) that learns to associate the va interval based on hemodynamic sensor temporal patterns. the associated va interval replaces the sensed atrial event signal during atrial fibrillation episodes. dated 2012-12-18
8335935,power management based on automatic workload detection,"an electronic device includes a kernel, a power manager to control power to a hardware component, and a neural network to monitor the kernel to recognize performance of a function of the electronic device. the neural network sends a signal to the power manager to reduce or turn off power to the hardware component based on information generated during monitoring of the kernel. the information may provide an indication of hit symbols for hardware components which are to be powered and/or hardware components which do not require power based on one or more operations performed by the kernel.",2012-12-18,"The title of the patent is power management based on automatic workload detection and its abstract is an electronic device includes a kernel, a power manager to control power to a hardware component, and a neural network to monitor the kernel to recognize performance of a function of the electronic device. the neural network sends a signal to the power manager to reduce or turn off power to the hardware component based on information generated during monitoring of the kernel. the information may provide an indication of hit symbols for hardware components which are to be powered and/or hardware components which do not require power based on one or more operations performed by the kernel. dated 2012-12-18"
8341068,method and apparatus for generating and evaluating ideas in an organization,"the present invention discloses a method and apparatus for generating and evaluating ideas within an organization through an idea market. it includes an automatic price-setting mechanism that modifies the share-price at the same instant that a trade is made, yet without the assistance of a market maker or a queue of orders. the instant price-setting is achieved by assuming a pre-defined relationship between the quantity of shares in the order and the average share price for that order. this relationship is the price-quantity function. the present invention also includes a mechanism of weighted parameters to modify the share-price in addition to pure supply and demand. the parameters can be adjusted over time in a neural network to optimize the relationship between the share price and the prediction of the actual idea value.",2012-12-25,"The title of the patent is method and apparatus for generating and evaluating ideas in an organization and its abstract is the present invention discloses a method and apparatus for generating and evaluating ideas within an organization through an idea market. it includes an automatic price-setting mechanism that modifies the share-price at the same instant that a trade is made, yet without the assistance of a market maker or a queue of orders. the instant price-setting is achieved by assuming a pre-defined relationship between the quantity of shares in the order and the average share price for that order. this relationship is the price-quantity function. the present invention also includes a mechanism of weighted parameters to modify the share-price in addition to pure supply and demand. the parameters can be adjusted over time in a neural network to optimize the relationship between the share price and the prediction of the actual idea value. dated 2012-12-25"
8341100,epithelial layer detector and related methods,"an epithelial detector and method for automatically identifying epithelial portions of a tissue sample, includes: staining the tissue sample with at least two dyes; applying a color transformation to a color image of the tissue sample to obtain one or more color channels; and applying a trained convolutional neural network to the color channels to obtain a decision for position in the tissue as to whether it is inside or outside an epithelial layer. also, a method for training the convolutional neural network.",2012-12-25,"The title of the patent is epithelial layer detector and related methods and its abstract is an epithelial detector and method for automatically identifying epithelial portions of a tissue sample, includes: staining the tissue sample with at least two dyes; applying a color transformation to a color image of the tissue sample to obtain one or more color channels; and applying a trained convolutional neural network to the color channels to obtain a decision for position in the tissue as to whether it is inside or outside an epithelial layer. also, a method for training the convolutional neural network. dated 2012-12-25"
8345962,transfer learning methods and systems for feed-forward visual recognition systems,"a method and system for training a neural network of a visual recognition computer system, extracts at least one feature of an image or video frame with a feature extractor; approximates the at least one feature of the image or video frame with an auxiliary output provided in the neural network; and measures a feature difference between the extracted at least one feature of the image or video frame and the approximated at least one feature of the image or video frame with an auxiliary error calculator. a joint learner of the method and system adjusts at least one parameter of the neural network to minimize the measured feature difference.",2013-01-01,"The title of the patent is transfer learning methods and systems for feed-forward visual recognition systems and its abstract is a method and system for training a neural network of a visual recognition computer system, extracts at least one feature of an image or video frame with a feature extractor; approximates the at least one feature of the image or video frame with an auxiliary output provided in the neural network; and measures a feature difference between the extracted at least one feature of the image or video frame and the approximated at least one feature of the image or video frame with an auxiliary error calculator. a joint learner of the method and system adjusts at least one parameter of the neural network to minimize the measured feature difference. dated 2013-01-01"
8346691,computer-implemented semi-supervised learning systems and methods,"computer-implemented systems and methods for determining a subset of unknown targets to investigate. for example, a method can be configured to receive a target data set, wherein the target data set includes known targets and unknown targets. a supervised model such as a neural network model is generated using the known targets. the unknown targets are used with the neural network model to generate values for the unknown targets. analysis with an unsupervised model is performed using the target data set in order to determine which of the unknown targets are outliers. a comparison of list of outlier unknown targets is performed with the values for the unknown targets that were generated by the neural network model. the subset of unknown targets to investigate is determined based upon the comparison.",2013-01-01,"The title of the patent is computer-implemented semi-supervised learning systems and methods and its abstract is computer-implemented systems and methods for determining a subset of unknown targets to investigate. for example, a method can be configured to receive a target data set, wherein the target data set includes known targets and unknown targets. a supervised model such as a neural network model is generated using the known targets. the unknown targets are used with the neural network model to generate values for the unknown targets. analysis with an unsupervised model is performed using the target data set in order to determine which of the unknown targets are outliers. a comparison of list of outlier unknown targets is performed with the values for the unknown targets that were generated by the neural network model. the subset of unknown targets to investigate is determined based upon the comparison. dated 2013-01-01"
8346692,spatio-temporal pattern recognition using a spiking neural network and processing thereof on a portable and/or distributed computer,"a spiking neural network has a layer of connected neurons exchanging signals. each neuron is connected to at least one other neuron. a neuron is active if it spikes at least once during a time interval. time-varying synaptic weights are computed between each neuron and at least one other neuron connected thereto. these weights are computed according to a number of active neurons that are connected to the neuron. the weights are also computed according to an activity of the spiking neural network during the time interval. spiking of each neuron is synchronized according to a number of active neurons connected to the neuron and according to the weights. a pattern is submitted to the spiking neural network for generating sequences of spikes, which are modulated over time by the spiking synchronization. the pattern is characterized according to the sequences of spikes generated in the spiking neural network.",2013-01-01,"The title of the patent is spatio-temporal pattern recognition using a spiking neural network and processing thereof on a portable and/or distributed computer and its abstract is a spiking neural network has a layer of connected neurons exchanging signals. each neuron is connected to at least one other neuron. a neuron is active if it spikes at least once during a time interval. time-varying synaptic weights are computed between each neuron and at least one other neuron connected thereto. these weights are computed according to a number of active neurons that are connected to the neuron. the weights are also computed according to an activity of the spiking neural network during the time interval. spiking of each neuron is synchronized according to a number of active neurons connected to the neuron and according to the weights. a pattern is submitted to the spiking neural network for generating sequences of spikes, which are modulated over time by the spiking synchronization. the pattern is characterized according to the sequences of spikes generated in the spiking neural network. dated 2013-01-01"
8346693,method for hammerstein modeling of steam generator plant,"the method for hammerstein modeling of a steam generator plant includes modeling of the linear dynamic part of a hammerstein model with a state-space model, and modeling the nonlinear part of the hammerstein model with a radial basis function neural network (rbfnn). particle swarm optimization (pso), typically known for its heuristic search capabilities, is used for estimating the parameters of the rbfnn. parameters of the linear part are estimated using a numerical algorithm for subspace state-space system identification (n4sid).",2013-01-01,"The title of the patent is method for hammerstein modeling of steam generator plant and its abstract is the method for hammerstein modeling of a steam generator plant includes modeling of the linear dynamic part of a hammerstein model with a state-space model, and modeling the nonlinear part of the hammerstein model with a radial basis function neural network (rbfnn). particle swarm optimization (pso), typically known for its heuristic search capabilities, is used for estimating the parameters of the rbfnn. parameters of the linear part are estimated using a numerical algorithm for subspace state-space system identification (n4sid). dated 2013-01-01"
8346711,method for identifying multi-input multi-output hammerstein models,"the method for the identifying of multiple input, multiple output (mimo) hammerstein models that includes modeling of the linear dynamic part of a hammerstein model with a state-space model, and modeling the nonlinear part of the hammerstein model with a radial basis function neural network (rbfnn).",2013-01-01,"The title of the patent is method for identifying multi-input multi-output hammerstein models and its abstract is the method for the identifying of multiple input, multiple output (mimo) hammerstein models that includes modeling of the linear dynamic part of a hammerstein model with a state-space model, and modeling the nonlinear part of the hammerstein model with a radial basis function neural network (rbfnn). dated 2013-01-01"
8346712,method for identifying hammerstein models,"the identification of hammerstein models relates to a computerized method for identifying hammerstein models in which the linear dynamic part is modeled by a space-state model and the static nonlinear part is modeled using a radial basis function neural network (rbfnn), and in which a particle swarm optimization (pso) algorithm is used to estimate the neural network parameters and a numerical algorithm for subspace state-space system identification (n4sid) is used for estimation of parameters of the linear part.",2013-01-01,"The title of the patent is method for identifying hammerstein models and its abstract is the identification of hammerstein models relates to a computerized method for identifying hammerstein models in which the linear dynamic part is modeled by a space-state model and the static nonlinear part is modeled using a radial basis function neural network (rbfnn), and in which a particle swarm optimization (pso) algorithm is used to estimate the neural network parameters and a numerical algorithm for subspace state-space system identification (n4sid) is used for estimation of parameters of the linear part. dated 2013-01-01"
8352216,system and method for advanced condition monitoring of an asset system,"a method for advanced condition monitoring of an asset system includes sensing actual values of an operating condition for an operating regime of the asset system using at least one sensor; estimating sensed values of the operating condition by using an auto-associative neural network; determining a residual vector between the estimated sensed values and the actual values; and performing a fault diagnostic on the residual vector. in another method, an operating space of the asset system is segmented into operating regimes; the auto-associative neural network determines estimates of actual measured values; a residual vector is determined from the auto-associative neural network; a fault diagnostic is performed on the residual vector; and a change of the operation of the asset system is determined by analysis of the residual vector. an alert is provided if necessary. a smart sensor system includes an on-board processing unit for performing the method of the invention.",2013-01-08,"The title of the patent is system and method for advanced condition monitoring of an asset system and its abstract is a method for advanced condition monitoring of an asset system includes sensing actual values of an operating condition for an operating regime of the asset system using at least one sensor; estimating sensed values of the operating condition by using an auto-associative neural network; determining a residual vector between the estimated sensed values and the actual values; and performing a fault diagnostic on the residual vector. in another method, an operating space of the asset system is segmented into operating regimes; the auto-associative neural network determines estimates of actual measured values; a residual vector is determined from the auto-associative neural network; a fault diagnostic is performed on the residual vector; and a change of the operation of the asset system is determined by analysis of the residual vector. an alert is provided if necessary. a smart sensor system includes an on-board processing unit for performing the method of the invention. dated 2013-01-08"
8352392,methods and system for modeling network traffic,"a method and system are provided for modeling network traffic in which an artificial neural network architecture is utilized in order to intelligently and adaptively model the capacity of a network. initially, the network traffic is decomposed into a plurality of categories, such as individual users, application usage or common usage groups. inputs to the artificial neural network are then defined such that a respective combination of inputs permits prediction of bandwidth capacity needs for that input condition. outputs of the artificial neural network are representative of the network traffic associated with the respective inputs. for example, a plurality of bandwidth profiles associated with respective categories may be defined. an artificial neural network may then be constructed and trained with those bandwidth profiles and then utilized to relate predict future bandwidth needs for the network.",2013-01-08,"The title of the patent is methods and system for modeling network traffic and its abstract is a method and system are provided for modeling network traffic in which an artificial neural network architecture is utilized in order to intelligently and adaptively model the capacity of a network. initially, the network traffic is decomposed into a plurality of categories, such as individual users, application usage or common usage groups. inputs to the artificial neural network are then defined such that a respective combination of inputs permits prediction of bandwidth capacity needs for that input condition. outputs of the artificial neural network are representative of the network traffic associated with the respective inputs. for example, a plurality of bandwidth profiles associated with respective categories may be defined. an artificial neural network may then be constructed and trained with those bandwidth profiles and then utilized to relate predict future bandwidth needs for the network. dated 2013-01-08"
8363950,combining online and offline recognizers in a handwriting recognition system,"described is a technology by which online recognition of handwritten input data is combined with offline recognition and processing to obtain a combined recognition result. in general, the combination improves overall recognition accuracy. in one aspect, online and offline recognition is separately performed to obtain online and offline character-level recognition scores for candidates (hypotheses). a statistical analysis-based combination algorithm, an adaboost algorithm, and/or a neural network-based combination may determine a combination function to combine the scores to produce a result set of one or more results. online and offline radical-level recognition may be performed. for example, a hmm recognizer may generate online radical scores used to build a radical graph, which is then rescored using the offline radical recognition scores. paths in the rescored graph are then searched to provide the combined recognition result, e.g., corresponding to the path with the highest score.",2013-01-29,"The title of the patent is combining online and offline recognizers in a handwriting recognition system and its abstract is described is a technology by which online recognition of handwritten input data is combined with offline recognition and processing to obtain a combined recognition result. in general, the combination improves overall recognition accuracy. in one aspect, online and offline recognition is separately performed to obtain online and offline character-level recognition scores for candidates (hypotheses). a statistical analysis-based combination algorithm, an adaboost algorithm, and/or a neural network-based combination may determine a combination function to combine the scores to produce a result set of one or more results. online and offline radical-level recognition may be performed. for example, a hmm recognizer may generate online radical scores used to build a radical graph, which is then rescored using the offline radical recognition scores. paths in the rescored graph are then searched to provide the combined recognition result, e.g., corresponding to the path with the highest score. dated 2013-01-29"
8374799,systems and methods for extending the dynamic range of mass spectrometry,"systems and methods are used to predict intensities of a saturated peak using a peak predictor. a set of data is selected from the plurality of intensity measurements that includes a saturated peak. confidence values are assigned to each data point in the set of data producing a plurality of confidence value weighted data points. a peak predictor is selected. the peak predictor is applied to the plurality of confidence value weighted data points of the saturated peak producing predicted intensities for the saturated peak. the confidence values can include system confidence values, predictor confidence values, or a combination of system confidence values and predictor confidence values. the peak predictor can be a theoretical model, a dynamic model, an artificial neural network, or an analytical function representing a best fit of a plurality of probability density functions to a first set of measured data that includes a representative non-saturated peak.",2013-02-12,"The title of the patent is systems and methods for extending the dynamic range of mass spectrometry and its abstract is systems and methods are used to predict intensities of a saturated peak using a peak predictor. a set of data is selected from the plurality of intensity measurements that includes a saturated peak. confidence values are assigned to each data point in the set of data producing a plurality of confidence value weighted data points. a peak predictor is selected. the peak predictor is applied to the plurality of confidence value weighted data points of the saturated peak producing predicted intensities for the saturated peak. the confidence values can include system confidence values, predictor confidence values, or a combination of system confidence values and predictor confidence values. the peak predictor can be a theoretical model, a dynamic model, an artificial neural network, or an analytical function representing a best fit of a plurality of probability density functions to a first set of measured data that includes a representative non-saturated peak. dated 2013-02-12"
8374974,neural network training data selection using memory reduced cluster analysis for field model development,"a system and method for selecting a training data set from a set of multidimensional geophysical input data samples for training a model to predict target data. the input data may be data sets produced by a pulsed neutron logging tool at multiple depth points in a cases well. target data may be responses of an open hole logging tool. the input data is divided into clusters. actual target data from the training well is linked to the clusters. the linked clusters are analyzed for variance, etc. and fuzzy inference is used to select a portion of each cluster to include in a training set. the reduced set is used to train a model, such as an artificial neural network. the trained model may then be used to produce synthetic open hole logs in response to inputs of cased hole log data.",2013-02-12,"The title of the patent is neural network training data selection using memory reduced cluster analysis for field model development and its abstract is a system and method for selecting a training data set from a set of multidimensional geophysical input data samples for training a model to predict target data. the input data may be data sets produced by a pulsed neutron logging tool at multiple depth points in a cases well. target data may be responses of an open hole logging tool. the input data is divided into clusters. actual target data from the training well is linked to the clusters. the linked clusters are analyzed for variance, etc. and fuzzy inference is used to select a portion of each cluster to include in a training set. the reduced set is used to train a model, such as an artificial neural network. the trained model may then be used to produce synthetic open hole logs in response to inputs of cased hole log data. dated 2013-02-12"
8380607,predicting economic trends via network communication mood tracking,"a method of investigating public mood from a multi-dimensional model approach and a method to predict economic market trends above chance level based on the multi-dimensional model approach are provided. the text-content of several large-scale collections of daily network communications are analyzed via mood assessment tools, measuring various mood dimensions. a granger causality analysis investigated the correlation between daily changes in public mood states via results of the daily mood time series of the mood assessment tools with changes in value of the dow jones industrial average (“djia”) over time. based on the above investigation, a self-organizing fuzzy neural network model was trained to predict next-day djia value based on a combination of past djia values and public mood state measurements across several specified mood dimensions, such as calm and a combination of calm and happy.",2013-02-19,"The title of the patent is predicting economic trends via network communication mood tracking and its abstract is a method of investigating public mood from a multi-dimensional model approach and a method to predict economic market trends above chance level based on the multi-dimensional model approach are provided. the text-content of several large-scale collections of daily network communications are analyzed via mood assessment tools, measuring various mood dimensions. a granger causality analysis investigated the correlation between daily changes in public mood states via results of the daily mood time series of the mood assessment tools with changes in value of the dow jones industrial average (“djia”) over time. based on the above investigation, a self-organizing fuzzy neural network model was trained to predict next-day djia value based on a combination of past djia values and public mood state measurements across several specified mood dimensions, such as calm and a combination of calm and happy. dated 2013-02-19"
8391575,automatic image analysis and quantification for fluorescence in situ hybridization,"an analysis system automatically analyzes and counts fluorescence signals present in biopsy tissue marked using fluorescence in situ hybridization (fish). the user of the system specifies classes of a class network and process steps of a process hierarchy. then pixel values in image slices of biopsy tissue are acquired in three dimensions. a computer-implemented network structure is generated by linking pixel values to objects of a data network according to the class network and process hierarchy. objects associated with pixel values at different depths of the biopsy tissue are used to determine the number, volume and distance between cell components. in one application, fluorescence signals that mark her2/neural genes and centromeres of chromosome seventeen are counted to diagnose breast cancer. her2/neural genes that overlap one another or that are covered by centromeres can be accurately counted. signal artifacts that do not mark genes can be identified by their excessive volume.",2013-03-05,"The title of the patent is automatic image analysis and quantification for fluorescence in situ hybridization and its abstract is an analysis system automatically analyzes and counts fluorescence signals present in biopsy tissue marked using fluorescence in situ hybridization (fish). the user of the system specifies classes of a class network and process steps of a process hierarchy. then pixel values in image slices of biopsy tissue are acquired in three dimensions. a computer-implemented network structure is generated by linking pixel values to objects of a data network according to the class network and process hierarchy. objects associated with pixel values at different depths of the biopsy tissue are used to determine the number, volume and distance between cell components. in one application, fluorescence signals that mark her2/neural genes and centromeres of chromosome seventeen are counted to diagnose breast cancer. her2/neural genes that overlap one another or that are covered by centromeres can be accurately counted. signal artifacts that do not mark genes can be identified by their excessive volume. dated 2013-03-05"
8391603,system and method for image segmentation,"a method of segmenting images receives an image (such as a medical image) and a segment in relation to the image, displays them to an observer, receives a modification to the segment from the observer, and generates a second segment in relation to a second image, responsive to the modification. an image segmentation system includes a learning scheme or model to take input from an observer feedback interface and to communicate with a means for drawing an image segment to permit adjustment of at least one image segmentation parameter (such as a threshold value). the learning scheme is provided with a knowledge base which may initially be created by processing offline images. the learning scheme may use any scheme such as a reinforcement learning agent, a fuzzy inference system or a neural network.",2013-03-05,"The title of the patent is system and method for image segmentation and its abstract is a method of segmenting images receives an image (such as a medical image) and a segment in relation to the image, displays them to an observer, receives a modification to the segment from the observer, and generates a second segment in relation to a second image, responsive to the modification. an image segmentation system includes a learning scheme or model to take input from an observer feedback interface and to communicate with a means for drawing an image segment to permit adjustment of at least one image segmentation parameter (such as a threshold value). the learning scheme is provided with a knowledge base which may initially be created by processing offline images. the learning scheme may use any scheme such as a reinforcement learning agent, a fuzzy inference system or a neural network. dated 2013-03-05"
8392347,"coating color database creating method, search method using the database, their system, program, and recording medium","the subject invention provides a method of creating a database for searching for a paint color having a desired texture, a search method using the database, and systems, programs, and recording mediums for carrying out the method and the search. the method for creating a database includes a step (s11) for storing spectral reflectance data and micro-brilliance data of a plurality of paint colors after associating each spectral reflectance data and each micro-brilliance data with a paint color code; a step (s13) for storing texture evaluation values of sample paint colors after associating the each texture evaluation value with the paint color code; a step (s14) for calculating characteristic quantities of the paint colors expressing textures using the spectral reflectance data and the micro-brilliance data, and storing the characteristic quantities after associating the each characteristic quantity with the paint color code; a step (s15) for carrying out a process for training a neural network using the characteristic quantities and the texture evaluation values of the sample paint colors as training data; and a step (s16) for inputting characteristic quantities of the paint colors other than the sample paint colors into the neural network after the training process, and storing output data after associating the each output data with the paint color code.",2013-03-05,"The title of the patent is coating color database creating method, search method using the database, their system, program, and recording medium and its abstract is the subject invention provides a method of creating a database for searching for a paint color having a desired texture, a search method using the database, and systems, programs, and recording mediums for carrying out the method and the search. the method for creating a database includes a step (s11) for storing spectral reflectance data and micro-brilliance data of a plurality of paint colors after associating each spectral reflectance data and each micro-brilliance data with a paint color code; a step (s13) for storing texture evaluation values of sample paint colors after associating the each texture evaluation value with the paint color code; a step (s14) for calculating characteristic quantities of the paint colors expressing textures using the spectral reflectance data and the micro-brilliance data, and storing the characteristic quantities after associating the each characteristic quantity with the paint color code; a step (s15) for carrying out a process for training a neural network using the characteristic quantities and the texture evaluation values of the sample paint colors as training data; and a step (s16) for inputting characteristic quantities of the paint colors other than the sample paint colors into the neural network after the training process, and storing output data after associating the each output data with the paint color code. dated 2013-03-05"
8392352,creation of neuro-fuzzy expert system from online analytical processing (olap) tools,"a method for automatic generation of a neuro-fuzzy expert system (fuzzy logic expert system implemented as a neural network) from data. the method comprising a data interface allowing description of location, type, and structure of the data. the interface also allows designation of input attributes and output attributes in the data structure; automatic neuro-fuzzy expert system generation driven by the data; training of the expert system's neural network on the data and the presentation of results which include new knowledge embedded in the parameters and structure of the trained neuro-fuzzy expert system to a user.",2013-03-05,"The title of the patent is creation of neuro-fuzzy expert system from online analytical processing (olap) tools and its abstract is a method for automatic generation of a neuro-fuzzy expert system (fuzzy logic expert system implemented as a neural network) from data. the method comprising a data interface allowing description of location, type, and structure of the data. the interface also allows designation of input attributes and output attributes in the data structure; automatic neuro-fuzzy expert system generation driven by the data; training of the expert system's neural network on the data and the presentation of results which include new knowledge embedded in the parameters and structure of the trained neuro-fuzzy expert system to a user. dated 2013-03-05"
8396689,method for analysis of the operation of a gas turbine,"a method for analyzing the operation of a gas turbine is provided. a neural network based upon a normal operation of the gas turbine is learned. a dynamic pressure signal is read by a pressure sensor in or on the compressor of the turbine, and an operating parameter is read by a further sensor. the dynamic pressure signal is subjected to a frequency analysis, a parameter of a frequency spectrum of the pressure signal being determined. based upon the measured operating parameter and the parameter of the frequency spectrum of the pressure signal, the neural network is learned. the measured operating parameter and the parameter of the frequency spectrum are input parameters, and a diagnostic characteristic value representing a probability of a presence of normal operation of the gas turbine as a function of the input parameters is output.",2013-03-12,"The title of the patent is method for analysis of the operation of a gas turbine and its abstract is a method for analyzing the operation of a gas turbine is provided. a neural network based upon a normal operation of the gas turbine is learned. a dynamic pressure signal is read by a pressure sensor in or on the compressor of the turbine, and an operating parameter is read by a further sensor. the dynamic pressure signal is subjected to a frequency analysis, a parameter of a frequency spectrum of the pressure signal being determined. based upon the measured operating parameter and the parameter of the frequency spectrum of the pressure signal, the neural network is learned. the measured operating parameter and the parameter of the frequency spectrum are input parameters, and a diagnostic characteristic value representing a probability of a presence of normal operation of the gas turbine as a function of the input parameters is output. dated 2013-03-12"
8396851,scalable associative text mining network and method,a text mining network that improves the performance of search engines by using a network of computer entities with autonomous neural networks. each neural network provides a weighted list of associated search terms for each search query. the lists of associated search terms from two or more computer entities are merged to a unique list of associated search terms by utilization of a virtual index algorithm. document result sets from the autonomous entities are merged to a unique result set by a weighted combination of two or more result sets.,2013-03-12,The title of the patent is scalable associative text mining network and method and its abstract is a text mining network that improves the performance of search engines by using a network of computer entities with autonomous neural networks. each neural network provides a weighted list of associated search terms for each search query. the lists of associated search terms from two or more computer entities are merged to a unique list of associated search terms by utilization of a virtual index algorithm. document result sets from the autonomous entities are merged to a unique result set by a weighted combination of two or more result sets. dated 2013-03-12
8401708,electric power system,"the electric power supply and demand control device judges whether or not electric power shortage is occurred or whether or not electric power surplus is occurred in the electric power supplier and demander provided with the electric power supply and demand control device based on data on total electric energy, the amount of maximum electric power demanded, and the amount of total electric power demanded of the following day in each electric power supplier and demander, predicted by a neural network; receives electric power from other electric power suppliers and demanders provided with the power generation devices and/or the electrical storage devices in the case where electric power shortage is occurred in the electric power supplier and demander; and controls to deliver electric power to other electric power suppliers and demanders in the case where electric power surplus is occurred in the electric power supplier and demander.",2013-03-19,"The title of the patent is electric power system and its abstract is the electric power supply and demand control device judges whether or not electric power shortage is occurred or whether or not electric power surplus is occurred in the electric power supplier and demander provided with the electric power supply and demand control device based on data on total electric energy, the amount of maximum electric power demanded, and the amount of total electric power demanded of the following day in each electric power supplier and demander, predicted by a neural network; receives electric power from other electric power suppliers and demanders provided with the power generation devices and/or the electrical storage devices in the case where electric power shortage is occurred in the electric power supplier and demander; and controls to deliver electric power to other electric power suppliers and demanders in the case where electric power surplus is occurred in the electric power supplier and demander. dated 2013-03-19"
8406523,"system, method and computer program product for detecting unwanted data using a rendered format","a system, method and computer program product are provided for detecting unwanted data. in use, data is rendered, after which it may be determined whether the rendered data is unwanted, utilizing either a neural network or optical character recognition.",2013-03-26,"The title of the patent is system, method and computer program product for detecting unwanted data using a rendered format and its abstract is a system, method and computer program product are provided for detecting unwanted data. in use, data is rendered, after which it may be determined whether the rendered data is unwanted, utilizing either a neural network or optical character recognition. dated 2013-03-26"
8412658,system and method for estimating long term characteristics of battery,"a system includes a learning data input unit for receiving initial and long term characteristic learning data of a battery to be a learning object; a measurement data input unit for receiving initial characteristic measurement data of a battery to be an object for long term characteristic estimation; an artificial neural network operation unit for converting the learning data into first and second data structures, allowing an artificial neural network to learn the learning data based on each data structure, converting the measurement data into first and second data structures, and individually applying the learned artificial neural network corresponding to each data structure to calculate and output long term characteristic estimation data based on each data structure; and a long term characteristic evaluation unit for calculating an error of the estimation data of each data structure and determining reliability of the estimation data depending on error.",2013-04-02,"The title of the patent is system and method for estimating long term characteristics of battery and its abstract is a system includes a learning data input unit for receiving initial and long term characteristic learning data of a battery to be a learning object; a measurement data input unit for receiving initial characteristic measurement data of a battery to be an object for long term characteristic estimation; an artificial neural network operation unit for converting the learning data into first and second data structures, allowing an artificial neural network to learn the learning data based on each data structure, converting the measurement data into first and second data structures, and individually applying the learned artificial neural network corresponding to each data structure to calculate and output long term characteristic estimation data based on each data structure; and a long term characteristic evaluation unit for calculating an error of the estimation data of each data structure and determining reliability of the estimation data depending on error. dated 2013-04-02"
8416120,method of sensor network localization through reconstruction of radiation pattern,"disclosed herein is a method of sensor network localization through reconstruction of a radiation pattern with a characteristic value of an antenna depending on orientation thereof. the method can minimize errors using an antenna characteristic value and a signal strength depending on the orientation. in addition, the method can minimize errors using an artificial neural network to characterize a distorted radiation pattern of an antenna and using it for the localization of a triangulation method. furthermore, the method can increases the localization rate even in a passive localization method by characterizing an asymmetric antenna radiation pattern and constructing the antenna characteristic through an artificial neural network.",2013-04-09,"The title of the patent is method of sensor network localization through reconstruction of radiation pattern and its abstract is disclosed herein is a method of sensor network localization through reconstruction of a radiation pattern with a characteristic value of an antenna depending on orientation thereof. the method can minimize errors using an antenna characteristic value and a signal strength depending on the orientation. in addition, the method can minimize errors using an artificial neural network to characterize a distorted radiation pattern of an antenna and using it for the localization of a triangulation method. furthermore, the method can increases the localization rate even in a passive localization method by characterizing an asymmetric antenna radiation pattern and constructing the antenna characteristic through an artificial neural network. dated 2013-04-09"
8417495,method of training neural network models and using same for drilling wellbores,"a method of creating and using a neural network model for wellbore operations is disclosed. the method, in one aspect, may include defining a plurality of a wellbore parameter; calculating a plurality of output values of a tool operating parameter using the plurality of values of the wellbore parameter as input to a preexisting model; and obtaining a neural network model by using the plurality of values of the wellbore parameter and the calculated plurality of output values of the tool operating parameter. the neural network may be utilized for any suitable wellbore operation, including in conjunction with a drilling assembly for drilling a wellbore.",2013-04-09,"The title of the patent is method of training neural network models and using same for drilling wellbores and its abstract is a method of creating and using a neural network model for wellbore operations is disclosed. the method, in one aspect, may include defining a plurality of a wellbore parameter; calculating a plurality of output values of a tool operating parameter using the plurality of values of the wellbore parameter as input to a preexisting model; and obtaining a neural network model by using the plurality of values of the wellbore parameter and the calculated plurality of output values of the tool operating parameter. the neural network may be utilized for any suitable wellbore operation, including in conjunction with a drilling assembly for drilling a wellbore. dated 2013-04-09"
8423490,method for computer-aided learning of a neural network and neural network,"there is described a method for computer-aided learning of a neural network, with a plurality of neurons in which the neurons of the neural network are divided into at least two layers, comprising a first layer and a second layer crosslinked with the first layer. in the first layer input information is respectively represented by one or more characteristic values from one or several characteristics, wherein every characteristic value comprises one or more neurons of the first layer. a plurality of categories is stored in the second layer, wherein every category comprises one or more neurons of the second layer. for one or several pieces of input information, respectively at least one category in the second layer is assigned to the characteristic values of the input information in the first layer. input information is entered into the first layer and subsequently at least one state variable of the neural network is determined and compared to the at least one category of this input information assigned in a preceding step. the crosslinking between the first and second layer is changed depending on the comparison result from a preceding step.",2013-04-16,"The title of the patent is method for computer-aided learning of a neural network and neural network and its abstract is there is described a method for computer-aided learning of a neural network, with a plurality of neurons in which the neurons of the neural network are divided into at least two layers, comprising a first layer and a second layer crosslinked with the first layer. in the first layer input information is respectively represented by one or more characteristic values from one or several characteristics, wherein every characteristic value comprises one or more neurons of the first layer. a plurality of categories is stored in the second layer, wherein every category comprises one or more neurons of the second layer. for one or several pieces of input information, respectively at least one category in the second layer is assigned to the characteristic values of the input information in the first layer. input information is entered into the first layer and subsequently at least one state variable of the neural network is determined and compared to the at least one category of this input information assigned in a preceding step. the crosslinking between the first and second layer is changed depending on the comparison result from a preceding step. dated 2013-04-16"
8424072,behavior-based security system,"described herein are techniques for operating a security server to determine behavioral profiles for entities in a network and to detect attacks or unauthorized traffic in a network based on those behavioral profiles. in one technique, a behavioral profile may be generated based on requests for security operations to be performed that are received at a security server from an entity in a network. the behavioral profile may be generated using learning techniques, including artificial intelligence techniques such as neural networks. when the security server receives from an entity one or more requests for security operations to be performed, the security server may compare properties of the requests to the behavioral profile for the entity and properties of requests commonly sent by the entity. the security server may determine a similarity score indicating how similar the request are to the behavioral profile and to requests commonly received from the entity.",2013-04-16,"The title of the patent is behavior-based security system and its abstract is described herein are techniques for operating a security server to determine behavioral profiles for entities in a network and to detect attacks or unauthorized traffic in a network based on those behavioral profiles. in one technique, a behavioral profile may be generated based on requests for security operations to be performed that are received at a security server from an entity in a network. the behavioral profile may be generated using learning techniques, including artificial intelligence techniques such as neural networks. when the security server receives from an entity one or more requests for security operations to be performed, the security server may compare properties of the requests to the behavioral profile for the entity and properties of requests commonly sent by the entity. the security server may determine a similarity score indicating how similar the request are to the behavioral profile and to requests commonly received from the entity. dated 2013-04-16"
8428348,image analysis through neural network using image average color,"architecture for comparing images by building an initial map from the average color and an inserted blackened area. accordingly, a map can be built that is more information-rich and smaller, thereby making the system more efficient. the architecture employs a kohonen neural network (or self-organizing map (som)) by guiding the learning of the som using characteristics of the images such as average color and a central area. a strong component of the average color of the image and the central area at the approximate center of the image are added to the uninitialized som, which allows related colors to converge toward the central area of the image. when input, the som organizes the color content of the image on a map, which can be used to compare the image with other images.",2013-04-23,"The title of the patent is image analysis through neural network using image average color and its abstract is architecture for comparing images by building an initial map from the average color and an inserted blackened area. accordingly, a map can be built that is more information-rich and smaller, thereby making the system more efficient. the architecture employs a kohonen neural network (or self-organizing map (som)) by guiding the learning of the som using characteristics of the images such as average color and a central area. a strong component of the average color of the image and the central area at the approximate center of the image are added to the uninitialized som, which allows related colors to converge toward the central area of the image. when input, the som organizes the color content of the image on a map, which can be used to compare the image with other images. dated 2013-04-23"
8428916,modeling of the radiation belt megnetosphere in decisional timeframes,"systems and methods for calculating l* in the magnetosphere with essentially the same accuracy as with a physics based model at many times the speed by developing a surrogate trained to be a surrogate for the physics-based model. the trained model can then beneficially process input data falling within the training range of the surrogate model. the surrogate model can be a feedforward neural network and the physics-based model can be the tsk03 model. operatively, the surrogate model can use parameters on which the physics-based model was based, and/or spatial data for the location where l* is to be calculated. surrogate models should be provided for each of a plurality of pitch angles. accordingly, a surrogate model having a closed drift shell can be used from the plurality of models. the feedforward neural network can have a plurality of input-layer units, there being at least one input-layer unit for each physics-based model parameter, a plurality of hidden layer units and at least one output unit for the value of l*.",2013-04-23,"The title of the patent is modeling of the radiation belt megnetosphere in decisional timeframes and its abstract is systems and methods for calculating l* in the magnetosphere with essentially the same accuracy as with a physics based model at many times the speed by developing a surrogate trained to be a surrogate for the physics-based model. the trained model can then beneficially process input data falling within the training range of the surrogate model. the surrogate model can be a feedforward neural network and the physics-based model can be the tsk03 model. operatively, the surrogate model can use parameters on which the physics-based model was based, and/or spatial data for the location where l* is to be calculated. surrogate models should be provided for each of a plurality of pitch angles. accordingly, a surrogate model having a closed drift shell can be used from the plurality of models. the feedforward neural network can have a plurality of input-layer units, there being at least one input-layer unit for each physics-based model parameter, a plurality of hidden layer units and at least one output unit for the value of l*. dated 2013-04-23"
8428935,neural network for classifying speech and textural data based on agglomerates in a taxonomy table,"a speech and textual analysis device and method for forming a search and/or classification catalog. the device is based on a linguistic database and includes a taxonomy table containing variable taxon nodes. the speech and textual analysis device includes a weighting module, a weighting parameter being additionally assigned to each stored taxon node to register recurrence frequency of terms in the linguistic and/or textual data that is to be classified and/or sorted. the speech and/or textual analysis device includes an integration module for determining a predefinable number of agglomerates based on the weighting parameters of the taxon nodes in the taxonomy table and at least one neuronal network module for classifying and/or sorting the speech and/or textual data based on the agglomerates in the taxonomy table.",2013-04-23,"The title of the patent is neural network for classifying speech and textural data based on agglomerates in a taxonomy table and its abstract is a speech and textual analysis device and method for forming a search and/or classification catalog. the device is based on a linguistic database and includes a taxonomy table containing variable taxon nodes. the speech and textual analysis device includes a weighting module, a weighting parameter being additionally assigned to each stored taxon node to register recurrence frequency of terms in the linguistic and/or textual data that is to be classified and/or sorted. the speech and/or textual analysis device includes an integration module for determining a predefinable number of agglomerates based on the weighting parameters of the taxon nodes in the taxonomy table and at least one neuronal network module for classifying and/or sorting the speech and/or textual data based on the agglomerates in the taxonomy table. dated 2013-04-23"
8433121,method for brightness level calculation in the area of interest of the digital x-ray image for medical applications,"the invention relates to methods for evaluation a level of brightness in the area of interest of the digital x-ray image for medical applications by means of the image histogram using a neural network. the calculations comprise of: image acquisition, image histogram calculation, converting histogram values into input arguments of the neural network and output values of the neural network acquiring. as input arguments of the neural network the histogram values calculated with the given bin width and normalized to unity are used. the level of brightness is calculated as a linear function of the output value of the neural network. neural network learning is performed using a learning set calculated on the base of the given image database; as a set of target values the levels of brightness calculated for each image over the area of interest and scaled to the range of the activation function of a neuron in the output layer of the neural network are used.",2013-04-30,"The title of the patent is method for brightness level calculation in the area of interest of the digital x-ray image for medical applications and its abstract is the invention relates to methods for evaluation a level of brightness in the area of interest of the digital x-ray image for medical applications by means of the image histogram using a neural network. the calculations comprise of: image acquisition, image histogram calculation, converting histogram values into input arguments of the neural network and output values of the neural network acquiring. as input arguments of the neural network the histogram values calculated with the given bin width and normalized to unity are used. the level of brightness is calculated as a linear function of the output value of the neural network. neural network learning is performed using a learning set calculated on the base of the given image database; as a set of target values the levels of brightness calculated for each image over the area of interest and scaled to the range of the activation function of a neuron in the output layer of the neural network are used. dated 2013-04-30"
8442667,applications of neural networks,"software for controlling processes in a heterogeneous semiconductor manufacturing environment may include a wafer-centric database, a real-time scheduler using a neural network, and a graphical user interface displaying simulated operation of the system. these features may be employed alone or in combination to offer improved usability and computational efficiency for real time control and monitoring of a semiconductor manufacturing process. more generally, these techniques may be usefully employed in a variety of real time control systems, particularly systems requiring complex scheduling decisions or heterogeneous systems constructed of hardware from numerous independent vendors.",2013-05-14,"The title of the patent is applications of neural networks and its abstract is software for controlling processes in a heterogeneous semiconductor manufacturing environment may include a wafer-centric database, a real-time scheduler using a neural network, and a graphical user interface displaying simulated operation of the system. these features may be employed alone or in combination to offer improved usability and computational efficiency for real time control and monitoring of a semiconductor manufacturing process. more generally, these techniques may be usefully employed in a variety of real time control systems, particularly systems requiring complex scheduling decisions or heterogeneous systems constructed of hardware from numerous independent vendors. dated 2013-05-14"
8442821,multi-frame prediction for hybrid neural network/hidden markov models,"a method and system for multi-frame prediction in a hybrid neural network/hidden markov model automatic speech recognition (asr) system is disclosed. an audio input signal may be transformed into a time sequence of feature vectors, each corresponding to respective temporal frame of a sequence of periodic temporal frames of the audio input signal. the time sequence of feature vectors may be concurrently input to a neural network, which may process them concurrently. in particular, the neural network may concurrently determine for the time sequence of feature vectors a set of emission probabilities for a plurality of hidden markov models of the asr system, where the set of emission probabilities are associated with the temporal frames. the set of emission probabilities may then be concurrently applied to the hidden markov models for determining speech content of the audio input signal.",2013-05-14,"The title of the patent is multi-frame prediction for hybrid neural network/hidden markov models and its abstract is a method and system for multi-frame prediction in a hybrid neural network/hidden markov model automatic speech recognition (asr) system is disclosed. an audio input signal may be transformed into a time sequence of feature vectors, each corresponding to respective temporal frame of a sequence of periodic temporal frames of the audio input signal. the time sequence of feature vectors may be concurrently input to a neural network, which may process them concurrently. in particular, the neural network may concurrently determine for the time sequence of feature vectors a set of emission probabilities for a plurality of hidden markov models of the asr system, where the set of emission probabilities are associated with the temporal frames. the set of emission probabilities may then be concurrently applied to the hidden markov models for determining speech content of the audio input signal. dated 2013-05-14"
8442825,biomimetic voice identifier,"a device for voice identification including a receiver, a segmenter, a resolver, two advancers, a buffer, and a plurality of iir resonator digital filters where each iir filter comprises a set of memory locations or functional equivalent to hold filter specifications, a memory location or functional equivalent to hold the arithmetic reciprocal of the filter's gain, a five cell controller array, several multipliers, an adder, a subtractor, and a logical non-shift register. each cell of the five cell controller array has five logical states, each acting as a five-position single-pole rotating switch that operates in unison with the four others. additionally, the device also includes an artificial neural network and a display means.",2013-05-14,"The title of the patent is biomimetic voice identifier and its abstract is a device for voice identification including a receiver, a segmenter, a resolver, two advancers, a buffer, and a plurality of iir resonator digital filters where each iir filter comprises a set of memory locations or functional equivalent to hold filter specifications, a memory location or functional equivalent to hold the arithmetic reciprocal of the filter's gain, a five cell controller array, several multipliers, an adder, a subtractor, and a logical non-shift register. each cell of the five cell controller array has five logical states, each acting as a five-position single-pole rotating switch that operates in unison with the four others. additionally, the device also includes an artificial neural network and a display means. dated 2013-05-14"
8442891,intermarket analysis,"a method and a system for performing intermarket analysis. the method can include, from a pool of available markets in which at least one key intermarket has been selected and removed, selecting at least one market as a general intermarket, and removing the market selected as the general intermarket from the pool of available markets. from the pool of available markets from which the general intermarket has been removed, at least one market can be selected as a predictive intermarket and removed from the pool of available markets. market data for each of the key intermarket, the general intermarket and the predictive intermarket can be processed to train a neural network. after training the neural network, market data for the primary market can be processed with the neural network to predict future market data for the primary market. the predicted future market data can be output.",2013-05-14,"The title of the patent is intermarket analysis and its abstract is a method and a system for performing intermarket analysis. the method can include, from a pool of available markets in which at least one key intermarket has been selected and removed, selecting at least one market as a general intermarket, and removing the market selected as the general intermarket from the pool of available markets. from the pool of available markets from which the general intermarket has been removed, at least one market can be selected as a predictive intermarket and removed from the pool of available markets. market data for each of the key intermarket, the general intermarket and the predictive intermarket can be processed to train a neural network. after training the neural network, market data for the primary market can be processed with the neural network to predict future market data for the primary market. the predicted future market data can be output. dated 2013-05-14"
8443169,interconnection network connecting operation-configurable nodes according to one or more levels of adjacency in multiple dimensions of communication in a multi-processor and a neural processor,"a wings array system for communicating between nodes using store and load instructions is described. couplings between nodes are made according to a 1 to n adjacency of connections in each dimension of a g×h matrix of nodes, where g≧n and h≧n and n is a positive odd integer. also, a 3d wings neural network processor is described as a 3d g×h×k network of neurons, each neuron with an n×n×n array of synaptic weight values stored in coupled memory nodes, where g≧n, h≧n, k≧n, and n is determined from a 1 to n adjacency of connections used in the g×h×k network. further, a hexagonal processor array is organized according to an inform coordinate system having axes at 60 degree spacing. nodes communicate on row paths parallel to an fm dimension of communication, column paths parallel to an io dimension of communication, and diagonal paths parallel to an nr dimension of communication.",2013-05-14,"The title of the patent is interconnection network connecting operation-configurable nodes according to one or more levels of adjacency in multiple dimensions of communication in a multi-processor and a neural processor and its abstract is a wings array system for communicating between nodes using store and load instructions is described. couplings between nodes are made according to a 1 to n adjacency of connections in each dimension of a g×h matrix of nodes, where g≧n and h≧n and n is a positive odd integer. also, a 3d wings neural network processor is described as a 3d g×h×k network of neurons, each neuron with an n×n×n array of synaptic weight values stored in coupled memory nodes, where g≧n, h≧n, k≧n, and n is determined from a 1 to n adjacency of connections used in the g×h×k network. further, a hexagonal processor array is organized according to an inform coordinate system having axes at 60 degree spacing. nodes communicate on row paths parallel to an fm dimension of communication, column paths parallel to an io dimension of communication, and diagonal paths parallel to an nr dimension of communication. dated 2013-05-14"
8447406,medical method and device for monitoring a neural brain network,"bioelectrical signals may be sensed within the brain by two or more electrodes to determine characteristics of a function of the brain. the signals obtained by the electrodes may be plotted over time to determine whether the brain function exhibits a normal or an abnormal pattern. if the brain function exhibits an abnormal pattern, an implantable medical device may dynamically determine based on the abnormal pattern and a previously-obtained plot associated with normal brain function, an appropriate electrical stimulation therapy. application of the appropriate electrical stimulation therapy causes the brain function to shift from the abnormal pattern to the normal pattern.",2013-05-21,"The title of the patent is medical method and device for monitoring a neural brain network and its abstract is bioelectrical signals may be sensed within the brain by two or more electrodes to determine characteristics of a function of the brain. the signals obtained by the electrodes may be plotted over time to determine whether the brain function exhibits a normal or an abnormal pattern. if the brain function exhibits an abnormal pattern, an implantable medical device may dynamically determine based on the abnormal pattern and a previously-obtained plot associated with normal brain function, an appropriate electrical stimulation therapy. application of the appropriate electrical stimulation therapy causes the brain function to shift from the abnormal pattern to the normal pattern. dated 2013-05-21"
8447431,method for sootblowing optimization,"a controller determines and adjusts system parameters, including cleanliness levels or sootblower operating settings, that are useful for maintaining the cleanliness of a fossil fuel boiler at an efficient level. some embodiments use a direct controller to determine cleanliness levels and/or sootblower operating settings. some embodiments use an indirect controller, with a system model, to determine cleanliness levels and/or sootblower settings. the controller may use a model that is, for example, a neural network, or a mass energy balance, or a genetically programmed model. the controller uses input about the actual performance or state of the boiler for adaptation. the controller may operate in conjunction with a sootblower optimization system that controls the actual settings of the sootblowers. the controller may coordinate cleanliness settings for multiple sootblowers and/or across a plurality of heat zones in the boiler.",2013-05-21,"The title of the patent is method for sootblowing optimization and its abstract is a controller determines and adjusts system parameters, including cleanliness levels or sootblower operating settings, that are useful for maintaining the cleanliness of a fossil fuel boiler at an efficient level. some embodiments use a direct controller to determine cleanliness levels and/or sootblower operating settings. some embodiments use an indirect controller, with a system model, to determine cleanliness levels and/or sootblower settings. the controller may use a model that is, for example, a neural network, or a mass energy balance, or a genetically programmed model. the controller uses input about the actual performance or state of the boiler for adaptation. the controller may operate in conjunction with a sootblower optimization system that controls the actual settings of the sootblowers. the controller may coordinate cleanliness settings for multiple sootblowers and/or across a plurality of heat zones in the boiler. dated 2013-05-21"
8447706,method for computer-aided control and/or regulation using two neural networks wherein the second neural network models a quality function and can be used to control a gas turbine,"a method for a computer-aided control of a technical system is provided. the method involves use of a cooperative learning method and artificial neural networks. in this context, feed-forward networks are linked to one another such that the architecture as a whole meets an optimality criterion. the network approximates the rewards observed to the expected rewards as an appraiser. in this way, exclusively observations which have actually been made are used in optimum fashion to determine a quality function. in the network, the optimum action in respect of the quality function is modeled by a neural network, the neural network supplying the optimum action selection rule for the given control problem. the method is specifically used to control a gas turbine.",2013-05-21,"The title of the patent is method for computer-aided control and/or regulation using two neural networks wherein the second neural network models a quality function and can be used to control a gas turbine and its abstract is a method for a computer-aided control of a technical system is provided. the method involves use of a cooperative learning method and artificial neural networks. in this context, feed-forward networks are linked to one another such that the architecture as a whole meets an optimality criterion. the network approximates the rewards observed to the expected rewards as an appraiser. in this way, exclusively observations which have actually been made are used in optimum fashion to determine a quality function. in the network, the optimum action in respect of the quality function is modeled by a neural network, the neural network supplying the optimum action selection rule for the given control problem. the method is specifically used to control a gas turbine. dated 2013-05-21"
8447713,automated legal evaluation using a neural network over a communications network,"a method for legal knowledge modeling and automated legal evaluation, such as for online, questionnaire-based legal analysis, is provided. information, such as facts and characteristics of a legal situation, as it relates to a legal conclusion, are modeled in an artificial neural network. the artificial neural network may comprise a plurality of nodes, wherein each node is associated with one or more weights and a function that calculates a legal conclusion based on one or more input values and the weights. the artificial neural network is automatically updated on a periodic basis to reflect new legislation or court decisions. using the artificial neural network, a legal conclusion based on the user's answers to a questionnaire may be determined. the legal conclusion is modified upon the input of evidence, which is in the form of answers to a set of questions designed to identify a legal conclusion.",2013-05-21,"The title of the patent is automated legal evaluation using a neural network over a communications network and its abstract is a method for legal knowledge modeling and automated legal evaluation, such as for online, questionnaire-based legal analysis, is provided. information, such as facts and characteristics of a legal situation, as it relates to a legal conclusion, are modeled in an artificial neural network. the artificial neural network may comprise a plurality of nodes, wherein each node is associated with one or more weights and a function that calculates a legal conclusion based on one or more input values and the weights. the artificial neural network is automatically updated on a periodic basis to reflect new legislation or court decisions. using the artificial neural network, a legal conclusion based on the user's answers to a questionnaire may be determined. the legal conclusion is modified upon the input of evidence, which is in the form of answers to a set of questions designed to identify a legal conclusion. dated 2013-05-21"
8457705,brain imaging system and methods for direct prosthesis control,"methods and systems for controlling a prosthesis using a brain imager that images a localized portion of the brain are provided according to one embodiment of the invention. the brain imager provides motor cortex activation data by illuminating the motor cortex with near infrared light (nir) and detecting the spectral changes of the nir light as passes through the brain. these spectral changes can be correlated with brain activity related to limbic control and may be provided to a neural network, for example, a fuzzy neural network that maps brain activity data to limbic control data. the limbic control data may then be used to control a prosthetic limb. other embodiments of the invention include fiber optics that provide light to and receive light from the surface of the scalp through hair.",2013-06-04,"The title of the patent is brain imaging system and methods for direct prosthesis control and its abstract is methods and systems for controlling a prosthesis using a brain imager that images a localized portion of the brain are provided according to one embodiment of the invention. the brain imager provides motor cortex activation data by illuminating the motor cortex with near infrared light (nir) and detecting the spectral changes of the nir light as passes through the brain. these spectral changes can be correlated with brain activity related to limbic control and may be provided to a neural network, for example, a fuzzy neural network that maps brain activity data to limbic control data. the limbic control data may then be used to control a prosthetic limb. other embodiments of the invention include fiber optics that provide light to and receive light from the surface of the scalp through hair. dated 2013-06-04"
8457706,estimation of a physiological parameter using a neural network,"a neural network is used to combine one or more estimates of a physiologic parameter with one or more associated signal quality metrics, creating a more accurate estimate of said physiologic parameter, as well as a second estimate of the accuracy of said physiologic parameter estimate.",2013-06-04,"The title of the patent is estimation of a physiological parameter using a neural network and its abstract is a neural network is used to combine one or more estimates of a physiologic parameter with one or more associated signal quality metrics, creating a more accurate estimate of said physiologic parameter, as well as a second estimate of the accuracy of said physiologic parameter estimate. dated 2013-06-04"
8457767,system and method for real-time industrial process modeling,"the present invention presents two new model types and a new method for evaluating a model used in the control application. these include a compound model, a hybrid model and a directional change coefficient model. the present invention allows the mixing of models with different inputs and outputs and the switching between these models based criteria for measuring optimization accuracy. the present invention allows switching between these models. the compound model is a model type that allows zooming in on the process to model parts of the data space with higher fidelity or resolution without loosing the capability to model the complete data space. the modeler does not loose any functionally over a regular neural network, but instead gains the ability to define the conditions when the model should use network weights best matched to the defined local conditions. the hybrid model is an extended version of a compound model. a hybrid model allows the combining of one or more models into a single model for purposes of interrogation or optimization. within the hybrid model may reside a compound model itself. the directional change model (dcc) allows better evaluation of the predictive capability of compound models. it may also be used with any other model type.",2013-06-04,"The title of the patent is system and method for real-time industrial process modeling and its abstract is the present invention presents two new model types and a new method for evaluating a model used in the control application. these include a compound model, a hybrid model and a directional change coefficient model. the present invention allows the mixing of models with different inputs and outputs and the switching between these models based criteria for measuring optimization accuracy. the present invention allows switching between these models. the compound model is a model type that allows zooming in on the process to model parts of the data space with higher fidelity or resolution without loosing the capability to model the complete data space. the modeler does not loose any functionally over a regular neural network, but instead gains the ability to define the conditions when the model should use network weights best matched to the defined local conditions. the hybrid model is an extended version of a compound model. a hybrid model allows the combining of one or more models into a single model for purposes of interrogation or optimization. within the hybrid model may reside a compound model itself. the directional change model (dcc) allows better evaluation of the predictive capability of compound models. it may also be used with any other model type. dated 2013-06-04"
8457795,energy-saving refrigeration through sensor-based prediction to hedge thermal and electromechanical inertia,"in one embodiment, the present invention is a retrofit to rapidly transition to existing consumer refrigerator-freezer product lines in order to greatly eliminate wasted energy. this occurs because spurious opening of the system doors allows heat to enter with the deleterious side effect of causing the compressor to cycle on and off. this, in turn, consumes more power than if such duty cycles could be predicted, which would allow for their smoothing. the invention takes advantage of existing sensor technologies and develops a computational framework for their fusion for the prediction of a dependency, which controls operation of the compressor. instances of a predictive schema are evolved and this approach allows for greater accuracy in less time than would be possible using competing neural network or support vector machine technologies. a novel evolutionary algorithm is included, which is so defined as to allow its execution on a lower-end computer.",2013-06-04,"The title of the patent is energy-saving refrigeration through sensor-based prediction to hedge thermal and electromechanical inertia and its abstract is in one embodiment, the present invention is a retrofit to rapidly transition to existing consumer refrigerator-freezer product lines in order to greatly eliminate wasted energy. this occurs because spurious opening of the system doors allows heat to enter with the deleterious side effect of causing the compressor to cycle on and off. this, in turn, consumes more power than if such duty cycles could be predicted, which would allow for their smoothing. the invention takes advantage of existing sensor technologies and develops a computational framework for their fusion for the prediction of a dependency, which controls operation of the compressor. instances of a predictive schema are evolved and this approach allows for greater accuracy in less time than would be possible using competing neural network or support vector machine technologies. a novel evolutionary algorithm is included, which is so defined as to allow its execution on a lower-end computer. dated 2013-06-04"
8460921,multinetwork nerve cell assay platform with parallel recording capability,"a neuronal network analysis plate having alternating rows of recording wells and amplifying wells. the recording wells contain a neural cell network and a series of electrodes for recording the action potential signals of the neurons. the electrodes are connected to amplifiers in adjacent amplifying wells. the close proximity of these amplifiers ideal because it permits the parallel, non-multiplexed recording of action potential signals from multiple different active nerve cell networks. the amplifiers in the amplifying wells can then be connected to external amplification equipment. the neuronal network analysis plate may be contained within a single commercially available 24 or 96 well plate. the neuronal network analysis plate can be used to detect and quantify pharmacological and toxicological responses of the neural cells to one or more agents in vitro.",2013-06-11,"The title of the patent is multinetwork nerve cell assay platform with parallel recording capability and its abstract is a neuronal network analysis plate having alternating rows of recording wells and amplifying wells. the recording wells contain a neural cell network and a series of electrodes for recording the action potential signals of the neurons. the electrodes are connected to amplifiers in adjacent amplifying wells. the close proximity of these amplifiers ideal because it permits the parallel, non-multiplexed recording of action potential signals from multiple different active nerve cell networks. the amplifiers in the amplifying wells can then be connected to external amplification equipment. the neuronal network analysis plate may be contained within a single commercially available 24 or 96 well plate. the neuronal network analysis plate can be used to detect and quantify pharmacological and toxicological responses of the neural cells to one or more agents in vitro. dated 2013-06-11"
8463025,distributed artificial intelligence services on a cell phone,a cell phone having distributed artificial intelligence services is provided. the cell phone includes a neural network for performing a first pass of object recognition on an image to identify objects of interest therein based on one or more criterion. the cell phone also includes a patch generator for deriving patches from the objects of interest. each of the patches includes a portion of a respective one of the objects of interest. the cell phone additionally includes a transmitter for transmitting the patches to a server for further processing in place of an entirety of the image to reduce network traffic.,2013-06-11,The title of the patent is distributed artificial intelligence services on a cell phone and its abstract is a cell phone having distributed artificial intelligence services is provided. the cell phone includes a neural network for performing a first pass of object recognition on an image to identify objects of interest therein based on one or more criterion. the cell phone also includes a patch generator for deriving patches from the objects of interest. each of the patches includes a portion of a respective one of the objects of interest. the cell phone additionally includes a transmitter for transmitting the patches to a server for further processing in place of an entirety of the image to reduce network traffic. dated 2013-06-11
8463678,generating method for transaction models with indicators for option,"the invention is to provide a generating method for transaction models with indicators for option. the method comprises: (s1) collecting a variety of indicators from the financial market; (s2) establishing an indicator pool saved with the indicators; (s3) distinguishing and classifying the indicators by neural network; (s4) determining a plurality of transaction models, the indicator of each transaction model in an independent classification; (s5) determining a date parameter for each indicator; (s6) deleting a part of the transaction models; (s7) determining a plurality of final transaction models from another transaction models; and (s8) determining a weight of each final transaction model.",2013-06-11,"The title of the patent is generating method for transaction models with indicators for option and its abstract is the invention is to provide a generating method for transaction models with indicators for option. the method comprises: (s1) collecting a variety of indicators from the financial market; (s2) establishing an indicator pool saved with the indicators; (s3) distinguishing and classifying the indicators by neural network; (s4) determining a plurality of transaction models, the indicator of each transaction model in an independent classification; (s5) determining a date parameter for each indicator; (s6) deleting a part of the transaction models; (s7) determining a plurality of final transaction models from another transaction models; and (s8) determining a weight of each final transaction model. dated 2013-06-11"
8463721,systems and methods for recognizing events,systems and methods for recognizing events include a processor for executing machine readable instructions. the processor may be electronically coupled to an electronic memory. a temporal sensor may be electronically coupled to the processor for generating a sequence of temporal signals relating to an unrecognized event. the temporal sensor may transmit the sequence of temporal signals to the processor. the processor may execute the machine readable instructions to: input the sequence of temporal signals relating to the unrecognized event to a recurrent neural network; transform the sequence of temporal signals relating to the unrecognized event to a neural output relating to the unrecognized event with the recurrent neural network; input the neural output relating to the unrecognized event into a random forest classifier; and recognize a recognized event based upon a transformation of the neural output relating to the unrecognized event with the random forest classifier.,2013-06-11,The title of the patent is systems and methods for recognizing events and its abstract is systems and methods for recognizing events include a processor for executing machine readable instructions. the processor may be electronically coupled to an electronic memory. a temporal sensor may be electronically coupled to the processor for generating a sequence of temporal signals relating to an unrecognized event. the temporal sensor may transmit the sequence of temporal signals to the processor. the processor may execute the machine readable instructions to: input the sequence of temporal signals relating to the unrecognized event to a recurrent neural network; transform the sequence of temporal signals relating to the unrecognized event to a neural output relating to the unrecognized event with the recurrent neural network; input the neural output relating to the unrecognized event into a random forest classifier; and recognize a recognized event based upon a transformation of the neural output relating to the unrecognized event with the random forest classifier. dated 2013-06-11
8463722,implementing a neural associative memory based on non-linear learning of discrete synapses,"this invention is in the field of machine learning and neural associative memory. in particular the invention discloses a neural associative memory structure for storing and maintaining associations between memory address patterns and memory content patterns using a neural network, as well as methods for retrieving such associations. a method for a non-linear synaptic learning of discrete synapses is disclosed, and its application on neural networks is laid out.",2013-06-11,"The title of the patent is implementing a neural associative memory based on non-linear learning of discrete synapses and its abstract is this invention is in the field of machine learning and neural associative memory. in particular the invention discloses a neural associative memory structure for storing and maintaining associations between memory address patterns and memory content patterns using a neural network, as well as methods for retrieving such associations. a method for a non-linear synaptic learning of discrete synapses is disclosed, and its application on neural networks is laid out. dated 2013-06-11"
8468108,modeling efficiency over a range of velocities in underwater vehicles,"a method of generating a model of propulsive efficiency for an autonomous underwater vehicle (auv) is based on a multilayer perception neural network model using data from aquatic species, such as undulatory fin propulsion in the knifefish (xenomystus nigri), and a sensitivity analysis is used to lower the number of required inputs. the model of propulsive efficiency allows an auv to achieve high values of propulsive efficiency over a range of forward velocity, giving a lowered energy drain on the battery. in an embodiment, externally monitored information, such as that on flow velocity, is conveyed to an apparatus residing in the vehicle's control unit, which in turn signals the locomotive unit to adopt kinematics, such as fin frequency and amplitude, associated with optimal propulsion efficiency. power savings could protract vehicle operational life and/or provide more power to other functions, such as communications.",2013-06-18,"The title of the patent is modeling efficiency over a range of velocities in underwater vehicles and its abstract is a method of generating a model of propulsive efficiency for an autonomous underwater vehicle (auv) is based on a multilayer perception neural network model using data from aquatic species, such as undulatory fin propulsion in the knifefish (xenomystus nigri), and a sensitivity analysis is used to lower the number of required inputs. the model of propulsive efficiency allows an auv to achieve high values of propulsive efficiency over a range of forward velocity, giving a lowered energy drain on the battery. in an embodiment, externally monitored information, such as that on flow velocity, is conveyed to an apparatus residing in the vehicle's control unit, which in turn signals the locomotive unit to adopt kinematics, such as fin frequency and amplitude, associated with optimal propulsion efficiency. power savings could protract vehicle operational life and/or provide more power to other functions, such as communications. dated 2013-06-18"
8468109,"architecture, system and method for artificial neural network implementation","systems and methods for a scalable artificial neural network, wherein the architecture includes: an input layer; at least one hidden layer; an output layer; and a parallelization subsystem configured to provide a variable degree of parallelization to the artificial neural network by providing scalability to neurons and layers. in a particular case, the systems and methods may include a back-propagation subsystem that is configured to scalably adjust weights in the artificial neural network in accordance with the variable degree of parallelization. systems and methods are also provided for selecting an appropriate degree of parallelization based on factors such as hardware resources and performance requirements.",2013-06-18,"The title of the patent is architecture, system and method for artificial neural network implementation and its abstract is systems and methods for a scalable artificial neural network, wherein the architecture includes: an input layer; at least one hidden layer; an output layer; and a parallelization subsystem configured to provide a variable degree of parallelization to the artificial neural network by providing scalability to neurons and layers. in a particular case, the systems and methods may include a back-propagation subsystem that is configured to scalably adjust weights in the artificial neural network in accordance with the variable degree of parallelization. systems and methods are also provided for selecting an appropriate degree of parallelization based on factors such as hardware resources and performance requirements. dated 2013-06-18"
8473270,methods and systems for controlling a semiconductor fabrication process,"software for controlling processes in a heterogeneous semiconductor manufacturing environment may include a wafer-centric database, a real-time scheduler using a neural network, and a graphical user interface displaying simulated operation of the system. these features may be employed alone or in combination to offer improved usability and computational efficiency for real time control and monitoring of a semiconductor manufacturing process. more generally, these techniques may be usefully employed in a variety of real time control systems, particularly systems requiring complex scheduling decisions or heterogeneous systems constructed of hardware from numerous independent vendors.",2013-06-25,"The title of the patent is methods and systems for controlling a semiconductor fabrication process and its abstract is software for controlling processes in a heterogeneous semiconductor manufacturing environment may include a wafer-centric database, a real-time scheduler using a neural network, and a graphical user interface displaying simulated operation of the system. these features may be employed alone or in combination to offer improved usability and computational efficiency for real time control and monitoring of a semiconductor manufacturing process. more generally, these techniques may be usefully employed in a variety of real time control systems, particularly systems requiring complex scheduling decisions or heterogeneous systems constructed of hardware from numerous independent vendors. dated 2013-06-25"
8489528,systems and methods for training neural networks based on concurrent use of current and recorded data,"various embodiments of the invention are neural network adaptive control systems and methods configured to concurrently consider both recorded and current data, so that persistent excitation is not required. a neural network adaptive control system of the present invention can specifically select and record data that has as many linearly independent elements as the dimension of the basis of the uncertainty. using this recorded data along with current data, the neural network adaptive control system can guarantee global exponential parameter convergence in adaptive parameter estimation problems. other embodiments of the neural network adaptive control system are also disclosed.",2013-07-16,"The title of the patent is systems and methods for training neural networks based on concurrent use of current and recorded data and its abstract is various embodiments of the invention are neural network adaptive control systems and methods configured to concurrently consider both recorded and current data, so that persistent excitation is not required. a neural network adaptive control system of the present invention can specifically select and record data that has as many linearly independent elements as the dimension of the basis of the uncertainty. using this recorded data along with current data, the neural network adaptive control system can guarantee global exponential parameter convergence in adaptive parameter estimation problems. other embodiments of the neural network adaptive control system are also disclosed. dated 2013-07-16"
8490194,"method and system for detecting malicious behavioral patterns in a computer, using machine learning","method for detecting malicious behavioral patterns which are related to malicious software such as a computer worm in computerized systems that include data exchange channels with other systems over a data network. accordingly, hardware and/or software parameters are determined in the computerized system that is can characterize known behavioral patterns thereof. known malicious code samples are learned by a machine learning process, such as decision trees and artificial neural networks, and the results of the machine learning process are analyzed in respect to the behavioral patterns of the computerized system. then known and unknown malicious code samples are identified according to the results of the machine learning process.",2013-07-16,"The title of the patent is method and system for detecting malicious behavioral patterns in a computer, using machine learning and its abstract is method for detecting malicious behavioral patterns which are related to malicious software such as a computer worm in computerized systems that include data exchange channels with other systems over a data network. accordingly, hardware and/or software parameters are determined in the computerized system that is can characterize known behavioral patterns thereof. known malicious code samples are learned by a machine learning process, such as decision trees and artificial neural networks, and the results of the machine learning process are analyzed in respect to the behavioral patterns of the computerized system. then known and unknown malicious code samples are identified according to the results of the machine learning process. dated 2013-07-16"
8494257,music score deconstruction,"data set generation and data set presentation for image processing are described. the processing determines a location for each of one or more musical artifacts in the image and identifies a corresponding label for each of the musical artifacts, generating a training file that associates the identified labels and determined locations of the musical artifacts with the image, and presenting the training file to a neural network for training.",2013-07-23,"The title of the patent is music score deconstruction and its abstract is data set generation and data set presentation for image processing are described. the processing determines a location for each of one or more musical artifacts in the image and identifies a corresponding label for each of the musical artifacts, generating a training file that associates the identified labels and determined locations of the musical artifacts with the image, and presenting the training file to a neural network for training. dated 2013-07-23"
8503797,automatic document classification using lexical and physical features,"an automatic document classification system is described that uses lexical and physical features to assign a class ciεc{c1, c2, . . . , ci} to a document d. the primary lexical features are the result of a feature selection method known as orthogonal centroid feature selection (ocfs). additional information may be gathered on character type frequencies (digits, letters, and symbols) within d. physical information is assembled through image analysis to yield physical attributes such as document dimensionality, text alignment, and color distribution. the resulting lexical and physical information is combined into an input vector x and is used to train a supervised neural network to perform the classification.",2013-08-06,"The title of the patent is automatic document classification using lexical and physical features and its abstract is an automatic document classification system is described that uses lexical and physical features to assign a class ciεc{c1, c2, . . . , ci} to a document d. the primary lexical features are the result of a feature selection method known as orthogonal centroid feature selection (ocfs). additional information may be gathered on character type frequencies (digits, letters, and symbols) within d. physical information is assembled through image analysis to yield physical attributes such as document dimensionality, text alignment, and color distribution. the resulting lexical and physical information is combined into an input vector x and is used to train a supervised neural network to perform the classification. dated 2013-08-06"
8504331,aerodynamic model identification process for aircraft simulation process,"a method is disclosed of identifying or fixing the value of the parameters (aerodynamic data) of a pre-determined set of equations forming the aerodynamic models of an aircraft in various configurations, so as to minimize the variance between the values anticipated by using these aerodynamic data, and reference data in cases of known configurations, includes the use of learning or optimization methods to determine at least one portion of the parameters' values. the method includes storing database data, choosing an optimization method or a neural network method, breaking the identification down into several partial identifications, carrying out successive partial identifications, validating the values found for the aerodynamic model's parameters or iterating the method according to a threshold criterion determined on the value of variances between the reference data and the values measured by the identified model.",2013-08-06,"The title of the patent is aerodynamic model identification process for aircraft simulation process and its abstract is a method is disclosed of identifying or fixing the value of the parameters (aerodynamic data) of a pre-determined set of equations forming the aerodynamic models of an aircraft in various configurations, so as to minimize the variance between the values anticipated by using these aerodynamic data, and reference data in cases of known configurations, includes the use of learning or optimization methods to determine at least one portion of the parameters' values. the method includes storing database data, choosing an optimization method or a neural network method, breaking the identification down into several partial identifications, carrying out successive partial identifications, validating the values found for the aerodynamic model's parameters or iterating the method according to a threshold criterion determined on the value of variances between the reference data and the values measured by the identified model. dated 2013-08-06"
8504361,deep neural networks and methods for using same,"a method and system for labeling a selected word of a sentence using a deep neural network includes, in one exemplary embodiment, determining an index term corresponding to each feature of the word, transforming the index term or terms of the word into a vector, and predicting a label for the word using the vector. the method and system, in another exemplary embodiment, includes determining, for each word in the sentence, an index term corresponding to each feature of the word, transforming the index term or terms of each word in the sentence into a vector, applying a convolution operation to the vector of the selected word and at least one of the vectors of the other words in the sentence, to transform the vectors into a matrix of vectors, each of the vectors in the matrix including a plurality of row values, constructing a single vector from the vectors in the matrix, and predicting a label for the selected word using the single vector.",2013-08-06,"The title of the patent is deep neural networks and methods for using same and its abstract is a method and system for labeling a selected word of a sentence using a deep neural network includes, in one exemplary embodiment, determining an index term corresponding to each feature of the word, transforming the index term or terms of the word into a vector, and predicting a label for the word using the vector. the method and system, in another exemplary embodiment, includes determining, for each word in the sentence, an index term corresponding to each feature of the word, transforming the index term or terms of each word in the sentence into a vector, applying a convolution operation to the vector of the selected word and at least one of the vectors of the other words in the sentence, to transform the vectors into a matrix of vectors, each of the vectors in the matrix including a plurality of row values, constructing a single vector from the vectors in the matrix, and predicting a label for the selected word using the single vector. dated 2013-08-06"
8504499,constant memory implementation of a phase-model neural network,"disclosed are systems, apparatuses, and methods for implementing a phase-model neural network using a fixed amount of memory. such a phase-model neural network includes a plurality of neurons, wherein each neuron is associated with two parameters—an activity and a phase. example methods include (i) generating a sequence of variables associated with a probability distribution of phases and (ii) sequentially sampling the probability distribution of phases using a fixed amount of memory, regardless of a number of phases used in the phase-model neural network.",2013-08-06,"The title of the patent is constant memory implementation of a phase-model neural network and its abstract is disclosed are systems, apparatuses, and methods for implementing a phase-model neural network using a fixed amount of memory. such a phase-model neural network includes a plurality of neurons, wherein each neuron is associated with two parameters—an activity and a phase. example methods include (i) generating a sequence of variables associated with a probability distribution of phases and (ii) sequentially sampling the probability distribution of phases using a fixed amount of memory, regardless of a number of phases used in the phase-model neural network. dated 2013-08-06"
8504500,"systems, methods, and apparatus for reconstruction of 3-d object morphology, position, orientation and texture using an array of tactile sensors","systems, methods, and apparatus are provided using signals from a set of tactile sensors mounted on a surface to determine the three-dimensional morphology (e.g., size, shape, orientation, and/or position) and texture of objects of arbitrary shape. analytical, numerical, and/or neural network approaches can be used to interpret the sensory data.",2013-08-06,"The title of the patent is systems, methods, and apparatus for reconstruction of 3-d object morphology, position, orientation and texture using an array of tactile sensors and its abstract is systems, methods, and apparatus are provided using signals from a set of tactile sensors mounted on a surface to determine the three-dimensional morphology (e.g., size, shape, orientation, and/or position) and texture of objects of arbitrary shape. analytical, numerical, and/or neural network approaches can be used to interpret the sensory data. dated 2013-08-06"
8509523,method of identifying an object in a visual scene,"a plurality of features determined from at least a portion of an image containing information about an object are processed with an inclusive neural network, and with a plurality of exclusive neural networks, so as to provide a plurality of inclusive probability values representing probabilities that the portion of the image corresponds to at least one of at least two different classes of objects, and for each exclusive neural network, so as to provide first and second exclusive probability values representing probabilities that the portion of the image respectively corresponds. or not. to at least one class of objects. the plurality of inclusive probability values, and the first and second exclusive probability values from each of the exclusive neural networks, provide for identifying whether the portion of the image corresponds, or not, to any of the at least two different classes of objects.",2013-08-13,"The title of the patent is method of identifying an object in a visual scene and its abstract is a plurality of features determined from at least a portion of an image containing information about an object are processed with an inclusive neural network, and with a plurality of exclusive neural networks, so as to provide a plurality of inclusive probability values representing probabilities that the portion of the image corresponds to at least one of at least two different classes of objects, and for each exclusive neural network, so as to provide first and second exclusive probability values representing probabilities that the portion of the image respectively corresponds. or not. to at least one class of objects. the plurality of inclusive probability values, and the first and second exclusive probability values from each of the exclusive neural networks, provide for identifying whether the portion of the image corresponds, or not, to any of the at least two different classes of objects. dated 2013-08-13"
8509895,ventricle pacing during atrial fibrillation episodes,"an adaptive dual chamber pacemaker and/or cardioverter defibrillator for delivering ventricular stimulation to the heart correlated with hemodynamic performance of the heart, including a hemodynamic sensor for monitoring the hemodynamic performance of the heart, an atrial electrode and a ventricular electrode for sensing ventricular and atrial signals, and a learning module having a spiking neural network processor for learning to associate the ventricular-atrial intervals sensed by the electrodes with the hemodynamic performance sensed by the hemodynamic sensor, calculating ventricular-atrial intervals, replacing the ventricular-atrial intervals calculated from the sensed ventricular and atrial signals with the learned associated ventricular-atrial intervals, and causing delivery according to the learned associated ventricular-atrial intervals of a ventricular stimulation to the heart during atrial fibrillation episodes.",2013-08-13,"The title of the patent is ventricle pacing during atrial fibrillation episodes and its abstract is an adaptive dual chamber pacemaker and/or cardioverter defibrillator for delivering ventricular stimulation to the heart correlated with hemodynamic performance of the heart, including a hemodynamic sensor for monitoring the hemodynamic performance of the heart, an atrial electrode and a ventricular electrode for sensing ventricular and atrial signals, and a learning module having a spiking neural network processor for learning to associate the ventricular-atrial intervals sensed by the electrodes with the hemodynamic performance sensed by the hemodynamic sensor, calculating ventricular-atrial intervals, replacing the ventricular-atrial intervals calculated from the sensed ventricular and atrial signals with the learned associated ventricular-atrial intervals, and causing delivery according to the learned associated ventricular-atrial intervals of a ventricular stimulation to the heart during atrial fibrillation episodes. dated 2013-08-13"
8510234,embedded health monitoring system based upon optimized neuro genetic fast estimator (ongfe),"a real time kernel for deploying health monitoring functions in condition base maintenance (cbm) and real time monitoring (rtm) systems is disclosed in this invention. the optimized neuro genetic fast estimator (ongfe) allows embedding failure detection, identification, and prognostics (fdi&p) capability by using intelligent software element (ise) based upon artificial neural network (ann). ongfe enables embedded fast and on-line training for designing anns, which perform very high performance fdi&p functions. an advantage is the optimization block based on pseudogenetic algorithms, which compensate for effects due to initial weight values and local minimums without the computational burden of genetic algorithms. it provides a synchronization block for communication with secondary diagnostic modules. also a scheme for conducting sensor data validation is embedded in smart sensors (ss). the algorithms are designed for a distributed, scalar, and modular deployment. the system electronics is built upon a network of smart sensors and a health monitoring computer for providing data acquisition capability and distributed computational power.",2013-08-13,"The title of the patent is embedded health monitoring system based upon optimized neuro genetic fast estimator (ongfe) and its abstract is a real time kernel for deploying health monitoring functions in condition base maintenance (cbm) and real time monitoring (rtm) systems is disclosed in this invention. the optimized neuro genetic fast estimator (ongfe) allows embedding failure detection, identification, and prognostics (fdi&p) capability by using intelligent software element (ise) based upon artificial neural network (ann). ongfe enables embedded fast and on-line training for designing anns, which perform very high performance fdi&p functions. an advantage is the optimization block based on pseudogenetic algorithms, which compensate for effects due to initial weight values and local minimums without the computational burden of genetic algorithms. it provides a synchronization block for communication with secondary diagnostic modules. also a scheme for conducting sensor data validation is embedded in smart sensors (ss). the algorithms are designed for a distributed, scalar, and modular deployment. the system electronics is built upon a network of smart sensors and a health monitoring computer for providing data acquisition capability and distributed computational power. dated 2013-08-13"
8510242,artificial neural network models for determining relative permeability of hydrocarbon reservoirs,"a system and method for modeling technology to predict accurately water-oil relative permeability uses a type of artificial neural network (ann) known as a generalized regression neural network (grnn) the ann models of relative permeability are developed using experimental data from waterflood core test samples collected from carbonate reservoirs of arabian oil fields three groups of data sets are used for training, verification, and testing the ann models analysis of the results of the testing data set show excellent correlation with the experimental data of relative permeability, and error analyses show these ann models outperform all published correlations",2013-08-13,"The title of the patent is artificial neural network models for determining relative permeability of hydrocarbon reservoirs and its abstract is a system and method for modeling technology to predict accurately water-oil relative permeability uses a type of artificial neural network (ann) known as a generalized regression neural network (grnn) the ann models of relative permeability are developed using experimental data from waterflood core test samples collected from carbonate reservoirs of arabian oil fields three groups of data sets are used for training, verification, and testing the ann models analysis of the results of the testing data set show excellent correlation with the experimental data of relative permeability, and error analyses show these ann models outperform all published correlations dated 2013-08-13"
8514392,spectrophotopolarimeter sensor and artificial neural network analytics for distant chemical and biological threat detection,"a system, apparatus, and method of generating stokes vectors, a mueller matrix, and polarized scattering from an aerosol aggregate includes providing an incident infrared laser beam; causing the incident infrared laser beam to be polarization-modulated using variable stress/strain birefringence imposed on a znse crystal; defining a stokes vector associated with the incident infrared laser beam; scattering the incident infrared laser beam from an aggregate aerosol comprising interferents and analyte particles; producing a scattered-beam reactant stokes vector by causing the scattered incident infrared laser beam to be polarization-modulated; generating a mueller matrix by taking a transformation of the stokes vector; and identifying the analyte using the mueller matrix. the mueller matrix may comprise m-elements that are functions of a wavelength of the infrared laser beam, backsattering orientation of the infrared laser beam, and a shape and size of the interferents and analyte particles.",2013-08-20,"The title of the patent is spectrophotopolarimeter sensor and artificial neural network analytics for distant chemical and biological threat detection and its abstract is a system, apparatus, and method of generating stokes vectors, a mueller matrix, and polarized scattering from an aerosol aggregate includes providing an incident infrared laser beam; causing the incident infrared laser beam to be polarization-modulated using variable stress/strain birefringence imposed on a znse crystal; defining a stokes vector associated with the incident infrared laser beam; scattering the incident infrared laser beam from an aggregate aerosol comprising interferents and analyte particles; producing a scattered-beam reactant stokes vector by causing the scattered incident infrared laser beam to be polarization-modulated; generating a mueller matrix by taking a transformation of the stokes vector; and identifying the analyte using the mueller matrix. the mueller matrix may comprise m-elements that are functions of a wavelength of the infrared laser beam, backsattering orientation of the infrared laser beam, and a shape and size of the interferents and analyte particles. dated 2013-08-20"
8515576,surgical robot and robotic controller,"the present invention was developed by a neurosurgeon and seeks to mimic the results of primate neurological research which is indicative of a human's actual neurological control structures and logic. specifically, the motor proprioceptive and tactile neurophysiology functioning of the surgeon's hands and internal hand control system from the muscular level through the intrafusal fiber system of the neural network is considered in creating the robot and method of operation of the present invention. therefore, the surgery is not slowed down as in the art, because the surgeon is in conscious and subconscious natural agreement and harmonization with the robotically actuated surgical instruments based on neurological mimicking of the surgeon's behavior with the functioning of the robot. therefore, the robot can enhance the surgeon's humanly limited senses while not introducing disruptive variables to the surgeon's naturally occurring operation of his neurophysiology. this is therefore also a new field, neurophysiological symbiotic robotics.",2013-08-20,"The title of the patent is surgical robot and robotic controller and its abstract is the present invention was developed by a neurosurgeon and seeks to mimic the results of primate neurological research which is indicative of a human's actual neurological control structures and logic. specifically, the motor proprioceptive and tactile neurophysiology functioning of the surgeon's hands and internal hand control system from the muscular level through the intrafusal fiber system of the neural network is considered in creating the robot and method of operation of the present invention. therefore, the surgery is not slowed down as in the art, because the surgeon is in conscious and subconscious natural agreement and harmonization with the robotically actuated surgical instruments based on neurological mimicking of the surgeon's behavior with the functioning of the robot. therefore, the robot can enhance the surgeon's humanly limited senses while not introducing disruptive variables to the surgeon's naturally occurring operation of his neurophysiology. this is therefore also a new field, neurophysiological symbiotic robotics. dated 2013-08-20"
8515676,method and apparatus for assessing the integrity of a rock mass,"a method for assessing the integrity of a rock mass, the method including impacting the rock mass, capturing an acoustic signal generated as a result of the impact, deriving a frequency distribution for the captured acoustic signal, processing data from the frequency distribution by means of a neural network process applying artificial intelligence to assess the inputted data, and presenting a signal from the neural network process which is indicative of the integrity of the rock mass.",2013-08-20,"The title of the patent is method and apparatus for assessing the integrity of a rock mass and its abstract is a method for assessing the integrity of a rock mass, the method including impacting the rock mass, capturing an acoustic signal generated as a result of the impact, deriving a frequency distribution for the captured acoustic signal, processing data from the frequency distribution by means of a neural network process applying artificial intelligence to assess the inputted data, and presenting a signal from the neural network process which is indicative of the integrity of the rock mass. dated 2013-08-20"
8515793,virtual production control system and method and computer program product thereof,"a virtual production control system (vpcs), and a virtual production control method and a computer program product thereof are provided. at first, the vpcs processes historical work-in-process (wip) information and a current shipping plan sent from a supplier side, thereby obtaining a plurality of sets of wip input/output historical data and a goods output schedule. then, the vpcs performs an integer programming (ip) method to find the latest output schedule in accordance to the current shipping plan; uses a genetic algorithm (ga) to fit the historical distributed-parameters; adopts a neural network (nn) method to predict the future distributed-parameters of production; and finally utilizes a petri nets to simulate and obtain a latest feasible input schedule and a latest feasible output schedule.",2013-08-20,"The title of the patent is virtual production control system and method and computer program product thereof and its abstract is a virtual production control system (vpcs), and a virtual production control method and a computer program product thereof are provided. at first, the vpcs processes historical work-in-process (wip) information and a current shipping plan sent from a supplier side, thereby obtaining a plurality of sets of wip input/output historical data and a goods output schedule. then, the vpcs performs an integer programming (ip) method to find the latest output schedule in accordance to the current shipping plan; uses a genetic algorithm (ga) to fit the historical distributed-parameters; adopts a neural network (nn) method to predict the future distributed-parameters of production; and finally utilizes a petri nets to simulate and obtain a latest feasible input schedule and a latest feasible output schedule. dated 2013-08-20"
8515884,neuro type-2 fuzzy based method for decision making,"according to a first aspect of the invention there is provided a method of decision-making comprising: a data input step to input data from a plurality of first data sources into a first data bank, analysing said input data by means of a first adaptive artificial neural network (ann), the neural network including a plurality of layers having at least an input layer, one or more hidden layers and an output layer, each layer comprising a plurality of interconnected neurons, the number of hidden neurons utilized being adaptive, the ann determining the most important input data and defining therefrom a second ann, deriving from the second ann a plurality of type-1 fuzzy sets for each first data source representing the data source, combining the type-1 fuzzy sets to create footprint of uncertainty (fou) for type-2 fuzzy sets, modelling the group decision of the combined first data sources; inputting data from a second data source, and assigning an aggregate score thereto, comparing the assigned aggregate score with a fuzzy set representing the group decision, and producing a decision therefrom. a method employing a developed ann as defined in claim 1 and extracting data from said ann, the data used to learn the parameters of a normal fuzzy logic system (fls).",2013-08-20,"The title of the patent is neuro type-2 fuzzy based method for decision making and its abstract is according to a first aspect of the invention there is provided a method of decision-making comprising: a data input step to input data from a plurality of first data sources into a first data bank, analysing said input data by means of a first adaptive artificial neural network (ann), the neural network including a plurality of layers having at least an input layer, one or more hidden layers and an output layer, each layer comprising a plurality of interconnected neurons, the number of hidden neurons utilized being adaptive, the ann determining the most important input data and defining therefrom a second ann, deriving from the second ann a plurality of type-1 fuzzy sets for each first data source representing the data source, combining the type-1 fuzzy sets to create footprint of uncertainty (fou) for type-2 fuzzy sets, modelling the group decision of the combined first data sources; inputting data from a second data source, and assigning an aggregate score thereto, comparing the assigned aggregate score with a fuzzy set representing the group decision, and producing a decision therefrom. a method employing a developed ann as defined in claim 1 and extracting data from said ann, the data used to learn the parameters of a normal fuzzy logic system (fls). dated 2013-08-20"
8515885,neuromorphic and synaptronic spiking neural network with synaptic weights learned using simulation,"embodiments of the invention provide neuromorphic-synaptronic systems, including neuromorphic-synaptronic circuits implementing spiking neural network with synaptic weights learned using simulation. one embodiment includes simulating a spiking neural network to generate synaptic weights learned via the simulation while maintaining one-to-one correspondence between the simulation and a digital circuit chip. the learned synaptic weights are loaded into the digital circuit chip implementing a spiking neural network, the digital circuit chip comprising a neuromorphic-synaptronic spiking neural network including plural synapse devices interconnecting multiple digital neurons.",2013-08-20,"The title of the patent is neuromorphic and synaptronic spiking neural network with synaptic weights learned using simulation and its abstract is embodiments of the invention provide neuromorphic-synaptronic systems, including neuromorphic-synaptronic circuits implementing spiking neural network with synaptic weights learned using simulation. one embodiment includes simulating a spiking neural network to generate synaptic weights learned via the simulation while maintaining one-to-one correspondence between the simulation and a digital circuit chip. the learned synaptic weights are loaded into the digital circuit chip implementing a spiking neural network, the digital circuit chip comprising a neuromorphic-synaptronic spiking neural network including plural synapse devices interconnecting multiple digital neurons. dated 2013-08-20"
8516568,neural network data filtering and monitoring systems and methods,"systems and methods are disclosed for filtering data in a neural network environment to filter out inappropriate content. in some embodiments, a data signal including a sensible representation is received. the sensible representation included in the data signal is produced in a sensible format. from the sensible representation in the sensible format, a clean copy of the sensible representation can be generated such that any inappropriate content present within the received data signal is not reproduced in the clean copy. optionally, additional filtering can occur before and/or after the generating of the clean copy. the (filtered) clean copy of the sensible representation is sent to a network. embodiments can permit the filtering of input to and/or output from a network.",2013-08-20,"The title of the patent is neural network data filtering and monitoring systems and methods and its abstract is systems and methods are disclosed for filtering data in a neural network environment to filter out inappropriate content. in some embodiments, a data signal including a sensible representation is received. the sensible representation included in the data signal is produced in a sensible format. from the sensible representation in the sensible format, a clean copy of the sensible representation can be generated such that any inappropriate content present within the received data signal is not reproduced in the clean copy. optionally, additional filtering can occur before and/or after the generating of the clean copy. the (filtered) clean copy of the sensible representation is sent to a network. embodiments can permit the filtering of input to and/or output from a network. dated 2013-08-20"
8516584,"method and system for detecting malicious behavioral patterns in a computer, using machine learning","method for detecting malicious behavioral patterns which are related to malicious software such as a computer worm in computerized systems that include data exchange channels with other systems over a data network. according to the proposed method, hardware and/or software parameters that can characterize known behavioral patterns in the computerized system are determined. known malicious code samples are learned by a machine learning process, such as decision trees, naïve bayes, bayesian networks, and artificial neural networks, and the results of the machine learning process are analyzed in respect to these behavioral patterns. then, known and unknown malicious code samples are identified according to the results of the machine learning process.",2013-08-20,"The title of the patent is method and system for detecting malicious behavioral patterns in a computer, using machine learning and its abstract is method for detecting malicious behavioral patterns which are related to malicious software such as a computer worm in computerized systems that include data exchange channels with other systems over a data network. according to the proposed method, hardware and/or software parameters that can characterize known behavioral patterns in the computerized system are determined. known malicious code samples are learned by a machine learning process, such as decision trees, naïve bayes, bayesian networks, and artificial neural networks, and the results of the machine learning process are analyzed in respect to these behavioral patterns. then, known and unknown malicious code samples are identified according to the results of the machine learning process. dated 2013-08-20"
8520933,method for searching and constructing 3d image database,"the present invention relates to methods for searching and constructing a 3d motif image database, wherein said 3d motif image database can be used to understand the connection relationship of a 3d network, e.g. a neural network comprising biological neural networks or artificial neural networks. the searching and constructing methods are applied on the 3d motif image database, a proper computer-aided graphic platform. the database not only facilitates the management of the huge amount of categorized data but also rationally excavates the hidden information cloaked within.",2013-08-27,"The title of the patent is method for searching and constructing 3d image database and its abstract is the present invention relates to methods for searching and constructing a 3d motif image database, wherein said 3d motif image database can be used to understand the connection relationship of a 3d network, e.g. a neural network comprising biological neural networks or artificial neural networks. the searching and constructing methods are applied on the 3d motif image database, a proper computer-aided graphic platform. the database not only facilitates the management of the huge amount of categorized data but also rationally excavates the hidden information cloaked within. dated 2013-08-27"
8521542,systems and methods for classifying account data using artificial neural networks,"systems, methods, and articles are provided for classifying account data using artificial neural networks. an example embodiment may include receiving account holder data for a plurality of account holders, identifying through computer automated operations relationships between the plurality of account holders and the account holder data, and analyzing the account holder data of the plurality of account holders to create one or more classifications based on the relationships between the plurality of account holders and the account holder data. another example embodiment may include classifying financial account holder data for a plurality of financial account holders using a kohonen network, and displaying a graphical representation of the classified financial account holder data to visualize one or more relationships between plurality of financial account holders and the financial account holder data. other embodiments may be described and claimed.",2013-08-27,"The title of the patent is systems and methods for classifying account data using artificial neural networks and its abstract is systems, methods, and articles are provided for classifying account data using artificial neural networks. an example embodiment may include receiving account holder data for a plurality of account holders, identifying through computer automated operations relationships between the plurality of account holders and the account holder data, and analyzing the account holder data of the plurality of account holders to create one or more classifications based on the relationships between the plurality of account holders and the account holder data. another example embodiment may include classifying financial account holder data for a plurality of financial account holders using a kohonen network, and displaying a graphical representation of the classified financial account holder data to visualize one or more relationships between plurality of financial account holders and the financial account holder data. other embodiments may be described and claimed. dated 2013-08-27"
8521669,neural associative memories based on optimal bayesian learning,"this invention is in the field of machine learning and neural associative memory. in particular the invention discloses a neural associative memory structure for storing and maintaining associations between memory address patterns and memory content patterns using a neural network, as well as methods for storing and retrieving such associations. bayesian learning is applied to achieve non-linear learning.",2013-08-27,"The title of the patent is neural associative memories based on optimal bayesian learning and its abstract is this invention is in the field of machine learning and neural associative memory. in particular the invention discloses a neural associative memory structure for storing and maintaining associations between memory address patterns and memory content patterns using a neural network, as well as methods for storing and retrieving such associations. bayesian learning is applied to achieve non-linear learning. dated 2013-08-27"
8521670,artificial neural network application for magnetic core width prediction and modeling for magnetic disk drive manufacture,"a method for predicting and optimizing magnetic core width of a write head using neural networks to analyze manufacturing parameters, and determining new manufacturing parameters that will provide more optimal magnetic core width results. the manufacturing parameters can include: write pole flare point; wrap around shield dimension; and side gap dimension.",2013-08-27,"The title of the patent is artificial neural network application for magnetic core width prediction and modeling for magnetic disk drive manufacture and its abstract is a method for predicting and optimizing magnetic core width of a write head using neural networks to analyze manufacturing parameters, and determining new manufacturing parameters that will provide more optimal magnetic core width results. the manufacturing parameters can include: write pole flare point; wrap around shield dimension; and side gap dimension. dated 2013-08-27"
8521671,neural network for clustering input data based on a gaussian mixture model,"disclosed are systems, apparatuses, and methods for clustering data. such a method includes providing input data to each of a plurality of cluster microcircuits of a neural network, wherein each cluster microcircuit includes a mean neural group and a variance neural group. the method also includes determining a response of each cluster microcircuit with respect to the input data. the method further includes modulating the mean neural group and the variance neural group of each cluster microcircuit responsive to a value system.",2013-08-27,"The title of the patent is neural network for clustering input data based on a gaussian mixture model and its abstract is disclosed are systems, apparatuses, and methods for clustering data. such a method includes providing input data to each of a plurality of cluster microcircuits of a neural network, wherein each cluster microcircuit includes a mean neural group and a variance neural group. the method also includes determining a response of each cluster microcircuit with respect to the input data. the method further includes modulating the mean neural group and the variance neural group of each cluster microcircuit responsive to a value system. dated 2013-08-27"
8521673,parallel processing device and parallel processing method,"a parallel processing device that computes a hierarchical neural network includes: a plurality of units identified by a characteristic unit numbers; a control section that outputs control data, including an input value and a selection unit number, to the plurality of units; and a storage section that stores a plurality of coupling weights, each of the coupling weights being associated with layer information. each of the units includes: a data input section that receives control data from the control section; a unit number match judgment section that judges whether the selection unit number matches the characteristic unit number; a unit processing section that computes the output value; and a data output section that outputs the output value to the control section when the unit number judgement section judges that the selection unit number matches the characteristic unit number.",2013-08-27,"The title of the patent is parallel processing device and parallel processing method and its abstract is a parallel processing device that computes a hierarchical neural network includes: a plurality of units identified by a characteristic unit numbers; a control section that outputs control data, including an input value and a selection unit number, to the plurality of units; and a storage section that stores a plurality of coupling weights, each of the coupling weights being associated with layer information. each of the units includes: a data input section that receives control data from the control section; a unit number match judgment section that judges whether the selection unit number matches the characteristic unit number; a unit processing section that computes the output value; and a data output section that outputs the output value to the control section when the unit number judgement section judges that the selection unit number matches the characteristic unit number. dated 2013-08-27"
8527037,reconstruction of a surface electrocardiogram from an endocardial electrogram using non-linear filtering,"the present invention relates to an active medical device that uses non-linear filtering for reconstructing a surface electrocardiogram from an endocardial electrogram. at least one endocardial egm electrogram signal is collected from of samples collected from at least one endocardial or epicardial derivation (71′, 72′, 73′), and at least one of a reconstructed surface electrocardiogram (ecg) signal through the processing of collected egm samples by a transfer function (tf) of a neural network (60′). the neural network (60′) is a time-delay-type network that simultaneously processes said at least one endocardial egm electrogram signal, formed by a first sequence of collected samples, and at least one delayed version of this egm signal, formed by a second sequence of collected samples distinct from the first sequence collected samples. the neural network (60′) provides said reconstructed ecg signal from the egm signal and its delayed version.",2013-09-03,"The title of the patent is reconstruction of a surface electrocardiogram from an endocardial electrogram using non-linear filtering and its abstract is the present invention relates to an active medical device that uses non-linear filtering for reconstructing a surface electrocardiogram from an endocardial electrogram. at least one endocardial egm electrogram signal is collected from of samples collected from at least one endocardial or epicardial derivation (71′, 72′, 73′), and at least one of a reconstructed surface electrocardiogram (ecg) signal through the processing of collected egm samples by a transfer function (tf) of a neural network (60′). the neural network (60′) is a time-delay-type network that simultaneously processes said at least one endocardial egm electrogram signal, formed by a first sequence of collected samples, and at least one delayed version of this egm signal, formed by a second sequence of collected samples distinct from the first sequence collected samples. the neural network (60′) provides said reconstructed ecg signal from the egm signal and its delayed version. dated 2013-09-03"
8527276,speech synthesis using deep neural networks,"a method and system for is disclosed for speech synthesis using deep neural networks. a neural network may be trained to map input phonetic transcriptions of training-time text strings into sequences of acoustic feature vectors, which yield predefined speech waveforms when processed by a signal generation module. the training-time text strings may correspond to written transcriptions of speech carried in the predefined speech waveforms. subsequent to training, a run-time text string may be translated to a run-time phonetic transcription, which may include a run-time sequence of phonetic-context descriptors, each of which contains a phonetic speech unit, data indicating phonetic context, and data indicating time duration of the respective phonetic speech unit. the trained neural network may then map the run-time sequence of the phonetic-context descriptors to run-time predicted feature vectors, which may in turn be translated into synthesized speech by the signal generation module.",2013-09-03,"The title of the patent is speech synthesis using deep neural networks and its abstract is a method and system for is disclosed for speech synthesis using deep neural networks. a neural network may be trained to map input phonetic transcriptions of training-time text strings into sequences of acoustic feature vectors, which yield predefined speech waveforms when processed by a signal generation module. the training-time text strings may correspond to written transcriptions of speech carried in the predefined speech waveforms. subsequent to training, a run-time text string may be translated to a run-time phonetic transcription, which may include a run-time sequence of phonetic-context descriptors, each of which contains a phonetic speech unit, data indicating phonetic context, and data indicating time duration of the respective phonetic speech unit. the trained neural network may then map the run-time sequence of the phonetic-context descriptors to run-time predicted feature vectors, which may in turn be translated into synthesized speech by the signal generation module. dated 2013-09-03"
8527776,synthesis of anomalous data to create artificial feature sets and use of same in computer network intrusion detection systems,detecting harmful or illegal intrusions into a computer network or into restricted portions of a computer network uses a process of synthesizing anomalous data to be used in training a neural network-based model for use in a computer network intrusion detection system. anomalous data for artificially creating a set of features reflecting anomalous behavior for a particular activity is performed. this is done in conjunction with the creation of normal-behavior feature values. a distribution of users of normal feature values and an expected distribution of users of anomalous feature values are then defined in the form of histograms. the anomalous-feature histogram is then sampled to produce anomalous-behavior feature values. these values are then used to train a model having a neural network training algorithm where the model is used in the computer network intrusion detection system. the model is trained such that it can efficiently recognize anomalous behavior by users in a dynamic computing environment where user behavior can change frequently.,2013-09-03,The title of the patent is synthesis of anomalous data to create artificial feature sets and use of same in computer network intrusion detection systems and its abstract is detecting harmful or illegal intrusions into a computer network or into restricted portions of a computer network uses a process of synthesizing anomalous data to be used in training a neural network-based model for use in a computer network intrusion detection system. anomalous data for artificially creating a set of features reflecting anomalous behavior for a particular activity is performed. this is done in conjunction with the creation of normal-behavior feature values. a distribution of users of normal feature values and an expected distribution of users of anomalous feature values are then defined in the form of histograms. the anomalous-feature histogram is then sampled to produce anomalous-behavior feature values. these values are then used to train a model having a neural network training algorithm where the model is used in the computer network intrusion detection system. the model is trained such that it can efficiently recognize anomalous behavior by users in a dynamic computing environment where user behavior can change frequently. dated 2013-09-03
8533130,use of neural networks for annotating search results,"a system for generating annotations of a document, including a plurality of neurons connected as a neural network, the neurons being associated with words, sentences and documents. an activity regulator regulates a minimum and/or maximum number of neurons of the neural network that are excited at any given time. the neurons are displayed to a user and identify the neurons that correspond to sentences containing a predetermined percentage of document meaning. the annotations can be also based on a context of the user's search query. the query can include keywords, documents considered relevant by the user, or both. positions of the neurons relative to each other can be changed on a display device, based on input from the user, with the change in position of one neuron changing the resulting annotations. the input from the user can also include changing a relevance of neurons relative to each other, or indicating relevance or irrelevance of a document or sentence.",2013-09-10,"The title of the patent is use of neural networks for annotating search results and its abstract is a system for generating annotations of a document, including a plurality of neurons connected as a neural network, the neurons being associated with words, sentences and documents. an activity regulator regulates a minimum and/or maximum number of neurons of the neural network that are excited at any given time. the neurons are displayed to a user and identify the neurons that correspond to sentences containing a predetermined percentage of document meaning. the annotations can be also based on a context of the user's search query. the query can include keywords, documents considered relevant by the user, or both. positions of the neurons relative to each other can be changed on a display device, based on input from the user, with the change in position of one neuron changing the resulting annotations. the input from the user can also include changing a relevance of neurons relative to each other, or indicating relevance or irrelevance of a document or sentence. dated 2013-09-10"
8533137,position resolved measurement apparatus and a method for acquiring space coordinates of a quantum beam incident thereon,"the position calculation of prior art position sensitive detector systems relies on a known geometry pattern of individual electrodes and the distribution of the charge parts. a heuristic estimation is made in order to calculate an initial coordinate of irradiation. in contrast, the present invention allows one to calculate the position of an incident particle in terms of direct mapping of the measured detector response into position coordinates detector surface. the device for estimating the space coordinates of an irradiation position onto a detector comprises a position sensitive detector; an irradiation source; means for measuring the response of detector generated upon irradiation by irradiation source; and an artificial neural network structure provided such that the measured detector response is the input to the artificial neural network structure and the initial space coordinates of irradiation are the output of the artificial neural network structure.",2013-09-10,"The title of the patent is position resolved measurement apparatus and a method for acquiring space coordinates of a quantum beam incident thereon and its abstract is the position calculation of prior art position sensitive detector systems relies on a known geometry pattern of individual electrodes and the distribution of the charge parts. a heuristic estimation is made in order to calculate an initial coordinate of irradiation. in contrast, the present invention allows one to calculate the position of an incident particle in terms of direct mapping of the measured detector response into position coordinates detector surface. the device for estimating the space coordinates of an irradiation position onto a detector comprises a position sensitive detector; an irradiation source; means for measuring the response of detector generated upon irradiation by irradiation source; and an artificial neural network structure provided such that the measured detector response is the input to the artificial neural network structure and the initial space coordinates of irradiation are the output of the artificial neural network structure. dated 2013-09-10"
8538541,subthreshold stimulation of a cochlea,"an implantable apparatus, such as a cochlear implant, for delivering electrical plasticity informative stimuli to a neural network of an implantee. the apparatus comprises a stimulator device (40) that generates stimulation signals, and an electrode array (20) that receives the stimulation signals and delivers the stimuli to the neural network of the implantee in response to the signals. the stimuli delivered to the implantee facilitates and/or controls the production and/or release of naturally occurring agents into the neural network to influence the functionality thereof.",2013-09-17,"The title of the patent is subthreshold stimulation of a cochlea and its abstract is an implantable apparatus, such as a cochlear implant, for delivering electrical plasticity informative stimuli to a neural network of an implantee. the apparatus comprises a stimulator device (40) that generates stimulation signals, and an electrode array (20) that receives the stimulation signals and delivers the stimuli to the neural network of the implantee in response to the signals. the stimuli delivered to the implantee facilitates and/or controls the production and/or release of naturally occurring agents into the neural network to influence the functionality thereof. dated 2013-09-17"
8538901,method for approximation of optimal control for nonlinear discrete time systems,"a method for approximation of optimal control for a nonlinear discrete time system in which the state variables are first obtained from a system model. control sequences are then iteratively generated for the network to optimize control variables for the network and in which the value for each control variable is independent of the other control variables. following optimization of the control variables, the control variables are then mapped onto a recurrent neural network utilizing conventional training methods.",2013-09-17,"The title of the patent is method for approximation of optimal control for nonlinear discrete time systems and its abstract is a method for approximation of optimal control for a nonlinear discrete time system in which the state variables are first obtained from a system model. control sequences are then iteratively generated for the network to optimize control variables for the network and in which the value for each control variable is independent of the other control variables. following optimization of the control variables, the control variables are then mapped onto a recurrent neural network utilizing conventional training methods. dated 2013-09-17"
8542899,automatic image analysis and quantification for fluorescence in situ hybridization,"an analysis system automatically analyzes and counts fluorescence signals present in biopsy tissue marked using fluorescence in situ hybridization (fish). the user of the system specifies classes of a class network and process steps of a process hierarchy. then pixel values in image slices of biopsy tissue are acquired in three dimensions. a computer-implemented network structure is generated by linking pixel values to objects of a data network according to the class network and process hierarchy. objects associated with pixel values at different depths of the biopsy tissue are used to determine the number, volume and distance between cell components. in one application, fluorescence signals that mark her2/neural genes and centromeres of chromosome seventeen are counted to diagnose breast cancer. her2/neural genes that overlap one another or that are covered by centromeres can be accurately counted. signal artifacts that do not mark genes can be identified by their excessive volume.",2013-09-24,"The title of the patent is automatic image analysis and quantification for fluorescence in situ hybridization and its abstract is an analysis system automatically analyzes and counts fluorescence signals present in biopsy tissue marked using fluorescence in situ hybridization (fish). the user of the system specifies classes of a class network and process steps of a process hierarchy. then pixel values in image slices of biopsy tissue are acquired in three dimensions. a computer-implemented network structure is generated by linking pixel values to objects of a data network according to the class network and process hierarchy. objects associated with pixel values at different depths of the biopsy tissue are used to determine the number, volume and distance between cell components. in one application, fluorescence signals that mark her2/neural genes and centromeres of chromosome seventeen are counted to diagnose breast cancer. her2/neural genes that overlap one another or that are covered by centromeres can be accurately counted. signal artifacts that do not mark genes can be identified by their excessive volume. dated 2013-09-24"
8543068,pulse coupled oscillator synchronization for wireless communications,"a transceiver node includes a pulse coupled oscillator in an integrated circuit, which can synchronize with other nodes to generate a global clock subsequently used to facilitate synchronous communications between individual nodes. known potential uses include a low power sensor node radio for an ad-hoc network for military applications and medical applications such as ingestible and implantable radios, self powered radios, and medical monitoring systems such as cardiac and neural monitoring patches.",2013-09-24,"The title of the patent is pulse coupled oscillator synchronization for wireless communications and its abstract is a transceiver node includes a pulse coupled oscillator in an integrated circuit, which can synchronize with other nodes to generate a global clock subsequently used to facilitate synchronous communications between individual nodes. known potential uses include a low power sensor node radio for an ad-hoc network for military applications and medical applications such as ingestible and implantable radios, self powered radios, and medical monitoring systems such as cardiac and neural monitoring patches. dated 2013-09-24"
8543343,method and apparatus for determining energy savings by using a baseline energy use model that incorporates an artificial intelligence algorithm,"a computer-based system, computer-implemented method, and computer program product facilitate determining energy cost savings in an energy-consuming facility, such as a commercial building, using an artificial intelligence model, for example a neural network model, that projects or estimates the amount of energy that would have been consumed by the facility but for the implementation of energy efficiency or conservation measures. energy savings are represented by the difference between the estimate of energy that would have been consumed but for the measures and the actual amount of energy consumed by the facility under actual conditions during a time interval after the measures have been implemented.",2013-09-24,"The title of the patent is method and apparatus for determining energy savings by using a baseline energy use model that incorporates an artificial intelligence algorithm and its abstract is a computer-based system, computer-implemented method, and computer program product facilitate determining energy cost savings in an energy-consuming facility, such as a commercial building, using an artificial intelligence model, for example a neural network model, that projects or estimates the amount of energy that would have been consumed by the facility but for the implementation of energy efficiency or conservation measures. energy savings are represented by the difference between the estimate of energy that would have been consumed but for the measures and the actual amount of energy consumed by the facility under actual conditions during a time interval after the measures have been implemented. dated 2013-09-24"
8543428,computerized system and method for estimating levels of obesity in an insured population,"a computerized system and method for estimating levels of obesity in an insured population using claims data. the model uses health risk assessment data comprising age, height, and weight information as well as information about health conditions and health behaviors for a member population. claims data is used to train a two-stage model on the member population. the first stage comprises a support vector machine, a rule-based module, and a generalized linear model that estimates the probability of obesity. the second stage comprises a regression neural network that operates on the output of the first stage and a subset of the input feature vector. cost and utilizations in these areas, along with overall health measures as well as demographics and social factors, are inputs to a set of pattern recognition engines that perform regression. the output is the estimated body mass index of the member.",2013-09-24,"The title of the patent is computerized system and method for estimating levels of obesity in an insured population and its abstract is a computerized system and method for estimating levels of obesity in an insured population using claims data. the model uses health risk assessment data comprising age, height, and weight information as well as information about health conditions and health behaviors for a member population. claims data is used to train a two-stage model on the member population. the first stage comprises a support vector machine, a rule-based module, and a generalized linear model that estimates the probability of obesity. the second stage comprises a regression neural network that operates on the output of the first stage and a subset of the input feature vector. cost and utilizations in these areas, along with overall health measures as well as demographics and social factors, are inputs to a set of pattern recognition engines that perform regression. the output is the estimated body mass index of the member. dated 2013-09-24"
8543522,automatic rule discovery from large-scale datasets to detect payment card fraud using classifiers,"a set of payment card transactions including a sparse set of fraudulent transactions is normalized, such that continuously valued literals in each of the set of transactions are transformed to discrete literals. the normalized transactions are used to train a classifier, such as a neural network, such that the classifier is trained to classify transactions as fraudulent or genuine. the fraudulent transactions in the set of payment card transactions are clustered to form a set of prototype transactions. each of the discrete literals in each of the prototype transactions is expanded using sensitivity analysis using the trained classifier as an oracle, and a rule for identifying fraudulent transactions is generated for each prototype transaction based on the transaction's respective expanded literals.",2013-09-24,"The title of the patent is automatic rule discovery from large-scale datasets to detect payment card fraud using classifiers and its abstract is a set of payment card transactions including a sparse set of fraudulent transactions is normalized, such that continuously valued literals in each of the set of transactions are transformed to discrete literals. the normalized transactions are used to train a classifier, such as a neural network, such that the classifier is trained to classify transactions as fraudulent or genuine. the fraudulent transactions in the set of payment card transactions are clustered to form a set of prototype transactions. each of the discrete literals in each of the prototype transactions is expanded using sensitivity analysis using the trained classifier as an oracle, and a rule for identifying fraudulent transactions is generated for each prototype transaction based on the transaction's respective expanded literals. dated 2013-09-24"
8543526,systems and methods using neural networks to reduce noise in audio signals,"systems, methods, and computer program products are provided to provide noise reduction for an input signal using a neural network. a feed-forward set of neuron groups is provided to enhance neuron activity within a particular frequency band based on prior reception of activity within that frequency band, and also to attenuate surrounding frequency bands. a surround-inhibition set of neuron groups further attenuates activity surrounding the stimulated frequency band.",2013-09-24,"The title of the patent is systems and methods using neural networks to reduce noise in audio signals and its abstract is systems, methods, and computer program products are provided to provide noise reduction for an input signal using a neural network. a feed-forward set of neuron groups is provided to enhance neuron activity within a particular frequency band based on prior reception of activity within that frequency band, and also to attenuate surrounding frequency bands. a surround-inhibition set of neuron groups further attenuates activity surrounding the stimulated frequency band. dated 2013-09-24"
8548231,predicate logic based image grammars for complex visual pattern recognition,"first order predicate logics are provided, extended with a bilattice based uncertainty handling formalism, as a means of formally encoding pattern grammars, to parse a set of image features, and detect the presence of different patterns of interest implemented on a processor. information from different sources and uncertainties from detections, are integrated within the bilattice framework. automated logical rule weight learning in the computer vision domain applies a rule weight optimization method which casts the instantiated inference tree as a knowledge-based neural network, to converge upon a set of rule weights that give optimal performance within the bilattice framework. applications are in (a) detecting the presence of humans under partial occlusions and (b) detecting large complex man made structures in satellite imagery (c) detection of spatio-temporal human and vehicular activities in video and (c) parsing of graphical user interfaces.",2013-10-01,"The title of the patent is predicate logic based image grammars for complex visual pattern recognition and its abstract is first order predicate logics are provided, extended with a bilattice based uncertainty handling formalism, as a means of formally encoding pattern grammars, to parse a set of image features, and detect the presence of different patterns of interest implemented on a processor. information from different sources and uncertainties from detections, are integrated within the bilattice framework. automated logical rule weight learning in the computer vision domain applies a rule weight optimization method which casts the instantiated inference tree as a knowledge-based neural network, to converge upon a set of rule weights that give optimal performance within the bilattice framework. applications are in (a) detecting the presence of humans under partial occlusions and (b) detecting large complex man made structures in satellite imagery (c) detection of spatio-temporal human and vehicular activities in video and (c) parsing of graphical user interfaces. dated 2013-10-01"
8548656,underwater vehicles with improved efficiency over a range of velocities,"an autonomous underwater vehicle (auv) uses a model of propulsive efficiency to achieve high values of propulsive efficiency over a range of forward velocities, giving a lowered energy drain on the battery. externally monitored information, such as that on flow velocity, is conveyed to an apparatus residing in the vehicle's control unit, which in turn signals the locomotive unit to adopt kinematics, such as fin frequency and amplitude, associated with optimal propulsion efficiency. in an embodiment, the model of propulsive efficiency is generated from a multilayer perception neural network model using data from aquatic species, such as undulatory fin propulsion in the knifefish (xenomystus nigri), and a sensitivity analysis is used to lower the number of required inputs. power savings could protract vehicle operational life and/or provide more power to other functions, such as communications.",2013-10-01,"The title of the patent is underwater vehicles with improved efficiency over a range of velocities and its abstract is an autonomous underwater vehicle (auv) uses a model of propulsive efficiency to achieve high values of propulsive efficiency over a range of forward velocities, giving a lowered energy drain on the battery. externally monitored information, such as that on flow velocity, is conveyed to an apparatus residing in the vehicle's control unit, which in turn signals the locomotive unit to adopt kinematics, such as fin frequency and amplitude, associated with optimal propulsion efficiency. in an embodiment, the model of propulsive efficiency is generated from a multilayer perception neural network model using data from aquatic species, such as undulatory fin propulsion in the knifefish (xenomystus nigri), and a sensitivity analysis is used to lower the number of required inputs. power savings could protract vehicle operational life and/or provide more power to other functions, such as communications. dated 2013-10-01"
8554555,method for automated training of a plurality of artificial neural networks,"the invention provides a method for automated training of a plurality of artificial neural networks for phoneme recognition using training data, wherein the training data comprises speech signals subdivided into frames, each frame associated with a phoneme label, wherein the phoneme label indicates a phoneme associated with the frame. a sequence of frames from the training data are provided, wherein the number of frames in the sequence of frames is at least equal to the number of artificial neural networks. each of the artificial neural networks is assigned a different subsequence of the provided sequence, wherein each subsequence comprises a predetermined number of frames. a common phoneme label for the sequence of frames is determined based on the phoneme labels of one or more frames of one or more subsequences of the provided sequence. each artificial neural network using the common phoneme label.",2013-10-08,"The title of the patent is method for automated training of a plurality of artificial neural networks and its abstract is the invention provides a method for automated training of a plurality of artificial neural networks for phoneme recognition using training data, wherein the training data comprises speech signals subdivided into frames, each frame associated with a phoneme label, wherein the phoneme label indicates a phoneme associated with the frame. a sequence of frames from the training data are provided, wherein the number of frames in the sequence of frames is at least equal to the number of artificial neural networks. each of the artificial neural networks is assigned a different subsequence of the provided sequence, wherein each subsequence comprises a predetermined number of frames. a common phoneme label for the sequence of frames is determined based on the phoneme labels of one or more frames of one or more subsequences of the provided sequence. each artificial neural network using the common phoneme label. dated 2013-10-08"
8554706,"power plant control device which uses a model, a learning signal, a correction signal, and a manipulation signal","a gas concentration estimation device of a coal-burning boiler adapted to estimate the concentration of the gas component included in an exhaust gas emitted from a coal-burning boiler using a neural network, including: a process database section adapted to store process data of a coal-burning boiler; a filtering processing section adapted to perform filtering processing for extracting data suitable for learning of a neural network from the process data stored in the process database section; a neural-network learning processing section adapted to perform learning processing of the neural network based on the data extracted by the filtering processing section and suitable for learning of the neural network; and a neural-network estimation processing section adapted to perform estimation processing of the co concentration or the nox concentration in the exhaust gas emitted from the coal-burning boiler based on the learning processing of the neural-network learning processing section.",2013-10-08,"The title of the patent is power plant control device which uses a model, a learning signal, a correction signal, and a manipulation signal and its abstract is a gas concentration estimation device of a coal-burning boiler adapted to estimate the concentration of the gas component included in an exhaust gas emitted from a coal-burning boiler using a neural network, including: a process database section adapted to store process data of a coal-burning boiler; a filtering processing section adapted to perform filtering processing for extracting data suitable for learning of a neural network from the process data stored in the process database section; a neural-network learning processing section adapted to perform learning processing of the neural network based on the data extracted by the filtering processing section and suitable for learning of the neural network; and a neural-network estimation processing section adapted to perform estimation processing of the co concentration or the nox concentration in the exhaust gas emitted from the coal-burning boiler based on the learning processing of the neural-network learning processing section. dated 2013-10-08"
8554707,method for the computer-assisted control and/or regulation of a technical system where the dynamic behavior of the technical system is modeled using a recurrent neural network,"a method for the computer-assisted control and/or regulation of a technical system is provided. the method includes two steps, namely modeling the dynamic behavior of the technical system with a recurrent neural network using training data, the recurrent neural network includes states and actions determined using a simulation model at different times and learning an action selection rule by the recurrent neural network to a further neural network. the method can be used with any technical system in order to control the system in an optimum computer-assisted manner. for example, the method can be used in the control of a gas turbine.",2013-10-08,"The title of the patent is method for the computer-assisted control and/or regulation of a technical system where the dynamic behavior of the technical system is modeled using a recurrent neural network and its abstract is a method for the computer-assisted control and/or regulation of a technical system is provided. the method includes two steps, namely modeling the dynamic behavior of the technical system with a recurrent neural network using training data, the recurrent neural network includes states and actions determined using a simulation model at different times and learning an action selection rule by the recurrent neural network to a further neural network. the method can be used with any technical system in order to control the system in an optimum computer-assisted manner. for example, the method can be used in the control of a gas turbine. dated 2013-10-08"
8565886,arousal state modulation with electrical stimulation,"in some examples, an arousal network of a brain of a patient can be activated to modify the arousal state of the patient, which may be useful in treating a cognitive disorder of the patient. in some examples, a bioelectrical brain signal indicative of electrical activity in a first portion of the brain is monitored to determine whether the patient is in a first arousal state, and, in response to determining the patient is in the first arousal state, electrical stimulation is delivered to a second portion of the brain to activate an arousal neural network in the first portion of the brain to induce a second arousal state to treat the cognitive disorder, where the second arousal state is different than the first arousal state.",2013-10-22,"The title of the patent is arousal state modulation with electrical stimulation and its abstract is in some examples, an arousal network of a brain of a patient can be activated to modify the arousal state of the patient, which may be useful in treating a cognitive disorder of the patient. in some examples, a bioelectrical brain signal indicative of electrical activity in a first portion of the brain is monitored to determine whether the patient is in a first arousal state, and, in response to determining the patient is in the first arousal state, electrical stimulation is delivered to a second portion of the brain to activate an arousal neural network in the first portion of the brain to induce a second arousal state to treat the cognitive disorder, where the second arousal state is different than the first arousal state. dated 2013-10-22"
8566265,combined spike domain and pulse domain signal processing,"a neural network has an array of interconnected processors, at least a first processor in the array operating in a pulse domain and at least a second processor in the array operating in a spike domain, and each said processor having: first inputs selectively coupled to other processors in the array of interconnected processors, each first input having an associated vccs (a 1 bit dac) coupled to a summing node, second inputs selectively coupled to inputs of the neural network, the second inputs having current generators associated therewith coupled to said summing node, a filter/integrator for generating an analog signal corresponding to current arriving at the summing node, and for processors operating in the pulse domain, an analog-to-pulse converter for converting an analog signal derived either directly from the filter/integrator or via a non-linear element, to the pulse domain, and providing the converted analog signal as an unquantized pulse domain signal at an output of each processor operating in the pulse domain and for processors operating in the spike domain, an analog-to-spike converter for converting an analog signal derived either directly from the filter/integrator or via a non-linear element, to the spike domain, and providing the converted analog signal as an unquantized spike domain signal at an output of each processor operating in the spike domain; wherein the array of interconnected processors are selectively interconnected with unquantized pulse domain and spike domain signals.",2013-10-22,"The title of the patent is combined spike domain and pulse domain signal processing and its abstract is a neural network has an array of interconnected processors, at least a first processor in the array operating in a pulse domain and at least a second processor in the array operating in a spike domain, and each said processor having: first inputs selectively coupled to other processors in the array of interconnected processors, each first input having an associated vccs (a 1 bit dac) coupled to a summing node, second inputs selectively coupled to inputs of the neural network, the second inputs having current generators associated therewith coupled to said summing node, a filter/integrator for generating an analog signal corresponding to current arriving at the summing node, and for processors operating in the pulse domain, an analog-to-pulse converter for converting an analog signal derived either directly from the filter/integrator or via a non-linear element, to the pulse domain, and providing the converted analog signal as an unquantized pulse domain signal at an output of each processor operating in the pulse domain and for processors operating in the spike domain, an analog-to-spike converter for converting an analog signal derived either directly from the filter/integrator or via a non-linear element, to the spike domain, and providing the converted analog signal as an unquantized spike domain signal at an output of each processor operating in the spike domain; wherein the array of interconnected processors are selectively interconnected with unquantized pulse domain and spike domain signals. dated 2013-10-22"
8577820,accurate and fast neural network training for library-based critical dimension (cd) metrology,"approaches for accurate neural network training for library-based critical dimension (cd) metrology are described. approaches for fast neural network training for library-based cd metrology are also described. in an example, a method includes optimizing a threshold for a principal component analysis (pca) of a spectrum data set to provide a principal component (pc) value, estimating a training target for one or more neural networks, training the one or more neural networks based both on the training target and on the pc value provided from optimizing the threshold for the pca, and providing a spectral library based on the one or more trained neural networks.",2013-11-05,"The title of the patent is accurate and fast neural network training for library-based critical dimension (cd) metrology and its abstract is approaches for accurate neural network training for library-based critical dimension (cd) metrology are described. approaches for fast neural network training for library-based cd metrology are also described. in an example, a method includes optimizing a threshold for a principal component analysis (pca) of a spectrum data set to provide a principal component (pc) value, estimating a training target for one or more neural networks, training the one or more neural networks based both on the training target and on the pc value provided from optimizing the threshold for the pca, and providing a spectral library based on the one or more trained neural networks. dated 2013-11-05"
8582860,signet ring cell detector and related methods,"a detector and method for automatically detecting signet ring cells in an image of a biopsy tissue sample, includes finding in the image, points about which cell membranes appear in radial symmetry; selecting as candidate points, at least ones of the points that have an adjacent nuclei with a predetermined shape feature; and applying a convolutional neural network to the candidate points to determine which of the candidate points are signet ring cells.",2013-11-12,"The title of the patent is signet ring cell detector and related methods and its abstract is a detector and method for automatically detecting signet ring cells in an image of a biopsy tissue sample, includes finding in the image, points about which cell membranes appear in radial symmetry; selecting as candidate points, at least ones of the points that have an adjacent nuclei with a predetermined shape feature; and applying a convolutional neural network to the candidate points to determine which of the candidate points are signet ring cells. dated 2013-11-12"
8583286,hybrid control device,"a brain-based device (bbd) for moving in a real-world environment has sensors that provide data about the environment, actuators to move the bbd, and a hybrid controller which includes a neural controller having a simulated nervous system being a model of selected areas of the human brain and a non-neural controller based on a computational algorithmic network. the neural controller and non-neural controller interact with one another to control movement of the bbd.",2013-11-12,"The title of the patent is hybrid control device and its abstract is a brain-based device (bbd) for moving in a real-world environment has sensors that provide data about the environment, actuators to move the bbd, and a hybrid controller which includes a neural controller having a simulated nervous system being a model of selected areas of the human brain and a non-neural controller based on a computational algorithmic network. the neural controller and non-neural controller interact with one another to control movement of the bbd. dated 2013-11-12"
8583565,brain imaging system and methods for direct prosthesis control,"methods and systems for controlling a prosthesis using a brain imager that images a localized portion of the brain are provided according to one embodiment of the invention. for example, the brain imager can provide motor cortex activation data using near infrared imaging techniques and eeg techniques among others. eeg and near infrared signals can be correlated with brain activity related to limbic control and may be provided to a neural network, for example, a fuzzy neural network that maps brain activity data to limbic control data. the limbic control data may then be used to control a prosthetic limb. other embodiments of the invention include fiber optics that provide light to and receive light from the surface of the scalp through hair.",2013-11-12,"The title of the patent is brain imaging system and methods for direct prosthesis control and its abstract is methods and systems for controlling a prosthesis using a brain imager that images a localized portion of the brain are provided according to one embodiment of the invention. for example, the brain imager can provide motor cortex activation data using near infrared imaging techniques and eeg techniques among others. eeg and near infrared signals can be correlated with brain activity related to limbic control and may be provided to a neural network, for example, a fuzzy neural network that maps brain activity data to limbic control data. the limbic control data may then be used to control a prosthetic limb. other embodiments of the invention include fiber optics that provide light to and receive light from the surface of the scalp through hair. dated 2013-11-12"
8583574,method of and apparatus for combining artificial intelligence (ai) concepts with event-driven security architectures and ideas,"user authentication apparatus controlling access to systems, inputs owner's login name and password and then extracts the owner's timing vectors from keystroke characteristics with which the owner forms a training set. a semantic network uses multiple links to indicate that different pattern components of user's behavioral access create different kinds of relationships and “symbolic representations”. a neural network is trained by using each of the owner's timing vectors in the training set as an input. when a user inputs the owner's login name and password, it's checked and the user's timing vector is extracted to type the user's password if checked and demoted in confidence level if otherwise. the user's timing vector is applied to neural network and difference between the input/output is compared with a predetermined threshold; and if the difference is greater than the threshold, is prohibited. preferably this is aided by response time to personal questions.",2013-11-12,"The title of the patent is method of and apparatus for combining artificial intelligence (ai) concepts with event-driven security architectures and ideas and its abstract is user authentication apparatus controlling access to systems, inputs owner's login name and password and then extracts the owner's timing vectors from keystroke characteristics with which the owner forms a training set. a semantic network uses multiple links to indicate that different pattern components of user's behavioral access create different kinds of relationships and “symbolic representations”. a neural network is trained by using each of the owner's timing vectors in the training set as an input. when a user inputs the owner's login name and password, it's checked and the user's timing vector is extracted to type the user's password if checked and demoted in confidence level if otherwise. the user's timing vector is applied to neural network and difference between the input/output is compared with a predetermined threshold; and if the difference is greater than the threshold, is prohibited. preferably this is aided by response time to personal questions. dated 2013-11-12"
8587140,estimating an achievable power production of a wind turbine by means of a neural network,"a method for estimating an achievable power production of a wind turbine, which is operated with a reduced power set point is provided. the method includes determining the values of at least two parameters, inputting the values of the at least two parameters into a neural network, and outputting an output value from the neural network. the at least two parameters are indicative of an operating condition of the wind turbine. thereby, the output value is an estimate of the achievable power production of the wind turbine. a control system which is adapted to carry out the described power estimation method is also provided. furthermore, a wind turbine which uses the control system adapted to carry out the described power estimation method is provided.",2013-11-19,"The title of the patent is estimating an achievable power production of a wind turbine by means of a neural network and its abstract is a method for estimating an achievable power production of a wind turbine, which is operated with a reduced power set point is provided. the method includes determining the values of at least two parameters, inputting the values of the at least two parameters into a neural network, and outputting an output value from the neural network. the at least two parameters are indicative of an operating condition of the wind turbine. thereby, the output value is an estimate of the achievable power production of the wind turbine. a control system which is adapted to carry out the described power estimation method is also provided. furthermore, a wind turbine which uses the control system adapted to carry out the described power estimation method is provided. dated 2013-11-19"
8589115,system and method for estimating torque and rotational speed of motor,"a system and a method for estimating a torque and a rotational speed of a motor are disclosed. the system includes a sound receiving device, a feature extraction device, and an artificial neural network module. in the method, at first, a plurality of training data are provided, wherein the training data includes a plurality of history sound feature values of the motor and history torque values or history rotation values corresponding thereto. thereafter, an artificial neural network stored in the artificial neural network module is trained by the history data to obtain a motor model of the motor. then, a motor sound signal made by the motor in a working state is received. thereafter, sound feature values of the motor sound signal are extracted. thereafter, the rotational speed value and the torque value are computed by the motor model in accordance with the at least one sound feature value.",2013-11-19,"The title of the patent is system and method for estimating torque and rotational speed of motor and its abstract is a system and a method for estimating a torque and a rotational speed of a motor are disclosed. the system includes a sound receiving device, a feature extraction device, and an artificial neural network module. in the method, at first, a plurality of training data are provided, wherein the training data includes a plurality of history sound feature values of the motor and history torque values or history rotation values corresponding thereto. thereafter, an artificial neural network stored in the artificial neural network module is trained by the history data to obtain a motor model of the motor. then, a motor sound signal made by the motor in a working state is received. thereafter, sound feature values of the motor sound signal are extracted. thereafter, the rotational speed value and the torque value are computed by the motor model in accordance with the at least one sound feature value. dated 2013-11-19"
8589317,human-assisted training of automated classifiers,"many computing scenarios involve the classification of content items within one or more categories. the content item set may be too large for humans to classify, but an automated classifier (e.g., an artificial neural network) may not be able to classify all content items with acceptable accuracy. instead, the automated classifier may calculate a classification confidence while classifying respective content items. content items having a low classification confidence may be sent to a human classifier, and may be added, along with the categories identified by the human classifier, to a training set. the automated classifier may then be retrained using the training set, thereby incrementally improving the classification confidence of the automated classifier while conserving the involvement of human classifiers. additionally, human classifiers may be rewarded for classifying the content items, and the costs of such rewards may be considered while selecting content items for the training set.",2013-11-19,"The title of the patent is human-assisted training of automated classifiers and its abstract is many computing scenarios involve the classification of content items within one or more categories. the content item set may be too large for humans to classify, but an automated classifier (e.g., an artificial neural network) may not be able to classify all content items with acceptable accuracy. instead, the automated classifier may calculate a classification confidence while classifying respective content items. content items having a low classification confidence may be sent to a human classifier, and may be added, along with the categories identified by the human classifier, to a training set. the automated classifier may then be retrained using the training set, thereby incrementally improving the classification confidence of the automated classifier while conserving the involvement of human classifiers. additionally, human classifiers may be rewarded for classifying the content items, and the costs of such rewards may be considered while selecting content items for the training set. dated 2013-11-19"
8589855,machine-learning based datapath extraction,"a datapath extraction tool uses machine-learning models to selectively classify clusters of cells in an integrated circuit design as either datapath logic or non-datapath logic based on cluster features. a support vector machine and a neural network can be used to build compact and run-time efficient models. a cluster is classified as datapath if both the support vector machine and the neural network indicate that it is datapath-like. the cluster features may include automorphism generators for the cell clusters, or physical information based on the cell locations from a previous (e.g., global) placement, such as a ratio of a total cell area for a given cluster to a half-perimeter of a bounding box for the given cluster.",2013-11-19,"The title of the patent is machine-learning based datapath extraction and its abstract is a datapath extraction tool uses machine-learning models to selectively classify clusters of cells in an integrated circuit design as either datapath logic or non-datapath logic based on cluster features. a support vector machine and a neural network can be used to build compact and run-time efficient models. a cluster is classified as datapath if both the support vector machine and the neural network indicate that it is datapath-like. the cluster features may include automorphism generators for the cell clusters, or physical information based on the cell locations from a previous (e.g., global) placement, such as a ratio of a total cell area for a given cluster to a half-perimeter of a bounding box for the given cluster. dated 2013-11-19"
8595162,robust controller for nonlinear mimo systems,"the robust controller for nonlinear mimo systems uses a radial basis function (rbf) neural network to generate optimal control signals abiding by constraints, if any, on the control signal or on the system output. the weights of the neural network are trained in the negative direction of the gradient of output squared error. nonlinearities in the system, as well as variations in system parameters, are handled by the robust controller. simulation results are included in the end to assess the performance of the proposed controller.",2013-11-26,"The title of the patent is robust controller for nonlinear mimo systems and its abstract is the robust controller for nonlinear mimo systems uses a radial basis function (rbf) neural network to generate optimal control signals abiding by constraints, if any, on the control signal or on the system output. the weights of the neural network are trained in the negative direction of the gradient of output squared error. nonlinearities in the system, as well as variations in system parameters, are handled by the robust controller. simulation results are included in the end to assess the performance of the proposed controller. dated 2013-11-26"
8595163,system for evaluating hyperdocuments using a trained artificial neural network,"an embodiment of a system for determining the disposition of a hyperdocument using a trained artificial neural network, including an information source, a requesting application, and a server containing a trained artificial neural network (ann) capable of evaluating the information and providing results reflecting the evaluation to a requesting application is described.",2013-11-26,"The title of the patent is system for evaluating hyperdocuments using a trained artificial neural network and its abstract is an embodiment of a system for determining the disposition of a hyperdocument using a trained artificial neural network, including an information source, a requesting application, and a server containing a trained artificial neural network (ann) capable of evaluating the information and providing results reflecting the evaluation to a requesting application is described. dated 2013-11-26"
8595164,wavelet modeling paradigms for cardiovascular physiological signal interpretation,"described herein is a method of processing a cardiovascular physiological signal, comprising: decomposing the cardiovascular physiological signal into a first plurality of wavelet coefficients using a wavelet transform; selecting a second plurality of wavelet coefficients from the first plurality of wavelet coefficients, the second plurality being a subset of the first plurality; classifying or clustering the cardiovascular physiological signal into one of a plurality of predetermined classes based on the second plurality of wavelet coefficients using an artificial neural network.",2013-11-26,"The title of the patent is wavelet modeling paradigms for cardiovascular physiological signal interpretation and its abstract is described herein is a method of processing a cardiovascular physiological signal, comprising: decomposing the cardiovascular physiological signal into a first plurality of wavelet coefficients using a wavelet transform; selecting a second plurality of wavelet coefficients from the first plurality of wavelet coefficients, the second plurality being a subset of the first plurality; classifying or clustering the cardiovascular physiological signal into one of a plurality of predetermined classes based on the second plurality of wavelet coefficients using an artificial neural network. dated 2013-11-26"
8595165,method for diagnosing urticaria and angioedema,"according to the invention there is provided a method for diagnosing urticaria or angioedema including: (a) asking a patient the following questions: are any nsaids or aspiring being taken; are symptoms triggered by aspirin, aspirin-containing drugs, orange juice, curry or high-aspirin content food; is tingling of the mouth or lips, swelling of the tongue, the inside of the mouth or throat, difficulty swallowing, or difficulty breathing experienced after other medications than those known to cause urticaria or angioedema; does urticaria or angioedema come on with physical stimuli such as cold, wet, wind and pressure; (b) carrying out one or more tests which includes a rast test to cat; (c) inputting the results of the questions and tests into a neural network that has been trained to diagnose urticaria or angioedema; and (d) producing an output indicative of urticaria or angioedema.",2013-11-26,"The title of the patent is method for diagnosing urticaria and angioedema and its abstract is according to the invention there is provided a method for diagnosing urticaria or angioedema including: (a) asking a patient the following questions: are any nsaids or aspiring being taken; are symptoms triggered by aspirin, aspirin-containing drugs, orange juice, curry or high-aspirin content food; is tingling of the mouth or lips, swelling of the tongue, the inside of the mouth or throat, difficulty swallowing, or difficulty breathing experienced after other medications than those known to cause urticaria or angioedema; does urticaria or angioedema come on with physical stimuli such as cold, wet, wind and pressure; (b) carrying out one or more tests which includes a rast test to cat; (c) inputting the results of the questions and tests into a neural network that has been trained to diagnose urticaria or angioedema; and (d) producing an output indicative of urticaria or angioedema. dated 2013-11-26"
8600068,systems and methods for inducing effects in a signal,"a system for inducing an effect in a raw audio signal comprises a computing device for receiving a first audio signal and a second audio signal from a signal source, and the second audio signal comprises the first audio signal induced with an effect. the system further comprises logic that parameterizes the effect in the second audio signal into an artificial neural network (ann).",2013-12-03,"The title of the patent is systems and methods for inducing effects in a signal and its abstract is a system for inducing an effect in a raw audio signal comprises a computing device for receiving a first audio signal and a second audio signal from a signal source, and the second audio signal comprises the first audio signal induced with an effect. the system further comprises logic that parameterizes the effect in the second audio signal into an artificial neural network (ann). dated 2013-12-03"
8606731,"coating color database creating method, search method using the database, their system, program, and recording medium","a method for creating a database for paint colors having a desired texture includes storing spectral reflectance data and micro-brilliance data of paint colors after associating each spectral reflectance data and each micro-brilliance data with a paint color code; storing texture evaluation values of sample paint colors after associating the each texture evaluation value with the paint color code; calculating characteristic quantities of the paint colors expressing textures using the spectral reflectance data and the micro-brilliance data, and storing the characteristic quantities after associating the each characteristic quantity with the paint color code; training a neural network using the characteristic quantities and the texture evaluation values of the sample paint colors as training data; and inputting characteristic quantities of the paint colors other than the sample paint colors into the neural network after the training, and storing output data after associating each output data with the paint color code.",2013-12-10,"The title of the patent is coating color database creating method, search method using the database, their system, program, and recording medium and its abstract is a method for creating a database for paint colors having a desired texture includes storing spectral reflectance data and micro-brilliance data of paint colors after associating each spectral reflectance data and each micro-brilliance data with a paint color code; storing texture evaluation values of sample paint colors after associating the each texture evaluation value with the paint color code; calculating characteristic quantities of the paint colors expressing textures using the spectral reflectance data and the micro-brilliance data, and storing the characteristic quantities after associating the each characteristic quantity with the paint color code; training a neural network using the characteristic quantities and the texture evaluation values of the sample paint colors as training data; and inputting characteristic quantities of the paint colors other than the sample paint colors into the neural network after the training, and storing output data after associating each output data with the paint color code. dated 2013-12-10"
8612198,methods and systems for controlling a semiconductor fabrication process,"software for controlling processes in a heterogeneous semiconductor manufacturing environment may include a wafer-centric database, a real-time scheduler using a neural network, and a graphical user interface displaying simulated operation of the system. these features may be employed alone or in combination to offer improved usability and computational efficiency for real time control and monitoring of a semiconductor manufacturing process. more generally, these techniques may be usefully employed in a variety of real time control systems, particularly systems requiring complex scheduling decisions or heterogeneous systems constructed of hardware from numerous independent vendors.",2013-12-17,"The title of the patent is methods and systems for controlling a semiconductor fabrication process and its abstract is software for controlling processes in a heterogeneous semiconductor manufacturing environment may include a wafer-centric database, a real-time scheduler using a neural network, and a graphical user interface displaying simulated operation of the system. these features may be employed alone or in combination to offer improved usability and computational efficiency for real time control and monitoring of a semiconductor manufacturing process. more generally, these techniques may be usefully employed in a variety of real time control systems, particularly systems requiring complex scheduling decisions or heterogeneous systems constructed of hardware from numerous independent vendors. dated 2013-12-17"
8613241,aquatic restraint device,"a restraining device for use in a water environment includes a plurality of tendrils that can be launched from a submerged device when an unauthorized swimmer is proximate the restraining device. data communication in a neural-network of restraining devices is facilitated by a central command that has the capability of directing restraining devices, normally, aquatic mines, to a target.",2013-12-24,"The title of the patent is aquatic restraint device and its abstract is a restraining device for use in a water environment includes a plurality of tendrils that can be launched from a submerged device when an unauthorized swimmer is proximate the restraining device. data communication in a neural-network of restraining devices is facilitated by a central command that has the capability of directing restraining devices, normally, aquatic mines, to a target. dated 2013-12-24"
8615476,protecting military perimeters from approaching human and vehicle using biologically realistic neural network,"an approaching human threat or vehicle, such as a suicide bomber nearing a secured zone such as a military base, may be detected and classified. a vibration recognition system may detect a systematic vibration event. the entity might be a medium, human, animal, or a passenger vehicle. the system may discriminate between such an event and a background or other vibration event, such as a falling tree limb. a seismic sensor may be employed to detect vibrations generated by footsteps and a vehicle. seismic waves may be processed locally where the sensor is located. the system may wirelessly communicate with a remote command center. temporal features of the vibration signals may be modeled by a biologically realistic neural network with good false recognition rates. the models may reject quadrupedal animal footsteps.",2013-12-24,"The title of the patent is protecting military perimeters from approaching human and vehicle using biologically realistic neural network and its abstract is an approaching human threat or vehicle, such as a suicide bomber nearing a secured zone such as a military base, may be detected and classified. a vibration recognition system may detect a systematic vibration event. the entity might be a medium, human, animal, or a passenger vehicle. the system may discriminate between such an event and a background or other vibration event, such as a falling tree limb. a seismic sensor may be employed to detect vibrations generated by footsteps and a vehicle. seismic waves may be processed locally where the sensor is located. the system may wirelessly communicate with a remote command center. temporal features of the vibration signals may be modeled by a biologically realistic neural network with good false recognition rates. the models may reject quadrupedal animal footsteps. dated 2013-12-24"
8615478,using affinity measures with supervised classifiers,"a non-binary affinity measure between any two data points for a supervised classifier may be determined. for example, affinity measures may be determined for tree, kernel-based, nearest neighbor-based and neural network supervised classifiers. by providing non-binary affinity measures using supervised classifiers, more information may be provided for clustering, analyzing and, particularly, for visualizing the results of data mining.",2013-12-24,"The title of the patent is using affinity measures with supervised classifiers and its abstract is a non-binary affinity measure between any two data points for a supervised classifier may be determined. for example, affinity measures may be determined for tree, kernel-based, nearest neighbor-based and neural network supervised classifiers. by providing non-binary affinity measures using supervised classifiers, more information may be provided for clustering, analyzing and, particularly, for visualizing the results of data mining. dated 2013-12-24"
8618761,method for controlling motor operation using information characterizing regions of motor operation,"a method for collecting operational parameters of a motor may include controlling the energization of a phase winding of the motor to establish an operating point, monitoring operational parameters of the motor that characterize a relationship between the energization control applied to the motor's phase winding and the motor's response to this control, and collecting information of the operational parameters for the operating point that characterizes the relationship between the applied energization control and the motor's response. the collected information characterizing the relationship between the applied energization control and the motor's response may be employed by a neural network to estimate the regions of operation of the motor. and a system for controlling the operation of motor may employ this information, the neural network, or both to regulate the energization of a motor's phase winding during a phase cycle.",2013-12-31,"The title of the patent is method for controlling motor operation using information characterizing regions of motor operation and its abstract is a method for collecting operational parameters of a motor may include controlling the energization of a phase winding of the motor to establish an operating point, monitoring operational parameters of the motor that characterize a relationship between the energization control applied to the motor's phase winding and the motor's response to this control, and collecting information of the operational parameters for the operating point that characterizes the relationship between the applied energization control and the motor's response. the collected information characterizing the relationship between the applied energization control and the motor's response may be employed by a neural network to estimate the regions of operation of the motor. and a system for controlling the operation of motor may employ this information, the neural network, or both to regulate the energization of a motor's phase winding during a phase cycle. dated 2013-12-31"
8620602,system for detecting leaks in single phase and multiphase fluid transport pipelines,"this patents refers to a system developed for detecting leaks in single-phase and multiphase fluid transport pipelines characterized by use measurements cells (3), sensors (4), locals processors (5) and neural models, where the measuring sensors (4) and the measurement cells (3) are installed at a number of locations along the pipeline with the purpose of monitoring the characteristic leak and normal operational pipeline transient waveforms. the local processors (5) are responsible for obtaining and sampling the signals supplied by the sensors (4), as well as their pre-processing, to make them compatible with the inputs to the neural model, this are associated to dynamic memory banks for analyzing the signals supplied by the sensors with the aim of emitting an alarm in the event that waveforms with the characteristics of a leak are detected. the local processors (5) are necessaries to implement and execute the neural models and, in the event that a leak is detected, carry out the localization calculations based on the different propagation velocities of the fluid dynamic transient caused by the leak. the system use of a communications network for transmitting data between the local processors with the aim of comparing the alarms originating from the local processors (5).",2013-12-31,"The title of the patent is system for detecting leaks in single phase and multiphase fluid transport pipelines and its abstract is this patents refers to a system developed for detecting leaks in single-phase and multiphase fluid transport pipelines characterized by use measurements cells (3), sensors (4), locals processors (5) and neural models, where the measuring sensors (4) and the measurement cells (3) are installed at a number of locations along the pipeline with the purpose of monitoring the characteristic leak and normal operational pipeline transient waveforms. the local processors (5) are responsible for obtaining and sampling the signals supplied by the sensors (4), as well as their pre-processing, to make them compatible with the inputs to the neural model, this are associated to dynamic memory banks for analyzing the signals supplied by the sensors with the aim of emitting an alarm in the event that waveforms with the characteristics of a leak are detected. the local processors (5) are necessaries to implement and execute the neural models and, in the event that a leak is detected, carry out the localization calculations based on the different propagation velocities of the fluid dynamic transient caused by the leak. the system use of a communications network for transmitting data between the local processors with the aim of comparing the alarms originating from the local processors (5). dated 2013-12-31"
8620844,"neuron device for simulating a nerve cell and neural network device, integer cluster device, feedback control device, and computer program product thereof","using variable neuron thresholds and extended hebb's rule in a neural network, a neuron device for simulating a nerve cell includes a threshold storage unit storing a threshold variable θ and threshold coefficients δθ1 and δθ2; an input reception unit receiving one or more input signal values at predetermined time intervals; an output unit outputting an output signal value “1” indicating that the neuron device is firing when the sum total s of received input signal values is equal to or greater than the value of the stored threshold variable θ, or a value “0” indicating that the neuron device is resting; and a threshold updating unit calculating δθ1x+δθ2(x−1) using the output signal value x and the stored threshold coefficients δθ1 and δθ2 and updating the value of the threshold variable θ stored in the threshold storage unit by increasing it by the calculation result.",2013-12-31,"The title of the patent is neuron device for simulating a nerve cell and neural network device, integer cluster device, feedback control device, and computer program product thereof and its abstract is using variable neuron thresholds and extended hebb's rule in a neural network, a neuron device for simulating a nerve cell includes a threshold storage unit storing a threshold variable θ and threshold coefficients δθ1 and δθ2; an input reception unit receiving one or more input signal values at predetermined time intervals; an output unit outputting an output signal value “1” indicating that the neuron device is firing when the sum total s of received input signal values is equal to or greater than the value of the stored threshold variable θ, or a value “0” indicating that the neuron device is resting; and a threshold updating unit calculating δθ1x+δθ2(x−1) using the output signal value x and the stored threshold coefficients δθ1 and δθ2 and updating the value of the threshold variable θ stored in the threshold storage unit by increasing it by the calculation result. dated 2013-12-31"
8626679,apparatus and method for estimating state of charge in battery using fuzzy algorithm implemented as neural network,"disclosed is an apparatus and method for estimating a state of charge (soc) in a battery, the apparatus including a detector unit; a soft computing unit for calculating and outputting a battery soc estimation value by processing a current, a voltage and a temperature detected by the detector unit using a computing algorithm, which is a fuzzy algorithm implemented as a neural network, the soft computing unit storing the battery soc estimation value in a memory, where the fuzzy algorithm has a form expressed as f=φ(p,x)w, where φ is one of a fuzzy radial function, a radial basis function, and an activation function in the neural network, p is a learning parameter, x is an input, and w is a weight to be updated during learning.",2014-01-07,"The title of the patent is apparatus and method for estimating state of charge in battery using fuzzy algorithm implemented as neural network and its abstract is disclosed is an apparatus and method for estimating a state of charge (soc) in a battery, the apparatus including a detector unit; a soft computing unit for calculating and outputting a battery soc estimation value by processing a current, a voltage and a temperature detected by the detector unit using a computing algorithm, which is a fuzzy algorithm implemented as a neural network, the soft computing unit storing the battery soc estimation value in a memory, where the fuzzy algorithm has a form expressed as f=φ(p,x)w, where φ is one of a fuzzy radial function, a radial basis function, and an activation function in the neural network, p is a learning parameter, x is an input, and w is a weight to be updated during learning. dated 2014-01-07"
8626684,"multi-modal neural network for universal, online learning","in one embodiment, the present invention provides a neural network comprising multiple modalities. each modality comprises multiple neurons. the neural network further comprises an interconnection lattice for cross-associating signaling between the neurons in different modalities. the interconnection lattice includes a plurality of perception neuron populations along a number of bottom-up signaling pathways, and a plurality of action neuron populations along a number of top-down signaling pathways. each perception neuron along a bottom-up signaling pathway has a corresponding action neuron along a reciprocal top-down signaling pathway. an input neuron population configured to receive sensory input drives perception neurons along a number of bottom-up signaling pathways. a first set of perception neurons along bottom-up signaling pathways drive a first set of action neurons along top-down signaling pathways. action neurons along a number of top-down signaling pathways drive an output neuron population configured to generate motor output.",2014-01-07,"The title of the patent is multi-modal neural network for universal, online learning and its abstract is in one embodiment, the present invention provides a neural network comprising multiple modalities. each modality comprises multiple neurons. the neural network further comprises an interconnection lattice for cross-associating signaling between the neurons in different modalities. the interconnection lattice includes a plurality of perception neuron populations along a number of bottom-up signaling pathways, and a plurality of action neuron populations along a number of top-down signaling pathways. each perception neuron along a bottom-up signaling pathway has a corresponding action neuron along a reciprocal top-down signaling pathway. an input neuron population configured to receive sensory input drives perception neurons along a number of bottom-up signaling pathways. a first set of perception neurons along bottom-up signaling pathways drive a first set of action neurons along top-down signaling pathways. action neurons along a number of top-down signaling pathways drive an output neuron population configured to generate motor output. dated 2014-01-07"
8630966,temporally dynamic artificial neural networks,"an apparatus, article and method containing an artificial neural network that, after training, produces new trainable nodes such that input data representative of a first event and input data representative of a second event both activate a subset of the new trainable nodes. the artificial neural network can generate an output that is influenced by the input data of both events. in various embodiments, the new trainable nodes are sequentially produced and show decreasing trainability over time such that, at a particular point in time, newer produced nodes are more trainable than earlier produced nodes. the artificial neural network can be included in various embodiments of methods, apparatus and articles for use in predicting or profiling events.",2014-01-14,"The title of the patent is temporally dynamic artificial neural networks and its abstract is an apparatus, article and method containing an artificial neural network that, after training, produces new trainable nodes such that input data representative of a first event and input data representative of a second event both activate a subset of the new trainable nodes. the artificial neural network can generate an output that is influenced by the input data of both events. in various embodiments, the new trainable nodes are sequentially produced and show decreasing trainability over time such that, at a particular point in time, newer produced nodes are more trainable than earlier produced nodes. the artificial neural network can be included in various embodiments of methods, apparatus and articles for use in predicting or profiling events. dated 2014-01-14"
8635328,determining time varying thresholds for monitored metrics,"a method and apparatus for determining time-varying thresholds for measured metrics are provided. with the method and apparatus, values of a given metric are captured over time. the behavior of the metric is analyzed to determine its seasonality. correlated historical values of the metric and additional related metrics (cross-correlation) are used as inputs to a feed-forward back propagation neural network, in order to train the network to generalize the behavior of the metric. from this generalized behavior, point-by-point threshold values are calculated. the metric is monitored and the monitored values are compared with the threshold values to determine if the metric has violated its normal time-varying behavior. if so, an event is generated to notify an administrator of the error condition.",2014-01-21,"The title of the patent is determining time varying thresholds for monitored metrics and its abstract is a method and apparatus for determining time-varying thresholds for measured metrics are provided. with the method and apparatus, values of a given metric are captured over time. the behavior of the metric is analyzed to determine its seasonality. correlated historical values of the metric and additional related metrics (cross-correlation) are used as inputs to a feed-forward back propagation neural network, in order to train the network to generalize the behavior of the metric. from this generalized behavior, point-by-point threshold values are calculated. the metric is monitored and the monitored values are compared with the threshold values to determine if the metric has violated its normal time-varying behavior. if so, an event is generated to notify an administrator of the error condition. dated 2014-01-21"
8639365,methods and systems for controlling a semiconductor fabrication process,"software for controlling processes in a heterogeneous semiconductor manufacturing environment may include a wafer-centric database, a real-time scheduler using a neural network, and a graphical user interface displaying simulated operation of the system. these features may be employed alone or in combination to offer improved usability and computational efficiency for real time control and monitoring of a semiconductor manufacturing process. more generally, these techniques may be usefully employed in a variety of real time control systems, particularly systems requiring complex scheduling decisions or heterogeneous systems constructed of hardware from numerous independent vendors.",2014-01-28,"The title of the patent is methods and systems for controlling a semiconductor fabrication process and its abstract is software for controlling processes in a heterogeneous semiconductor manufacturing environment may include a wafer-centric database, a real-time scheduler using a neural network, and a graphical user interface displaying simulated operation of the system. these features may be employed alone or in combination to offer improved usability and computational efficiency for real time control and monitoring of a semiconductor manufacturing process. more generally, these techniques may be usefully employed in a variety of real time control systems, particularly systems requiring complex scheduling decisions or heterogeneous systems constructed of hardware from numerous independent vendors. dated 2014-01-28"
8639489,methods and systems for controlling a semiconductor fabrication process,"software for controlling processes in a heterogeneous semiconductor manufacturing environment may include a wafer-centric database, a real-time scheduler using a neural network, and a graphical user interface displaying simulated operation of the system. these features may be employed alone or in combination to offer improved usability and computational efficiency for real time control and monitoring of a semiconductor manufacturing process. more generally, these techniques may be usefully employed in a variety of real time control systems, particularly systems requiring complex scheduling decisions or heterogeneous systems constructed of hardware from numerous independent vendors.",2014-01-28,"The title of the patent is methods and systems for controlling a semiconductor fabrication process and its abstract is software for controlling processes in a heterogeneous semiconductor manufacturing environment may include a wafer-centric database, a real-time scheduler using a neural network, and a graphical user interface displaying simulated operation of the system. these features may be employed alone or in combination to offer improved usability and computational efficiency for real time control and monitoring of a semiconductor manufacturing process. more generally, these techniques may be usefully employed in a variety of real time control systems, particularly systems requiring complex scheduling decisions or heterogeneous systems constructed of hardware from numerous independent vendors. dated 2014-01-28"
8639637,intelligent control toolkit,a neuro-fuzzy controller is provided. the neuro-fuzzy controller includes a predictor that receives inputs and makes prediction inputs. the prediction inputs are passed to a fuzzy cluster module that includes a neural network fuzzifing said prediction inputs and passing the result to an inference engine. the output of the inference engine is defuzzified and provided as an output of the controller. the fuzzifier and defuzzifier preferably represent a neural network employing a trigonometrical series. the inference engine preferably employs rules that are determined using genetic programming.,2014-01-28,The title of the patent is intelligent control toolkit and its abstract is a neuro-fuzzy controller is provided. the neuro-fuzzy controller includes a predictor that receives inputs and makes prediction inputs. the prediction inputs are passed to a fuzzy cluster module that includes a neural network fuzzifing said prediction inputs and passing the result to an inference engine. the output of the inference engine is defuzzified and provided as an output of the controller. the fuzzifier and defuzzifier preferably represent a neural network employing a trigonometrical series. the inference engine preferably employs rules that are determined using genetic programming. dated 2014-01-28
8642349,artificial neural network proteomic tumor classification,"here the inventors describe a tumor classifier based on protein expression. also disclosed is the use of proteomics to construct a highly accurate artificial neural network (ann)-based classifier for the detection of an individual tumor type, as well as distinguishing between six common tumor types in an unknown primary diagnosis setting. discriminating sets of proteins are also identified and are used as biomarkers for six carcinomas. a leave-one-out cross validation (loocv) method was used to test the ability of the constructed network to predict the single held out sample from each iteration with a maximum predictive accuracy of 87% and an average predictive accuracy of 82% over the range of proteins chosen for its construction.",2014-02-04,"The title of the patent is artificial neural network proteomic tumor classification and its abstract is here the inventors describe a tumor classifier based on protein expression. also disclosed is the use of proteomics to construct a highly accurate artificial neural network (ann)-based classifier for the detection of an individual tumor type, as well as distinguishing between six common tumor types in an unknown primary diagnosis setting. discriminating sets of proteins are also identified and are used as biomarkers for six carcinomas. a leave-one-out cross validation (loocv) method was used to test the ability of the constructed network to predict the single held out sample from each iteration with a maximum predictive accuracy of 87% and an average predictive accuracy of 82% over the range of proteins chosen for its construction. dated 2014-02-04"
8644961,model based control and estimation of mercury emissions,"a method and apparatus for estimating and/or controlling mercury emissions in a steam generating unit. a model of the steam generating unit is used to predict mercury emissions. in one embodiment of the invention, the model is a neural network (nn) model. an optimizer may be used in connection with the model to determine optimal setpoint values for manipulated variables associated with operation of the steam generating unit.",2014-02-04,"The title of the patent is model based control and estimation of mercury emissions and its abstract is a method and apparatus for estimating and/or controlling mercury emissions in a steam generating unit. a model of the steam generating unit is used to predict mercury emissions. in one embodiment of the invention, the model is a neural network (nn) model. an optimizer may be used in connection with the model to determine optimal setpoint values for manipulated variables associated with operation of the steam generating unit. dated 2014-02-04"
8655813,synaptic weight normalized spiking neuronal networks,"neuronal networks of electronic neurons interconnected via electronic synapses with synaptic weight normalization. the synaptic weights are based on learning rules for the neuronal network, such that a synaptic weight for a synapse determines the effect of a spiking source neuron on a target neuron connected via the synapse. each synaptic weight is maintained within a predetermined range by performing synaptic weight normalization for neural network stability.",2014-02-18,"The title of the patent is synaptic weight normalized spiking neuronal networks and its abstract is neuronal networks of electronic neurons interconnected via electronic synapses with synaptic weight normalization. the synaptic weights are based on learning rules for the neuronal network, such that a synaptic weight for a synapse determines the effect of a spiking source neuron on a target neuron connected via the synapse. each synaptic weight is maintained within a predetermined range by performing synaptic weight normalization for neural network stability. dated 2014-02-18"
8655814,modeling efficiency over a range of velocities in underwater vehicles,"a method of generating a model of propulsive efficiency for an autonomous underwater vehicle (auv) is based on a multilayer perception neural network model using data from aquatic species, such as undulatory fin propulsion in the knifefish (xenomystus nigri), and a sensitivity analysis is used to lower the number of required inputs. the model of propulsive efficiency allows an auv to achieve high values of propulsive efficiency over a range of forward velocity, giving a lowered energy drain on the battery. in an embodiment, externally monitored information, such as that on flow velocity, is conveyed to an apparatus residing in the vehicle's control unit, which in turn signals the locomotive unit to adopt kinematics, such as fin frequency and amplitude, associated with optimal propulsion efficiency. power savings could protract vehicle operational life and/or provide more power to other functions, such as communications.",2014-02-18,"The title of the patent is modeling efficiency over a range of velocities in underwater vehicles and its abstract is a method of generating a model of propulsive efficiency for an autonomous underwater vehicle (auv) is based on a multilayer perception neural network model using data from aquatic species, such as undulatory fin propulsion in the knifefish (xenomystus nigri), and a sensitivity analysis is used to lower the number of required inputs. the model of propulsive efficiency allows an auv to achieve high values of propulsive efficiency over a range of forward velocity, giving a lowered energy drain on the battery. in an embodiment, externally monitored information, such as that on flow velocity, is conveyed to an apparatus residing in the vehicle's control unit, which in turn signals the locomotive unit to adopt kinematics, such as fin frequency and amplitude, associated with optimal propulsion efficiency. power savings could protract vehicle operational life and/or provide more power to other functions, such as communications. dated 2014-02-18"
8660796,method and system of processing gamma count rate curves using neural networks,"processing gamma count rate decay curves using neural networks. at least some of the illustrative embodiments are methods comprising obtaining a gamma count rate decay curve one each for a plurality of gamma detectors of a nuclear logging tool (the gamma count rate decay curves recorded at a particular borehole depth), applying the gamma count rate decay curves to input nodes of a neural network, predicting by the neural network a geophysical parameter of the formation surrounding the borehole, repeating the obtaining, applying and predicting for a plurality of borehole depths, and producing a plot of the geophysical parameter of the formation as a function of borehole depth.",2014-02-25,"The title of the patent is method and system of processing gamma count rate curves using neural networks and its abstract is processing gamma count rate decay curves using neural networks. at least some of the illustrative embodiments are methods comprising obtaining a gamma count rate decay curve one each for a plurality of gamma detectors of a nuclear logging tool (the gamma count rate decay curves recorded at a particular borehole depth), applying the gamma count rate decay curves to input nodes of a neural network, predicting by the neural network a geophysical parameter of the formation surrounding the borehole, repeating the obtaining, applying and predicting for a plurality of borehole depths, and producing a plot of the geophysical parameter of the formation as a function of borehole depth. dated 2014-02-25"
8668644,method of predicting acute cardiopulmonary events and survivability of a patient,"a method of producing an artificial neural network capable of predicting the survivability of a patient, including: storing in an electronic database patient health data comprising a plurality of sets of data, each set having at least one of a first parameter relating to heart rate variability data and a second parameter relating to vital sign data, each set further having a third parameter relating to patient survivability; providing a network of nodes interconnected to form an artificial neural network, the nodes comprising a plurality of artificial neurons, each artificial neuron having at least one input with an associated weight; and training the artificial neural network using the patient health data such that the associated weight of the at least one input of each artificial neuron is adjusted in response to respective first, second and third parameters of different sets of data from the patient health data.",2014-03-11,"The title of the patent is method of predicting acute cardiopulmonary events and survivability of a patient and its abstract is a method of producing an artificial neural network capable of predicting the survivability of a patient, including: storing in an electronic database patient health data comprising a plurality of sets of data, each set having at least one of a first parameter relating to heart rate variability data and a second parameter relating to vital sign data, each set further having a third parameter relating to patient survivability; providing a network of nodes interconnected to form an artificial neural network, the nodes comprising a plurality of artificial neurons, each artificial neuron having at least one input with an associated weight; and training the artificial neural network using the patient health data such that the associated weight of the at least one input of each artificial neuron is adjusted in response to respective first, second and third parameters of different sets of data from the patient health data. dated 2014-03-11"
8676721,"method, system and apparatus for intelligent management of oil and gas platform surface equipment","a method, system, apparatus (and related computer program) for intelligent management of oil and gas offshore and onshore platform surface equipment over a computer network is disclosed. the system utilizes a data aggregator for gathering real-time data streams from surface equipment located on such platform(s), such surface equipment containing one or more sensors for monitoring in real time the performance of equipment operational parameters of interest. the data analysis engine is in network communication with the data aggregator, and comprises a trained neural network capable of generating self organizing maps, and creating predictive operational parameters regarding such surface equipment. an interface is provided for inputting into the neural network various data including, for example, the published performance operational parameters for such equipment. a network user interface is also provided for transmitting such predictive operational input to one or more end user terminals equipped with end user dashboard display software.",2014-03-18,"The title of the patent is method, system and apparatus for intelligent management of oil and gas platform surface equipment and its abstract is a method, system, apparatus (and related computer program) for intelligent management of oil and gas offshore and onshore platform surface equipment over a computer network is disclosed. the system utilizes a data aggregator for gathering real-time data streams from surface equipment located on such platform(s), such surface equipment containing one or more sensors for monitoring in real time the performance of equipment operational parameters of interest. the data analysis engine is in network communication with the data aggregator, and comprises a trained neural network capable of generating self organizing maps, and creating predictive operational parameters regarding such surface equipment. an interface is provided for inputting into the neural network various data including, for example, the published performance operational parameters for such equipment. a network user interface is also provided for transmitting such predictive operational input to one or more end user terminals equipped with end user dashboard display software. dated 2014-03-18"
8676728,sound localization with artificial neural network,the location of a sound within a given spatial volume may be used in applications such as augmented reality environments. an artificial neural network processes time-difference-of-arrival data (tdoa) from a known microphone array to determine a spatial location of the sound. the neural network may be located locally or available as a cloud service. the artificial neural network is trained with perturbed and non-perturbed tdoa data.,2014-03-18,The title of the patent is sound localization with artificial neural network and its abstract is the location of a sound within a given spatial volume may be used in applications such as augmented reality environments. an artificial neural network processes time-difference-of-arrival data (tdoa) from a known microphone array to determine a spatial location of the sound. the neural network may be located locally or available as a cloud service. the artificial neural network is trained with perturbed and non-perturbed tdoa data. dated 2014-03-18
8682820,on demand multi-objective network optimization,"implementations of the present disclosure include methods, systems, and computer-readable storage mediums for selecting an on-demand technology configuration, including receiving a request, the request including a plurality of properties, processing the request using a plurality of classifiers to generate a plurality of request classes associated with the plurality of properties, processing the plurality of request classes using a first neural network to identify one or more technologies relevant to the request, processing each of the one or more technologies using a second neural network to identify one or more technology configurations for each of the one or more technologies, and processing each of the one or more technology configurations to identify a target technology configuration.",2014-03-25,"The title of the patent is on demand multi-objective network optimization and its abstract is implementations of the present disclosure include methods, systems, and computer-readable storage mediums for selecting an on-demand technology configuration, including receiving a request, the request including a plurality of properties, processing the request using a plurality of classifiers to generate a plurality of request classes associated with the plurality of properties, processing the plurality of request classes using a first neural network to identify one or more technologies relevant to the request, processing each of the one or more technologies using a second neural network to identify one or more technology configurations for each of the one or more technologies, and processing each of the one or more technology configurations to identify a target technology configuration. dated 2014-03-25"
8690748,apparatus for measurement and treatment of a patient,"a signal processing method and system combines multi-scale decomposition, such as wavelet, pre-processing together with a compression technique, such as an auto-associative artificial neural network, operating in the multi-scale decomposition domain for signal denoising and extraction. all compressions are performed in the decomposed domain. a reverse decomposition such as an inverse discrete wavelet transform is performed on the combined outputs from all the compression modules to recover a clean signal back in the time domain. a low-cost, non-drug, non-invasive, on-demand therapy braincap system and method are pharmaceutically non-intrusive to the body for the purpose of disease diagnosis, treatment therapy, and direct mind control of external devices and systems. it is based on recognizing abnormal brainwave signatures and intervenes at the earliest moment, using magnetic and/or electric stimulations to reset the brainwaves back to normality. the feedback system is self-regulatory and the treatment stops when the brainwaves return to normal. the braincap contains multiple sensing electrodes and microcoils; the microcoils are pairs of crossed microcoils or 3-axis triple crossed microcoils.",2014-04-08,"The title of the patent is apparatus for measurement and treatment of a patient and its abstract is a signal processing method and system combines multi-scale decomposition, such as wavelet, pre-processing together with a compression technique, such as an auto-associative artificial neural network, operating in the multi-scale decomposition domain for signal denoising and extraction. all compressions are performed in the decomposed domain. a reverse decomposition such as an inverse discrete wavelet transform is performed on the combined outputs from all the compression modules to recover a clean signal back in the time domain. a low-cost, non-drug, non-invasive, on-demand therapy braincap system and method are pharmaceutically non-intrusive to the body for the purpose of disease diagnosis, treatment therapy, and direct mind control of external devices and systems. it is based on recognizing abnormal brainwave signatures and intervenes at the earliest moment, using magnetic and/or electric stimulations to reset the brainwaves back to normality. the feedback system is self-regulatory and the treatment stops when the brainwaves return to normal. the braincap contains multiple sensing electrodes and microcoils; the microcoils are pairs of crossed microcoils or 3-axis triple crossed microcoils. dated 2014-04-08"
8694451,neural network system,"a neural network system that can minimize circuit resources for constituting a self-learning mechanism and be reconfigured into network configurations suitable for various purposes includes a neural network engine that operates in a first and a second operation mode and performs an operation representing a characteristic determined by setting network configuration information and weight information with respect to the network configuration, and a von neumann-type microprocessor that is connected to the neural network engine and performs a cooperative operation in accordance with the first or the second operation mode together with the neural network engine. the von neumann-type microprocessor recalculates the weight information or remakes the configuration information as a cooperative operation according to the first operation mode, and sets or updates the configuration information or the weight information set in the neural network engine, as a cooperative operation according to the second operation mode.",2014-04-08,"The title of the patent is neural network system and its abstract is a neural network system that can minimize circuit resources for constituting a self-learning mechanism and be reconfigured into network configurations suitable for various purposes includes a neural network engine that operates in a first and a second operation mode and performs an operation representing a characteristic determined by setting network configuration information and weight information with respect to the network configuration, and a von neumann-type microprocessor that is connected to the neural network engine and performs a cooperative operation in accordance with the first or the second operation mode together with the neural network engine. the von neumann-type microprocessor recalculates the weight information or remakes the configuration information as a cooperative operation according to the first operation mode, and sets or updates the configuration information or the weight information set in the neural network engine, as a cooperative operation according to the second operation mode. dated 2014-04-08"
8700235,"estimation of a criterion of load to which a structural component of an aircraft is subjected, and assistance for the detection of a so-called “hard” landing by virtue of such a criterion","a method for estimating a loading criterion relating to the load experienced by a structural component of an aircraft, and assistance with detecting a so-called “hard” landing. the method includes measuring parameters of the aircraft and calculating at least one loading criterion for the loading of the structural component using at least one neural network receiving the parameters as input. assistance with detecting a hard landing then includes determining of a time of impact of the aircraft on a landing strip from the measured parameters, then estimating a plurality of the parameters at the determined time of impact so as to calculate the at least one loading criterion relating to the loading of the structural component.",2014-04-15,"The title of the patent is estimation of a criterion of load to which a structural component of an aircraft is subjected, and assistance for the detection of a so-called “hard” landing by virtue of such a criterion and its abstract is a method for estimating a loading criterion relating to the load experienced by a structural component of an aircraft, and assistance with detecting a so-called “hard” landing. the method includes measuring parameters of the aircraft and calculating at least one loading criterion for the loading of the structural component using at least one neural network receiving the parameters as input. assistance with detecting a hard landing then includes determining of a time of impact of the aircraft on a landing strip from the measured parameters, then estimating a plurality of the parameters at the determined time of impact so as to calculate the at least one loading criterion relating to the loading of the structural component. dated 2014-04-15"
8700541,modeling method of neuro-fuzzy system,"a modeling method of neuro-fuzzy system including a rule-defining process and a network-building process is disclosed. the rule-defining process divides a plurality of training data into a plurality of groups to accordingly define a plurality of fuzzy rules, and the network-building process constructs a fuzzy neural network based on the fuzzy rules obtained by the rule-defining process. the provided modeling method of neuro-fuzzy system is capable of building a neuro-fuzzy system extremely similar to an original function that generates training data of the neuro-fuzzy system.",2014-04-15,"The title of the patent is modeling method of neuro-fuzzy system and its abstract is a modeling method of neuro-fuzzy system including a rule-defining process and a network-building process is disclosed. the rule-defining process divides a plurality of training data into a plurality of groups to accordingly define a plurality of fuzzy rules, and the network-building process constructs a fuzzy neural network based on the fuzzy rules obtained by the rule-defining process. the provided modeling method of neuro-fuzzy system is capable of building a neuro-fuzzy system extremely similar to an original function that generates training data of the neuro-fuzzy system. dated 2014-04-15"
8700552,exploiting sparseness in training deep neural networks,"deep neural network (dnn) training technique embodiments are presented that train a dnn while exploiting the sparseness of non-zero hidden layer interconnection weight values. generally, a fully connected dnn is initially trained by sweeping through a full training set a number of times. then, for the most part, only the interconnections whose weight magnitudes exceed a minimum weight threshold are considered in further training. this minimum weight threshold can be established as a value that results in only a prescribed maximum number of interconnections being considered when setting interconnection weight values via an error back-propagation procedure during the training. it is noted that the continued dnn training tends to converge much faster than the initial training.",2014-04-15,"The title of the patent is exploiting sparseness in training deep neural networks and its abstract is deep neural network (dnn) training technique embodiments are presented that train a dnn while exploiting the sparseness of non-zero hidden layer interconnection weight values. generally, a fully connected dnn is initially trained by sweeping through a full training set a number of times. then, for the most part, only the interconnections whose weight magnitudes exceed a minimum weight threshold are considered in further training. this minimum weight threshold can be established as a value that results in only a prescribed maximum number of interconnections being considered when setting interconnection weight values via an error back-propagation procedure during the training. it is noted that the continued dnn training tends to converge much faster than the initial training. dated 2014-04-15"
8705849,method and system for object recognition based on a trainable dynamic system,a system for object recognition in which a multi-dimensional scanner generates a temporal sequence of multi-dimensional output data of a scanned object. that data is then coupled as an input signal to a trainable dynamic system. the system exemplified by a general-purpose recurrent neural network is previously trained to generate an output signal representative of the class of the object in response to a temporal sequence of multi-dimensional data.,2014-04-22,The title of the patent is method and system for object recognition based on a trainable dynamic system and its abstract is a system for object recognition in which a multi-dimensional scanner generates a temporal sequence of multi-dimensional output data of a scanned object. that data is then coupled as an input signal to a trainable dynamic system. the system exemplified by a general-purpose recurrent neural network is previously trained to generate an output signal representative of the class of the object in response to a temporal sequence of multi-dimensional data. dated 2014-04-22
8706464,"health data dynamics, its sources and linkage with genetic/molecular tests","method and system for the analysis and source localization of the dynamical patterns in medical and health data, and linking such dynamical patterns with the individual's genetic and/or molecular data. the invention makes use of optimally positioned sensors (sensor arrays) providing input data for signal processing, time-series analysis, pattern recognition and mathematical modeling to facilitate dynamical tracking of systemic arterial pressure without a pressure cuff, local vascular activity, electrocardiographic (ecg), respiratory, physical, muscular, gastrointestinal and neural activity, temperature and other physiological/health data. the invention also facilitates separation of local signals (such as local aneurisms or local vascular activity) from non-local, central or systemic patterns (e.g. systemic blood pressure). in addition, the invention improves identification of dynamical patterns associated with a specific genotype/disorder for screening, personalized risk assessment, diagnosis and treatment control. the system can be implemented in a specialized processor, such as an ambulatory blood pressure monitor, electrocardiograph, holter monitor located outside subject's body or implanted inside the body, mobile/cell phone or smart phone/personal digital assistant, computer or computer network (the internet), including wireless or mobile network. the system can be also linked to the electronic health/medical records and other databases.",2014-04-22,"The title of the patent is health data dynamics, its sources and linkage with genetic/molecular tests and its abstract is method and system for the analysis and source localization of the dynamical patterns in medical and health data, and linking such dynamical patterns with the individual's genetic and/or molecular data. the invention makes use of optimally positioned sensors (sensor arrays) providing input data for signal processing, time-series analysis, pattern recognition and mathematical modeling to facilitate dynamical tracking of systemic arterial pressure without a pressure cuff, local vascular activity, electrocardiographic (ecg), respiratory, physical, muscular, gastrointestinal and neural activity, temperature and other physiological/health data. the invention also facilitates separation of local signals (such as local aneurisms or local vascular activity) from non-local, central or systemic patterns (e.g. systemic blood pressure). in addition, the invention improves identification of dynamical patterns associated with a specific genotype/disorder for screening, personalized risk assessment, diagnosis and treatment control. the system can be implemented in a specialized processor, such as an ambulatory blood pressure monitor, electrocardiograph, holter monitor located outside subject's body or implanted inside the body, mobile/cell phone or smart phone/personal digital assistant, computer or computer network (the internet), including wireless or mobile network. the system can be also linked to the electronic health/medical records and other databases. dated 2014-04-22"
8712549,method and system for monitoring and treating hemodynamic parameters,"a multiplexed medical carrier provides for sensing one or more patient parameters and/or delivering energy via separately identifiable effectors. the carrier includes a body and at least two electrical conductors coupled with at least two effectors. effectors may be any combination of sensors, actuators or both. sensors may measure such parameters as pressure, oxygen content, volume, conductivity, fluid flow rate, or any other chemical or physical parameters. actuators may be used, for example, to pace a heart, stimulate muscle or neural tissue, broadcast ultrasonic energy, emit light, heat or other forms of radiation, or deliver any form of energy or substance. a method for collecting medical data from a patient includes interrogating a network of multiplexed sensors residing on parallel conductors in the patient, including addressing a first addressable sensor in the network to obtain data and addressing a second addressable sensor in the network to obtain data.",2014-04-29,"The title of the patent is method and system for monitoring and treating hemodynamic parameters and its abstract is a multiplexed medical carrier provides for sensing one or more patient parameters and/or delivering energy via separately identifiable effectors. the carrier includes a body and at least two electrical conductors coupled with at least two effectors. effectors may be any combination of sensors, actuators or both. sensors may measure such parameters as pressure, oxygen content, volume, conductivity, fluid flow rate, or any other chemical or physical parameters. actuators may be used, for example, to pace a heart, stimulate muscle or neural tissue, broadcast ultrasonic energy, emit light, heat or other forms of radiation, or deliver any form of energy or substance. a method for collecting medical data from a patient includes interrogating a network of multiplexed sensors residing on parallel conductors in the patient, including addressing a first addressable sensor in the network to obtain data and addressing a second addressable sensor in the network to obtain data. dated 2014-04-29"
8712940,structural plasticity in spiking neural networks with symmetric dual of an electronic neuron,"a neural system comprises multiple neurons interconnected via synapse devices. each neuron integrates input signals arriving on its dendrite, generates a spike in response to the integrated input signals exceeding a threshold, and sends the spike to the interconnected neurons via its axon. the system further includes multiple noruens, each noruen is interconnected via the interconnect network with those neurons that the noruen's corresponding neuron sends its axon to. each noruen integrates input spikes from connected spiking neurons and generates a spike in response to the integrated input spikes exceeding a threshold. there can be one noruen for every corresponding neuron. for a first neuron connected via its axon via a synapse to dendrite of a second neuron, a noruen corresponding to the second neuron is connected via its axon through the same synapse to dendrite of the noruen corresponding to the first neuron.",2014-04-29,"The title of the patent is structural plasticity in spiking neural networks with symmetric dual of an electronic neuron and its abstract is a neural system comprises multiple neurons interconnected via synapse devices. each neuron integrates input signals arriving on its dendrite, generates a spike in response to the integrated input signals exceeding a threshold, and sends the spike to the interconnected neurons via its axon. the system further includes multiple noruens, each noruen is interconnected via the interconnect network with those neurons that the noruen's corresponding neuron sends its axon to. each noruen integrates input spikes from connected spiking neurons and generates a spike in response to the integrated input spikes exceeding a threshold. there can be one noruen for every corresponding neuron. for a first neuron connected via its axon via a synapse to dendrite of a second neuron, a noruen corresponding to the second neuron is connected via its axon through the same synapse to dendrite of the noruen corresponding to the first neuron. dated 2014-04-29"
8712941,elementary network description for efficient link between neuronal models and neuromorphic systems,"a simple format is disclosed and referred to as elementary network description (end). the format can fully describe a large-scale neuronal model and embodiments of software or hardware engines to simulate such a model efficiently. the architecture of such neuromorphic engines is optimal for high-performance parallel processing of spiking networks with spike-timing dependent plasticity. the format is specifically tuned for neural systems and specialized neuromorphic hardware, thereby serving as a bridge between developers of brain models and neuromorphic hardware manufactures.",2014-04-29,"The title of the patent is elementary network description for efficient link between neuronal models and neuromorphic systems and its abstract is a simple format is disclosed and referred to as elementary network description (end). the format can fully describe a large-scale neuronal model and embodiments of software or hardware engines to simulate such a model efficiently. the architecture of such neuromorphic engines is optimal for high-performance parallel processing of spiking networks with spike-timing dependent plasticity. the format is specifically tuned for neural systems and specialized neuromorphic hardware, thereby serving as a bridge between developers of brain models and neuromorphic hardware manufactures. dated 2014-04-29"
8717223,classification of subsurface objects using singular values derived from signal frames,"the classification system represents a detected object with a feature vector derived from the return signals acquired by an array of n transceivers operating in multistatic mode. the classification system generates the feature vector by transforming the real-valued return signals into complex-valued spectra, using, for example, a fast fourier transform. the classification system then generates a feature vector of singular values for each user-designated spectral sub-band by applying a singular value decomposition (svd) to the n×n square complex-valued matrix formed from sub-band samples associated with all possible transmitter-receiver pairs. the resulting feature vector of singular values may be transformed into a feature vector of singular value likelihoods and then subjected to a multi-category linear or neural network classifier for object classification.",2014-05-06,"The title of the patent is classification of subsurface objects using singular values derived from signal frames and its abstract is the classification system represents a detected object with a feature vector derived from the return signals acquired by an array of n transceivers operating in multistatic mode. the classification system generates the feature vector by transforming the real-valued return signals into complex-valued spectra, using, for example, a fast fourier transform. the classification system then generates a feature vector of singular values for each user-designated spectral sub-band by applying a singular value decomposition (svd) to the n×n square complex-valued matrix formed from sub-band samples associated with all possible transmitter-receiver pairs. the resulting feature vector of singular values may be transformed into a feature vector of singular value likelihoods and then subjected to a multi-category linear or neural network classifier for object classification. dated 2014-05-06"
8718271,call routing methods and systems based on multiple variable standardized scoring,"systems and methods are disclosed for routing callers to agents in a contact center, along with an intelligent routing system. an exemplary method includes combining multiple output variables of a pattern matching algorithm (for matching callers and agents) into a single metric for use in the routing system. the pattern matching algorithm may include a neural network architecture, where the exemplary method combines output variables from multiple neural networks. the method may include determining a z-score of the variable outputs and determining a linear combination of the determined z-scores for a desired output. callers may be routed to agents via the pattern matching algorithm to maximize the output value or score of the linear combination. the output variables may include revenue generation, cost, customer satisfaction performance, first call resolution, cancellation, or other variable outputs from the pattern matching algorithm of the system.",2014-05-06,"The title of the patent is call routing methods and systems based on multiple variable standardized scoring and its abstract is systems and methods are disclosed for routing callers to agents in a contact center, along with an intelligent routing system. an exemplary method includes combining multiple output variables of a pattern matching algorithm (for matching callers and agents) into a single metric for use in the routing system. the pattern matching algorithm may include a neural network architecture, where the exemplary method combines output variables from multiple neural networks. the method may include determining a z-score of the variable outputs and determining a linear combination of the determined z-scores for a desired output. callers may be routed to agents via the pattern matching algorithm to maximize the output value or score of the linear combination. the output variables may include revenue generation, cost, customer satisfaction performance, first call resolution, cancellation, or other variable outputs from the pattern matching algorithm of the system. dated 2014-05-06"
8718792,system and method of repairing of neural networks,"a method and system for re-establishing a pathway in a damaged or severed neural network includes an imaging device, an alignment device and a treatment device. an accurate image of the damaged neural network is created. an alignment device imparts wave energy into a damaged region of the neural network to direct re-growth axons into a remaining endoneurial tube to direct axon growth back to the correct targets to re-establish the severed neural network.",2014-05-06,"The title of the patent is system and method of repairing of neural networks and its abstract is a method and system for re-establishing a pathway in a damaged or severed neural network includes an imaging device, an alignment device and a treatment device. an accurate image of the damaged neural network is created. an alignment device imparts wave energy into a damaged region of the neural network to direct re-growth axons into a remaining endoneurial tube to direct axon growth back to the correct targets to re-establish the severed neural network. dated 2014-05-06"
8719199,systems and methods for providing a neural network having an elementary network description for efficient implementation of event-triggered plasticity rules,"a simple format is disclosed and referred to as elementary network description (end). the format can fully describe a large-scale neuronal model and embodiments of software or hardware engines to simulate such a model efficiently. the architecture of such neuromorphic engines is optimal for high-performance parallel processing of spiking networks with spike-timing dependent plasticity. the software and hardware engines are optimized to take into account short-term and long-term synaptic plasticity in the form of ltd, ltp, and stdp.",2014-05-06,"The title of the patent is systems and methods for providing a neural network having an elementary network description for efficient implementation of event-triggered plasticity rules and its abstract is a simple format is disclosed and referred to as elementary network description (end). the format can fully describe a large-scale neuronal model and embodiments of software or hardware engines to simulate such a model efficiently. the architecture of such neuromorphic engines is optimal for high-performance parallel processing of spiking networks with spike-timing dependent plasticity. the software and hardware engines are optimized to take into account short-term and long-term synaptic plasticity in the form of ltd, ltp, and stdp. dated 2014-05-06"
8725662,apparatus and method for partial evaluation of synaptic updates based on system events,"apparatus and methods for partial evaluation of synaptic updates in neural networks. in one embodiment, a pre-synaptic unit is connected to a several post synaptic units via communication channels. information related to a plurality of post-synaptic pulses generated by the post-synaptic units is stored by the network in response to a system event. synaptic channel updates are performed by the network using the time intervals between a pre-synaptic pulse, which is being generated prior to the system event, and at least a portion of the plurality of the post synaptic pulses. the system event enables removal of the information related to the portion of the post-synaptic pulses from the storage device. a shared memory block within the storage device is used to store data related to post-synaptic pulses generated by different post-synaptic nodes. this configuration enables memory use optimization of post-synaptic units with different firing rates.",2014-05-13,"The title of the patent is apparatus and method for partial evaluation of synaptic updates based on system events and its abstract is apparatus and methods for partial evaluation of synaptic updates in neural networks. in one embodiment, a pre-synaptic unit is connected to a several post synaptic units via communication channels. information related to a plurality of post-synaptic pulses generated by the post-synaptic units is stored by the network in response to a system event. synaptic channel updates are performed by the network using the time intervals between a pre-synaptic pulse, which is being generated prior to the system event, and at least a portion of the plurality of the post synaptic pulses. the system event enables removal of the information related to the portion of the post-synaptic pulses from the storage device. a shared memory block within the storage device is used to store data related to post-synaptic pulses generated by different post-synaptic nodes. this configuration enables memory use optimization of post-synaptic units with different firing rates. dated 2014-05-13"
8725669,signal processing method and apparatus,"a signal processing method and system combines multi-scale decomposition, such as wavelet, pre-processing together with a compression technique, such as an auto-associative artificial neural network, operating in the multi-scale decomposition domain for signal denoising and extraction. all compressions are performed in the decomposed domain. a reverse decomposition such as an inverse discrete wavelet transform is performed on the combined outputs from all the compression modules to recover a clean signal back in the time domain. a low-cost, non-drug, non-invasive, on-demand therapy braincap system and method are pharmaceutically non-intrusive to the body for the purpose of disease diagnosis, treatment therapy, and direct mind control of external devices and systems. it is based on recognizing abnormal brainwave signatures and intervenes at the earliest moment, using magnetic and/or electric stimulations to reset the brainwaves back to normality. the feedback system is self-regulatory and the treatment stops when the brainwaves return to normal. the braincap contains multiple sensing electrodes and microcoils; the microcoils are pairs of crossed microcoils or 3-axis triple crossed microcoils.",2014-05-13,"The title of the patent is signal processing method and apparatus and its abstract is a signal processing method and system combines multi-scale decomposition, such as wavelet, pre-processing together with a compression technique, such as an auto-associative artificial neural network, operating in the multi-scale decomposition domain for signal denoising and extraction. all compressions are performed in the decomposed domain. a reverse decomposition such as an inverse discrete wavelet transform is performed on the combined outputs from all the compression modules to recover a clean signal back in the time domain. a low-cost, non-drug, non-invasive, on-demand therapy braincap system and method are pharmaceutically non-intrusive to the body for the purpose of disease diagnosis, treatment therapy, and direct mind control of external devices and systems. it is based on recognizing abnormal brainwave signatures and intervenes at the earliest moment, using magnetic and/or electric stimulations to reset the brainwaves back to normality. the feedback system is self-regulatory and the treatment stops when the brainwaves return to normal. the braincap contains multiple sensing electrodes and microcoils; the microcoils are pairs of crossed microcoils or 3-axis triple crossed microcoils. dated 2014-05-13"
8732758,television system with aided user program searching,a system having an adaptive browse feature and an adaptive flip feature is provided. the adaptative browse and flip features may be selected to receive program viewing suggestions. the system may provide a suggestion by displaying an adaptive browse region or adaptative flip region including a program suggestion. the system identifies programs to suggest based on a user=s viewing activity. the system uses different algorithms that are user-selectable and user-adjustable to identify program suggestions. the system may query a program guide database to build a list of programs having attributes similar to the attributes of the current program or the last viewed program. the system may use an adaptive learning algorithm such as a neural network. the neural network may be trained by the program guide by monitoring user-viewing activity.each algorithm may be personalized for multiple users.,2014-05-20,The title of the patent is television system with aided user program searching and its abstract is a system having an adaptive browse feature and an adaptive flip feature is provided. the adaptative browse and flip features may be selected to receive program viewing suggestions. the system may provide a suggestion by displaying an adaptive browse region or adaptative flip region including a program suggestion. the system identifies programs to suggest based on a user=s viewing activity. the system uses different algorithms that are user-selectable and user-adjustable to identify program suggestions. the system may query a program guide database to build a list of programs having attributes similar to the attributes of the current program or the last viewed program. the system may use an adaptive learning algorithm such as a neural network. the neural network may be trained by the program guide by monitoring user-viewing activity.each algorithm may be personalized for multiple users. dated 2014-05-20
8738032,hybrid location using a weighted average of location readings and signal strengths of wireless access points,"a query device scans radio frequencies for visible transmitting devices. the querying device receives at least a signal strength and identifier information associated with each of the transmitting devices. the list of visible devices is used to query a database containing location information for a plurality of visible devices. the list may be sent to a locationing system that may perform a location analysis on the resulting data to return a location to the query device. the weighted average of the locations returned in the database query may be computed to determine the location of the querying device, with the weight for each of the locations being the current signal strength detected by the querying device. neural network analysis may also be used to determine the location of the querying device. learning and seeding operations many also be used to populate the database with location information for transmitting devices.",2014-05-27,"The title of the patent is hybrid location using a weighted average of location readings and signal strengths of wireless access points and its abstract is a query device scans radio frequencies for visible transmitting devices. the querying device receives at least a signal strength and identifier information associated with each of the transmitting devices. the list of visible devices is used to query a database containing location information for a plurality of visible devices. the list may be sent to a locationing system that may perform a location analysis on the resulting data to return a location to the query device. the weighted average of the locations returned in the database query may be computed to determine the location of the querying device, with the weight for each of the locations being the current signal strength detected by the querying device. neural network analysis may also be used to determine the location of the querying device. learning and seeding operations many also be used to populate the database with location information for transmitting devices. dated 2014-05-27"
8738554,event-driven universal neural network circuit,the present invention provides an event-driven universal neural network circuit. the circuit comprises a plurality of neural modules. each neural module comprises multiple digital neurons such that each neuron in a neural module has a corresponding neuron in another neural module. an interconnection network comprising a plurality of digital synapses interconnects the neural modules. each synapse interconnects a first neural module to a second neural module by interconnecting a neuron in the first neural module to a corresponding neuron in the second neural module. corresponding neurons in the first neural module and the second neural module communicate via the synapses. each synapse comprises a learning rule associating a neuron in the first neural module with a corresponding neuron in the second neural module. a control module generates signals which define a set of time steps for event-driven operation of the neurons and event communication via the interconnection network.,2014-05-27,The title of the patent is event-driven universal neural network circuit and its abstract is the present invention provides an event-driven universal neural network circuit. the circuit comprises a plurality of neural modules. each neural module comprises multiple digital neurons such that each neuron in a neural module has a corresponding neuron in another neural module. an interconnection network comprising a plurality of digital synapses interconnects the neural modules. each synapse interconnects a first neural module to a second neural module by interconnecting a neuron in the first neural module to a corresponding neuron in the second neural module. corresponding neurons in the first neural module and the second neural module communicate via the synapses. each synapse comprises a learning rule associating a neuron in the first neural module with a corresponding neuron in the second neural module. a control module generates signals which define a set of time steps for event-driven operation of the neurons and event communication via the interconnection network. dated 2014-05-27
8744986,effort estimation using text analysis,"a system, method and program product for estimating effort of implementing a system based on a use case specification document. a system is provided that includes: a volumetrics processor that quantifies a structure of the document and evaluates a format of the document; a domain processor that identifies a domain of the system associated with the document; a complexity processor that defines a set of complexity variables associated with the document based on the structure of the document, a format of the document and a domain of the document; and a neural network that estimates an effort based on the set of complexity variables.",2014-06-03,"The title of the patent is effort estimation using text analysis and its abstract is a system, method and program product for estimating effort of implementing a system based on a use case specification document. a system is provided that includes: a volumetrics processor that quantifies a structure of the document and evaluates a format of the document; a domain processor that identifies a domain of the system associated with the document; a complexity processor that defines a set of complexity variables associated with the document based on the structure of the document, a format of the document and a domain of the document; and a neural network that estimates an effort based on the set of complexity variables. dated 2014-06-03"
8755958,underwater vehicles with improved efficiency over a range of velocities,"an underwater vehicle uses an undulatory fin propulsion system and a control apparatus to receive sensor data and control the operation of the vehicle in real time over a wide range of forward velocities. in an embodiment, a model of propulsive efficiency is used to achieve high values of propulsive efficiency, giving a lowered energy drain on the battery. externally monitored information, such as that on flow velocity, is conveyed to the control apparatus residing in the vehicle's control unit, which in turn signals the propulsion system to adopt kinematics, such as fin frequency and amplitude, associated with optimal propulsion efficiency. in an embodiment, the model of propulsive efficiency is generated from a multilayer perception neural network model using data from aquatic species, such as undulatory fin propulsion in the knifefish (xenomystus nigri), and a sensitivity analysis is used to lower the number of required inputs. power savings could protract vehicle operational life and/or provide more power to other functions, such as communications.",2014-06-17,"The title of the patent is underwater vehicles with improved efficiency over a range of velocities and its abstract is an underwater vehicle uses an undulatory fin propulsion system and a control apparatus to receive sensor data and control the operation of the vehicle in real time over a wide range of forward velocities. in an embodiment, a model of propulsive efficiency is used to achieve high values of propulsive efficiency, giving a lowered energy drain on the battery. externally monitored information, such as that on flow velocity, is conveyed to the control apparatus residing in the vehicle's control unit, which in turn signals the propulsion system to adopt kinematics, such as fin frequency and amplitude, associated with optimal propulsion efficiency. in an embodiment, the model of propulsive efficiency is generated from a multilayer perception neural network model using data from aquatic species, such as undulatory fin propulsion in the knifefish (xenomystus nigri), and a sensitivity analysis is used to lower the number of required inputs. power savings could protract vehicle operational life and/or provide more power to other functions, such as communications. dated 2014-06-17"
8761514,character recognition apparatus and method based on character orientation,"a character recognition apparatus and method based on a character orientation are provided, in which an input image is binarized, at least one character area is extracted from the binarized image, a slope value of the extracted at least one character area is calculated, the calculated slope value is set as a character feature value, and a character is recognized by using a neural network for recognizing a plurality of characters by receiving the set character feature value. accordingly, the probability of wrongly recognizing a similar character decreases, and a recognition ratio of each character increases.",2014-06-24,"The title of the patent is character recognition apparatus and method based on character orientation and its abstract is a character recognition apparatus and method based on a character orientation are provided, in which an input image is binarized, at least one character area is extracted from the binarized image, a slope value of the extracted at least one character area is calculated, the calculated slope value is set as a character feature value, and a character is recognized by using a neural network for recognizing a plurality of characters by receiving the set character feature value. accordingly, the probability of wrongly recognizing a similar character decreases, and a recognition ratio of each character increases. dated 2014-06-24"
8762061,process for generating spatially continuous wind profiles from wind profiler measurements,"a neural network process for improving wind retrievals from wind profiler measurements is described. in this invention, a neural network is trained to retrieve (missing or incomplete) upper level winds from ground based wind profiler measurements. radiosonde measurements in conjunction with wind profiler ground measurements for specific geographical locations are used as training sets for the neural network. the idea is to retrieve timely and spatially continuous upper level wind information from (fragmented or incomplete) wind profiler measurements.",2014-06-24,"The title of the patent is process for generating spatially continuous wind profiles from wind profiler measurements and its abstract is a neural network process for improving wind retrievals from wind profiler measurements is described. in this invention, a neural network is trained to retrieve (missing or incomplete) upper level winds from ground based wind profiler measurements. radiosonde measurements in conjunction with wind profiler ground measurements for specific geographical locations are used as training sets for the neural network. the idea is to retrieve timely and spatially continuous upper level wind information from (fragmented or incomplete) wind profiler measurements. dated 2014-06-24"
8762306,neural network for glucose therapy recommendation,a multifunctional neural network system for prediction which includes memory components to store previous values of data within a network. the memory components provide the system with the ability to learn relationships/patterns existent in the data over time.,2014-06-24,The title of the patent is neural network for glucose therapy recommendation and its abstract is a multifunctional neural network system for prediction which includes memory components to store previous values of data within a network. the memory components provide the system with the ability to learn relationships/patterns existent in the data over time. dated 2014-06-24
8762307,control system for plant,"a control system for a plant e.g. as a non-linear system, which is capable of properly suppressing interaction occurring between a plurality of control inputs and a plurality of controlled variables, thereby making it possible to properly control the controlled variables and easily design the control system. in the control system, each of a plurality of interaction suppression parameters for correcting the control inputs, respectively, such that the interaction is suppressed is calculated using a neural network constructed by using, out of the plurality of control inputs, a control input other than a control input corrected by a calculated interaction suppression parameter, as an input, and the interaction suppression parameter as an output.",2014-06-24,"The title of the patent is control system for plant and its abstract is a control system for a plant e.g. as a non-linear system, which is capable of properly suppressing interaction occurring between a plurality of control inputs and a plurality of controlled variables, thereby making it possible to properly control the controlled variables and easily design the control system. in the control system, each of a plurality of interaction suppression parameters for correcting the control inputs, respectively, such that the interaction is suppressed is calculated using a neural network constructed by using, out of the plurality of control inputs, a control input other than a control input corrected by a calculated interaction suppression parameter, as an input, and the interaction suppression parameter as an output. dated 2014-06-24"
8768556,protection envelope switching,"an apparatus defines a protection envelope in an aircraft, including a processor and at least one sensor, each sensor being coupled with the processor, the processor executing at least one neural network based algorithm. the at least one sensor monitors flight parameters of the aircraft thereby generating monitored flight parameters. the processor divides the performance envelope of the aircraft into predefined flight regimes, wherein for each predefined flight regime, the processor defines and stores a suitable protection envelope. the processor determines an estimated flight regime of the aircraft using the neural network based algorithm based on the monitored flight parameters. the processor selects a respective suitable protection envelope for the aircraft based on the estimated flight regime.",2014-07-01,"The title of the patent is protection envelope switching and its abstract is an apparatus defines a protection envelope in an aircraft, including a processor and at least one sensor, each sensor being coupled with the processor, the processor executing at least one neural network based algorithm. the at least one sensor monitors flight parameters of the aircraft thereby generating monitored flight parameters. the processor divides the performance envelope of the aircraft into predefined flight regimes, wherein for each predefined flight regime, the processor defines and stores a suitable protection envelope. the processor determines an estimated flight regime of the aircraft using the neural network based algorithm based on the monitored flight parameters. the processor selects a respective suitable protection envelope for the aircraft based on the estimated flight regime. dated 2014-07-01"
8774830,training pattern recognition systems for determining user device locations,"a query device scans radio frequencies for visible transmitting devices. the querying device receives at least a signal strength and identifier information associated with each of the transmitting devices. the list of visible devices is used to query a database containing location information for a plurality of visible devices. the list may be sent to a locationing system that may perform a location analysis on the resulting data to return a location to the query device. the weighted average of the locations returned in the database query may be computed to determine the location of the querying device, with the weight for each of the locations being the current signal strength detected by the querying device. neural network analysis may also be used to determine the location of the querying device. learning and seeding operations many also be used to populate the database with location information for transmitting devices.",2014-07-08,"The title of the patent is training pattern recognition systems for determining user device locations and its abstract is a query device scans radio frequencies for visible transmitting devices. the querying device receives at least a signal strength and identifier information associated with each of the transmitting devices. the list of visible devices is used to query a database containing location information for a plurality of visible devices. the list may be sent to a locationing system that may perform a location analysis on the resulting data to return a location to the query device. the weighted average of the locations returned in the database query may be computed to determine the location of the querying device, with the weight for each of the locations being the current signal strength detected by the querying device. neural network analysis may also be used to determine the location of the querying device. learning and seeding operations many also be used to populate the database with location information for transmitting devices. dated 2014-07-08"
8774831,database seeding with location information for wireless access points,"a query device scans radio frequencies for visible transmitting devices. the querying device receives at least a signal strength and identifier information associated with each of the transmitting devices. the list of visible devices is used to query a database containing location information for a plurality of visible devices. the list may be sent to a locationing system that may perform a location analysis on the resulting data to return a location to the query device. the weighted average of the locations returned in the database query may be computed to determine the location of the querying device, with the weight for each of the locations being the current signal strength detected by the querying device. neural network analysis may also be used to determine the location of the querying device. learning and seeding operations many also be used to populate the database with location information for transmitting devices.",2014-07-08,"The title of the patent is database seeding with location information for wireless access points and its abstract is a query device scans radio frequencies for visible transmitting devices. the querying device receives at least a signal strength and identifier information associated with each of the transmitting devices. the list of visible devices is used to query a database containing location information for a plurality of visible devices. the list may be sent to a locationing system that may perform a location analysis on the resulting data to return a location to the query device. the weighted average of the locations returned in the database query may be computed to determine the location of the querying device, with the weight for each of the locations being the current signal strength detected by the querying device. neural network analysis may also be used to determine the location of the querying device. learning and seeding operations many also be used to populate the database with location information for transmitting devices. dated 2014-07-08"
8775148,methods and systems for controlling a semiconductor fabrication process,"software for controlling processes in a heterogeneous semiconductor manufacturing environment may include a wafer-centric database, a real-time scheduler using a neural network, and a graphical user interface displaying simulated operation of the system. these features may be employed alone or in combination to offer improved usability and computational efficiency for real time control and monitoring of a semiconductor manufacturing process. more generally, these techniques may be usefully employed in a variety of real time control systems, particularly systems requiring complex scheduling decisions or heterogeneous systems constructed of hardware from numerous independent vendors.",2014-07-08,"The title of the patent is methods and systems for controlling a semiconductor fabrication process and its abstract is software for controlling processes in a heterogeneous semiconductor manufacturing environment may include a wafer-centric database, a real-time scheduler using a neural network, and a graphical user interface displaying simulated operation of the system. these features may be employed alone or in combination to offer improved usability and computational efficiency for real time control and monitoring of a semiconductor manufacturing process. more generally, these techniques may be usefully employed in a variety of real time control systems, particularly systems requiring complex scheduling decisions or heterogeneous systems constructed of hardware from numerous independent vendors. dated 2014-07-08"
8775340,detection and prediction of physiological events in people with sleep disordered breathing using a lamstar neural network,apparatus and methods are disclosed for generating and outputting physiological event results from physiological data related to a patient. physiological event results include results predicting and/or detecting individual physiological events related to a medical condition of the patient.,2014-07-08,The title of the patent is detection and prediction of physiological events in people with sleep disordered breathing using a lamstar neural network and its abstract is apparatus and methods are disclosed for generating and outputting physiological event results from physiological data related to a patient. physiological event results include results predicting and/or detecting individual physiological events related to a medical condition of the patient. dated 2014-07-08
8775341,intelligent control with hierarchical stacked neural networks,"a system and method of detecting an aberrant message is provided. an ordered set of words within the message is detected. the set of words found within the message is linked to a corresponding set of expected words, the set of expected words having semantic attributes. a set of grammatical structures represented in the message is detected, based on the ordered set of words and the semantic attributes of the corresponding set of expected words. a cognitive noise vector comprising a quantitative measure of a deviation between grammatical structures represented in the message and an expected measure of grammatical structures for a message of the type is then determined. the cognitive noise vector may be processed by higher levels of the neural network and/or an external processor.",2014-07-08,"The title of the patent is intelligent control with hierarchical stacked neural networks and its abstract is a system and method of detecting an aberrant message is provided. an ordered set of words within the message is detected. the set of words found within the message is linked to a corresponding set of expected words, the set of expected words having semantic attributes. a set of grammatical structures represented in the message is detected, based on the ordered set of words and the semantic attributes of the corresponding set of expected words. a cognitive noise vector comprising a quantitative measure of a deviation between grammatical structures represented in the message and an expected measure of grammatical structures for a message of the type is then determined. the cognitive noise vector may be processed by higher levels of the neural network and/or an external processor. dated 2014-07-08"
8775346,learning method of neural network circuit,"a neuron circuit in a neural network circuit element includes a waveform generating circuit for generating a bipolar sawtooth pulse voltage, and a first input signal has a bipolar sawtooth pulse waveform. for a period during which the first input signal is permitted to be input to a first electrode of a variable resistance element, the bipolar sawtooth pulse voltage generated within the neural network circuit element including the variable resistance element which is applied with the first input signal from another neural network circuit element is input to a control electrode of the variable resistance element. the resistance value of the variable resistance element changes due to an electric potential difference between the first electrode and the control electrode, the electric potential difference being generated depending on an input timing difference between a voltage applied to the first electrode and the voltage applied to the control electrode.",2014-07-08,"The title of the patent is learning method of neural network circuit and its abstract is a neuron circuit in a neural network circuit element includes a waveform generating circuit for generating a bipolar sawtooth pulse voltage, and a first input signal has a bipolar sawtooth pulse waveform. for a period during which the first input signal is permitted to be input to a first electrode of a variable resistance element, the bipolar sawtooth pulse voltage generated within the neural network circuit element including the variable resistance element which is applied with the first input signal from another neural network circuit element is input to a control electrode of the variable resistance element. the resistance value of the variable resistance element changes due to an electric potential difference between the first electrode and the control electrode, the electric potential difference being generated depending on an input timing difference between a voltage applied to the first electrode and the voltage applied to the control electrode. dated 2014-07-08"
8781632,hydrate monitoring system,a method for analyzing a fluid containing one or more analytes of interest includes; measuring a plurality of properties of a sample fluid with unknown concentrations of the one or more analytes of interest; and using the measurements and a model of the relationship between the plurality of properties and concentrations of the one or more analytes to calculate the concentration of at least one of the analytes of interest. the model may be an artificial neural network. the method may be used to monitor the concentration of inhibitors of gas hydrate formation in a fluid. apparatus for use in the method is also provided.,2014-07-15,The title of the patent is hydrate monitoring system and its abstract is a method for analyzing a fluid containing one or more analytes of interest includes; measuring a plurality of properties of a sample fluid with unknown concentrations of the one or more analytes of interest; and using the measurements and a model of the relationship between the plurality of properties and concentrations of the one or more analytes to calculate the concentration of at least one of the analytes of interest. the model may be an artificial neural network. the method may be used to monitor the concentration of inhibitors of gas hydrate formation in a fluid. apparatus for use in the method is also provided. dated 2014-07-15
8781982,system and method for estimating remaining useful life,mechanisms for predicting a remaining useful life of a cutter head of a milling machine that includes a plurality of flutes are disclosed. features are extracted from reference data associated with a plurality of reference cutter heads. the reference data includes reference vibration data and reference wear data. at least two neural network predictive models are trained in parallel for predicting the remaining life of a new cutter head based upon the extracted features. operational data associated with the new cutter head is obtained. the operational data includes operational vibration data and operational wear data. features extracted from the operational data are input into an optimal predictive model of the at least two neural network predictive models. a remaining useful life of the new cutter head is estimated by the optimal predictive model.,2014-07-15,The title of the patent is system and method for estimating remaining useful life and its abstract is mechanisms for predicting a remaining useful life of a cutter head of a milling machine that includes a plurality of flutes are disclosed. features are extracted from reference data associated with a plurality of reference cutter heads. the reference data includes reference vibration data and reference wear data. at least two neural network predictive models are trained in parallel for predicting the remaining life of a new cutter head based upon the extracted features. operational data associated with the new cutter head is obtained. the operational data includes operational vibration data and operational wear data. features extracted from the operational data are input into an optimal predictive model of the at least two neural network predictive models. a remaining useful life of the new cutter head is estimated by the optimal predictive model. dated 2014-07-15
8786288,concentric buttons of different sizes for imaging and standoff correction,disclosed is a method of estimating a property of an earth formation penetrated by a borehole. the method includes conveying a carrier through the borehole and performing a plurality of electrical measurements on the formation using a sensor disposed at the carrier and having a plurality of electrodes disposed in a concentric arrangement wherein a standoff distance between the sensor and a wall of the borehole has an influence on each electrical measurement in the plurality of electrical measurements. the method further includes determining an impedance for each electrical measurement in the plurality of electrical measurements and inputting the determined impedances into an artificial neural network implemented by a processor. the artificial neural network outputs the property wherein the outputted property compensates for the influence of sensor standoff distance on each electrical measurement in the plurality of electrical measurements.,2014-07-22,The title of the patent is concentric buttons of different sizes for imaging and standoff correction and its abstract is disclosed is a method of estimating a property of an earth formation penetrated by a borehole. the method includes conveying a carrier through the borehole and performing a plurality of electrical measurements on the formation using a sensor disposed at the carrier and having a plurality of electrodes disposed in a concentric arrangement wherein a standoff distance between the sensor and a wall of the borehole has an influence on each electrical measurement in the plurality of electrical measurements. the method further includes determining an impedance for each electrical measurement in the plurality of electrical measurements and inputting the determined impedances into an artificial neural network implemented by a processor. the artificial neural network outputs the property wherein the outputted property compensates for the influence of sensor standoff distance on each electrical measurement in the plurality of electrical measurements. dated 2014-07-22
8788441,intelligent control with hierarchical stacked neural networks,"an intelligent control system based on an explicit model of cognitive development (table 1) performs high-level functions. it comprises up to o hierarchically stacked neural networks, nm, . . . , nm+(o−1), where m denotes the stage/order tasks performed in the first neural network, nm, and o denotes the highest stage/order tasks performed in the highest-level neural network. the type of processing actions performed in a network, nm, corresponds to the complexity for stage/order m. thus n1 performs tasks at the level corresponding to stage/order 1. n5 processes information at the level corresponding to stage/order 5. stacked neural networks begin and end at any stage/order, but information must be processed by each stage in ascending order sequence. stages/orders cannot be skipped. each neural network in a stack may use different architectures, interconnections, algorithms, and training methods, depending on the stage/order of the neural network and the type of intelligent control system implemented.",2014-07-22,"The title of the patent is intelligent control with hierarchical stacked neural networks and its abstract is an intelligent control system based on an explicit model of cognitive development (table 1) performs high-level functions. it comprises up to o hierarchically stacked neural networks, nm, . . . , nm+(o−1), where m denotes the stage/order tasks performed in the first neural network, nm, and o denotes the highest stage/order tasks performed in the highest-level neural network. the type of processing actions performed in a network, nm, corresponds to the complexity for stage/order m. thus n1 performs tasks at the level corresponding to stage/order 1. n5 processes information at the level corresponding to stage/order 5. stacked neural networks begin and end at any stage/order, but information must be processed by each stage in ascending order sequence. stages/orders cannot be skipped. each neural network in a stack may use different architectures, interconnections, algorithms, and training methods, depending on the stage/order of the neural network and the type of intelligent control system implemented. dated 2014-07-22"
8788444,data analysis method and system,"the present invention relates to the analysis of data to identify relationships between the input data and one or more conditions. one method of analyzing such data is by the use of neural networks which are non-linear statistical data modelling tools, the structure of which may be changed based on information that is passed through the network during a training phase. a known problem that affects neural networks is the issue of overtraining which arises in overcomplex or overspecified systems when the capacity of the network significantly exceeds the needed parameters. the present invention provides a method of analyzing data using a neural network with a constrained architecture that mitigates the problems associated with the prior art.",2014-07-22,"The title of the patent is data analysis method and system and its abstract is the present invention relates to the analysis of data to identify relationships between the input data and one or more conditions. one method of analyzing such data is by the use of neural networks which are non-linear statistical data modelling tools, the structure of which may be changed based on information that is passed through the network during a training phase. a known problem that affects neural networks is the issue of overtraining which arises in overcomplex or overspecified systems when the capacity of the network significantly exceeds the needed parameters. the present invention provides a method of analyzing data using a neural network with a constrained architecture that mitigates the problems associated with the prior art. dated 2014-07-22"
8793203,effort estimation using text analysis,"a system, method and program product for estimating effort of implementing a system based on a use case specification document. a method is provided that includes: quantifying a structure of the document and evaluating a format of the document using a computing device; identifying a domain of an application associated with the document; defining a set of complexity variables associated with the document based on the structure of the document, a format of the document and a domain of the document; using a neural network to estimate an effort based on the set of complexity variables; and outputting the effort via a tangible medium.",2014-07-29,"The title of the patent is effort estimation using text analysis and its abstract is a system, method and program product for estimating effort of implementing a system based on a use case specification document. a method is provided that includes: quantifying a structure of the document and evaluating a format of the document using a computing device; identifying a domain of an application associated with the document; defining a set of complexity variables associated with the document based on the structure of the document, a format of the document and a domain of the document; using a neural network to estimate an effort based on the set of complexity variables; and outputting the effort via a tangible medium. dated 2014-07-29"
8798345,"diagnosis processing device, diagnosis processing system, diagnosis processing method, diagnosis processing program and computer-readable recording medium, and classification processing device","a diagnosis processing device is provided in which diagnosis is realizable by a simple arrangement. a diagnosis processing device (1) of the present invention includes: a learning pattern creating section (10a) for creating a learning pattern by sampling data from a learning image in which abnormality information indicating a substantive feature of abnormality of a target is pre-known; a learning processing section (12) for causing a neural network (17) to learn, by using learning patterns; a diagnostic pattern creating section (10b) for creating a diagnostic pattern by sampling data from a diagnostic image in which abnormality information is unknown; a determination processing section (18) for determining a substantive feature of the abnormality of the target indicated in the abnormality information in the diagnostic image, based on an output value outputted, in response to an input of the diagnostic pattern, from a learned neural network (17) which is a neural network subjected to learning.",2014-08-05,"The title of the patent is diagnosis processing device, diagnosis processing system, diagnosis processing method, diagnosis processing program and computer-readable recording medium, and classification processing device and its abstract is a diagnosis processing device is provided in which diagnosis is realizable by a simple arrangement. a diagnosis processing device (1) of the present invention includes: a learning pattern creating section (10a) for creating a learning pattern by sampling data from a learning image in which abnormality information indicating a substantive feature of abnormality of a target is pre-known; a learning processing section (12) for causing a neural network (17) to learn, by using learning patterns; a diagnostic pattern creating section (10b) for creating a diagnostic pattern by sampling data from a diagnostic image in which abnormality information is unknown; a determination processing section (18) for determining a substantive feature of the abnormality of the target indicated in the abnormality information in the diagnostic image, based on an output value outputted, in response to an input of the diagnostic pattern, from a learned neural network (17) which is a neural network subjected to learning. dated 2014-08-05"
8799198,borehole drilling optimization with multiple cutting structures,a method of optimizing a drilling operating parameter or a drilling system parameter for a drilling assembly employing at least first and second distinct cutting structures includes entering at least one design parameter for each of the cutting structures into a trained artificial neural network. at least one of the design parameters of the first cutting structure may be optionally combined with at least one of the design parameters of the second cutting structure. the combined design parameter may also be entered into the artificial neural network.,2014-08-05,The title of the patent is borehole drilling optimization with multiple cutting structures and its abstract is a method of optimizing a drilling operating parameter or a drilling system parameter for a drilling assembly employing at least first and second distinct cutting structures includes entering at least one design parameter for each of the cutting structures into a trained artificial neural network. at least one of the design parameters of the first cutting structure may be optionally combined with at least one of the design parameters of the second cutting structure. the combined design parameter may also be entered into the artificial neural network. dated 2014-08-05
8799199,"universal, online learning in multi-modal perception-action semilattices","in one embodiment, the present invention provides a method for interconnecting neurons in a neural network. at least one node among a first set of nodes is interconnected to at least one node among a second set of nodes, and nodes of the first and second set are arranged in a lattice. at least one node of the first set represents a sensory-motor modality of the neural network. at least one node of the second set is a union of at least two nodes of the first set. each node in the lattice has an acyclic digraph comprising multiple vertices and directed edges. each vertex represents a neuron population. each directed edge comprises multiple synaptic connections. vertices in different acyclic digraphs are interconnected using an acyclic bottom-up digraph. the bottom-up digraph has a corresponding acyclic top-down digraph. vertices in the bottom-up digraph are interconnected to vertices in the top-down digraph.",2014-08-05,"The title of the patent is universal, online learning in multi-modal perception-action semilattices and its abstract is in one embodiment, the present invention provides a method for interconnecting neurons in a neural network. at least one node among a first set of nodes is interconnected to at least one node among a second set of nodes, and nodes of the first and second set are arranged in a lattice. at least one node of the first set represents a sensory-motor modality of the neural network. at least one node of the second set is a union of at least two nodes of the first set. each node in the lattice has an acyclic digraph comprising multiple vertices and directed edges. each vertex represents a neuron population. each directed edge comprises multiple synaptic connections. vertices in different acyclic digraphs are interconnected using an acyclic bottom-up digraph. the bottom-up digraph has a corresponding acyclic top-down digraph. vertices in the bottom-up digraph are interconnected to vertices in the top-down digraph. dated 2014-08-05"
8801626,flexible neural localization devices and methods,"methods for determining if a nerve is nearby a device. the neural stimulation tools described herein are configured to be flexible and low-profile, so that they can be used within body regions that may be tortuous or difficult to reach, such as within a compressed or partially occluded neural foramen. in most cases, these tools described herein are ribbon-shaped and adapted to be manipulated bimanually, applying force to the ends of the devices from separate locations outside of the patient's body. thus, the distal end region of the device may be configured to couple to the proximal end of a guidewire. one or more surfaces of the devices may include an electrode or multi-polar network of electrodes configured to stimulate only nerves within a predetermined distance of a particular face of the device. methods of using these devices are described.",2014-08-12,"The title of the patent is flexible neural localization devices and methods and its abstract is methods for determining if a nerve is nearby a device. the neural stimulation tools described herein are configured to be flexible and low-profile, so that they can be used within body regions that may be tortuous or difficult to reach, such as within a compressed or partially occluded neural foramen. in most cases, these tools described herein are ribbon-shaped and adapted to be manipulated bimanually, applying force to the ends of the devices from separate locations outside of the patient's body. thus, the distal end region of the device may be configured to couple to the proximal end of a guidewire. one or more surfaces of the devices may include an electrode or multi-polar network of electrodes configured to stimulate only nerves within a predetermined distance of a particular face of the device. methods of using these devices are described. dated 2014-08-12"
8805581,procedural memory learning and robot control,"methods and apparatus for procedural memory learning to control a robot by demonstrating a task action to the robot and having the robot learn the action according to a similarity matrix of correlated values, attributes, and parameters obtained from the robot as the robot performs the demonstrated action. learning is done by an artificial neural network associated with the robot controller, so that the robot learns to perform the task associated with the similarity matrix. extended similarity matrices can contain integrated and differentiated values of variables. procedural memory learning reduces overhead in instructing robots to perform tasks. continued learning improves performance and provides automatic compensation for changes in robot condition and environmental factors.",2014-08-12,"The title of the patent is procedural memory learning and robot control and its abstract is methods and apparatus for procedural memory learning to control a robot by demonstrating a task action to the robot and having the robot learn the action according to a similarity matrix of correlated values, attributes, and parameters obtained from the robot as the robot performs the demonstrated action. learning is done by an artificial neural network associated with the robot controller, so that the robot learns to perform the task associated with the similarity matrix. extended similarity matrices can contain integrated and differentiated values of variables. procedural memory learning reduces overhead in instructing robots to perform tasks. continued learning improves performance and provides automatic compensation for changes in robot condition and environmental factors. dated 2014-08-12"
8805587,method for optimizing and controlling pressure in gas-oil separation plants,the method for optimizing and controlling pressure in gas-oil separation plants utilizes a genetic algorithm-based control method for controlling pressure in each stage of a multi-stage gas-oil separation plant to optimize oil production parameters. a neural network simulation model is used with an optimization procedure to provide on-line operation optimization of the multi-stage gas-oil separation plant. pressure set points of each stage are automatically and continuously adjusted in the presence of fluctuating ambient temperatures and production rates to ensure optimal oil recovery and optimal quality of the produced oil.,2014-08-12,The title of the patent is method for optimizing and controlling pressure in gas-oil separation plants and its abstract is the method for optimizing and controlling pressure in gas-oil separation plants utilizes a genetic algorithm-based control method for controlling pressure in each stage of a multi-stage gas-oil separation plant to optimize oil production parameters. a neural network simulation model is used with an optimization procedure to provide on-line operation optimization of the multi-stage gas-oil separation plant. pressure set points of each stage are automatically and continuously adjusted in the presence of fluctuating ambient temperatures and production rates to ensure optimal oil recovery and optimal quality of the produced oil. dated 2014-08-12
8811427,method for data transmission in a communication network,a method for data transmission from a first node to a second node in a communication network includes: analyzing data arriving in a buffer of the first node to generate an arrival history curve describing the amounts of data arriving in the buffer over a past time period; processing the arrival history curve to generate a plurality of signals corresponding to filtered components of the wavelet transform; processing each signal in a separate neural network trained with training patterns based on previously received and buffered data to generate forecast signals for future data arrivals; recombining the forecast signals to generate an arrival forecast curve describing the amounts of data arriving in the buffer in a future period; the first node generating bandwidth reservation requests based on the arrival forecast curve and transmitting the bandwidth requests to the second node; the second node allocating bandwidth based on the bandwidth requests; and transmitting data from the first node to the second node within the allocated bandwidth.,2014-08-19,The title of the patent is method for data transmission in a communication network and its abstract is a method for data transmission from a first node to a second node in a communication network includes: analyzing data arriving in a buffer of the first node to generate an arrival history curve describing the amounts of data arriving in the buffer over a past time period; processing the arrival history curve to generate a plurality of signals corresponding to filtered components of the wavelet transform; processing each signal in a separate neural network trained with training patterns based on previously received and buffered data to generate forecast signals for future data arrivals; recombining the forecast signals to generate an arrival forecast curve describing the amounts of data arriving in the buffer in a future period; the first node generating bandwidth reservation requests based on the arrival forecast curve and transmitting the bandwidth requests to the second node; the second node allocating bandwidth based on the bandwidth requests; and transmitting data from the first node to the second node within the allocated bandwidth. dated 2014-08-19
8812414,low-power event-driven neural computing architecture in neural networks,"a neural network includes an electronic synapse array of multiple digital synapses interconnecting a plurality of digital electronic neurons. each synapse interconnects an axon of a pre-synaptic neuron with a dendrite of a post-synaptic neuron. each neuron integrates input spikes and generates a spike event in response to the integrated input spikes exceeding a threshold. a decoder receives spike events sequentially and transmits the spike events to selected axons in the synapse array. an encoder transmits spike events corresponding to spiking neurons. a controller coordinates events from the synapse array to the neurons, and signals when neurons may compute their spike events within each time step, ensuring one-to-one correspondence with an equivalent software model. the synapse array includes an interconnecting crossbar that sequentially receives spike events from axons, wherein one axon at a time drives the crossbar, and the crossbar transmits synaptic events in parallel to multiple neurons.",2014-08-19,"The title of the patent is low-power event-driven neural computing architecture in neural networks and its abstract is a neural network includes an electronic synapse array of multiple digital synapses interconnecting a plurality of digital electronic neurons. each synapse interconnects an axon of a pre-synaptic neuron with a dendrite of a post-synaptic neuron. each neuron integrates input spikes and generates a spike event in response to the integrated input spikes exceeding a threshold. a decoder receives spike events sequentially and transmits the spike events to selected axons in the synapse array. an encoder transmits spike events corresponding to spiking neurons. a controller coordinates events from the synapse array to the neurons, and signals when neurons may compute their spike events within each time step, ensuring one-to-one correspondence with an equivalent software model. the synapse array includes an interconnecting crossbar that sequentially receives spike events from axons, wherein one axon at a time drives the crossbar, and the crossbar transmits synaptic events in parallel to multiple neurons. dated 2014-08-19"
8812415,neuromorphic and synaptronic spiking neural network crossbar circuits with synaptic weights learned using a one-to-one correspondence with a simulation,"embodiments of the invention provide neuromorphic-synaptronic systems, including neuromorphic-synaptronic circuit chips implementing spiking neural network with synaptic weights learned using simulation. one embodiment includes simulating a spiking neural network to generate synaptic weights learned via the simulation while maintaining one-to-one correspondence between the simulation and a digital circuit chip. the learned synaptic weights are loaded into the digital circuit chip implementing a spiking neural network, the digital circuit chip comprising a neuromorphic-synaptronic spiking neural network including plural synapse devices interconnecting multiple digital neurons.",2014-08-19,"The title of the patent is neuromorphic and synaptronic spiking neural network crossbar circuits with synaptic weights learned using a one-to-one correspondence with a simulation and its abstract is embodiments of the invention provide neuromorphic-synaptronic systems, including neuromorphic-synaptronic circuit chips implementing spiking neural network with synaptic weights learned using simulation. one embodiment includes simulating a spiking neural network to generate synaptic weights learned via the simulation while maintaining one-to-one correspondence between the simulation and a digital circuit chip. the learned synaptic weights are loaded into the digital circuit chip implementing a spiking neural network, the digital circuit chip comprising a neuromorphic-synaptronic spiking neural network including plural synapse devices interconnecting multiple digital neurons. dated 2014-08-19"
8812418,memristive adaptive resonance networks,a method for implementing an artificial neural network includes connecting a plurality of receiving neurons to a plurality of transmitting neurons through memristive synapses. each memristive synapse has a weight which is initialized into a conductive state. a binary input vector is presented through the memristive synapses to the plurality of receiving neurons and the state of one or more of the memristive synapses modified based on the binary input vector.,2014-08-19,The title of the patent is memristive adaptive resonance networks and its abstract is a method for implementing an artificial neural network includes connecting a plurality of receiving neurons to a plurality of transmitting neurons through memristive synapses. each memristive synapse has a weight which is initialized into a conductive state. a binary input vector is presented through the memristive synapses to the plurality of receiving neurons and the state of one or more of the memristive synapses modified based on the binary input vector. dated 2014-08-19
8812493,search results ranking using editing distance and document information,"architecture for extracting document information from documents received as search results based on a query string, and computing an edit distance between the data string and the query string. the edit distance is employed in determining relevance of the document as part of result ranking by detecting near-matches of a whole query or part of the query. the edit distance evaluates how close the query string is to a given data stream that includes document information such as tauc (title, anchor text, url, clicks) information, etc. the architecture includes the index-time splitting of compound terms in the url to allow the more effective discovery of query terms. additionally, index-time filtering of anchor text is utilized to find the top n anchors of one or more of the document results. the tauc information can be input to a neural network (e.g., 2-layer) to improve relevance metrics for ranking the search results.",2014-08-19,"The title of the patent is search results ranking using editing distance and document information and its abstract is architecture for extracting document information from documents received as search results based on a query string, and computing an edit distance between the data string and the query string. the edit distance is employed in determining relevance of the document as part of result ranking by detecting near-matches of a whole query or part of the query. the edit distance evaluates how close the query string is to a given data stream that includes document information such as tauc (title, anchor text, url, clicks) information, etc. the architecture includes the index-time splitting of compound terms in the url to allow the more effective discovery of query terms. additionally, index-time filtering of anchor text is utilized to find the top n anchors of one or more of the document results. the tauc information can be input to a neural network (e.g., 2-layer) to improve relevance metrics for ranking the search results. dated 2014-08-19"
8818676,redundant torque security path,an engine control system includes a torque request control module to determine a first engine torque request. an artificial neural network (ann) torque request module determines a second engine torque request using an ann model. a torque security check module that selectively generates a malfunction signal based on the difference between the first engine torque request and the second engine torque request.,2014-08-26,The title of the patent is redundant torque security path and its abstract is an engine control system includes a torque request control module to determine a first engine torque request. an artificial neural network (ann) torque request module determines a second engine torque request using an ann model. a torque security check module that selectively generates a malfunction signal based on the difference between the first engine torque request and the second engine torque request. dated 2014-08-26
8818923,neural network device with engineered delays for pattern storage and matching,"described is a system for searching a continuous data stream for exact matches with a priori stored data sequences. the system includes a neural network with an input and an output layer. the input layer has one neuron for each possible character or number in the data stream, and the output layer has one neuron for each stored pattern. importantly, the delays of the connections from input to output layer are engineered to match the temporal occurrence of an input character within a stored sequence. thus, if an input sequence has the proper time gaps between characters, matching a stored pattern, then the delayed neural signals result in a simultaneous activation at the receiving neuron, which indicates a detected pattern. for storing a pattern, only one connection for each pair of input character and output neuron has to be specified resulting in sparse coding and quick storage.",2014-08-26,"The title of the patent is neural network device with engineered delays for pattern storage and matching and its abstract is described is a system for searching a continuous data stream for exact matches with a priori stored data sequences. the system includes a neural network with an input and an output layer. the input layer has one neuron for each possible character or number in the data stream, and the output layer has one neuron for each stored pattern. importantly, the delays of the connections from input to output layer are engineered to match the temporal occurrence of an input character within a stored sequence. thus, if an input sequence has the proper time gaps between characters, matching a stored pattern, then the delayed neural signals result in a simultaneous activation at the receiving neuron, which indicates a detected pattern. for storing a pattern, only one connection for each pair of input character and output neuron has to be specified resulting in sparse coding and quick storage. dated 2014-08-26"
8825350,systems and methods involving features of adaptive and/or autonomous traffic control,"systems and method are disclosed for adaptive and/or autonomous traffic control. in one illustrative implementation, there is provided a method for processing traffic information. moreover, the method may include receiving data regarding travel of vehicles associated with an intersection, using neural network technology to recognize types and/or states of traffic, and using the neural network technology to process/determine/memorize optimal traffic flow decisions as a function of experience information. exemplary implementations may also include using the neural network technology to achieve efficient traffic flow via recognition of the optimal traffic flow decisions.",2014-09-02,"The title of the patent is systems and methods involving features of adaptive and/or autonomous traffic control and its abstract is systems and method are disclosed for adaptive and/or autonomous traffic control. in one illustrative implementation, there is provided a method for processing traffic information. moreover, the method may include receiving data regarding travel of vehicles associated with an intersection, using neural network technology to recognize types and/or states of traffic, and using the neural network technology to process/determine/memorize optimal traffic flow decisions as a function of experience information. exemplary implementations may also include using the neural network technology to achieve efficient traffic flow via recognition of the optimal traffic flow decisions. dated 2014-09-02"
8827716,method and device for automation of the interpretative analysis of a digital line drawn by a person,"the method for the automation of the interpretative analysis of a digital line drawn by a person is provided. the method includes a preliminary step for the acquisition of an archive of digital lines drawn by persons; a preliminary step for the acquisition of graphic evaluations of each of the lines of the archive; a preliminary step for the acquisition of psychological evaluations of the persons who have drawn each of the lines of the archive. the method further includes a preliminary step of providing a first and/or at least a second supervised neural network and relative training to associate, for each of the lines of the archive, the graphic evaluations with the corresponding psychological evaluations. the method further includes a step for the acquisition of a new digital line drawn by a person for the execution of its interpretative analysis; a step for the acquisition of the graphic evaluations for the new line; and an association step wherein the first and/or second trained supervised neural network automatically associates the corresponding psychological evaluations with the graphic evaluations of the new line.",2014-09-09,"The title of the patent is method and device for automation of the interpretative analysis of a digital line drawn by a person and its abstract is the method for the automation of the interpretative analysis of a digital line drawn by a person is provided. the method includes a preliminary step for the acquisition of an archive of digital lines drawn by persons; a preliminary step for the acquisition of graphic evaluations of each of the lines of the archive; a preliminary step for the acquisition of psychological evaluations of the persons who have drawn each of the lines of the archive. the method further includes a preliminary step of providing a first and/or at least a second supervised neural network and relative training to associate, for each of the lines of the archive, the graphic evaluations with the corresponding psychological evaluations. the method further includes a step for the acquisition of a new digital line drawn by a person for the execution of its interpretative analysis; a step for the acquisition of the graphic evaluations for the new line; and an association step wherein the first and/or second trained supervised neural network automatically associates the corresponding psychological evaluations with the graphic evaluations of the new line. dated 2014-09-09"
8838446,method and apparatus of transforming speech feature vectors using an auto-associative neural network,provided is a method and apparatus for transforming a speech feature vector. the method includes extracting a feature vector required for speech recognition from a speech signal and transforming the extracted feature vector using an auto-associative neural network (aann).,2014-09-16,The title of the patent is method and apparatus of transforming speech feature vectors using an auto-associative neural network and its abstract is provided is a method and apparatus for transforming a speech feature vector. the method includes extracting a feature vector required for speech recognition from a speech signal and transforming the extracted feature vector using an auto-associative neural network (aann). dated 2014-09-16
8838514,optimal technique search method and system that creates a virtual cell division space to create/form a neural network,"disclosed are an optimal technique search method and system that can enable a more effective search for optimal techniques for problem solutions than in the past through the use of a neural network employing genetic algorithm. provided therein are an execution unit (1) that uses a neural network employing a genetic algorithm to search for an optimal technique and which executes operations using said technique, and an evaluation unit (2) that, along with creating initial setting to transmit to said execution unit, evaluates the content of the operations of the execution unit after the operations have been executed and has the execution unit (1) execute operations a plurality of times, and thereby derives as the optimal technique the initial settings that executed the most effective operation when transmitted to the execution unit (1) out of the results derived from said plurality of operation executions. as a result, a small scale and effective optimal technique search becomes possible when using a neural network, as described in [0024] and [0025].",2014-09-16,"The title of the patent is optimal technique search method and system that creates a virtual cell division space to create/form a neural network and its abstract is disclosed are an optimal technique search method and system that can enable a more effective search for optimal techniques for problem solutions than in the past through the use of a neural network employing genetic algorithm. provided therein are an execution unit (1) that uses a neural network employing a genetic algorithm to search for an optimal technique and which executes operations using said technique, and an evaluation unit (2) that, along with creating initial setting to transmit to said execution unit, evaluates the content of the operations of the execution unit after the operations have been executed and has the execution unit (1) execute operations a plurality of times, and thereby derives as the optimal technique the initial settings that executed the most effective operation when transmitted to the execution unit (1) out of the results derived from said plurality of operation executions. as a result, a small scale and effective optimal technique search becomes possible when using a neural network, as described in [0024] and [0025]. dated 2014-09-16"
8839441,method and system for adaptive vulnerability scanning of an application,"a method and system for adaptive vulnerability scanning (avs) of an application is provided. the adaptive vulnerability scanning of an application assists in identifying new vulnerabilities dynamically. the endpoints of an application are scanned using a predefined set of rules. subsequently, one or more possible vulnerabilities are presented. the vulnerabilities are analyzed and predefined rules are modified. the steps of scanning the application and modification of rules are iteratively repeated till the adaptive vulnerability scanning capability is achieved. a neural network is used for training the adaptive vulnerability scanner. this neural network is made to learn some rules based on predefined set of rules while undergoing the training phase. at least one weight in neural networks is altered while imparting the self learning capability.",2014-09-16,"The title of the patent is method and system for adaptive vulnerability scanning of an application and its abstract is a method and system for adaptive vulnerability scanning (avs) of an application is provided. the adaptive vulnerability scanning of an application assists in identifying new vulnerabilities dynamically. the endpoints of an application are scanned using a predefined set of rules. subsequently, one or more possible vulnerabilities are presented. the vulnerabilities are analyzed and predefined rules are modified. the steps of scanning the application and modification of rules are iteratively repeated till the adaptive vulnerability scanning capability is achieved. a neural network is used for training the adaptive vulnerability scanner. this neural network is made to learn some rules based on predefined set of rules while undergoing the training phase. at least one weight in neural networks is altered while imparting the self learning capability. dated 2014-09-16"
8840562,signal processing warping technique,"methods and systems are provided for using time-frequency warping to analyze a physiological signal. one embodiment includes applying a warping operator to the physiological signal based on the energy density of the signal. the warped physiological signal may be analyzed to determine whether non-physiological signal components are present. further, the same warping operator may be applied to signal quality indicators, and the warped physiological signal may be analyzed based on the warped signal quality indicators. non-physiological signal components, or types of non-physiological noise sources, may be identified based on a comparison of the physiological signal with the signal quality indicators. non-physiological signal components may also be identified based on a neural network of known noise functions. in some embodiments, the non-physiological signal components may be removed to increase accuracy in estimating physiological parameters.",2014-09-23,"The title of the patent is signal processing warping technique and its abstract is methods and systems are provided for using time-frequency warping to analyze a physiological signal. one embodiment includes applying a warping operator to the physiological signal based on the energy density of the signal. the warped physiological signal may be analyzed to determine whether non-physiological signal components are present. further, the same warping operator may be applied to signal quality indicators, and the warped physiological signal may be analyzed based on the warped signal quality indicators. non-physiological signal components, or types of non-physiological noise sources, may be identified based on a comparison of the physiological signal with the signal quality indicators. non-physiological signal components may also be identified based on a neural network of known noise functions. in some embodiments, the non-physiological signal components may be removed to increase accuracy in estimating physiological parameters. dated 2014-09-23"
8843425,hierarchical routing for two-way information flow and structural plasticity in neural networks,"hierarchical routing for two-way information flow and structural plasticity in a neural network is provided. in one embodiment the network includes multiple core modules, wherein each core module has a plurality of incoming connections with predetermined addresses. each core module also has a plurality of outgoing connections such that each outgoing connection targets an incoming connection in a core module among the multiple core modules. the network also has a routing system that selectively routes signals among the core modules based on a reconfigurable hierarchical organization of the core modules. the network approximates a fully connected network such that each outgoing connection on any core module can target and reach any incoming connection on any core module without requiring a fully connected network. the routing system provides two-way information flow between neurons utilizing hierarchical routing.",2014-09-23,"The title of the patent is hierarchical routing for two-way information flow and structural plasticity in neural networks and its abstract is hierarchical routing for two-way information flow and structural plasticity in a neural network is provided. in one embodiment the network includes multiple core modules, wherein each core module has a plurality of incoming connections with predetermined addresses. each core module also has a plurality of outgoing connections such that each outgoing connection targets an incoming connection in a core module among the multiple core modules. the network also has a routing system that selectively routes signals among the core modules based on a reconfigurable hierarchical organization of the core modules. the network approximates a fully connected network such that each outgoing connection on any core module can target and reach any incoming connection on any core module without requiring a fully connected network. the routing system provides two-way information flow between neurons utilizing hierarchical routing. dated 2014-09-23"
8849573,method and apparatus for neutron porosity measurement using a neural network,"a method of estimating formation porosity using a neural network for neutron porosity tools. in the training stage, the near-to-far ratio, environmental variables, such as mineralogy, borehole size, standoff etc., are fed to the inputs and the neural network is trained for obtaining the related true porosity (the output). the trained neural network is implanted into tool's firmware for the real time porosity measurement, accounting for the environmental effects considered during training.",2014-09-30,"The title of the patent is method and apparatus for neutron porosity measurement using a neural network and its abstract is a method of estimating formation porosity using a neural network for neutron porosity tools. in the training stage, the near-to-far ratio, environmental variables, such as mineralogy, borehole size, standoff etc., are fed to the inputs and the neural network is trained for obtaining the related true porosity (the output). the trained neural network is implanted into tool's firmware for the real time porosity measurement, accounting for the environmental effects considered during training. dated 2014-09-30"
8855387,system for detecting bone cancer metastases,"the invention relates to a detection system for automatic detection of bone cancer metastases from a set of isotope bone scan images of a patients skeleton, the system comprising a shape identifier unit, a hotspot detection unit, a hotspot feature extraction unit, a first artificial neural network unit, a patient feature extraction unit, and a second artificial neural network unit.",2014-10-07,"The title of the patent is system for detecting bone cancer metastases and its abstract is the invention relates to a detection system for automatic detection of bone cancer metastases from a set of isotope bone scan images of a patients skeleton, the system comprising a shape identifier unit, a hotspot detection unit, a hotspot feature extraction unit, a first artificial neural network unit, a patient feature extraction unit, and a second artificial neural network unit. dated 2014-10-07"
8855749,determination of a physiological parameter,"methods and systems are provided for analyzing a physiological signal by applying a continuous wavelet transform on the signal and comparing the wavelet transformation to a library of wavelet signatures corresponding to one or more physiological conditions and/or patient conditions. a pulse oximeter system may relate the wavelet transformation with one or more of the wavelet signatures based on filters and/or thresholds, and may determine that the wavelet transformation indicates that the patient of the physiological signal has a physiological condition indicated by the related wavelet signature. in some embodiments, the pulse oximeter system may use previous analyses in a neural network to update the library. further, non-physiological components of the wavelet transformation may also be identified and removed.",2014-10-07,"The title of the patent is determination of a physiological parameter and its abstract is methods and systems are provided for analyzing a physiological signal by applying a continuous wavelet transform on the signal and comparing the wavelet transformation to a library of wavelet signatures corresponding to one or more physiological conditions and/or patient conditions. a pulse oximeter system may relate the wavelet transformation with one or more of the wavelet signatures based on filters and/or thresholds, and may determine that the wavelet transformation indicates that the patient of the physiological signal has a physiological condition indicated by the related wavelet signature. in some embodiments, the pulse oximeter system may use previous analyses in a neural network to update the library. further, non-physiological components of the wavelet transformation may also be identified and removed. dated 2014-10-07"
8856055,reconfigurable and customizable general-purpose circuits for neural networks,"a reconfigurable neural network circuit is provided. the reconfigurable neural network circuit comprises an electronic synapse array including multiple synapses interconnecting a plurality of digital electronic neurons. each neuron comprises an integrator that integrates input spikes and generates a signal when the integrated inputs exceed a threshold. the circuit further comprises a control module for reconfiguring the synapse array. the control module comprises a global final state machine that controls timing for operation of the circuit, and a priority encoder that allows spiking neurons to sequentially access the synapse array.",2014-10-07,"The title of the patent is reconfigurable and customizable general-purpose circuits for neural networks and its abstract is a reconfigurable neural network circuit is provided. the reconfigurable neural network circuit comprises an electronic synapse array including multiple synapses interconnecting a plurality of digital electronic neurons. each neuron comprises an integrator that integrates input spikes and generates a signal when the integrated inputs exceed a threshold. the circuit further comprises a control module for reconfiguring the synapse array. the control module comprises a global final state machine that controls timing for operation of the circuit, and a priority encoder that allows spiking neurons to sequentially access the synapse array. dated 2014-10-07"
8858929,optically sensitive cell network,a neural network is disclosed. the neural network comprises a plurality of optogenetically modified neural cells being three-dimensionally distributed in a hydrogel medium and being disconnected from any solid support having a shear modulus above 1 gpa.,2014-10-14,The title of the patent is optically sensitive cell network and its abstract is a neural network is disclosed. the neural network comprises a plurality of optogenetically modified neural cells being three-dimensionally distributed in a hydrogel medium and being disconnected from any solid support having a shear modulus above 1 gpa. dated 2014-10-14
8862194,method for improved oxygen saturation estimation in the presence of noise,"the present disclosure relates, according to some embodiments, to devices, systems, and methods for estimating a physiological parameter in the presence of noise. for example, the disclosure relates, in some embodiments, to devices, systems, and methods for assessing (e.g., estimating, measuring, calculating) oxygen saturation (spo2). methods of assessing spo2 may include assessing a noise metric associated with motion artifact. in some embodiments, a percentage (e.g., an empirically determined percentage) of a noise metric may be simply added to the spo2 estimate to produce a corrected spo2 estimate. an oximetry algorithm may include, according to some embodiments, combining multiple internal spo2 estimates and associated noise and/or signal quality metrics (e.g., using a radial basis neural network) to produce a modified (e.g., corrected) spo2 estimate (e.g., rather than merely selecting the estimate from a finite number of candidates). a modified spo2 estimate may include little or no movement-based error.",2014-10-14,"The title of the patent is method for improved oxygen saturation estimation in the presence of noise and its abstract is the present disclosure relates, according to some embodiments, to devices, systems, and methods for estimating a physiological parameter in the presence of noise. for example, the disclosure relates, in some embodiments, to devices, systems, and methods for assessing (e.g., estimating, measuring, calculating) oxygen saturation (spo2). methods of assessing spo2 may include assessing a noise metric associated with motion artifact. in some embodiments, a percentage (e.g., an empirically determined percentage) of a noise metric may be simply added to the spo2 estimate to produce a corrected spo2 estimate. an oximetry algorithm may include, according to some embodiments, combining multiple internal spo2 estimates and associated noise and/or signal quality metrics (e.g., using a radial basis neural network) to produce a modified (e.g., corrected) spo2 estimate (e.g., rather than merely selecting the estimate from a finite number of candidates). a modified spo2 estimate may include little or no movement-based error. dated 2014-10-14"
8862527,neural networks and method for training neural networks,"methods (30) for training an artificial neural network (nn) are disclosed. an example method (30) includes: initializing the nn by selecting an output of the nn to be trained and connecting an output neuron of the nn to input neuron(s) in an input layer of the nn for the selected output; preparing a data set to be learnt by the nn; and, applying the prepared data set to the nn to be learnt by applying an input vector of the prepared data set to the first hidden layer of the nn, or the output layer of the nn if the nn has no hidden layer(s), and determining whether at least one neuron for the selected output in each layer of the nn can learn to produce the associated output for the input vector.",2014-10-14,"The title of the patent is neural networks and method for training neural networks and its abstract is methods (30) for training an artificial neural network (nn) are disclosed. an example method (30) includes: initializing the nn by selecting an output of the nn to be trained and connecting an output neuron of the nn to input neuron(s) in an input layer of the nn for the selected output; preparing a data set to be learnt by the nn; and, applying the prepared data set to the nn to be learnt by applying an input vector of the prepared data set to the first hidden layer of the nn, or the output layer of the nn if the nn has no hidden layer(s), and determining whether at least one neuron for the selected output in each layer of the nn can learn to produce the associated output for the input vector. dated 2014-10-14"
8868477,multi-compartment neurons with neural cores,embodiments of the invention provide a neural core circuit comprising a synaptic interconnect network including plural electronic synapses for interconnecting one or more source electronic neurons with one or more target electronic neurons. the interconnect network further includes multiple axon paths and multiple dendrite paths. each synapse is at a cross-point junction of the interconnect network between a dendrite path and an axon path. the core circuit further comprises a routing module maintaining routing information. the routing module routes output from a source electronic neuron to one or more selected axon paths. each synapse provides a configurable level of signal conduction from an axon path of a source electronic neuron to a dendrite path of a target electronic neuron.,2014-10-21,The title of the patent is multi-compartment neurons with neural cores and its abstract is embodiments of the invention provide a neural core circuit comprising a synaptic interconnect network including plural electronic synapses for interconnecting one or more source electronic neurons with one or more target electronic neurons. the interconnect network further includes multiple axon paths and multiple dendrite paths. each synapse is at a cross-point junction of the interconnect network between a dendrite path and an axon path. the core circuit further comprises a routing module maintaining routing information. the routing module routes output from a source electronic neuron to one or more selected axon paths. each synapse provides a configurable level of signal conduction from an axon path of a source electronic neuron to a dendrite path of a target electronic neuron. dated 2014-10-21
8873826,method for brightness level calculation of the digital x-ray image for medical applications,"the invention relates to methods for evaluation a brightness level of the digital x-ray image for medical applications by means of the image histogram using a neural network. the calculations comprise of: image acquisition, image histogram calculation, transformation the frequencies of the histogram into input arguments of the neural network and calculation the brightness level as linear transform of the output value of the neural network. training of the neural network is performed over a learning set calculated over the given image database. the transformed frequencies of histograms of these images are used as a set of input arguments of the neural network. the brightness levels calculated for each image over the region of interest and scaled to the range of output values of the neuron network are used as a set of target values.",2014-10-28,"The title of the patent is method for brightness level calculation of the digital x-ray image for medical applications and its abstract is the invention relates to methods for evaluation a brightness level of the digital x-ray image for medical applications by means of the image histogram using a neural network. the calculations comprise of: image acquisition, image histogram calculation, transformation the frequencies of the histogram into input arguments of the neural network and calculation the brightness level as linear transform of the output value of the neural network. training of the neural network is performed over a learning set calculated over the given image database. the transformed frequencies of histograms of these images are used as a set of input arguments of the neural network. the brightness levels calculated for each image over the region of interest and scaled to the range of output values of the neuron network are used as a set of target values. dated 2014-10-28"
8874276,event-based control system for wind turbine generators,"the present invention relates to a control system comprising a control interface between one or more wind turbine generators and a power grid, where the wind turbine generators are coupled to the power grid and contribute to the power production of the grid. the control interface is arranged to receive a set of event data. in embodiments, the set of event data may be any data available to a scada system. the set of event data is analyzed in terms of predetermined event rules comprising at least one predefined event condition and a set of adaptive event conditions. based on the analysis an event output is provided in order to control a parameter of the one or more wind turbine generators. in embodiments, the control system may be implemented in, or in connection with a scada system, moreover, the event output may be based on fuzzy logic, a neural network or statistical analysis.",2014-10-28,"The title of the patent is event-based control system for wind turbine generators and its abstract is the present invention relates to a control system comprising a control interface between one or more wind turbine generators and a power grid, where the wind turbine generators are coupled to the power grid and contribute to the power production of the grid. the control interface is arranged to receive a set of event data. in embodiments, the set of event data may be any data available to a scada system. the set of event data is analyzed in terms of predetermined event rules comprising at least one predefined event condition and a set of adaptive event conditions. based on the analysis an event output is provided in order to control a parameter of the one or more wind turbine generators. in embodiments, the control system may be implemented in, or in connection with a scada system, moreover, the event output may be based on fuzzy logic, a neural network or statistical analysis. dated 2014-10-28"
8874434,method and apparatus for full natural language parsing,"the method and apparatus for discriminative natural language parsing, uses a deep convolutional neural network adapted for text and a structured tag inference in a graph. in the method and apparatus, a trained recursive convolutional graph transformer network, formed by the deep convolutional neural network and the graph, predicts “levels” of a parse tree based on predictions of previous levels.",2014-10-28,"The title of the patent is method and apparatus for full natural language parsing and its abstract is the method and apparatus for discriminative natural language parsing, uses a deep convolutional neural network adapted for text and a structured tag inference in a graph. in the method and apparatus, a trained recursive convolutional graph transformer network, formed by the deep convolutional neural network and the graph, predicts “levels” of a parse tree based on predictions of previous levels. dated 2014-10-28"
8874496,encoding and decoding machine with recurrent neural networks,"techniques for reconstructing a signal encoded with a time encoding machine (tem) using a recurrent neural network including receiving a tem-encoded signal, processing the tem-encoded signal, and reconstructing the tem-encoded signal with a recurrent neural network.",2014-10-28,"The title of the patent is encoding and decoding machine with recurrent neural networks and its abstract is techniques for reconstructing a signal encoded with a time encoding machine (tem) using a recurrent neural network including receiving a tem-encoded signal, processing the tem-encoded signal, and reconstructing the tem-encoded signal with a recurrent neural network. dated 2014-10-28"
8874498,"unsupervised, supervised, and reinforced learning via spiking computation","the present invention relates to unsupervised, supervised and reinforced learning via spiking computation. the neural network comprises a plurality of neural modules. each neural module comprises multiple digital neurons such that each neuron in a neural module has a corresponding neuron in another neural module. an interconnection network comprising a plurality of edges interconnects the plurality of neural modules. each edge interconnects a first neural module to a second neural module, and each edge comprises a weighted synaptic connection between every neuron in the first neural module and a corresponding neuron in the second neural module.",2014-10-28,"The title of the patent is unsupervised, supervised, and reinforced learning via spiking computation and its abstract is the present invention relates to unsupervised, supervised and reinforced learning via spiking computation. the neural network comprises a plurality of neural modules. each neural module comprises multiple digital neurons such that each neuron in a neural module has a corresponding neuron in another neural module. an interconnection network comprising a plurality of edges interconnects the plurality of neural modules. each edge interconnects a first neural module to a second neural module, and each edge comprises a weighted synaptic connection between every neuron in the first neural module and a corresponding neuron in the second neural module. dated 2014-10-28"
8880448,predicting odor pleasantness with an electronic nose,"apparatus and method for assessing odors, comprises an electronic nose, to be applied to an odor and to output a structure identifying the odor; a neural network which maps an extracted structure to a first location on a pre-learned axis of odor pleasantness; and an output for outputting an assessment of an applied odor based on said first location. the assessment may be a prediction of how pleasant a user will consider the odor.",2014-11-04,"The title of the patent is predicting odor pleasantness with an electronic nose and its abstract is apparatus and method for assessing odors, comprises an electronic nose, to be applied to an odor and to output a structure identifying the odor; a neural network which maps an extracted structure to a first location on a pre-learned axis of odor pleasantness; and an output for outputting an assessment of an applied odor based on said first location. the assessment may be a prediction of how pleasant a user will consider the odor. dated 2014-11-04"
8880450,systems and methods for predicting characteristics of an artificial heart using an artificial neural network,"a system configured to predict characteristics of an artificial heart is described. the system includes a processor and memory in electronic communication with the processor, and an artificial neural network configured to receive an input vector of a predetermined length to train the artificial neural network, produce an output vector based on the input vector, and compare the output vector with a target vector of the predetermined length. when the output vector does not match the target vector within a predetermined error rate, the network is configured to adjust at least one weight, and when the output vector matches the target vector within the predetermined error rate, the network is configured to execute the input vector to produce an estimate at least one characteristic of the artificial heart.",2014-11-04,"The title of the patent is systems and methods for predicting characteristics of an artificial heart using an artificial neural network and its abstract is a system configured to predict characteristics of an artificial heart is described. the system includes a processor and memory in electronic communication with the processor, and an artificial neural network configured to receive an input vector of a predetermined length to train the artificial neural network, produce an output vector based on the input vector, and compare the output vector with a target vector of the predetermined length. when the output vector does not match the target vector within a predetermined error rate, the network is configured to adjust at least one weight, and when the output vector matches the target vector within the predetermined error rate, the network is configured to execute the input vector to produce an estimate at least one characteristic of the artificial heart. dated 2014-11-04"
8885842,methods and apparatus to determine locations of audience members,a disclosed example method to determine a location of an audience member involves generating a correlation analysis result based on correlating first audio samples from a stationary audio detector with second audio samples from a portable audio detector carried by the audience member and determining via a neural network the location of the audience member based on the correlation analysis result.,2014-11-11,The title of the patent is methods and apparatus to determine locations of audience members and its abstract is a disclosed example method to determine a location of an audience member involves generating a correlation analysis result based on correlating first audio samples from a stationary audio detector with second audio samples from a portable audio detector carried by the audience member and determining via a neural network the location of the audience member based on the correlation analysis result. dated 2014-11-11
8885927,user emotion detection method and associated handwriting input electronic device,"a user emotion detection method for a handwriting input electronic device is provided. the method includes steps of: obtaining at least one handwriting input characteristic parameter; determining a user emotion parameter by an artificial neural network of the handwriting input electronic device according to the handwriting input characteristic value and at least one associated linkage value; displaying the user emotion parameter on a touch display panel of the handwriting input electronic device; receiving a user feedback parameter; determining whether to adjust the at least one associated linkage value and if yes, adjusting the at least one associated linkage value according to the user feedback parameter to construct and adjust the artificial neural network.",2014-11-11,"The title of the patent is user emotion detection method and associated handwriting input electronic device and its abstract is a user emotion detection method for a handwriting input electronic device is provided. the method includes steps of: obtaining at least one handwriting input characteristic parameter; determining a user emotion parameter by an artificial neural network of the handwriting input electronic device according to the handwriting input characteristic value and at least one associated linkage value; displaying the user emotion parameter on a touch display panel of the handwriting input electronic device; receiving a user feedback parameter; determining whether to adjust the at least one associated linkage value and if yes, adjusting the at least one associated linkage value according to the user feedback parameter to construct and adjust the artificial neural network. dated 2014-11-11"
8886579,"methods, apparatus and products for semantic processing of text","a computer-implemented method of training a neural network includes training a first neural network of a self organizing map type with a first set of first text documents each containing one or more keywords in a semantic context to map each document to a point in the self organizing map y semantic clustering; determining, for each keyword in the first set, all points in the self organizing map to which first documents containing said keyword are mapped, as a pattern and storing said pattern for said keyword in a pattern dictionary; forming at least one sequence of keywords from a second set of second text documents each containing one or more keywords in a semantic context; translating said at least one sequence of keywords into at least one sequence of patterns using the pattern dictionary; and training a second neural network with the at least one sequence of patterns.",2014-11-11,"The title of the patent is methods, apparatus and products for semantic processing of text and its abstract is a computer-implemented method of training a neural network includes training a first neural network of a self organizing map type with a first set of first text documents each containing one or more keywords in a semantic context to map each document to a point in the self organizing map y semantic clustering; determining, for each keyword in the first set, all points in the self organizing map to which first documents containing said keyword are mapped, as a pattern and storing said pattern for said keyword in a pattern dictionary; forming at least one sequence of keywords from a second set of second text documents each containing one or more keywords in a semantic context; translating said at least one sequence of keywords into at least one sequence of patterns using the pattern dictionary; and training a second neural network with the at least one sequence of patterns. dated 2014-11-11"
8892486,"processor node, artificial neural network and method operation of an artificial neural network","there is provided a temporal processor node for use as an input node in the input layer of a class network in an artificial neural network, the class network being operable to generate an output signal based on a network input vector component received by the input layer, the temporal processor node being operable to receive observation data representing the observed state of a monitored entity as a component of the network input vector. the temporal processor node comprises a memory module operable to store a most recently observed state of the monitored entity in the memory module as a current state, a modification module having a timer, the timer being operable to output a value representing time elapsed since observation of the current state, the modification module being operable to modify the current state with a modification factor dependent on the value output by the timer, wherein when triggered, the temporal processor node is operable to output the modified current state as a representation of the current state.",2014-11-18,"The title of the patent is processor node, artificial neural network and method operation of an artificial neural network and its abstract is there is provided a temporal processor node for use as an input node in the input layer of a class network in an artificial neural network, the class network being operable to generate an output signal based on a network input vector component received by the input layer, the temporal processor node being operable to receive observation data representing the observed state of a monitored entity as a component of the network input vector. the temporal processor node comprises a memory module operable to store a most recently observed state of the monitored entity in the memory module as a current state, a modification module having a timer, the timer being operable to output a value representing time elapsed since observation of the current state, the modification module being operable to modify the current state with a modification factor dependent on the value output by the timer, wherein when triggered, the temporal processor node is operable to output the modified current state as a representation of the current state. dated 2014-11-18"
8897925,heat dissipation control system and control method thereof,"a heat dissipation control system comprises a sensing unit, an artificial neural network computing unit, and two heat dissipation units. the artificial neural network computing unit performs computation for controlling based on a plurality of electronic-device temperatures sent out by the sensing unit. the computation for controlling performs a back propagation algorithm on an objective function which is defined a as the square of an error function. accordingly, cooling effects suitable for the heat dissipation units are generated in order to achieve an optimum heat dissipation effect.",2014-11-25,"The title of the patent is heat dissipation control system and control method thereof and its abstract is a heat dissipation control system comprises a sensing unit, an artificial neural network computing unit, and two heat dissipation units. the artificial neural network computing unit performs computation for controlling based on a plurality of electronic-device temperatures sent out by the sensing unit. the computation for controlling performs a back propagation algorithm on an objective function which is defined a as the square of an error function. accordingly, cooling effects suitable for the heat dissipation units are generated in order to achieve an optimum heat dissipation effect. dated 2014-11-25"
8898045,system and method of predicting gas saturation of a formation using neural networks,"predicting gas saturation of a formation using neural networks. at least some of the illustrative embodiments include obtaining a gamma count rate decay curve one each for a plurality of gamma detectors of a nuclear logging tool (the gamma count rate decay curves recorded at a particular borehole depth), applying at least a portion of each gamma count rate decay curve to input nodes of a neural network, predicting a value indicative of gas saturation of a formation (the predicting by the neural network in the absence of a formation porosity value supplied to the neural network), and producing a plot of the value indicative of gas saturation of the formation as a function of borehole depth.",2014-11-25,"The title of the patent is system and method of predicting gas saturation of a formation using neural networks and its abstract is predicting gas saturation of a formation using neural networks. at least some of the illustrative embodiments include obtaining a gamma count rate decay curve one each for a plurality of gamma detectors of a nuclear logging tool (the gamma count rate decay curves recorded at a particular borehole depth), applying at least a portion of each gamma count rate decay curve to input nodes of a neural network, predicting a value indicative of gas saturation of a formation (the predicting by the neural network in the absence of a formation porosity value supplied to the neural network), and producing a plot of the value indicative of gas saturation of the formation as a function of borehole depth. dated 2014-11-25"
8898097,reconfigurable and customizable general-purpose circuits for neural networks,"a reconfigurable neural network circuit is provided. the reconfigurable neural network circuit comprises an electronic synapse array including multiple synapses interconnecting a plurality of digital electronic neurons. each neuron comprises an integrator that integrates input spikes and generates a signal when the integrated inputs exceed a threshold. the circuit further comprises a control module for reconfiguring the synapse array. the control module comprises a global final state machine that controls timing for operation of the circuit, and a priority encoder that allows spiking neurons to sequentially access the synapse array.",2014-11-25,"The title of the patent is reconfigurable and customizable general-purpose circuits for neural networks and its abstract is a reconfigurable neural network circuit is provided. the reconfigurable neural network circuit comprises an electronic synapse array including multiple synapses interconnecting a plurality of digital electronic neurons. each neuron comprises an integrator that integrates input spikes and generates a signal when the integrated inputs exceed a threshold. the circuit further comprises a control module for reconfiguring the synapse array. the control module comprises a global final state machine that controls timing for operation of the circuit, and a priority encoder that allows spiking neurons to sequentially access the synapse array. dated 2014-11-25"
8903746,"system and method for viewing, modifying, storing, and running artificial neural network components","a system and method for artificial neural network processing includes, for example, modifying, by a computer processor, a value of a charge of a node of an artificial neural depending on a number of elapsed steps since a prior predefined significant event. a system and method includes, for example, providing by a processor a real-time representation of an artificial neural network and of graphical effects of a running of the neural network. a system and method includes, for example, automatically modifying the behavior of network nodes based on simultaneous occurrences of events.",2014-12-02,"The title of the patent is system and method for viewing, modifying, storing, and running artificial neural network components and its abstract is a system and method for artificial neural network processing includes, for example, modifying, by a computer processor, a value of a charge of a node of an artificial neural depending on a number of elapsed steps since a prior predefined significant event. a system and method includes, for example, providing by a processor a real-time representation of an artificial neural network and of graphical effects of a running of the neural network. a system and method includes, for example, automatically modifying the behavior of network nodes based on simultaneous occurrences of events. dated 2014-12-02"
8903753,steam turbine performance testing,"a steam turbine performance testing system, including at least one computer hardware device, including a neural network created using a dynamic steam turbine thermodynamic model and preliminary data collected from a steam turbine; a network tester for testing the neural network with testing data; a current performance calculator for calculating a current performance of the steam turbine from operation data of the steam turbine; and a projected performance calculator for calculating a projected performance of the steam turbine from the current performance.",2014-12-02,"The title of the patent is steam turbine performance testing and its abstract is a steam turbine performance testing system, including at least one computer hardware device, including a neural network created using a dynamic steam turbine thermodynamic model and preliminary data collected from a steam turbine; a network tester for testing the neural network with testing data; a current performance calculator for calculating a current performance of the steam turbine from operation data of the steam turbine; and a projected performance calculator for calculating a projected performance of the steam turbine from the current performance. dated 2014-12-02"
8909574,systems for matching sparkle appearance of coatings,this disclosure is directed to a process for producing one or more predicted target sparkle values of a target coating composition. an artificial neural network can be used in the process. the process disclosed herein can be used for color and appearance matching in the coating industry including vehicle original equipment manufacturing (oem) coatings and refinish coatings. a system for producing one or more predicted target sparkle values of a target coating composition is also disclosed.,2014-12-09,The title of the patent is systems for matching sparkle appearance of coatings and its abstract is this disclosure is directed to a process for producing one or more predicted target sparkle values of a target coating composition. an artificial neural network can be used in the process. the process disclosed herein can be used for color and appearance matching in the coating industry including vehicle original equipment manufacturing (oem) coatings and refinish coatings. a system for producing one or more predicted target sparkle values of a target coating composition is also disclosed. dated 2014-12-09
8909576,neuromorphic event-driven neural computing architecture in a scalable neural network,an event-driven neural network includes a plurality of interconnected core circuits is provided. each core circuit includes an electronic synapse array has multiple digital synapses interconnecting a plurality of digital electronic neurons. a synapse interconnects an axon of a pre-synaptic neuron with a dendrite of a post-synaptic neuron. a neuron integrates input spikes and generates a spike event in response to the integrated input spikes exceeding a threshold. each core circuit also has a scheduler that receives a spike event and delivers the spike event to a selected axon in the synapse array based on a schedule for deterministic event delivery.,2014-12-09,The title of the patent is neuromorphic event-driven neural computing architecture in a scalable neural network and its abstract is an event-driven neural network includes a plurality of interconnected core circuits is provided. each core circuit includes an electronic synapse array has multiple digital synapses interconnecting a plurality of digital electronic neurons. a synapse interconnects an axon of a pre-synaptic neuron with a dendrite of a post-synaptic neuron. a neuron integrates input spikes and generates a spike event in response to the integrated input spikes exceeding a threshold. each core circuit also has a scheduler that receives a spike event and delivers the spike event to a selected axon in the synapse array based on a schedule for deterministic event delivery. dated 2014-12-09
8910508,early detection of overheating devices,"a sensor module is provided that monitors the odor within the physical enclosure of a computing device that includes one or more components. a recognition module determines whether the odor within the physical enclosure is indicative of an overheating component that is overheating within the physical enclosure of the computing device. the recognition module may use an artificial neural network (ann) to determine whether the odor is indicative of an overheating component. an alert module initiates an overheating protocol in response to determining that the odor within the physical enclosure is indicative of an overheating component. the alert module may, for example, alert the user and/or applications that a component is overheating.",2014-12-16,"The title of the patent is early detection of overheating devices and its abstract is a sensor module is provided that monitors the odor within the physical enclosure of a computing device that includes one or more components. a recognition module determines whether the odor within the physical enclosure is indicative of an overheating component that is overheating within the physical enclosure of the computing device. the recognition module may use an artificial neural network (ann) to determine whether the odor is indicative of an overheating component. an alert module initiates an overheating protocol in response to determining that the odor within the physical enclosure is indicative of an overheating component. the alert module may, for example, alert the user and/or applications that a component is overheating. dated 2014-12-16"
8918351,providing transposable access to a synapse array using column aggregation,"embodiments of the invention relate to providing transposable access to a synapse array using column aggregation. one embodiment comprises a neural network including a plurality of electronic axons, a plurality of electronic neurons, and a crossbar for interconnecting the axons with the neurons. the crossbar comprises a plurality of electronic synapses. each synapse interconnects an axon with a neuron. the neural network further comprises a column aggregation module for transposable access to one or more synapses of the crossbar using column aggregation.",2014-12-23,"The title of the patent is providing transposable access to a synapse array using column aggregation and its abstract is embodiments of the invention relate to providing transposable access to a synapse array using column aggregation. one embodiment comprises a neural network including a plurality of electronic axons, a plurality of electronic neurons, and a crossbar for interconnecting the axons with the neurons. the crossbar comprises a plurality of electronic synapses. each synapse interconnects an axon with a neuron. the neural network further comprises a column aggregation module for transposable access to one or more synapses of the crossbar using column aggregation. dated 2014-12-23"
8918352,learning processes for single hidden layer neural networks with linear output units,"learning processes for a single hidden layer neural network, including linear input units, nonlinear hidden units, and linear output units, calculate the lower-layer network parameter gradients by taking into consideration a solution for the upper-layer network parameters. the upper-layer network parameters are calculated by a closed form formula given the lower-layer network parameters. an accelerated gradient algorithm can be used to update the lower-layer network parameters. a weighted gradient also can be used. with the combination of these techniques, accelerated training with faster convergence, to a point with a lower error rate, can be obtained.",2014-12-23,"The title of the patent is learning processes for single hidden layer neural networks with linear output units and its abstract is learning processes for a single hidden layer neural network, including linear input units, nonlinear hidden units, and linear output units, calculate the lower-layer network parameter gradients by taking into consideration a solution for the upper-layer network parameters. the upper-layer network parameters are calculated by a closed form formula given the lower-layer network parameters. an accelerated gradient algorithm can be used to update the lower-layer network parameters. a weighted gradient also can be used. with the combination of these techniques, accelerated training with faster convergence, to a point with a lower error rate, can be obtained. dated 2014-12-23"
8920327,method for determining cardiac output,"in a method for determining cardiac output from an arterial blood pressure curve measured at the periphery, in which the blood pressure curve measured at the periphery is arithmetically transformed into the corresponding central blood pressure curve and the cardiac output is calculated from the central blood pressure curve, the transformation of the blood pressure curve measured at the periphery into the corresponding central blood pressure curve is performed by the aid of an artificial neural network whose weighting values are determined by learning.",2014-12-30,"The title of the patent is method for determining cardiac output and its abstract is in a method for determining cardiac output from an arterial blood pressure curve measured at the periphery, in which the blood pressure curve measured at the periphery is arithmetically transformed into the corresponding central blood pressure curve and the cardiac output is calculated from the central blood pressure curve, the transformation of the blood pressure curve measured at the periphery into the corresponding central blood pressure curve is performed by the aid of an artificial neural network whose weighting values are determined by learning. dated 2014-12-30"
8924024,method for sootblowing optimization,"a controller determines and adjusts system parameters, including cleanliness levels or sootblower operating settings, that are useful for maintaining the cleanliness of a fossil fuel boiler at an efficient level. some embodiments use a direct controller to determine cleanliness levels and/or sootblower operating settings. some embodiments use an indirect controller, with a system model, to determine cleanliness levels and/or sootblower settings. the controller may use a model that is, for example, a neural network, or a mass energy balance, or a genetically programmed model. the controller uses input about the actual performance or state of the boiler for adaptation. the controller may operate in conjunction with a sootblower optimization system that controls the actual settings of the sootblowers. the controller may coordinate cleanliness settings for multiple sootblowers and/or across a plurality of heat zones in the boiler.",2014-12-30,"The title of the patent is method for sootblowing optimization and its abstract is a controller determines and adjusts system parameters, including cleanliness levels or sootblower operating settings, that are useful for maintaining the cleanliness of a fossil fuel boiler at an efficient level. some embodiments use a direct controller to determine cleanliness levels and/or sootblower operating settings. some embodiments use an indirect controller, with a system model, to determine cleanliness levels and/or sootblower settings. the controller may use a model that is, for example, a neural network, or a mass energy balance, or a genetically programmed model. the controller uses input about the actual performance or state of the boiler for adaptation. the controller may operate in conjunction with a sootblower optimization system that controls the actual settings of the sootblowers. the controller may coordinate cleanliness settings for multiple sootblowers and/or across a plurality of heat zones in the boiler. dated 2014-12-30"
8924321,three-layered neuron devices for neural network with reset voltage pulse,"a neuron device includes a bottom electrode, a top electrode, and a layer of metal oxide variable resistance material sandwiched between the bottom electrode and the top electrode, in which the neuron device is switched to a normal state upon application of reset pulse, and is switched to an excitation state upon application of stimulus pulses. the neuron device has a comprehensive response to different amplitude, different width of a stimulus voltage pulse and different number of a sequence of stimulus pulses, and provides functionalities of a weighting section and a computing section. the neuron device has a simple structure, excellent scalability, quick speed, low operation voltage, and is compatible with the conventional silicon-based cmos fabrication process, and thus suitable for mass production. the neuron device is capable of performing many biological functions and complex logic operations.",2014-12-30,"The title of the patent is three-layered neuron devices for neural network with reset voltage pulse and its abstract is a neuron device includes a bottom electrode, a top electrode, and a layer of metal oxide variable resistance material sandwiched between the bottom electrode and the top electrode, in which the neuron device is switched to a normal state upon application of reset pulse, and is switched to an excitation state upon application of stimulus pulses. the neuron device has a comprehensive response to different amplitude, different width of a stimulus voltage pulse and different number of a sequence of stimulus pulses, and provides functionalities of a weighting section and a computing section. the neuron device has a simple structure, excellent scalability, quick speed, low operation voltage, and is compatible with the conventional silicon-based cmos fabrication process, and thus suitable for mass production. the neuron device is capable of performing many biological functions and complex logic operations. dated 2014-12-30"
8930247,system and methods for content-based financial decision making support,"robust content-based decision-making support is enabled by software with a customizable knowledge base. utilizing proprietary information contained within a knowledge base, the software enables users to search the indexed database by feature, example firm, or pattern and update the knowledge base based on the results. the information contained in the knowledge base enables results to be ranked by relevance and enables other feedback to be provided. the system and methods provide process support by helping financial professionals identify, analyze, and construct data analysis patterns based on individual domain knowledge and preferences. the system and methods automatically detect abnormal patterns and automatically analyze their correlations to market events to provide further process support to financial professionals. using the results of any searching, analysis, and processing, the system and methods provide a neural network or other learning algorithm to provide content-based decision-making support.",2015-01-06,"The title of the patent is system and methods for content-based financial decision making support and its abstract is robust content-based decision-making support is enabled by software with a customizable knowledge base. utilizing proprietary information contained within a knowledge base, the software enables users to search the indexed database by feature, example firm, or pattern and update the knowledge base based on the results. the information contained in the knowledge base enables results to be ranked by relevance and enables other feedback to be provided. the system and methods provide process support by helping financial professionals identify, analyze, and construct data analysis patterns based on individual domain knowledge and preferences. the system and methods automatically detect abnormal patterns and automatically analyze their correlations to market events to provide further process support to financial professionals. using the results of any searching, analysis, and processing, the system and methods provide a neural network or other learning algorithm to provide content-based decision-making support. dated 2015-01-06"
8930292,learning and auditory scene analysis in gradient frequency nonlinear oscillator networks,"a method for learning connections between nonlinear oscillators in a neural network comprising the steps of providing a plurality of nonlinear oscillators, with each respective oscillator producing an oscillation distinct from the others in response to an input and detecting an input at an at least first oscillator of the plurality of nonlinear oscillators. detecting an input at an at least a second oscillator of the plurality of nonlinear oscillators, comparing the oscillation of the at least first oscillator to the oscillation of the at least second oscillator at a point in time, and determining whether there is coherency between the oscillation of the at least first oscillator and the oscillation of the at least second oscillator. changing at least one of the amplitude and phase of a connection between the at least first oscillator and the at least second least oscillator as a function coherency between the at least first oscillator and the oscillation of the at least second oscillator.",2015-01-06,"The title of the patent is learning and auditory scene analysis in gradient frequency nonlinear oscillator networks and its abstract is a method for learning connections between nonlinear oscillators in a neural network comprising the steps of providing a plurality of nonlinear oscillators, with each respective oscillator producing an oscillation distinct from the others in response to an input and detecting an input at an at least first oscillator of the plurality of nonlinear oscillators. detecting an input at an at least a second oscillator of the plurality of nonlinear oscillators, comparing the oscillation of the at least first oscillator to the oscillation of the at least second oscillator at a point in time, and determining whether there is coherency between the oscillation of the at least first oscillator and the oscillation of the at least second oscillator. changing at least one of the amplitude and phase of a connection between the at least first oscillator and the at least second least oscillator as a function coherency between the at least first oscillator and the oscillation of the at least second oscillator. dated 2015-01-06"
8930299,systems and methods for wind forecasting and grid management,"in one embodiment, a wind power ramp event nowcasting system includes a wind condition analyzer for detecting a wind power ramp signal; a sensor array, situated in an area relative to a wind farm, the sensor array providing data to the wind condition analyzer; a mesoscale numerical model; a neural network pattern recognizer; and a statistical forecast model, wherein the statistical model receives input from the wind condition analyzer, the mesoscale numerical model, and the neural network pattern recognizer; and the statistical forecast model outputs a time and duration for the wind power ramp event (wpre) for the wind farm.",2015-01-06,"The title of the patent is systems and methods for wind forecasting and grid management and its abstract is in one embodiment, a wind power ramp event nowcasting system includes a wind condition analyzer for detecting a wind power ramp signal; a sensor array, situated in an area relative to a wind farm, the sensor array providing data to the wind condition analyzer; a mesoscale numerical model; a neural network pattern recognizer; and a statistical forecast model, wherein the statistical model receives input from the wind condition analyzer, the mesoscale numerical model, and the neural network pattern recognizer; and the statistical forecast model outputs a time and duration for the wind power ramp event (wpre) for the wind farm. dated 2015-01-06"
8932220,method of predicting acute cardiopulmonary events and survivability of a patient,"a method of producing an artificial neural network capable of predicting the survivability of a patient, including: storing in an electronic database patient health data comprising a plurality of sets of data, each set having at least one of a first parameter relating to heart rate variability data and a second parameter relating to vital sign data, each set further having a third parameter relating to patient survivability; providing a network of nodes interconnected to form an artificial neural network, the nodes comprising a plurality of artificial neurons, each artificial neuron having at least one input with an associated weight; and training the artificial neural network using the patient health data such that the associated weight of the at least one input of each artificial neuron is adjusted in response to respective first, second and third parameters of different sets of data from the patient health data.",2015-01-13,"The title of the patent is method of predicting acute cardiopulmonary events and survivability of a patient and its abstract is a method of producing an artificial neural network capable of predicting the survivability of a patient, including: storing in an electronic database patient health data comprising a plurality of sets of data, each set having at least one of a first parameter relating to heart rate variability data and a second parameter relating to vital sign data, each set further having a third parameter relating to patient survivability; providing a network of nodes interconnected to form an artificial neural network, the nodes comprising a plurality of artificial neurons, each artificial neuron having at least one input with an associated weight; and training the artificial neural network using the patient health data such that the associated weight of the at least one input of each artificial neuron is adjusted in response to respective first, second and third parameters of different sets of data from the patient health data. dated 2015-01-13"
8933572,adaptive superconductive magnetic energy storage (smes) control method and system,"the adaptive superconductive magnetic energy storage (smes) control method and system control a smes device connected to a power generation system. a radial basis function neural network (rbfnn) connected to the controller adaptively adjusts gain constants of the controller. a processor executes an improved particle swarm optimization (ipso) procedure to train the rbfnn from input-output training data created by the ipso, and thereafter generate starting weights for the neural network. tests carried out show that the proposed adaptive smes controller maintains the dc capacitor voltage constant, thus improving the efficiency of wind energy transfer. the power output (reactive and real) of the smes device improves the voltage profile following large voltage dips and provides added damping to the system.",2015-01-13,"The title of the patent is adaptive superconductive magnetic energy storage (smes) control method and system and its abstract is the adaptive superconductive magnetic energy storage (smes) control method and system control a smes device connected to a power generation system. a radial basis function neural network (rbfnn) connected to the controller adaptively adjusts gain constants of the controller. a processor executes an improved particle swarm optimization (ipso) procedure to train the rbfnn from input-output training data created by the ipso, and thereafter generate starting weights for the neural network. tests carried out show that the proposed adaptive smes controller maintains the dc capacitor voltage constant, thus improving the efficiency of wind energy transfer. the power output (reactive and real) of the smes device improves the voltage profile following large voltage dips and provides added damping to the system. dated 2015-01-13"
8943007,"spike tagging for debugging, querying, and causal analysis","embodiments of the invention relate to spike tagging for a neural network. one embodiment comprises a neural network including multiple electronic neurons and a plurality of weighted synaptic connections interconnecting the neurons. an originating neuron of the neural network generates a spike event and a message tag that includes information relating to said originating neuron. a neuron of the neural network receives a spike event and a message tag from an interconnected neuron. in response to one or more received spike events, a receiving neuron spikes and sends a message tag selected from received message tags to an interconnected neuron.",2015-01-27,"The title of the patent is spike tagging for debugging, querying, and causal analysis and its abstract is embodiments of the invention relate to spike tagging for a neural network. one embodiment comprises a neural network including multiple electronic neurons and a plurality of weighted synaptic connections interconnecting the neurons. an originating neuron of the neural network generates a spike event and a message tag that includes information relating to said originating neuron. a neuron of the neural network receives a spike event and a message tag from an interconnected neuron. in response to one or more received spike events, a receiving neuron spikes and sends a message tag selected from received message tags to an interconnected neuron. dated 2015-01-27"
8943008,apparatus and methods for reinforcement learning in artificial neural networks,"neural network apparatus and methods for implementing reinforcement learning. in one implementation, the neural network is a spiking neural network, and the apparatus and methods may be used for example to enable an adaptive signal processing system to effect focused exploration by associative adaptation, including providing a negative reward signal to the network, which may increase excitability of the neurons in combination with decrease in excitability of active neurons. in certain implementations, the increase is gradual and of smaller magnitude, compared to the excitability decrease. in some implementations, the increase/decrease of the neuron excitability is effectuated by increasing/decreasing an efficacy of the respective synaptic connections delivering presynaptic inputs into the neuron. the focused exploration may be achieved for instance by non-associative potentiation configured based at least on the input spike rate. the non-associative potentiation may further comprise depression of connections that provide input in excess of a desired limit.",2015-01-27,"The title of the patent is apparatus and methods for reinforcement learning in artificial neural networks and its abstract is neural network apparatus and methods for implementing reinforcement learning. in one implementation, the neural network is a spiking neural network, and the apparatus and methods may be used for example to enable an adaptive signal processing system to effect focused exploration by associative adaptation, including providing a negative reward signal to the network, which may increase excitability of the neurons in combination with decrease in excitability of active neurons. in certain implementations, the increase is gradual and of smaller magnitude, compared to the excitability decrease. in some implementations, the increase/decrease of the neuron excitability is effectuated by increasing/decreasing an efficacy of the respective synaptic connections delivering presynaptic inputs into the neuron. the focused exploration may be achieved for instance by non-associative potentiation configured based at least on the input spike rate. the non-associative potentiation may further comprise depression of connections that provide input in excess of a desired limit. dated 2015-01-27"
8951193,method of predicting acute cardiopulmonary events and survivability of a patient,"a method of producing an artificial neural network capable of predicting the survivability of a patient, including: storing in an electronic database patient health data comprising a plurality of sets of data, each set having at least one of a first parameter relating to heart rate variability data and a second parameter relating to vital sign data, each set further having a third parameter relating to patient survivability; providing a network of nodes interconnected to form an artificial neural network, the nodes comprising a plurality of artificial neurons, each artificial neuron having at least one input with an associated weight; and training the artificial neural network using the patient health data such that the associated weight of the at least one input of each artificial neuron is adjusted in response to respective first, second and third parameters of different sets of data from the patient health data.",2015-02-10,"The title of the patent is method of predicting acute cardiopulmonary events and survivability of a patient and its abstract is a method of producing an artificial neural network capable of predicting the survivability of a patient, including: storing in an electronic database patient health data comprising a plurality of sets of data, each set having at least one of a first parameter relating to heart rate variability data and a second parameter relating to vital sign data, each set further having a third parameter relating to patient survivability; providing a network of nodes interconnected to form an artificial neural network, the nodes comprising a plurality of artificial neurons, each artificial neuron having at least one input with an associated weight; and training the artificial neural network using the patient health data such that the associated weight of the at least one input of each artificial neuron is adjusted in response to respective first, second and third parameters of different sets of data from the patient health data. dated 2015-02-10"
8953436,automotive neural network,network node modules within a vehicle are arranged to form a reconfigurable automotive neural network. each network node module includes one or more subsystems for performing one or more operations and a local processing module for communicating with the one or more subsystems. a switch coupled between the one or more subsystems and the processing module re-routes traffic from the one or more subsystems to an external processing module upon failure of the local processing module.,2015-02-10,The title of the patent is automotive neural network and its abstract is network node modules within a vehicle are arranged to form a reconfigurable automotive neural network. each network node module includes one or more subsystems for performing one or more operations and a local processing module for communicating with the one or more subsystems. a switch coupled between the one or more subsystems and the processing module re-routes traffic from the one or more subsystems to an external processing module upon failure of the local processing module. dated 2015-02-10
8954304,neural net for use in drilling simulation,"a method of optimizing a drilling tool assembly including inputting well data into an optimization system, the optimization system having an experience data set and an artificial neural network. the method further including comparing the well data to the experience data set and developing an initial drilling tool assembly based on the comparing the well data to the experience data, wherein the drilling tool assembly is developed using the artificial neural network. additionally, the method including simulating the initial drilling tool assembly in the optimization system and creating result data in the optimization system based on the simulating.",2015-02-10,"The title of the patent is neural net for use in drilling simulation and its abstract is a method of optimizing a drilling tool assembly including inputting well data into an optimization system, the optimization system having an experience data set and an artificial neural network. the method further including comparing the well data to the experience data set and developing an initial drilling tool assembly based on the comparing the well data to the experience data, wherein the drilling tool assembly is developed using the artificial neural network. additionally, the method including simulating the initial drilling tool assembly in the optimization system and creating result data in the optimization system based on the simulating. dated 2015-02-10"
8955383,ultrasonic gas leak detector with false alarm discrimination,"an ultrasonic gas leak detector is configured to discriminate the ultrasound generated by a pressurized gas leak into the atmosphere from false alarm ultrasound. an exemplary embodiment includes a sensor for detecting ultrasonic energy and providing sensor signals, and an electronic controller responsive to the sensor signals. in one exemplary embodiment, the electronic controller is configured to provide a threshold comparator function to compare a sensor signal value representative of sensed ultrasonic energy to a gas detection threshold value, and an artificial neural network (ann) function for processing signals derived from the digital sensor signals and applying ann coefficients configured to discriminate false alarm sources from gas leaks. an output function generates detector outputs in dependence on the threshold comparator output and the ann output.",2015-02-17,"The title of the patent is ultrasonic gas leak detector with false alarm discrimination and its abstract is an ultrasonic gas leak detector is configured to discriminate the ultrasound generated by a pressurized gas leak into the atmosphere from false alarm ultrasound. an exemplary embodiment includes a sensor for detecting ultrasonic energy and providing sensor signals, and an electronic controller responsive to the sensor signals. in one exemplary embodiment, the electronic controller is configured to provide a threshold comparator function to compare a sensor signal value representative of sensed ultrasonic energy to a gas detection threshold value, and an artificial neural network (ann) function for processing signals derived from the digital sensor signals and applying ann coefficients configured to discriminate false alarm sources from gas leaks. an output function generates detector outputs in dependence on the threshold comparator output and the ann output. dated 2015-02-17"
8965112,sequence transcription with deep neural networks,"systems and methods for sequence transcription with neural networks are provided. more particularly, a neural network can be implemented to map a plurality of training images received by the neural network into a probabilistic model of sequences comprising p(s|x) by maximizing log p(s|x) on the plurality of training images. x represents an input image and s represents an output sequence of characters for the input image. the trained neural network can process a received image containing characters associated with building numbers. the trained neural network can generate a predicted sequence of characters by processing the received image.",2015-02-24,"The title of the patent is sequence transcription with deep neural networks and its abstract is systems and methods for sequence transcription with neural networks are provided. more particularly, a neural network can be implemented to map a plurality of training images received by the neural network into a probabilistic model of sequences comprising p(s|x) by maximizing log p(s|x) on the plurality of training images. x represents an input image and s represents an output sequence of characters for the input image. the trained neural network can process a received image containing characters associated with building numbers. the trained neural network can generate a predicted sequence of characters by processing the received image. dated 2015-02-24"
8965664,controller for plant,"a controller for a plant that controls a controlled variable for the plant in accordance with estimated values, allowing to reduce any error in the estimated values that is caused by solid variation or aging of the plant. a controller for an exhaust emission control system has an estimated inert-egr value calculation section (711) to calculate the estimated value iegrhat for the inert-egr amount on the basis of an input vector u through a neural network, an estimated laf sensor output value calculation section (712) to calculate the estimated value φhat for an exhaust air-fuel ratio correlating with the inert-egr amount on the basis of the input vector u through the neural network, an laf sensor (34) to detect the exhaust air-fuel ratio, and a nonlinear adaptive corrector (713) to calculate the adaptive input uvns such that the estimated error ehat between the detected value φact from the laf sensor (34) and the estimated output value φhat of the laf sensor (34) is minimized.",2015-02-24,"The title of the patent is controller for plant and its abstract is a controller for a plant that controls a controlled variable for the plant in accordance with estimated values, allowing to reduce any error in the estimated values that is caused by solid variation or aging of the plant. a controller for an exhaust emission control system has an estimated inert-egr value calculation section (711) to calculate the estimated value iegrhat for the inert-egr amount on the basis of an input vector u through a neural network, an estimated laf sensor output value calculation section (712) to calculate the estimated value φhat for an exhaust air-fuel ratio correlating with the inert-egr amount on the basis of the input vector u through the neural network, an laf sensor (34) to detect the exhaust air-fuel ratio, and a nonlinear adaptive corrector (713) to calculate the adaptive input uvns such that the estimated error ehat between the detected value φact from the laf sensor (34) and the estimated output value φhat of the laf sensor (34) is minimized. dated 2015-02-24"
8965819,system and method for effective caching using neural networks,"systems and methods for selecting an appropriate caching algorithm to be used when temporarily storing data accessed by an executing application using a neural network may dynamically and/or iteratively replace an initial caching algorithm being used for the application. an input layer of the neural network may gather values of performance related parameters, such as cache hit rates, data throughput rates, or memory access request response times. the neural network may detect a pattern or change in a pattern of accesses, or a change in a workload, a hardware component, or an operating system parameter. dependent on these and/or other inputs, the neural network may select and apply a caching algorithm likely to improve performance of the application. other inputs to the neural network may include values of hardware configuration parameters and/or operating system parameters. the neural network may perform a training exercise or may be self-training, e.g., using reinforcement learning.",2015-02-24,"The title of the patent is system and method for effective caching using neural networks and its abstract is systems and methods for selecting an appropriate caching algorithm to be used when temporarily storing data accessed by an executing application using a neural network may dynamically and/or iteratively replace an initial caching algorithm being used for the application. an input layer of the neural network may gather values of performance related parameters, such as cache hit rates, data throughput rates, or memory access request response times. the neural network may detect a pattern or change in a pattern of accesses, or a change in a workload, a hardware component, or an operating system parameter. dependent on these and/or other inputs, the neural network may select and apply a caching algorithm likely to improve performance of the application. other inputs to the neural network may include values of hardware configuration parameters and/or operating system parameters. the neural network may perform a training exercise or may be self-training, e.g., using reinforcement learning. dated 2015-02-24"
8965821,learning method of neural network circuit,"a neuron circuit in a neural network circuit element includes a waveform generating circuit for generating a predetermined pulse voltage, and a first input signal has a waveform of the predetermined pulse voltage. for a period having a predetermined duration of the predetermined pulse voltage generated within the neural network circuit element including the variable resistance element which is applied with the first input signal from another neural network circuit element, the first input signal is permitted to be input to the control electrode of the variable resistance element, to change the resistance value of the variable resistance element due to an electric potential difference generated between the first electrode and the control electrode which occurs depending on an input timing of the first input signal with respect to the period during which the first input signal is permitted to be input to the control electrode.",2015-02-24,"The title of the patent is learning method of neural network circuit and its abstract is a neuron circuit in a neural network circuit element includes a waveform generating circuit for generating a predetermined pulse voltage, and a first input signal has a waveform of the predetermined pulse voltage. for a period having a predetermined duration of the predetermined pulse voltage generated within the neural network circuit element including the variable resistance element which is applied with the first input signal from another neural network circuit element, the first input signal is permitted to be input to the control electrode of the variable resistance element, to change the resistance value of the variable resistance element due to an electric potential difference generated between the first electrode and the control electrode which occurs depending on an input timing of the first input signal with respect to the period during which the first input signal is permitted to be input to the control electrode. dated 2015-02-24"
8972029,methods and systems for controlling a semiconductor fabrication process,"software for controlling processes in a heterogeneous semiconductor manufacturing environment may include a wafer-centric database, a real-time scheduler using a neural network, and a graphical user interface displaying simulated operation of the system. these features may be employed alone or in combination to offer improved usability and computational efficiency for real time control and monitoring of a semiconductor manufacturing process. more generally, these techniques may be usefully employed in a variety of real time control systems, particularly systems requiring complex scheduling decisions or heterogeneous systems constructed of hardware from numerous independent vendors.",2015-03-03,"The title of the patent is methods and systems for controlling a semiconductor fabrication process and its abstract is software for controlling processes in a heterogeneous semiconductor manufacturing environment may include a wafer-centric database, a real-time scheduler using a neural network, and a graphical user interface displaying simulated operation of the system. these features may be employed alone or in combination to offer improved usability and computational efficiency for real time control and monitoring of a semiconductor manufacturing process. more generally, these techniques may be usefully employed in a variety of real time control systems, particularly systems requiring complex scheduling decisions or heterogeneous systems constructed of hardware from numerous independent vendors. dated 2015-03-03"
8976893,predistortion according to an artificial neural network (ann)-based model,"embodiments include a method for predistorting an input signal at a predistorter to compensate for distortion introduced by a non-linear electronic device operating on the input signal to produce an output signal. the method entails generating first and second signal samples for each of a plurality of sampling time instances. the first and second signal samples represent the input and output signals, and are spaced at unit-delay intervals. the method further entails calculating, from the first and second signal samples, parameters for an ann-based model. the ann-based model includes a tapped delay line configured to dynamically model memory effects of the distortion introduced by the device, or of the response of the predistorter, with a multi-unit delay interval between at least one pair of adjacent delays. the method also includes predistorting the input signal according to the ann-based model, to produce a predistorted input signal for input to the device.",2015-03-10,"The title of the patent is predistortion according to an artificial neural network (ann)-based model and its abstract is embodiments include a method for predistorting an input signal at a predistorter to compensate for distortion introduced by a non-linear electronic device operating on the input signal to produce an output signal. the method entails generating first and second signal samples for each of a plurality of sampling time instances. the first and second signal samples represent the input and output signals, and are spaced at unit-delay intervals. the method further entails calculating, from the first and second signal samples, parameters for an ann-based model. the ann-based model includes a tapped delay line configured to dynamically model memory effects of the distortion introduced by the device, or of the response of the predistorter, with a multi-unit delay interval between at least one pair of adjacent delays. the method also includes predistorting the input signal according to the ann-based model, to produce a predistorted input signal for input to the device. dated 2015-03-10"
8976929,automatic generation of patient-specific radiation therapy planning parameters,"an apparatus and method for automatically generating radiation treatment planning parameters are disclosed. in accordance with the illustrative embodiment, a database is constructed that stores: (i) patient data and past treatment plans by expert human planners for these patients, and (ii) optimal treatment plans that are generated using multi-objective optimization and pareto front search and that represent the best tradeoff opportunities of the patient case, and a predictive model (e.g., a neural network, a decision tree, a support vector machine [svm], etc.) is then trained via a learning algorithm on a plurality of input/output mappings derived from the contents of the database. during training, the predictive model is trained to identify and infer patterns in the treatment plan data through a process of generalization. once trained, the predictive model can then be used to automatically generate radiation treatment planning parameters for new patients.",2015-03-10,"The title of the patent is automatic generation of patient-specific radiation therapy planning parameters and its abstract is an apparatus and method for automatically generating radiation treatment planning parameters are disclosed. in accordance with the illustrative embodiment, a database is constructed that stores: (i) patient data and past treatment plans by expert human planners for these patients, and (ii) optimal treatment plans that are generated using multi-objective optimization and pareto front search and that represent the best tradeoff opportunities of the patient case, and a predictive model (e.g., a neural network, a decision tree, a support vector machine [svm], etc.) is then trained via a learning algorithm on a plurality of input/output mappings derived from the contents of the database. during training, the predictive model is trained to identify and infer patterns in the treatment plan data through a process of generalization. once trained, the predictive model can then be used to automatically generate radiation treatment planning parameters for new patients. dated 2015-03-10"
8977578,synaptic time multiplexing neuromorphic network that forms subsets of connections during different time slots,"a synaptic time-multiplexed (stm) neuromorphic network includes a neural fabric that includes nodes and switches to define inter-nodal connections between selected nodes of the neural fabric. the stm neuromorphic network further includes a neuromorphic controller to form subsets of a set of the inter-nodal connections representing a fully connected neural network. each subset is formed during a different time slot of a plurality of time slots of a time multiplexing cycle of the stm neuromorphic network. in combination, the inter-nodal connection subsets implement the fully connected neural network. a method of synaptic time multiplexing a neuromorphic network includes providing the neural fabric and forming the subsets of the set of inter-nodal connections.",2015-03-10,"The title of the patent is synaptic time multiplexing neuromorphic network that forms subsets of connections during different time slots and its abstract is a synaptic time-multiplexed (stm) neuromorphic network includes a neural fabric that includes nodes and switches to define inter-nodal connections between selected nodes of the neural fabric. the stm neuromorphic network further includes a neuromorphic controller to form subsets of a set of the inter-nodal connections representing a fully connected neural network. each subset is formed during a different time slot of a plurality of time slots of a time multiplexing cycle of the stm neuromorphic network. in combination, the inter-nodal connection subsets implement the fully connected neural network. a method of synaptic time multiplexing a neuromorphic network includes providing the neural fabric and forming the subsets of the set of inter-nodal connections. dated 2015-03-10"
8977583,"synaptic, dendritic, somatic, and axonal plasticity in a network of neural cores using a plastic multi-stage crossbar switching","embodiments of the invention provide a neural network comprising multiple functional neural core circuits, and a dynamically reconfigurable switch interconnect between the functional neural core circuits. the interconnect comprises multiple connectivity neural core circuits. each functional neural core circuit comprises a first and a second core module. each core module comprises a plurality of electronic neurons, a plurality of incoming electronic axons, and multiple electronic synapses interconnecting the incoming axons to the neurons. each neuron has a corresponding outgoing electronic axon. in one embodiment, zero or more sets of connectivity neural core circuits interconnect outgoing axons in a functional neural core circuit to incoming axons in the same functional neural core circuit. in another embodiment, zero or more sets of connectivity neural core circuits interconnect outgoing and incoming axons in a functional neural core circuit to incoming and outgoing axons in a different functional neural core circuit, respectively.",2015-03-10,"The title of the patent is synaptic, dendritic, somatic, and axonal plasticity in a network of neural cores using a plastic multi-stage crossbar switching and its abstract is embodiments of the invention provide a neural network comprising multiple functional neural core circuits, and a dynamically reconfigurable switch interconnect between the functional neural core circuits. the interconnect comprises multiple connectivity neural core circuits. each functional neural core circuit comprises a first and a second core module. each core module comprises a plurality of electronic neurons, a plurality of incoming electronic axons, and multiple electronic synapses interconnecting the incoming axons to the neurons. each neuron has a corresponding outgoing electronic axon. in one embodiment, zero or more sets of connectivity neural core circuits interconnect outgoing axons in a functional neural core circuit to incoming axons in the same functional neural core circuit. in another embodiment, zero or more sets of connectivity neural core circuits interconnect outgoing and incoming axons in a functional neural core circuit to incoming and outgoing axons in a different functional neural core circuit, respectively. dated 2015-03-10"
8983885,prospective media content generation using neural network modeling,"a system for prospectively identifying media characteristics for inclusion in media content is disclosed. a neural network database including media characteristic information and feature information may associate relationships among the media characteristic information and feature information. personal characteristic information associated with target media consumers may be used to select a subset of the neural network database. a first set of nodes, representing selected feature information, may be activated. the node interactions may be calculated to detect the activation of a second set of nodes, the second set of nodes representing media characteristic information. generally, a node is activated when an activation value of the node exceeds a threshold value. media characteristic information may be identified for inclusion in media content based on the second set of nodes.",2015-03-17,"The title of the patent is prospective media content generation using neural network modeling and its abstract is a system for prospectively identifying media characteristics for inclusion in media content is disclosed. a neural network database including media characteristic information and feature information may associate relationships among the media characteristic information and feature information. personal characteristic information associated with target media consumers may be used to select a subset of the neural network database. a first set of nodes, representing selected feature information, may be activated. the node interactions may be calculated to detect the activation of a second set of nodes, the second set of nodes representing media characteristic information. generally, a node is activated when an activation value of the node exceeds a threshold value. media characteristic information may be identified for inclusion in media content based on the second set of nodes. dated 2015-03-17"
8989394,active delay method and a improved wireless binaural hearing device using the same method,"disclosed are an active delay method and an improved wireless binaural hearing device. the binaural hearing device includes: a first hearing device including a first microphone, an amplifier and a wireless transmitter; and a second hearing device including a second microphone, an amplifier, a wireless transmitter, a wireless receiver which receives a signal from the wireless transmitter, an active delay circuit which synchronizes the received signal with a signal acquired by the second microphone, a neural network which synchronizes the delayed signal, and a speaker which converts the synchronized signal into a voice signal. with this configuration, it is possible to prevent incorrect detection of the position of the sound source or paralalia due to a time delay which is produced in the wireless binaural hearing device and reduce noises due to a time difference between both hearing devices, thereby providing a binaural hearing device with high quality.",2015-03-24,"The title of the patent is active delay method and a improved wireless binaural hearing device using the same method and its abstract is disclosed are an active delay method and an improved wireless binaural hearing device. the binaural hearing device includes: a first hearing device including a first microphone, an amplifier and a wireless transmitter; and a second hearing device including a second microphone, an amplifier, a wireless transmitter, a wireless receiver which receives a signal from the wireless transmitter, an active delay circuit which synchronizes the received signal with a signal acquired by the second microphone, a neural network which synchronizes the delayed signal, and a speaker which converts the synchronized signal into a voice signal. with this configuration, it is possible to prevent incorrect detection of the position of the sound source or paralalia due to a time delay which is produced in the wireless binaural hearing device and reduce noises due to a time difference between both hearing devices, thereby providing a binaural hearing device with high quality. dated 2015-03-24"
8990130,consolidating multiple neurosynaptic cores into one memory,"embodiments of the invention relate to a neural network system comprising a single memory block for multiple neurosynaptic core modules. one embodiment comprises a neural network system including a memory array that maintains information for multiple neurosynaptic core modules. each neurosynaptic core module comprises multiple neurons. the neural network system further comprises at least one logic circuit. each logic circuit receives neuronal firing events targeting a neurosynaptic core module of the neural network system, and said logic circuit integrates the firing events received based on information maintained in said memory for said neurosynaptic core module.",2015-03-24,"The title of the patent is consolidating multiple neurosynaptic cores into one memory and its abstract is embodiments of the invention relate to a neural network system comprising a single memory block for multiple neurosynaptic core modules. one embodiment comprises a neural network system including a memory array that maintains information for multiple neurosynaptic core modules. each neurosynaptic core module comprises multiple neurons. the neural network system further comprises at least one logic circuit. each logic circuit receives neuronal firing events targeting a neurosynaptic core module of the neural network system, and said logic circuit integrates the firing events received based on information maintained in said memory for said neurosynaptic core module. dated 2015-03-24"
8990131,"bottom sediment determination device, ultrasonic finder, and method and program for setting parameters","this disclosure provide a bottom sediment determining device, which is inputted with an echo signal corresponding to an ultrasonic wave outputted underwater, and determines water bottom sediment using a neural network. the device includes a memory for storing two or more parameters to be used in the neural network so as to be associated with positional information, a receiver for receiving the positional information, an acquisition module for acquiring the parameters corresponding to the positional information, and a setting module for setting the parameters to the neural network.",2015-03-24,"The title of the patent is bottom sediment determination device, ultrasonic finder, and method and program for setting parameters and its abstract is this disclosure provide a bottom sediment determining device, which is inputted with an echo signal corresponding to an ultrasonic wave outputted underwater, and determines water bottom sediment using a neural network. the device includes a memory for storing two or more parameters to be used in the neural network so as to be associated with positional information, a receiver for receiving the positional information, an acquisition module for acquiring the parameters corresponding to the positional information, and a setting module for setting the parameters to the neural network. dated 2015-03-24"
8990132,artificial neural networks based on a low-order model of biological neural networks,"a low-order model (lom) of biological neural networks and its mathematical equivalents including the clusterer interpreter probabilistic associative memory (cipam) are disclosed. they are artificial neural networks (anns) organized as networks of processing units (pus), each pu comprising artificial neuronal encoders, synapses, spiking/nonspiking neurons, and a scheme for maximal generalization. if the weights in the artificial synapses in a pu have been learned (and then fixed) or can be adjusted by the unsupervised accumulation rule and the unsupervised covariance rule (or supervised covariance rule), the pu is called unsupervised (or supervised) pu. the disclosed anns, with these hebbian-type learning rules, can learn large numbers of large input vectors with temporally/spatially hierarchical causes with ease and recognize such causes with maximal generalization despite corruption, distortion and occlusion. an ann with a network of unsupervised pus (called clusterer) and offshoot supervised pus (called interpreter) is an architecture for many applications.",2015-03-24,"The title of the patent is artificial neural networks based on a low-order model of biological neural networks and its abstract is a low-order model (lom) of biological neural networks and its mathematical equivalents including the clusterer interpreter probabilistic associative memory (cipam) are disclosed. they are artificial neural networks (anns) organized as networks of processing units (pus), each pu comprising artificial neuronal encoders, synapses, spiking/nonspiking neurons, and a scheme for maximal generalization. if the weights in the artificial synapses in a pu have been learned (and then fixed) or can be adjusted by the unsupervised accumulation rule and the unsupervised covariance rule (or supervised covariance rule), the pu is called unsupervised (or supervised) pu. the disclosed anns, with these hebbian-type learning rules, can learn large numbers of large input vectors with temporally/spatially hierarchical causes with ease and recognize such causes with maximal generalization despite corruption, distortion and occlusion. an ann with a network of unsupervised pus (called clusterer) and offshoot supervised pus (called interpreter) is an architecture for many applications. dated 2015-03-24"
8992435,system and method for classifying a heart sound,"a method and system for electronically classifying a pre-processed heart sound signal of a patient as functional (normal) or pathological is provided. the pre-processed patient heart sound signal is segmentised and features are extracted therefrom (104) to build up a feature vector which is representative of the heart sound signal. the feature vector is then fed to a diagnostic decision support network (105) comprising a plurality of artificial neural networks, each relating to a known heart pathology, which is in turn used to conduct the classification.",2015-03-31,"The title of the patent is system and method for classifying a heart sound and its abstract is a method and system for electronically classifying a pre-processed heart sound signal of a patient as functional (normal) or pathological is provided. the pre-processed patient heart sound signal is segmentised and features are extracted therefrom (104) to build up a feature vector which is representative of the heart sound signal. the feature vector is then fed to a diagnostic decision support network (105) comprising a plurality of artificial neural networks, each relating to a known heart pathology, which is in turn used to conduct the classification. dated 2015-03-31"
8996430,hierarchical scalable neuromorphic synaptronic system for synaptic and structural plasticity,"in one embodiment, the present invention provides a neural network circuit comprising multiple symmetric core circuits. each symmetric core circuit comprises a first core module and a second core module. each core module comprises a plurality of electronic neurons, a plurality of electronic axons, and an interconnection network comprising multiple electronic synapses interconnecting the axons to the neurons. each synapse interconnects an axon to a neuron. the first core module and the second core module are logically overlayed on one another such that neurons in the first core module are proximal to axons in the second core module, and axons in the first core module are proximal to neurons in the second core module. each neuron in each core module receives axonal firing events via interconnected axons and generates a neuronal firing event according to a neuronal activation function.",2015-03-31,"The title of the patent is hierarchical scalable neuromorphic synaptronic system for synaptic and structural plasticity and its abstract is in one embodiment, the present invention provides a neural network circuit comprising multiple symmetric core circuits. each symmetric core circuit comprises a first core module and a second core module. each core module comprises a plurality of electronic neurons, a plurality of electronic axons, and an interconnection network comprising multiple electronic synapses interconnecting the axons to the neurons. each synapse interconnects an axon to a neuron. the first core module and the second core module are logically overlayed on one another such that neurons in the first core module are proximal to axons in the second core module, and axons in the first core module are proximal to neurons in the second core module. each neuron in each core module receives axonal firing events via interconnected axons and generates a neuronal firing event according to a neuronal activation function. dated 2015-03-31"
9000918,security barriers with automated reconnaissance,"an intrusion delaying barrier includes primary and secondary physical structures and can be instrumented with multiple sensors incorporated into an electronic monitoring and alarm system. such an instrumented intrusion delaying barrier may be used as a perimeter intrusion defense and assessment system (pidas). problems with not providing effective delay to breaches by intentional intruders and/or terrorists who would otherwise evade detection are solved by attaching the secondary structures to the primary structure, and attaching at least some of the sensors to the secondary structures. by having multiple sensors of various types physically interconnected serves to enable sensors on different parts of the overall structure to respond to common disturbances and thereby provide effective corroboration that a disturbance is not merely a nuisance or false alarm. use of a machine learning network such as a neural network exploits such corroboration.",2015-04-07,"The title of the patent is security barriers with automated reconnaissance and its abstract is an intrusion delaying barrier includes primary and secondary physical structures and can be instrumented with multiple sensors incorporated into an electronic monitoring and alarm system. such an instrumented intrusion delaying barrier may be used as a perimeter intrusion defense and assessment system (pidas). problems with not providing effective delay to breaches by intentional intruders and/or terrorists who would otherwise evade detection are solved by attaching the secondary structures to the primary structure, and attaching at least some of the sensors to the secondary structures. by having multiple sensors of various types physically interconnected serves to enable sensors on different parts of the overall structure to respond to common disturbances and thereby provide effective corroboration that a disturbance is not merely a nuisance or false alarm. use of a machine learning network such as a neural network exploits such corroboration. dated 2015-04-07"
9008440,component recognizing apparatus and component recognizing method,"disclosed are a component recognizing apparatus and a component recognizing method. the component recognizing apparatus includes: an image preprocessing unit configured to extract component edges from an input component image by using a plurality of edge detecting techniques, and detect a component region by using the extracted component edges; a feature extracting unit configured to extract a component feature from the detected component region, and create a feature vector by using the component feature; and a component recognizing unit configured to input the created feature vector to an artificial neural network which has learned in advance to recognize a component category through a plurality of component image samples, and recognize the component category according to a result.",2015-04-14,"The title of the patent is component recognizing apparatus and component recognizing method and its abstract is disclosed are a component recognizing apparatus and a component recognizing method. the component recognizing apparatus includes: an image preprocessing unit configured to extract component edges from an input component image by using a plurality of edge detecting techniques, and detect a component region by using the extracted component edges; a feature extracting unit configured to extract a component feature from the detected component region, and create a feature vector by using the component feature; and a component recognizing unit configured to input the created feature vector to an artificial neural network which has learned in advance to recognize a component category through a plurality of component image samples, and recognize the component category according to a result. dated 2015-04-14"
9009038,method and system for analyzing digital sound audio signal associated with baby cry,"a method for analyzing a digital audio signal associated with a baby cry, comprising the steps of: (a) processing the digital audio signal using a spectral analysis to generate a spectral data; (b) processing the digital audio signal using a time-frequency analysis to generate a time-frequency characteristic; (c) categorizing the baby cry into one of a basic type and a special type based on the spectral data; (d) if the baby cry is of the basic type, determining a basic need based on the time-frequency characteristic and a predetermined lookup table; and (e) if the baby cry is of the special type, determining a special need by inputting the time-frequency characteristic into a pre-trained artificial neural network.",2015-04-14,"The title of the patent is method and system for analyzing digital sound audio signal associated with baby cry and its abstract is a method for analyzing a digital audio signal associated with a baby cry, comprising the steps of: (a) processing the digital audio signal using a spectral analysis to generate a spectral data; (b) processing the digital audio signal using a time-frequency analysis to generate a time-frequency characteristic; (c) categorizing the baby cry into one of a basic type and a special type based on the spectral data; (d) if the baby cry is of the basic type, determining a basic need based on the time-frequency characteristic and a predetermined lookup table; and (e) if the baby cry is of the special type, determining a special need by inputting the time-frequency characteristic into a pre-trained artificial neural network. dated 2015-04-14"
9015086,learnable contextual network,a method and apparatus for detection of relationships between objects in a meta-model semantic network is described. semantic objects and semantic relations of a meta-model of business objects are generated from a meta-model semantic network. the semantic relations are based on connections between the semantic objects. a neural network is formed based on usage of the semantic objects and the semantic relations. the neural network is integrated with the semantic objects and the semantic relations to generate a contextual network. a statistical analysis of the connections between the semantic objects in the contextual network is performed to identify stronger semantic relations. the identified stronger semantic relations are used to update the neural network. the updated neural network is integrated into the contextual network.,2015-04-21,The title of the patent is learnable contextual network and its abstract is a method and apparatus for detection of relationships between objects in a meta-model semantic network is described. semantic objects and semantic relations of a meta-model of business objects are generated from a meta-model semantic network. the semantic relations are based on connections between the semantic objects. a neural network is formed based on usage of the semantic objects and the semantic relations. the neural network is integrated with the semantic objects and the semantic relations to generate a contextual network. a statistical analysis of the connections between the semantic objects in the contextual network is performed to identify stronger semantic relations. the identified stronger semantic relations are used to update the neural network. the updated neural network is integrated into the contextual network. dated 2015-04-21
9015093,intelligent control with hierarchical stacked neural networks,"a method of processing information is provided. the method involves receiving a message; processing the message with a trained artificial neural network based processor, having at least one set of outputs which represent information in a non-arbitrary organization of actions based on an architecture of the artificial neural network based processor and the training; representing as a noise vector at least one data pattern in the message which is incompletely represented in the non-arbitrary organization of actions; analyzing the noise vector distinctly from the trained artificial neural network; searching at least one database; and generating an output in dependence on said analyzing and said searching.",2015-04-21,"The title of the patent is intelligent control with hierarchical stacked neural networks and its abstract is a method of processing information is provided. the method involves receiving a message; processing the message with a trained artificial neural network based processor, having at least one set of outputs which represent information in a non-arbitrary organization of actions based on an architecture of the artificial neural network based processor and the training; representing as a noise vector at least one data pattern in the message which is incompletely represented in the non-arbitrary organization of actions; analyzing the noise vector distinctly from the trained artificial neural network; searching at least one database; and generating an output in dependence on said analyzing and said searching. dated 2015-04-21"
9015095,neural network designing method and digital-to-analog fitting method,"a neural network designing method forms a rnn (recurrent neural network) circuit to include a plurality of oscillating rnn circuits configured to output natural oscillations, and an adding circuit configured to obtain a sum of outputs of the plurality of oscillating rnn circuits, and inputs discrete data to the plurality of oscillating rnn circuits in order to compute a fitting curve with respect to the discrete data output from the adding circuit.",2015-04-21,"The title of the patent is neural network designing method and digital-to-analog fitting method and its abstract is a neural network designing method forms a rnn (recurrent neural network) circuit to include a plurality of oscillating rnn circuits configured to output natural oscillations, and an adding circuit configured to obtain a sum of outputs of the plurality of oscillating rnn circuits, and inputs discrete data to the plurality of oscillating rnn circuits in order to compute a fitting curve with respect to the discrete data output from the adding circuit. dated 2015-04-21"
9015096,continuous time spiking neural network event-based simulation that schedules co-pending events using an indexable list of nodes,"certain aspects of the present disclosure provide methods and apparatus for a continuous-time neural network event-based simulation that includes a multi-dimensional multi-schedule architecture with ordered and unordered schedules and accelerators to provide for faster event sorting; and a formulation of modeling event operations as anticipating (the future) and advancing (update/jump ahead/catch up) rules or methods to provide a continuous-time neural network model. in this manner, the advantages include faster simulation of spiking neural networks (order(s) of magnitude); and a method for describing and modeling continuous time neurons, synapses, and general neural network behaviors.",2015-04-21,"The title of the patent is continuous time spiking neural network event-based simulation that schedules co-pending events using an indexable list of nodes and its abstract is certain aspects of the present disclosure provide methods and apparatus for a continuous-time neural network event-based simulation that includes a multi-dimensional multi-schedule architecture with ordered and unordered schedules and accelerators to provide for faster event sorting; and a formulation of modeling event operations as anticipating (the future) and advancing (update/jump ahead/catch up) rules or methods to provide a continuous-time neural network model. in this manner, the advantages include faster simulation of spiking neural networks (order(s) of magnitude); and a method for describing and modeling continuous time neurons, synapses, and general neural network behaviors. dated 2015-04-21"
9020302,method for producing super-resolution images and nonlinear digital filter for implementing same,"a method and a digital filter, for use with photo and video images, includes using a camera or video camera equipped with sensors and an electronic shutter to capture a plurality of frames of low resolution and producing one frame of high resolution. a plurality of frames are exposed. initial images are in the form of a continuous sequence of frames with high-speed capture. the frequency of the frames is inversely proportional to the magnitude of that part of the light-sensitive region of the sensor that is being scanned. the initial images are aligned and an enhanced image is produced. the enhanced image is filtered using a nonlinear filter which includes a neural network that is pretrained using a test image including radial and sinusoidal test charts, as well as reference points.",2015-04-28,"The title of the patent is method for producing super-resolution images and nonlinear digital filter for implementing same and its abstract is a method and a digital filter, for use with photo and video images, includes using a camera or video camera equipped with sensors and an electronic shutter to capture a plurality of frames of low resolution and producing one frame of high resolution. a plurality of frames are exposed. initial images are in the form of a continuous sequence of frames with high-speed capture. the frequency of the frames is inversely proportional to the magnitude of that part of the light-sensitive region of the sensor that is being scanned. the initial images are aligned and an enhanced image is produced. the enhanced image is filtered using a nonlinear filter which includes a neural network that is pretrained using a test image including radial and sinusoidal test charts, as well as reference points. dated 2015-04-28"
9020867,cortical simulator for object-oriented simulation of a neural network,embodiments of the invention relate to a function-level simulator for modeling a neurosynaptic chip. one embodiment comprises simulating a neural network using an object-oriented framework including a plurality of object-oriented classes. each class corresponds to a component of a neural network. running a simulation model of the neural network includes instantiating multiple simulation objects from the classes. each simulation object is an instance of one of the classes.,2015-04-28,The title of the patent is cortical simulator for object-oriented simulation of a neural network and its abstract is embodiments of the invention relate to a function-level simulator for modeling a neurosynaptic chip. one embodiment comprises simulating a neural network using an object-oriented framework including a plurality of object-oriented classes. each class corresponds to a component of a neural network. running a simulation model of the neural network includes instantiating multiple simulation objects from the classes. each simulation object is an instance of one of the classes. dated 2015-04-28
9022140,methods and systems for improved drilling operations using real-time and historical drilling data,"methods and systems are described for improved drilling operations through the use of real-time drilling data to predict bit wear, lithology, pore pressure, a rotating friction coefficient, permeability, and cost in real-time and to adjust drilling parameters in real-time based on the predictions. the real-time lithology prediction is made by processing the real-time drilling data through a multilayer neural network. the real-time bit wear prediction is made by using the real-time drilling data to predict a bit efficiency factor and to detect changes in the bit efficiency factor over time. these predictions may be used to adjust drilling parameters in the drilling operation in real-time, subject to override by the operator. the methods and systems may also include determining various downhole hydraulics parameters and a rotary friction factor. historical data may be used in combination with real-time data to provide expert system assistance and to identify safety concerns.",2015-05-05,"The title of the patent is methods and systems for improved drilling operations using real-time and historical drilling data and its abstract is methods and systems are described for improved drilling operations through the use of real-time drilling data to predict bit wear, lithology, pore pressure, a rotating friction coefficient, permeability, and cost in real-time and to adjust drilling parameters in real-time based on the predictions. the real-time lithology prediction is made by processing the real-time drilling data through a multilayer neural network. the real-time bit wear prediction is made by using the real-time drilling data to predict a bit efficiency factor and to detect changes in the bit efficiency factor over time. these predictions may be used to adjust drilling parameters in the drilling operation in real-time, subject to override by the operator. the methods and systems may also include determining various downhole hydraulics parameters and a rotary friction factor. historical data may be used in combination with real-time data to provide expert system assistance and to identify safety concerns. dated 2015-05-05"
9026964,intelligent metamodel integrated verilog-ams for fast and accurate analog block design exploration,a method for modeling a circuit comprising storing a plurality of design variable ranges for a circuit component in a non-transient electronic data memory. performing transistor-level simulations at a plurality of sample points for the circuit component to generate a plurality of design variable samples for the circuit component. storing a neural network architecture in the non-transient electronic data memory that models the plurality of design variable samples for the circuit component. storing a performance metric metamodel and a circuit parameter metamodel generated using verilog-ams.,2015-05-05,The title of the patent is intelligent metamodel integrated verilog-ams for fast and accurate analog block design exploration and its abstract is a method for modeling a circuit comprising storing a plurality of design variable ranges for a circuit component in a non-transient electronic data memory. performing transistor-level simulations at a plurality of sample points for the circuit component to generate a plurality of design variable samples for the circuit component. storing a neural network architecture in the non-transient electronic data memory that models the plurality of design variable samples for the circuit component. storing a performance metric metamodel and a circuit parameter metamodel generated using verilog-ams. dated 2015-05-05
9028416,method for measuring intracranial elasticity,a novel method to noninvasively measure intracranial pressure (icp) and more generally brain elasticity is disclosed. icp is determined using an algorithm coupled on a simulated artificial neural network (sann) that calculates icp based on a determination of a set of interacted ultrasound signals (iuss) generated from multiple ultrasound pulses. the methods and systems of the present invention are capable of rapidly determining icp without manual review of epg waves by a technician.,2015-05-12,The title of the patent is method for measuring intracranial elasticity and its abstract is a novel method to noninvasively measure intracranial pressure (icp) and more generally brain elasticity is disclosed. icp is determined using an algorithm coupled on a simulated artificial neural network (sann) that calculates icp based on a determination of a set of interacted ultrasound signals (iuss) generated from multiple ultrasound pulses. the methods and systems of the present invention are capable of rapidly determining icp without manual review of epg waves by a technician. dated 2015-05-12
9036745,use of neural network based matched filter for fast response time in high-speed communications channels,a neural network is used within a receiver to discriminate a large set of input waveforms without using a very large set of conventional matched filters. the neural network is trained under actual line conditions as opposed to the requirement for ideal signals when using matched filters. the finite waveforms are based on digital modulation principles. a best match is made between a received waveform from the noisy channel and that of previously trained waveforms in order to extract data. neural network based matched filter allows data be discriminated separately for each sub-carrier channel in the receiver. the neural network system allows fast processing and is suitable for high-speed data communications systems.,2015-05-19,The title of the patent is use of neural network based matched filter for fast response time in high-speed communications channels and its abstract is a neural network is used within a receiver to discriminate a large set of input waveforms without using a very large set of conventional matched filters. the neural network is trained under actual line conditions as opposed to the requirement for ideal signals when using matched filters. the finite waveforms are based on digital modulation principles. a best match is made between a received waveform from the noisy channel and that of previously trained waveforms in order to extract data. neural network based matched filter allows data be discriminated separately for each sub-carrier channel in the receiver. the neural network system allows fast processing and is suitable for high-speed data communications systems. dated 2015-05-19
9037224,apparatus for treating a patient,"a signal processing method and system combines multi-scale decomposition, such as wavelet, pre-processing together with a compression technique, such as an auto-associative artificial neural network, operating in the multi-scale decomposition domain for signal denoising and extraction. all compressions are performed in the decomposed domain. a reverse decomposition such as an inverse discrete wavelet transform is performed on the combined outputs from all the compression modules to recover a clean signal back in the time domain. a low-cost, non-drug, non-invasive, on-demand therapy braincap system and method are pharmaceutically non-intrusive to the body for the purpose of disease diagnosis, treatment therapy, and direct mind control of external devices and systems. it is based on recognizing abnormal brainwave signatures and intervenes at the earliest moment, using magnetic and/or electric stimulations to reset the brainwaves back to normality. the feedback system is self-regulatory and the treatment stops when the brainwaves return to normal. the braincap contains multiple sensing electrodes and microcoils; the microcoils are pairs of crossed microcoils or 3-axis triple crossed microcoils.",2015-05-19,"The title of the patent is apparatus for treating a patient and its abstract is a signal processing method and system combines multi-scale decomposition, such as wavelet, pre-processing together with a compression technique, such as an auto-associative artificial neural network, operating in the multi-scale decomposition domain for signal denoising and extraction. all compressions are performed in the decomposed domain. a reverse decomposition such as an inverse discrete wavelet transform is performed on the combined outputs from all the compression modules to recover a clean signal back in the time domain. a low-cost, non-drug, non-invasive, on-demand therapy braincap system and method are pharmaceutically non-intrusive to the body for the purpose of disease diagnosis, treatment therapy, and direct mind control of external devices and systems. it is based on recognizing abnormal brainwave signatures and intervenes at the earliest moment, using magnetic and/or electric stimulations to reset the brainwaves back to normality. the feedback system is self-regulatory and the treatment stops when the brainwaves return to normal. the braincap contains multiple sensing electrodes and microcoils; the microcoils are pairs of crossed microcoils or 3-axis triple crossed microcoils. dated 2015-05-19"
9043255,optimally configuring an information landscape,"according to an embodiment of the present invention, a system optimizes an information processing environment, and comprises at least one processor. the system collects information pertaining to operational behavior of the information processing environment and including a plurality of parameters. a neural network structure is established to associate the parameters to a desired operational performance characteristic for the information processing environment. the neural network structure is trained with the collected information from the information processing environment to produce a model for the information processing environment. the model is optimized to determine values for the parameters and the information processing environment is adjusted based on the determined parameter values to attain the desired operational performance of the information processing environment. embodiments of the present invention further include a method and computer program product for optimizing an information processing environment in substantially the same manner described above.",2015-05-26,"The title of the patent is optimally configuring an information landscape and its abstract is according to an embodiment of the present invention, a system optimizes an information processing environment, and comprises at least one processor. the system collects information pertaining to operational behavior of the information processing environment and including a plurality of parameters. a neural network structure is established to associate the parameters to a desired operational performance characteristic for the information processing environment. the neural network structure is trained with the collected information from the information processing environment to produce a model for the information processing environment. the model is optimized to determine values for the parameters and the information processing environment is adjusted based on the determined parameter values to attain the desired operational performance of the information processing environment. embodiments of the present invention further include a method and computer program product for optimizing an information processing environment in substantially the same manner described above. dated 2015-05-26"
9043738,machine-learning based datapath extraction,"a datapath extraction tool uses machine-learning models to selectively classify clusters of cells in an integrated circuit design as either datapath logic or non-datapath logic based on cluster features. a support vector machine and a neural network can be used to build compact and run-time efficient models. a cluster is classified as datapath if both the support vector machine and the neural network indicate that it is datapath-like. the cluster features may include automorphism generators for the cell clusters, or physical information based on the cell locations from a previous (e.g., global) placement, such as a ratio of a total cell area for a given cluster to a half-perimeter of a bounding box for the given cluster.",2015-05-26,"The title of the patent is machine-learning based datapath extraction and its abstract is a datapath extraction tool uses machine-learning models to selectively classify clusters of cells in an integrated circuit design as either datapath logic or non-datapath logic based on cluster features. a support vector machine and a neural network can be used to build compact and run-time efficient models. a cluster is classified as datapath if both the support vector machine and the neural network indicate that it is datapath-like. the cluster features may include automorphism generators for the cell clusters, or physical information based on the cell locations from a previous (e.g., global) placement, such as a ratio of a total cell area for a given cluster to a half-perimeter of a bounding box for the given cluster. dated 2015-05-26"
9047566,quadratic regularization for neural network with skip-layer connections,"according to one aspect of the invention, target data comprising observations is received. a neural network comprising input neurons, output neurons, hidden neurons, skip-layer connections, and non-skip-layer connections is used to analyze the target data based on an overall objective function that comprises a linear regression part, the neural network's unregularized objective function, and a regularization term. an overall optimized first vector value of a first vector and an overall optimized second vector value of a second vector are determined based on the target data and the overall objective function. the first vector comprises skip-layer weights for the skip-layer connections and output neuron biases, whereas the second vector comprises non-skip-layer weights for the non-skip-layer connections.",2015-06-02,"The title of the patent is quadratic regularization for neural network with skip-layer connections and its abstract is according to one aspect of the invention, target data comprising observations is received. a neural network comprising input neurons, output neurons, hidden neurons, skip-layer connections, and non-skip-layer connections is used to analyze the target data based on an overall objective function that comprises a linear regression part, the neural network's unregularized objective function, and a regularization term. an overall optimized first vector value of a first vector and an overall optimized second vector value of a second vector are determined based on the target data and the overall objective function. the first vector comprises skip-layer weights for the skip-layer connections and output neuron biases, whereas the second vector comprises non-skip-layer weights for the non-skip-layer connections. dated 2015-06-02"
9052896,adjusting mobile device state based on user intentions and/or identity,"in one embodiment, when a computing system is in a first state, a first set of inputs from one or more first sensors is detected. a first sensor value array is generated, and the first value array is fed as input to a first function generated by a first neural network. one or more first output values are calculated based on the first function, and a determination is made based on these first output values if a first action has occurred. if a first action has occurred, a second sensor value array is generated from a second set of inputs from one or more second sensors. the second sensor value array is fed as input to a second function generated by a second neural network. one or more second output values are calculated based on the second function, and the first state is exited based on these second output values.",2015-06-09,"The title of the patent is adjusting mobile device state based on user intentions and/or identity and its abstract is in one embodiment, when a computing system is in a first state, a first set of inputs from one or more first sensors is detected. a first sensor value array is generated, and the first value array is fed as input to a first function generated by a first neural network. one or more first output values are calculated based on the first function, and a determination is made based on these first output values if a first action has occurred. if a first action has occurred, a second sensor value array is generated from a second set of inputs from one or more second sensors. the second sensor value array is fed as input to a second function generated by a second neural network. one or more second output values are calculated based on the second function, and the first state is exited based on these second output values. dated 2015-06-09"
9053431,intelligent control with hierarchical stacked neural networks,"a system and method of detecting an aberrant message is provided. an ordered set of words within the message is detected. the set of words found within the message is linked to a corresponding set of expected words, the set of expected words having semantic attributes. a set of grammatical structures represented in the message is detected, based on the ordered set of words and the semantic attributes of the corresponding set of expected words. a cognitive noise vector comprising a quantitative measure of a deviation between grammatical structures represented in the message and an expected measure of grammatical structures for a message of the type is then determined. the cognitive noise vector may be processed by higher levels of the neural network and/or an external processor.",2015-06-09,"The title of the patent is intelligent control with hierarchical stacked neural networks and its abstract is a system and method of detecting an aberrant message is provided. an ordered set of words within the message is detected. the set of words found within the message is linked to a corresponding set of expected words, the set of expected words having semantic attributes. a set of grammatical structures represented in the message is detected, based on the ordered set of words and the semantic attributes of the corresponding set of expected words. a cognitive noise vector comprising a quantitative measure of a deviation between grammatical structures represented in the message and an expected measure of grammatical structures for a message of the type is then determined. the cognitive noise vector may be processed by higher levels of the neural network and/or an external processor. dated 2015-06-09"
9063032,signal monitoring system for monitoring strain applied to a composite component,a system for estimating a strain of a component and method of estimating strain is provided. the system includes a signal generator configured to transmit a signal toward the component. a sensor is coupled to the component and configured to receive the signal and to generate a reflected signal. the system includes a fiber bragg grating filter coupled to the sensor and configured to filter the reflected signal and to generate a filtered signal. a detector is coupled to the filter and configured to convert the filtered signal to a time domain signal. the system includes an artificial neural network coupled to the detector and configured to process the time domain signal to facilitate estimating the strain of the component.,2015-06-23,The title of the patent is signal monitoring system for monitoring strain applied to a composite component and its abstract is a system for estimating a strain of a component and method of estimating strain is provided. the system includes a signal generator configured to transmit a signal toward the component. a sensor is coupled to the component and configured to receive the signal and to generate a reflected signal. the system includes a fiber bragg grating filter coupled to the sensor and configured to filter the reflected signal and to generate a filtered signal. a detector is coupled to the filter and configured to convert the filtered signal to a time domain signal. the system includes an artificial neural network coupled to the detector and configured to process the time domain signal to facilitate estimating the strain of the component. dated 2015-06-23
9064215,learning spike timing precision,"certain aspects of the present disclosure provide methods and apparatus for learning or determining delays between neuron models so that the uncertainty in input spike timing is accounted for in the margin of time between a delayed pre-synaptic input spike and a post-synaptic spike. in this manner, a neural network can correctly match patterns (even in the presence of significant jitter) and correctly distinguish between different noisy patterns. one example method generally includes determining an uncertainty associated with a first pre-synaptic spike time of a first neuron model for a pattern to be learned; and determining a delay based on the uncertainty, such that the delay added to a second pre-synaptic spike time of the first neuron model results in a causal margin of time between the delayed second pre-synaptic spike time and a post-synaptic spike time of a second neuron model.",2015-06-23,"The title of the patent is learning spike timing precision and its abstract is certain aspects of the present disclosure provide methods and apparatus for learning or determining delays between neuron models so that the uncertainty in input spike timing is accounted for in the margin of time between a delayed pre-synaptic input spike and a post-synaptic spike. in this manner, a neural network can correctly match patterns (even in the presence of significant jitter) and correctly distinguish between different noisy patterns. one example method generally includes determining an uncertainty associated with a first pre-synaptic spike time of a first neuron model for a pattern to be learned; and determining a delay based on the uncertainty, such that the delay added to a second pre-synaptic spike time of the first neuron model results in a causal margin of time between the delayed second pre-synaptic spike time and a post-synaptic spike time of a second neuron model. dated 2015-06-23"
9064498,apparatus and method for processing an audio signal for speech enhancement using a feature extraction,"an apparatus for processing an audio signal to obtain control information for a speech enhancement filter has a feature extractor for extracting at least one feature per frequency band of a plurality of frequency bands of a short-time spectral representation of a plurality of short-time spectral representations, where the at least one feature represents a spectral shape of the short-time spectral representation in the frequency band. the apparatus additionally has a feature combiner for combining the at least one feature for each frequency band using combination parameters to obtain the control information for the speech enhancement filter for a time portion of the audio signal. the feature combiner can use a neural network regression method, which is based on combination parameters determined in a training phase for the neural network.",2015-06-23,"The title of the patent is apparatus and method for processing an audio signal for speech enhancement using a feature extraction and its abstract is an apparatus for processing an audio signal to obtain control information for a speech enhancement filter has a feature extractor for extracting at least one feature per frequency band of a plurality of frequency bands of a short-time spectral representation of a plurality of short-time spectral representations, where the at least one feature represents a spectral shape of the short-time spectral representation in the frequency band. the apparatus additionally has a feature combiner for combining the at least one feature for each frequency band using combination parameters to obtain the control information for the speech enhancement filter for a time portion of the audio signal. the feature combiner can use a neural network regression method, which is based on combination parameters determined in a training phase for the neural network. dated 2015-06-23"
9070455,memristor device with resistance adjustable by moving a magnetic wall by spin transfer and use of said memristor in a neural network,"a device with adjustable resistance includes two magnetic elements separated by an insulating or semi-conductor element. the resistance of the device depends on the position of a magnetic wall in one of the magnetic elements, the magnetic wall separating two areas of said magnetic element each having a separate homogeneous direction of magnetization. the device comprises means for moving the magnetic wall in the magnetic element by applying a spin-polarized electric current, such that the resistance of the device is adjustable in a continuous range of values. the invention is useful in neuromimetic circuits, neural networks and bio-inspired computers.",2015-06-30,"The title of the patent is memristor device with resistance adjustable by moving a magnetic wall by spin transfer and use of said memristor in a neural network and its abstract is a device with adjustable resistance includes two magnetic elements separated by an insulating or semi-conductor element. the resistance of the device depends on the position of a magnetic wall in one of the magnetic elements, the magnetic wall separating two areas of said magnetic element each having a separate homogeneous direction of magnetization. the device comprises means for moving the magnetic wall in the magnetic element by applying a spin-polarized electric current, such that the resistance of the device is adjustable in a continuous range of values. the invention is useful in neuromimetic circuits, neural networks and bio-inspired computers. dated 2015-06-30"
9075941,method for optimizing electrodeposition process of a plurality of vias in wafer,"the presently claimed invention provides a method for optimizing an electrodeposition process of a plurality of vias in a wafer. instead of simulating a large number of via on the wafer for via filling, a representative via is selected with the maximum value of critical factor, which is a function of process parameters. the filling of the representative via is simulated with different sampling points to find out the filling goodness in order to find out the optimized process windows of process parameters. an optimizer is also disclosed, which either provides sampling points or reduces sampling points under artificial neural network method. calculation of filling goodness is used for evaluating via filling quality and further comparing among via fillings simulated at different sampling points. consequently, the method of present invention is able to shorten the simulation time for via filling as well as provide a process window with high accuracy.",2015-07-07,"The title of the patent is method for optimizing electrodeposition process of a plurality of vias in wafer and its abstract is the presently claimed invention provides a method for optimizing an electrodeposition process of a plurality of vias in a wafer. instead of simulating a large number of via on the wafer for via filling, a representative via is selected with the maximum value of critical factor, which is a function of process parameters. the filling of the representative via is simulated with different sampling points to find out the filling goodness in order to find out the optimized process windows of process parameters. an optimizer is also disclosed, which either provides sampling points or reduces sampling points under artificial neural network method. calculation of filling goodness is used for evaluating via filling quality and further comparing among via fillings simulated at different sampling points. consequently, the method of present invention is able to shorten the simulation time for via filling as well as provide a process window with high accuracy. dated 2015-07-07"
9076107,neural network system and uses thereof,a multifunctional neural network system for prediction which includes memory components to store previous values of data within a network. the memory components provide the system with the ability to learn relationships/patterns existent in the data over time.,2015-07-07,The title of the patent is neural network system and uses thereof and its abstract is a multifunctional neural network system for prediction which includes memory components to store previous values of data within a network. the memory components provide the system with the ability to learn relationships/patterns existent in the data over time. dated 2015-07-07
9077491,three layer cascade adaptive neural fuzzy inference system (anfis) based intelligent controller scheme and device,"intelligent technique is an effective method to perform the network resource management. a three layer cascade adaptive neural fuzzy inference system (anfis) based intelligent controller is proposed for the mobile wireless network to optimize the maximum average throughput, minimum transmit power and interference for multimedia call services. the proposed intelligent controller is designed with a three layer cascade architecture, which mainly contains an anfis rate controller (arc) in the first layer, an anfis power controller (apc) in the second layer and an anfis interference controller (aic) in the third layer. the design aim of the proposed three layer cascade anfis cognitive engine is maximizing the average throughput of the mobile wireless network, while minimizing the transmit power and interference power.",2015-07-07,"The title of the patent is three layer cascade adaptive neural fuzzy inference system (anfis) based intelligent controller scheme and device and its abstract is intelligent technique is an effective method to perform the network resource management. a three layer cascade adaptive neural fuzzy inference system (anfis) based intelligent controller is proposed for the mobile wireless network to optimize the maximum average throughput, minimum transmit power and interference for multimedia call services. the proposed intelligent controller is designed with a three layer cascade architecture, which mainly contains an anfis rate controller (arc) in the first layer, an anfis power controller (apc) in the second layer and an anfis interference controller (aic) in the third layer. the design aim of the proposed three layer cascade anfis cognitive engine is maximizing the average throughput of the mobile wireless network, while minimizing the transmit power and interference power. dated 2015-07-07"
9082078,neural processing engine and architecture using the same,a neural processing engine may perform processing within a neural processing system and/or artificial neural network. the neural processing engine may be configured to effectively and efficiently perform the type of processing required in implementing a neural processing system and/or an artificial neural network. this configuration may facilitate such processing with neural processing engines having an enhanced computational density and/or processor density with respect to conventional processing units.,2015-07-14,The title of the patent is neural processing engine and architecture using the same and its abstract is a neural processing engine may perform processing within a neural processing system and/or artificial neural network. the neural processing engine may be configured to effectively and efficiently perform the type of processing required in implementing a neural processing system and/or an artificial neural network. this configuration may facilitate such processing with neural processing engines having an enhanced computational density and/or processor density with respect to conventional processing units. dated 2015-07-14
9087301,hardware architecture for simulating a neural network of neurons,"embodiments of the invention relate to a neural network system for simulating neurons of a neural model. one embodiment comprises a memory device that maintains neuronal states for multiple neurons, a lookup table that maintains state transition information for multiple neuronal states, and a controller unit that manages the memory device. the controller unit updates a neuronal state for each neuron based on incoming spike events targeting said neuron and state transition information corresponding to said neuronal state.",2015-07-21,"The title of the patent is hardware architecture for simulating a neural network of neurons and its abstract is embodiments of the invention relate to a neural network system for simulating neurons of a neural model. one embodiment comprises a memory device that maintains neuronal states for multiple neurons, a lookup table that maintains state transition information for multiple neuronal states, and a controller unit that manages the memory device. the controller unit updates a neuronal state for each neuron based on incoming spike events targeting said neuron and state transition information corresponding to said neuronal state. dated 2015-07-21"
9091613,multi-spectral ultrasonic gas leak detector,"an ultrasonic gas leak detector is configured to discriminate the ultrasound generated by a pressurized gas leak into the atmosphere from false alarm ultrasound. an exemplary embodiment includes multiple acoustic sensors for detecting acoustic energy and providing sensor signals, including a broadband sensor and at least one narrowband sensor, and an electronic controller responsive to the sensor signals. in one exemplary embodiment, the electronic controller is configured to provide a threshold comparator function to compare a sensor signal value representative of sensed ultrasonic energy to a gas detection threshold value, and an artificial neural network (ann) function for processing signals derived from the multitude of sensor signals and applying ann coefficients configured to discriminate false alarm sources from gas leaks. an output function generates detector outputs in dependence on the threshold comparator output and the ann output.",2015-07-28,"The title of the patent is multi-spectral ultrasonic gas leak detector and its abstract is an ultrasonic gas leak detector is configured to discriminate the ultrasound generated by a pressurized gas leak into the atmosphere from false alarm ultrasound. an exemplary embodiment includes multiple acoustic sensors for detecting acoustic energy and providing sensor signals, including a broadband sensor and at least one narrowband sensor, and an electronic controller responsive to the sensor signals. in one exemplary embodiment, the electronic controller is configured to provide a threshold comparator function to compare a sensor signal value representative of sensed ultrasonic energy to a gas detection threshold value, and an artificial neural network (ann) function for processing signals derived from the multitude of sensor signals and applying ann coefficients configured to discriminate false alarm sources from gas leaks. an output function generates detector outputs in dependence on the threshold comparator output and the ann output. dated 2015-07-28"
9092726,neural network frequency control,"systems and methods for controlling frequency output of an electronic oscillator to compensate for effects of a parameter experienced by the oscillator incorporate artificial neural network processing functionality for generating correction signals. a neural network processing module includes one or more neurons which receive one or more inputs corresponding to a parameter of an electronic oscillator, such as temperature. weights are calculated and applied to inputs to the neurons of the neural network as part of a training process, wherein the weights help shape the output of the neural network processing module. the neural network may include a linear summation module configured to provide an output signal that is at least partially based on outputs of the one or more neurons.",2015-07-28,"The title of the patent is neural network frequency control and its abstract is systems and methods for controlling frequency output of an electronic oscillator to compensate for effects of a parameter experienced by the oscillator incorporate artificial neural network processing functionality for generating correction signals. a neural network processing module includes one or more neurons which receive one or more inputs corresponding to a parameter of an electronic oscillator, such as temperature. weights are calculated and applied to inputs to the neurons of the neural network as part of a training process, wherein the weights help shape the output of the neural network processing module. the neural network may include a linear summation module configured to provide an output signal that is at least partially based on outputs of the one or more neurons. dated 2015-07-28"
9092729,trim effect compensation using an artificial neural network,"systems and methods for controlling frequency output of an electronic oscillator to compensate for effects of one or more parameters experienced by the oscillator incorporate artificial neural network processing functionality for generating correction signals. a neural network processing module includes one or more neurons which receive one or more inputs corresponding to parameters of an electronic oscillator, such as temperature and control voltage (or correction voltage). one or more sets of weights are calculated and applied to inputs to the neurons of the neural network as part of a training process, wherein the weights help shape the output of the neural network processing module. the neural network may include a linear summation module configured to provide an output signal that is at least partially based on outputs of the one or more neurons.",2015-07-28,"The title of the patent is trim effect compensation using an artificial neural network and its abstract is systems and methods for controlling frequency output of an electronic oscillator to compensate for effects of one or more parameters experienced by the oscillator incorporate artificial neural network processing functionality for generating correction signals. a neural network processing module includes one or more neurons which receive one or more inputs corresponding to parameters of an electronic oscillator, such as temperature and control voltage (or correction voltage). one or more sets of weights are calculated and applied to inputs to the neurons of the neural network as part of a training process, wherein the weights help shape the output of the neural network processing module. the neural network may include a linear summation module configured to provide an output signal that is at least partially based on outputs of the one or more neurons. dated 2015-07-28"
9092730,neural network frequency control and compensation of control voltage linearity,systems and methods of using an artificial neural network processing module to compensate a control voltage and create a linear output response for an electronic oscillator to produce a target frequency. the artificial neural network processing module includes one or more neurons which receive one or more inputs corresponding to the control voltage. the artificial neural network processing module is configured to provide a correction based at least in part on the control voltage and pre-calculated dac values. the pre-calculated dac values are determined in part by predetermined or predefined pull ranges and linear control voltage transfer functions. the artificial neural network processing module can preferably achieve a control voltage tuning linearity better than 0.5% linearity over an entire tuning range of +−75 ppm.,2015-07-28,The title of the patent is neural network frequency control and compensation of control voltage linearity and its abstract is systems and methods of using an artificial neural network processing module to compensate a control voltage and create a linear output response for an electronic oscillator to produce a target frequency. the artificial neural network processing module includes one or more neurons which receive one or more inputs corresponding to the control voltage. the artificial neural network processing module is configured to provide a correction based at least in part on the control voltage and pre-calculated dac values. the pre-calculated dac values are determined in part by predetermined or predefined pull ranges and linear control voltage transfer functions. the artificial neural network processing module can preferably achieve a control voltage tuning linearity better than 0.5% linearity over an entire tuning range of +−75 ppm. dated 2015-07-28
9092737,"systems, methods, and apparatus for 3-d surface mapping, compliance mapping, and spatial registration with an array of cantilevered tactile hair or whisker sensors","systems, methods, and apparatus are provided using signals from a set of tactile sensors mounted on a surface to determine a surface topography. an example method includes receiving a set of moment and force input data from one or more identified topographies. the example method includes using a neural network to receive input from a training data set based on the first set of moment and force input data from the one or more identified topographies. network weights to be used by the neural network to produce the training data set are modified via an evolutionary algorithm that tests vectors of candidate network weights. the example method includes receiving a moment and force input from a test object surface and reconstructing the surface topology based on the neural network outputs.",2015-07-28,"The title of the patent is systems, methods, and apparatus for 3-d surface mapping, compliance mapping, and spatial registration with an array of cantilevered tactile hair or whisker sensors and its abstract is systems, methods, and apparatus are provided using signals from a set of tactile sensors mounted on a surface to determine a surface topography. an example method includes receiving a set of moment and force input data from one or more identified topographies. the example method includes using a neural network to receive input from a training data set based on the first set of moment and force input data from the one or more identified topographies. network weights to be used by the neural network to produce the training data set are modified via an evolutionary algorithm that tests vectors of candidate network weights. the example method includes receiving a moment and force input from a test object surface and reconstructing the surface topology based on the neural network outputs. dated 2015-07-28"
9095266,method for treating a patient,"a signal processing method and system combines multi-scale decomposition, such as wavelet, pre-processing together with a compression technique, such as an auto-associative artificial neural network, operating in the multi-scale decomposition domain for signal denoising and extraction. all compressions are performed in the decomposed domain. a reverse decomposition such as an inverse discrete wavelet transform is performed on the combined outputs from all the compression modules to recover a clean signal back in the time domain. a low-cost, non-drug, non-invasive, on-demand therapy braincap system and method are pharmaceutically non-intrusive to the body for the purpose of disease diagnosis, treatment therapy, and direct mind control of external devices and systems. it is based on recognizing abnormal brainwave signatures and intervenes at the earliest moment, using magnetic and/or electric stimulations to reset the brainwaves back to normality. the feedback system is self-regulatory and the treatment stops when the brainwaves return to normal.",2015-08-04,"The title of the patent is method for treating a patient and its abstract is a signal processing method and system combines multi-scale decomposition, such as wavelet, pre-processing together with a compression technique, such as an auto-associative artificial neural network, operating in the multi-scale decomposition domain for signal denoising and extraction. all compressions are performed in the decomposed domain. a reverse decomposition such as an inverse discrete wavelet transform is performed on the combined outputs from all the compression modules to recover a clean signal back in the time domain. a low-cost, non-drug, non-invasive, on-demand therapy braincap system and method are pharmaceutically non-intrusive to the body for the purpose of disease diagnosis, treatment therapy, and direct mind control of external devices and systems. it is based on recognizing abnormal brainwave signatures and intervenes at the earliest moment, using magnetic and/or electric stimulations to reset the brainwaves back to normality. the feedback system is self-regulatory and the treatment stops when the brainwaves return to normal. dated 2015-08-04"
9095303,"system and apparatus for early detection, prevention, containment or abatement of spread abnormal brain activity","a method, comprising detecting an epileptic event in a neural network, wherein the event occurs in a first node; identifying a second node; and applying a therapy to the second node or any connection. a method, comprising determining a first body index indicative of epileptic activity; monitoring a second body index; detecting an indication of activity spread, based upon at least the second body index; and taking a responsive action, such as delivering therapy, modifying therapy, logging the indication, or warning. a method, comprising detecting an epileptic event in a first node of a neural network; applying a first therapy to a first neural structure for treating the event; and applying a second therapy to a second neural structure of the patient based on an event spread proclivity to a third neural structure. a non-transitive, computer-readable storage device for storing data that when executed by a processor, perform a method.",2015-08-04,"The title of the patent is system and apparatus for early detection, prevention, containment or abatement of spread abnormal brain activity and its abstract is a method, comprising detecting an epileptic event in a neural network, wherein the event occurs in a first node; identifying a second node; and applying a therapy to the second node or any connection. a method, comprising determining a first body index indicative of epileptic activity; monitoring a second body index; detecting an indication of activity spread, based upon at least the second body index; and taking a responsive action, such as delivering therapy, modifying therapy, logging the indication, or warning. a method, comprising detecting an epileptic event in a first node of a neural network; applying a first therapy to a first neural structure for treating the event; and applying a second therapy to a second neural structure of the patient based on an event spread proclivity to a third neural structure. a non-transitive, computer-readable storage device for storing data that when executed by a processor, perform a method. dated 2015-08-04"
9098811,spiking neuron network apparatus and methods,"apparatus and methods for heterosynaptic plasticity in a spiking neural network having multiple neurons configured to process sensory input. in one exemplary approach, a heterosynaptic plasticity mechanism is configured to select alternate plasticity rules when performing neuronal updates. the selection mechanism is adapted based on recent post-synaptic activity of neighboring neurons. when neighbor activity is low, a regular stdp update rule is effectuated. when neighbor activity is high, an alternate stdp update rule, configured to reduce probability of post-synaptic spike generation by the neuron associated with the update, is used. the heterosynaptic mechanism impedes that neuron to respond to (or learn) features within the sensory input that have been detected by neighboring neurons, thereby forcing the neuron to learn a different feature or feature set. the heterosynaptic methodology advantageously introduces competition among neighboring neurons, in order to increase receptive field diversity and improve feature detection capabilities of the network.",2015-08-04,"The title of the patent is spiking neuron network apparatus and methods and its abstract is apparatus and methods for heterosynaptic plasticity in a spiking neural network having multiple neurons configured to process sensory input. in one exemplary approach, a heterosynaptic plasticity mechanism is configured to select alternate plasticity rules when performing neuronal updates. the selection mechanism is adapted based on recent post-synaptic activity of neighboring neurons. when neighbor activity is low, a regular stdp update rule is effectuated. when neighbor activity is high, an alternate stdp update rule, configured to reduce probability of post-synaptic spike generation by the neuron associated with the update, is used. the heterosynaptic mechanism impedes that neuron to respond to (or learn) features within the sensory input that have been detected by neighboring neurons, thereby forcing the neuron to learn a different feature or feature set. the heterosynaptic methodology advantageously introduces competition among neighboring neurons, in order to increase receptive field diversity and improve feature detection capabilities of the network. dated 2015-08-04"
9104977,systems and methods for predicting characteristics of an artificial heart using an artificial neural network,"a system configured to predict characteristics of an artificial heart is described. the system includes a processor and memory in electronic communication with the processor, and an artificial neural network configured to receive an input vector of a predetermined length to train the artificial neural network, produce an output vector based on the input vector, and compare the output vector with a target vector of the predetermined length. when the output vector does not match the target vector within a predetermined error rate, the network is configured to adjust at least one weight, and when the output vector matches the target vector within the predetermined error rate, the network is configured to execute the input vector to produce an estimate at least one characteristic of the artificial heart.",2015-08-11,"The title of the patent is systems and methods for predicting characteristics of an artificial heart using an artificial neural network and its abstract is a system configured to predict characteristics of an artificial heart is described. the system includes a processor and memory in electronic communication with the processor, and an artificial neural network configured to receive an input vector of a predetermined length to train the artificial neural network, produce an output vector based on the input vector, and compare the output vector with a target vector of the predetermined length. when the output vector does not match the target vector within a predetermined error rate, the network is configured to adjust at least one weight, and when the output vector matches the target vector within the predetermined error rate, the network is configured to execute the input vector to produce an estimate at least one characteristic of the artificial heart. dated 2015-08-11"
9107595,node excitation driving function measures for cerebral cortex network analysis of electroencephalograms,"methods and apparatuses for estimating brain activity of a human subject from the measurement of electroencephalograms (eeg) are disclosed. in one method, cortical neural sources in the cerebral cortex of the brain of the subject are specified. next, using a model of the human brain which treats the cortical neural sources as nodes in a cortical source network, cortical source activations are estimated from the measured electroencephalograms for each of the cortical neural sources in the network for the subject. source network modulation control signals are then determined for the subject from the cortical source activations which are assumed to correspond to control modulators in the human brain. and a network activity classification is computed from determined modulation control signals for the subject. the innovative technology may be included in an automated aiding system in the electronic aiding of tasks performed by human operators.",2015-08-18,"The title of the patent is node excitation driving function measures for cerebral cortex network analysis of electroencephalograms and its abstract is methods and apparatuses for estimating brain activity of a human subject from the measurement of electroencephalograms (eeg) are disclosed. in one method, cortical neural sources in the cerebral cortex of the brain of the subject are specified. next, using a model of the human brain which treats the cortical neural sources as nodes in a cortical source network, cortical source activations are estimated from the measured electroencephalograms for each of the cortical neural sources in the network for the subject. source network modulation control signals are then determined for the subject from the cortical source activations which are assumed to correspond to control modulators in the human brain. and a network activity classification is computed from determined modulation control signals for the subject. the innovative technology may be included in an automated aiding system in the electronic aiding of tasks performed by human operators. dated 2015-08-18"
9108009,system and method for diagnosis and treatment of a breathing pattern of a patient,"described is a system including a sensor and a processing arrangement. the sensor measures data corresponding to a patient's breathing patterns. the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of a rem sleep state. in another embodiment, the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of one of the following states: (i) a wake state and (ii) a sleep state. in another embodiment, a neural network analyzes the data to determine whether the breathing patterns are indicative of one of the following states: (i) a rem sleep state, (ii) a wake state and (iii) a sleep state. in another embodiment, the processing arrangement analyzes the data to determine whether the breathing pattern is indicative of an arousal.",2015-08-18,"The title of the patent is system and method for diagnosis and treatment of a breathing pattern of a patient and its abstract is described is a system including a sensor and a processing arrangement. the sensor measures data corresponding to a patient's breathing patterns. the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of a rem sleep state. in another embodiment, the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of one of the following states: (i) a wake state and (ii) a sleep state. in another embodiment, a neural network analyzes the data to determine whether the breathing patterns are indicative of one of the following states: (i) a rem sleep state, (ii) a wake state and (iii) a sleep state. in another embodiment, the processing arrangement analyzes the data to determine whether the breathing pattern is indicative of an arousal. dated 2015-08-18"
9111215,conditional plasticity spiking neuron network apparatus and methods,"apparatus and methods for conditional plasticity in a neural network. in one approach, conditional plasticity mechanism is configured to select alternate plasticity rules when performing connection updates. the selection mechanism is adapted based on a comparison of actual connection efficiency and target efficiency. for instance, when actual efficiency is below the target value, the stdp rule may be modulated to increase long term potentiation. similarly, when actual efficiency is above the target value, the stdp rule may be modulated to increase long term connection depression. the conditional plasticity mechanism dynamically adjusts connection efficacy, and prevents uncontrolled increase of connection weights, thereby improving network operation when processing information of a varying nature.",2015-08-18,"The title of the patent is conditional plasticity spiking neuron network apparatus and methods and its abstract is apparatus and methods for conditional plasticity in a neural network. in one approach, conditional plasticity mechanism is configured to select alternate plasticity rules when performing connection updates. the selection mechanism is adapted based on a comparison of actual connection efficiency and target efficiency. for instance, when actual efficiency is below the target value, the stdp rule may be modulated to increase long term potentiation. similarly, when actual efficiency is above the target value, the stdp rule may be modulated to increase long term connection depression. the conditional plasticity mechanism dynamically adjusts connection efficacy, and prevents uncontrolled increase of connection weights, thereby improving network operation when processing information of a varying nature. dated 2015-08-18"
9111222,method and apparatus for switching the binary state of a location in memory in a probabilistic manner to store synaptic weights of a neural network,certain aspects of the present disclosure support a technique for utilizing a memory in probabilistic manner to store information about weights of synapses of a neural network.,2015-08-18,The title of the patent is method and apparatus for switching the binary state of a location in memory in a probabilistic manner to store synaptic weights of a neural network and its abstract is certain aspects of the present disclosure support a technique for utilizing a memory in probabilistic manner to store information about weights of synapses of a neural network. dated 2015-08-18
9117169,methods and apparatuses for modeling shale characteristics in wellbore servicing fluids using an artificial neural network,"an apparatus and method for determining a formation/fluid interaction of a target formation and a target drilling fluid is described herein. the method may include training an artificial neural network using a training data set. the training data set may include a formation characteristic of a source formation and a fluid characteristic of a source drilling fluid and experimental data on source formation/fluid interaction. once the artificial neural network is trained, a formation characteristic of the target formation and fluid characteristic of target drilling fluid may be input. the formation characteristic of the target formation may correspond to the formation characteristic of the source formation. the fluid characteristic of the target drilling fluid may correspond to the fluid characteristic of the source drilling fluid. a formation/fluid interaction of the target formation and the target drilling fluid may be determined using a value output by the artificial neural network.",2015-08-25,"The title of the patent is methods and apparatuses for modeling shale characteristics in wellbore servicing fluids using an artificial neural network and its abstract is an apparatus and method for determining a formation/fluid interaction of a target formation and a target drilling fluid is described herein. the method may include training an artificial neural network using a training data set. the training data set may include a formation characteristic of a source formation and a fluid characteristic of a source drilling fluid and experimental data on source formation/fluid interaction. once the artificial neural network is trained, a formation characteristic of the target formation and fluid characteristic of target drilling fluid may be input. the formation characteristic of the target formation may correspond to the formation characteristic of the source formation. the fluid characteristic of the target drilling fluid may correspond to the fluid characteristic of the source drilling fluid. a formation/fluid interaction of the target formation and the target drilling fluid may be determined using a value output by the artificial neural network. dated 2015-08-25"
9120356,load estimation system and method for a vehicle tire,"estimating a load bearing on a vehicle tire includes an inflation pressure measuring sensor for measuring tire inflation pressure and generating a measured tire inflation pressure signal; a deformation measuring sensor mounted in a tire region of the vehicle tire, the deformation measuring sensor in the form of a piezoelectric bending sensor generating a deformation signal estimating a length of a tire contact patch length as the senor rolls through a tire footprint. tire rolling speed is estimated from the deformation signal and a peak to peak amplitude variation within the deformation signal is detected and measured. an artificial neural network receives the tire rolling speed estimation, the contact patch length estimation, the amplitude variation within the deformation signal, and the measured inflation pressure of the tire. the artificial neural network adaptively interprets the input data and generates an output load estimation based thereon.",2015-09-01,"The title of the patent is load estimation system and method for a vehicle tire and its abstract is estimating a load bearing on a vehicle tire includes an inflation pressure measuring sensor for measuring tire inflation pressure and generating a measured tire inflation pressure signal; a deformation measuring sensor mounted in a tire region of the vehicle tire, the deformation measuring sensor in the form of a piezoelectric bending sensor generating a deformation signal estimating a length of a tire contact patch length as the senor rolls through a tire footprint. tire rolling speed is estimated from the deformation signal and a peak to peak amplitude variation within the deformation signal is detected and measured. an artificial neural network receives the tire rolling speed estimation, the contact patch length estimation, the amplitude variation within the deformation signal, and the measured inflation pressure of the tire. the artificial neural network adaptively interprets the input data and generates an output load estimation based thereon. dated 2015-09-01"
9125590,"medical ventilator capable of early detecting and recognizing types of pneumonia, gas recognition chip, and method for recognizing gas thereof","a medical ventilator capable of early detecting and recognizing types of pneumonia, a gas recognition chip, and a method for recognizing gas thereof are disclosed. the gas recognition chip of the medical ventilator comprises a sensor array, a sensor interface circuit, a stochastic neural network chip, a memory and a microcontroller. the sensor array receives a plurality of multiple types of gases to produce odor signals corresponding to each type of gas. the sensor interface circuit analyzes the odor signals to produce gas pattern signals corresponding to each type of gas. the stochastic neural network chip amplifies the differences between the gas pattern signals and performs dimensional reduction on the gas pattern signals to aid the analysis. the memory stores training data. the microcontroller performs a mixed gas recognizing algorithm to early detect and recognize the type of the pneumonia according to the gas training data.",2015-09-08,"The title of the patent is medical ventilator capable of early detecting and recognizing types of pneumonia, gas recognition chip, and method for recognizing gas thereof and its abstract is a medical ventilator capable of early detecting and recognizing types of pneumonia, a gas recognition chip, and a method for recognizing gas thereof are disclosed. the gas recognition chip of the medical ventilator comprises a sensor array, a sensor interface circuit, a stochastic neural network chip, a memory and a microcontroller. the sensor array receives a plurality of multiple types of gases to produce odor signals corresponding to each type of gas. the sensor interface circuit analyzes the odor signals to produce gas pattern signals corresponding to each type of gas. the stochastic neural network chip amplifies the differences between the gas pattern signals and performs dimensional reduction on the gas pattern signals to aid the analysis. the memory stores training data. the microcontroller performs a mixed gas recognizing algorithm to early detect and recognize the type of the pneumonia according to the gas training data. dated 2015-09-08"
9129190,identifying objects in images,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for identifying objects in images. one of the methods includes obtaining a first training image; down-sampling the first training image to generate a low-resolution first training image; processing the low-resolution first training image using a first neural network to generate a plurality of features of the low-resolution first training image and first scores for the low-resolution first training image; processing the first scores and the features of the low-resolution first training image using an initial patch locator neural network to generate an initial location of an initial patch of the first training image; locally perturbing the initial location to select an adjusted location for the initial patch of the first training image; and updating the current values of the parameters of the initial patch locator neural network to generate updated values using the adjusted location.",2015-09-08,"The title of the patent is identifying objects in images and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for identifying objects in images. one of the methods includes obtaining a first training image; down-sampling the first training image to generate a low-resolution first training image; processing the low-resolution first training image using a first neural network to generate a plurality of features of the low-resolution first training image and first scores for the low-resolution first training image; processing the first scores and the features of the low-resolution first training image using an initial patch locator neural network to generate an initial location of an initial patch of the first training image; locally perturbing the initial location to select an adjusted location for the initial patch of the first training image; and updating the current values of the parameters of the initial patch locator neural network to generate updated values using the adjusted location. dated 2015-09-08"
9129218,intelligent control with hierarchical stacked neural networks,"an intelligent control system based on an explicit model of cognitive development (table 1) performs high-level functions. it comprises up to o hierarchically stacked neural networks, nm, . . . , nm+(o−1), where m denotes the stage/order tasks performed in the first neural network, nm, and o denotes the highest stage/order tasks performed in the highest-level neural network. the type of processing actions performed in a network, nm, corresponds to the complexity for stage/order m. thus n1 performs tasks at the level corresponding to stage/order 1. n5 processes information at the level corresponding to stage/order 5. stacked neural networks begin and end at any stage/order, but information must be processed by each stage in ascending order sequence. stages/orders cannot be skipped. each neural network in a stack may use different architectures, interconnections, algorithms, and training methods, depending on the stage/order of the neural network and the type of intelligent control system implemented.",2015-09-08,"The title of the patent is intelligent control with hierarchical stacked neural networks and its abstract is an intelligent control system based on an explicit model of cognitive development (table 1) performs high-level functions. it comprises up to o hierarchically stacked neural networks, nm, . . . , nm+(o−1), where m denotes the stage/order tasks performed in the first neural network, nm, and o denotes the highest stage/order tasks performed in the highest-level neural network. the type of processing actions performed in a network, nm, corresponds to the complexity for stage/order m. thus n1 performs tasks at the level corresponding to stage/order 1. n5 processes information at the level corresponding to stage/order 5. stacked neural networks begin and end at any stage/order, but information must be processed by each stage in ascending order sequence. stages/orders cannot be skipped. each neural network in a stack may use different architectures, interconnections, algorithms, and training methods, depending on the stage/order of the neural network and the type of intelligent control system implemented. dated 2015-09-08"
9129221,spiking neural network feedback apparatus and methods,"in one approach, spiking neurons receive sensory stimulus and context signal that correspond to the same context. when the stimulus provides sufficient excitation, neurons generate response. context connections are adjusted according to inverse spike-timing dependent plasticity. when the context signal precedes the post synaptic spike, context synaptic connections are depressed. conversely, whenever the context signal follows the post synaptic spike, the connections are potentiated. the inverse stdp connection adjustment ensures precise control of feedback-induced firing, eliminates runaway positive feedback loops, and enables self-stabilizing network operation. in another aspect, the connection adjustment methodology facilitates robust context switching when processing visual information. when a context (such an object) becomes intermittently absent, prior context connection potentiation enables firing for a period of time. if the object remains absent, the connection becomes depressed thereby preventing further firing.",2015-09-08,"The title of the patent is spiking neural network feedback apparatus and methods and its abstract is in one approach, spiking neurons receive sensory stimulus and context signal that correspond to the same context. when the stimulus provides sufficient excitation, neurons generate response. context connections are adjusted according to inverse spike-timing dependent plasticity. when the context signal precedes the post synaptic spike, context synaptic connections are depressed. conversely, whenever the context signal follows the post synaptic spike, the connections are potentiated. the inverse stdp connection adjustment ensures precise control of feedback-induced firing, eliminates runaway positive feedback loops, and enables self-stabilizing network operation. in another aspect, the connection adjustment methodology facilitates robust context switching when processing visual information. when a context (such an object) becomes intermittently absent, prior context connection potentiation enables firing for a period of time. if the object remains absent, the connection becomes depressed thereby preventing further firing. dated 2015-09-08"
9129223,sound localization with artificial neural network,the location of a sound within a given spatial volume may be used in applications such as augmented reality environments. an artificial neural network processes time-difference-of-arrival data (tdoa) from a known microphone array to determine a spatial location of the sound. the neural network may be located locally or available as a cloud service. the artificial neural network is trained with perturbed and non-perturbed tdoa data.,2015-09-08,The title of the patent is sound localization with artificial neural network and its abstract is the location of a sound within a given spatial volume may be used in applications such as augmented reality environments. an artificial neural network processes time-difference-of-arrival data (tdoa) from a known microphone array to determine a spatial location of the sound. the neural network may be located locally or available as a cloud service. the artificial neural network is trained with perturbed and non-perturbed tdoa data. dated 2015-09-08
9135103,hybrid memory failure bitmap classification,"aspects of the invention relate to techniques for classifying memory failure bitmaps using both rule-based classification and artificial neural network-based classification methods. the rule-based classification method employs classification rules comprising those for global failure patterns. the artificial neural network-based classification method classifies local failure patterns. one of the artificial neural network models is the kohonen self-organizing map model. the input vector for a failure pattern may contain four elements: pattern aspect ratio, failing bit ratio, dominant failing column number and dominant failing row number.",2015-09-15,"The title of the patent is hybrid memory failure bitmap classification and its abstract is aspects of the invention relate to techniques for classifying memory failure bitmaps using both rule-based classification and artificial neural network-based classification methods. the rule-based classification method employs classification rules comprising those for global failure patterns. the artificial neural network-based classification method classifies local failure patterns. one of the artificial neural network models is the kohonen self-organizing map model. the input vector for a failure pattern may contain four elements: pattern aspect ratio, failing bit ratio, dominant failing column number and dominant failing row number. dated 2015-09-15"
9141877,method for context aware text recognition,"a method for context-aware text recognition employing two neuromorphic computing models, auto-associative neural network and cogent confabulation. the neural network model performs the character recognition from input image and produces one or more candidates for each character in the text image input. the confabulation models perform the context-aware text extraction and completion, based on the character recognition outputs and the word and sentence knowledge bases.",2015-09-22,"The title of the patent is method for context aware text recognition and its abstract is a method for context-aware text recognition employing two neuromorphic computing models, auto-associative neural network and cogent confabulation. the neural network model performs the character recognition from input image and produces one or more candidates for each character in the text image input. the confabulation models perform the context-aware text extraction and completion, based on the character recognition outputs and the word and sentence knowledge bases. dated 2015-09-22"
9147032,machine-learning based datapath extraction,"a datapath extraction tool uses machine-learning models to selectively classify clusters of cells in an integrated circuit design as either datapath logic or non-datapath logic based on cluster features. a support vector machine and a neural network can be used to build compact and run-time efficient models. a cluster is classified as datapath if both the support vector machine and the neural network indicate that it is datapath-like. the cluster features may include automorphism generators for the cell clusters, or physical information based on the cell locations from a previous (e.g., global) placement, such as a ratio of a total cell area for a given cluster to a half-perimeter of a bounding box for the given cluster.",2015-09-29,"The title of the patent is machine-learning based datapath extraction and its abstract is a datapath extraction tool uses machine-learning models to selectively classify clusters of cells in an integrated circuit design as either datapath logic or non-datapath logic based on cluster features. a support vector machine and a neural network can be used to build compact and run-time efficient models. a cluster is classified as datapath if both the support vector machine and the neural network indicate that it is datapath-like. the cluster features may include automorphism generators for the cell clusters, or physical information based on the cell locations from a previous (e.g., global) placement, such as a ratio of a total cell area for a given cluster to a half-perimeter of a bounding box for the given cluster. dated 2015-09-29"
9147154,classifying resources using a deep network,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for scoring concept terms using a deep network. one of the methods includes receiving an input comprising a plurality of features of a resource, wherein each feature is a value of a respective attribute of the resource; processing each of the features using a respective embedding function to generate one or more numeric values; processing the numeric values using one or more neural network layers to generate an alternative representation of the features, wherein processing the floating point values comprises applying one or more non-linear transformations to the floating point values; and processing the alternative representation of the input using a classifier to generate a respective category score for each category in a pre-determined set of categories, wherein each of the respective category scores measure a predicted likelihood that the resource belongs to the corresponding category.",2015-09-29,"The title of the patent is classifying resources using a deep network and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for scoring concept terms using a deep network. one of the methods includes receiving an input comprising a plurality of features of a resource, wherein each feature is a value of a respective attribute of the resource; processing each of the features using a respective embedding function to generate one or more numeric values; processing the numeric values using one or more neural network layers to generate an alternative representation of the features, wherein processing the floating point values comprises applying one or more non-linear transformations to the floating point values; and processing the alternative representation of the input using a classifier to generate a respective category score for each category in a pre-determined set of categories, wherein each of the respective category scores measure a predicted likelihood that the resource belongs to the corresponding category. dated 2015-09-29"
9147156,apparatus and methods for synaptic update in a pulse-coded network,"apparatus and methods for efficient synaptic update in a network such as a spiking neural network. in one embodiment, the post-synaptic updates, in response to generation of a post-synaptic pulse by a post-synaptic unit, are delayed until a subsequent pre-synaptic pulse is received by the unit. pre-synaptic updates are performed first following by the post-synaptic update, thus ensuring synaptic connection status is up-to-date. the delay update mechanism is used in conjunction with system “flush” events in order to ensure accurate network operation, and prevent loss of information under a variety of pre-synaptic and post-synaptic unit firing rates. a large network partition mechanism is used in one variant with network processing apparatus in order to enable processing of network signals in a limited functionality embedded hardware environment.",2015-09-29,"The title of the patent is apparatus and methods for synaptic update in a pulse-coded network and its abstract is apparatus and methods for efficient synaptic update in a network such as a spiking neural network. in one embodiment, the post-synaptic updates, in response to generation of a post-synaptic pulse by a post-synaptic unit, are delayed until a subsequent pre-synaptic pulse is received by the unit. pre-synaptic updates are performed first following by the post-synaptic update, thus ensuring synaptic connection status is up-to-date. the delay update mechanism is used in conjunction with system “flush” events in order to ensure accurate network operation, and prevent loss of information under a variety of pre-synaptic and post-synaptic unit firing rates. a large network partition mechanism is used in one variant with network processing apparatus in order to enable processing of network signals in a limited functionality embedded hardware environment. dated 2015-09-29"
9147159,extracting predictive segments from sampled data,"a system and method is disclosed which predicts the relative occurrence or presence of an event or item based on sample data consisting of samples which contain and samples which do not contain the event or item. the samples also consist of any number of descriptive attributes, which may be continuous variables, binary variables, or categorical variables. given the sampled data, the system automatically creates statistically optimal segments from which a functional input/output relationship can be derived. these segments can either be used directly in the form of a lookup table or in some cases as input data to a secondary modeling system such as a linear regression module, a neural network, or other predictive system.",2015-09-29,"The title of the patent is extracting predictive segments from sampled data and its abstract is a system and method is disclosed which predicts the relative occurrence or presence of an event or item based on sample data consisting of samples which contain and samples which do not contain the event or item. the samples also consist of any number of descriptive attributes, which may be continuous variables, binary variables, or categorical variables. given the sampled data, the system automatically creates statistically optimal segments from which a functional input/output relationship can be derived. these segments can either be used directly in the form of a lookup table or in some cases as input data to a secondary modeling system such as a linear regression module, a neural network, or other predictive system. dated 2015-09-29"
9152827,apparatus for performing matrix vector multiplication approximation using crossbar arrays of resistive memory devices,"an apparatus that performs the mathematical matrix-vector multiplication approximation operations using crossbar arrays of resistive memory devices (e.g. memristor, resistive random-access memory, spintronics, etc.). a crossbar array formed by resistive memory devices serves as a memory array that stores the coefficients of a matrix. combined with input and output analog circuits, the crossbar array system realizes the method of performing matrix-vector multiplication approximation operations with significant performance, area and energy advantages over existing methods and designs. this invention also includes an extended method that realizes the auto-associative neural network recall function using the resistive memory crossbar architecture.",2015-10-06,"The title of the patent is apparatus for performing matrix vector multiplication approximation using crossbar arrays of resistive memory devices and its abstract is an apparatus that performs the mathematical matrix-vector multiplication approximation operations using crossbar arrays of resistive memory devices (e.g. memristor, resistive random-access memory, spintronics, etc.). a crossbar array formed by resistive memory devices serves as a memory array that stores the coefficients of a matrix. combined with input and output analog circuits, the crossbar array system realizes the method of performing matrix-vector multiplication approximation operations with significant performance, area and energy advantages over existing methods and designs. this invention also includes an extended method that realizes the auto-associative neural network recall function using the resistive memory crossbar architecture. dated 2015-10-06"
9152916,multi-compartment neurons with neural cores,embodiments of the invention provide a neural core circuit comprising a synaptic interconnect network including plural electronic synapses for interconnecting one or more source electronic neurons with one or more target electronic neurons. the interconnect network further includes multiple axon paths and multiple dendrite paths. each synapse is at a cross-point junction of the interconnect network between a dendrite path and an axon path. the core circuit further comprises a routing module maintaining routing information. the routing module routes output from a source electronic neuron to one or more selected axon paths. each synapse provides a configurable level of signal conduction from an axon path of a source electronic neuron to a dendrite path of a target electronic neuron.,2015-10-06,The title of the patent is multi-compartment neurons with neural cores and its abstract is embodiments of the invention provide a neural core circuit comprising a synaptic interconnect network including plural electronic synapses for interconnecting one or more source electronic neurons with one or more target electronic neurons. the interconnect network further includes multiple axon paths and multiple dendrite paths. each synapse is at a cross-point junction of the interconnect network between a dendrite path and an axon path. the core circuit further comprises a routing module maintaining routing information. the routing module routes output from a source electronic neuron to one or more selected axon paths. each synapse provides a configurable level of signal conduction from an axon path of a source electronic neuron to a dendrite path of a target electronic neuron. dated 2015-10-06
9153230,mobile speech recognition hardware accelerator,a method for executing a mobile speech recognition software application based on a multi-layer neural network model includes providing to a hardware accelerator in the mobile device to classify one or more frames of an audio signal. the hardware accelerator includes a multiplier-accumulator (mac) unit to perform matrix multiplication operations involved in computing the neural network output.,2015-10-06,The title of the patent is mobile speech recognition hardware accelerator and its abstract is a method for executing a mobile speech recognition software application based on a multi-layer neural network model includes providing to a hardware accelerator in the mobile device to classify one or more frames of an audio signal. the hardware accelerator includes a multiplier-accumulator (mac) unit to perform matrix multiplication operations involved in computing the neural network output. dated 2015-10-06
9153231,adaptive neural network speech recognition models,"neural networks may be used in certain automatic speech recognition systems. to improve performance of these neural networks, they may be updated/retrained during run time by training the neural network based on the output of a speech recognition system or based on the output of the neural networks themselves. the outputs may include weighted outputs, lattices, weighted n-best lists, or the like. the neural networks may be acoustic model neural networks or language model neural networks. the neural networks may be retrained after each pass through the network, after each utterance, or in varying time scales.",2015-10-06,"The title of the patent is adaptive neural network speech recognition models and its abstract is neural networks may be used in certain automatic speech recognition systems. to improve performance of these neural networks, they may be updated/retrained during run time by training the neural network based on the output of a speech recognition system or based on the output of the neural networks themselves. the outputs may include weighted outputs, lattices, weighted n-best lists, or the like. the neural networks may be acoustic model neural networks or language model neural networks. the neural networks may be retrained after each pass through the network, after each utterance, or in varying time scales. dated 2015-10-06"
9155861,neural drug delivery system with fluidic threads,"a neural drug delivery system with fluidic threads implantable into tissue, including: a plurality of fluid delivery conduits configured to transport fluid and having an array of fluid delivery ports through which the fluid is selectively released; a plurality of port gates each including a mesh structure coupled to a corresponding fluid delivery port and coated with an electroactive polymer; a voltage source providing a conductive signal; and an interconnect network that carries the conductive signal to the port gates. in response of the electroactive polymer to the conductive signal, each port gate is selectively operable between a closed mode that prevents transfer of fluid through its corresponding fluid delivery port to the tissue, and an open mode that allows transfer of the fluid through its corresponding fluid delivery port to the tissue.",2015-10-13,"The title of the patent is neural drug delivery system with fluidic threads and its abstract is a neural drug delivery system with fluidic threads implantable into tissue, including: a plurality of fluid delivery conduits configured to transport fluid and having an array of fluid delivery ports through which the fluid is selectively released; a plurality of port gates each including a mesh structure coupled to a corresponding fluid delivery port and coated with an electroactive polymer; a voltage source providing a conductive signal; and an interconnect network that carries the conductive signal to the port gates. in response of the electroactive polymer to the conductive signal, each port gate is selectively operable between a closed mode that prevents transfer of fluid through its corresponding fluid delivery port to the tissue, and an open mode that allows transfer of the fluid through its corresponding fluid delivery port to the tissue. dated 2015-10-13"
9159137,probabilistic neural network based moving object detection method and an apparatus using the same,"the present disclosure proposes a method of moving object detection in variable bit-rate video steams based on probabilistic neural networks, and the method features a background generation module and a moving object detection module. the background generation module produces a model of background images which express properties of variable bit-rate video streams. the moving object detection module distinguishes a moving object in both low and high bit-rate video steams in an efficient manner. the detection result is generated by calculating the output value of the probabilistic neural networks.",2015-10-13,"The title of the patent is probabilistic neural network based moving object detection method and an apparatus using the same and its abstract is the present disclosure proposes a method of moving object detection in variable bit-rate video steams based on probabilistic neural networks, and the method features a background generation module and a moving object detection module. the background generation module produces a model of background images which express properties of variable bit-rate video streams. the moving object detection module distinguishes a moving object in both low and high bit-rate video steams in an efficient manner. the detection result is generated by calculating the output value of the probabilistic neural networks. dated 2015-10-13"
9165213,"information processing apparatus, information processing method, and program","an information processing apparatus includes a network learning portion that performs learning of an appearance/position recognition network by constraining first to third weights and using a learning image, wherein the appearance/position recognition network has a foreground layer including a position node, a background layer including a background node, and an image layer including a pixel node, and is a neural network in which the position node, the background node, and the pixel node are connected to each other, and wherein the first weight is a connection weight between the position node and the pixel node, the second weight is a connection weight between the position node and the background node, and the third weight is a connection weight between the background node and the pixel node.",2015-10-20,"The title of the patent is information processing apparatus, information processing method, and program and its abstract is an information processing apparatus includes a network learning portion that performs learning of an appearance/position recognition network by constraining first to third weights and using a learning image, wherein the appearance/position recognition network has a foreground layer including a position node, a background layer including a background node, and an image layer including a pixel node, and is a neural network in which the position node, the background node, and the pixel node are connected to each other, and wherein the first weight is a connection weight between the position node and the pixel node, the second weight is a connection weight between the position node and the background node, and the third weight is a connection weight between the background node and the pixel node. dated 2015-10-20"
9165219,image distortion correction method and image distortion correction device using the same,"an image distortion correction method and an image distortion correction device are provided. the image distortion correction method uses a neural network model to perform a correcting operation on an original image so as to obtain a correction image with plural correction points. firstly, a position coordinate of the correction point is inputted into the neural network model, so that a first direction coordinate correction amount is outputted from the neural network model. then, the position coordinate of the correction point is inputted into the neural network model, so that a second direction coordinate correction amount is outputted from the neural network model. afterwards, a pixel value of the original image corresponding to the first direction coordinate correction amount and the second direction coordinate correction amount is used as a pixel value of the correction point.",2015-10-20,"The title of the patent is image distortion correction method and image distortion correction device using the same and its abstract is an image distortion correction method and an image distortion correction device are provided. the image distortion correction method uses a neural network model to perform a correcting operation on an original image so as to obtain a correction image with plural correction points. firstly, a position coordinate of the correction point is inputted into the neural network model, so that a first direction coordinate correction amount is outputted from the neural network model. then, the position coordinate of the correction point is inputted into the neural network model, so that a second direction coordinate correction amount is outputted from the neural network model. afterwards, a pixel value of the original image corresponding to the first direction coordinate correction amount and the second direction coordinate correction amount is used as a pixel value of the correction point. dated 2015-10-20"
9165243,tensor deep stacked neural network,"a tensor deep stacked neural (t-dsn) network for obtaining predictions for discriminative modeling problems. the t-dsn network and method use bilinear modeling with a tensor representation to map a hidden layer to the predication layer. the t-dsn network is constructed by stacking blocks of a single hidden layer tensor neural network (shltnn) on top of each other. the single hidden layer for each block then is separated or divided into a plurality of two or more sections. in some embodiments, the hidden layer is separated into a first hidden layer section and a second hidden layer section. these multiple sections of the hidden layer are combined using a product operator to obtain an implicit hidden layer having a single section. in some embodiments the product operator is a khatri-rao product. a prediction is made using the implicit hidden layer and weights, and the output prediction layer is consequently obtained.",2015-10-20,"The title of the patent is tensor deep stacked neural network and its abstract is a tensor deep stacked neural (t-dsn) network for obtaining predictions for discriminative modeling problems. the t-dsn network and method use bilinear modeling with a tensor representation to map a hidden layer to the predication layer. the t-dsn network is constructed by stacking blocks of a single hidden layer tensor neural network (shltnn) on top of each other. the single hidden layer for each block then is separated or divided into a plurality of two or more sections. in some embodiments, the hidden layer is separated into a first hidden layer section and a second hidden layer section. these multiple sections of the hidden layer are combined using a product operator to obtain an implicit hidden layer having a single section. in some embodiments the product operator is a khatri-rao product. a prediction is made using the implicit hidden layer and weights, and the output prediction layer is consequently obtained. dated 2015-10-20"
9165245,apparatus and method for partial evaluation of synaptic updates based on system events,"apparatus and methods for partial evaluation of synaptic updates in neural networks. in one embodiment, a pre-synaptic unit is connected to a several post synaptic units via communication channels. information related to a plurality of post-synaptic pulses generated by the post-synaptic units is stored by the network in response to a system event. synaptic channel updates are performed by the network using the time intervals between a pre-synaptic pulse, which is being generated prior to the system event, and at least a portion of the plurality of the post synaptic pulses. the system event enables removal of the information related to the portion of the post-synaptic pulses from the storage device. a shared memory block within the storage device is used to store data related to post-synaptic pulses generated by different post-synaptic nodes. this configuration enables memory use optimization of post-synaptic units with different firing rates.",2015-10-20,"The title of the patent is apparatus and method for partial evaluation of synaptic updates based on system events and its abstract is apparatus and methods for partial evaluation of synaptic updates in neural networks. in one embodiment, a pre-synaptic unit is connected to a several post synaptic units via communication channels. information related to a plurality of post-synaptic pulses generated by the post-synaptic units is stored by the network in response to a system event. synaptic channel updates are performed by the network using the time intervals between a pre-synaptic pulse, which is being generated prior to the system event, and at least a portion of the plurality of the post synaptic pulses. the system event enables removal of the information related to the portion of the post-synaptic pulses from the storage device. a shared memory block within the storage device is used to store data related to post-synaptic pulses generated by different post-synaptic nodes. this configuration enables memory use optimization of post-synaptic units with different firing rates. dated 2015-10-20"
9168018,system and method for classifying a heart sound,"a method and system for electronically classifying a pre-processed heart sound signal of a patient as functional (normal) or pathological is provided. the pre-processed patient heart sound signal is segmentised and features are extracted therefrom (104) to build up a feature vector which is representative of the heart sound signal. the feature vector is then fed to a diagnostic decision support network (105) comprising a plurality of artificial neural networks, each relating to a known heart pathology, which is in turn used to conduct the classification.",2015-10-27,"The title of the patent is system and method for classifying a heart sound and its abstract is a method and system for electronically classifying a pre-processed heart sound signal of a patient as functional (normal) or pathological is provided. the pre-processed patient heart sound signal is segmentised and features are extracted therefrom (104) to build up a feature vector which is representative of the heart sound signal. the feature vector is then fed to a diagnostic decision support network (105) comprising a plurality of artificial neural networks, each relating to a known heart pathology, which is in turn used to conduct the classification. dated 2015-10-27"
9168344,system and method for diagnosis and treatment of a breathing pattern of a patient,"described is a system including a sensor and a processing arrangement. the sensor measures data corresponding to a patient's breathing patterns. the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of a rem sleep state. in another embodiment, the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of one of the following states: (i) a wake state and (ii) a sleep state. in another embodiment, a neural network analyzes the data to determine whether the breathing patterns are indicative of one of the following states: (i) a rem sleep state, (ii) a wake state and (iii) a sleep state. in another embodiment, the processing arrangement analyzes the data to determine whether the breathing pattern is indicative of an arousal.",2015-10-27,"The title of the patent is system and method for diagnosis and treatment of a breathing pattern of a patient and its abstract is described is a system including a sensor and a processing arrangement. the sensor measures data corresponding to a patient's breathing patterns. the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of a rem sleep state. in another embodiment, the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of one of the following states: (i) a wake state and (ii) a sleep state. in another embodiment, a neural network analyzes the data to determine whether the breathing patterns are indicative of one of the following states: (i) a rem sleep state, (ii) a wake state and (iii) a sleep state. in another embodiment, the processing arrangement analyzes the data to determine whether the breathing pattern is indicative of an arousal. dated 2015-10-27"
9170256,device and method for erythrocyte morphology analysis,"the disclosure provides a device and a method for performing morphological analysis for erythrocytes, wherein the method for performing morphological analysis for erythrocytes comprises: collecting a morphological image of each of cells in a sample through a charge coupled device (ccd) after amplifying the sample through an automatic microscope; segmenting and positioning the image and extracting target feature parameters after digitizing the image through an image-digital converter; isolating morphological feature parameters of each of the erythrocytes through a classifier established on the basis of the neural network, and normalizing each type of the morphological feature parameters of the erythrocytes through a feature fusion device established on the basis of fuzzy clustering; performing a statistical analysis on each type of normalized parameters obtained or performing a comprehensive statistical analysis according to a plurality of types of parameters, and expressing the result of the statistical analysis or the comprehensive statistical analysis in the form of graph or numerical table, thereby judging whether the morphology of the erythrocyte is normal. the source and property of the erythrocytes can be identified according to the detection for each type of the erythrocytes with the abnormal morphology.",2015-10-27,"The title of the patent is device and method for erythrocyte morphology analysis and its abstract is the disclosure provides a device and a method for performing morphological analysis for erythrocytes, wherein the method for performing morphological analysis for erythrocytes comprises: collecting a morphological image of each of cells in a sample through a charge coupled device (ccd) after amplifying the sample through an automatic microscope; segmenting and positioning the image and extracting target feature parameters after digitizing the image through an image-digital converter; isolating morphological feature parameters of each of the erythrocytes through a classifier established on the basis of the neural network, and normalizing each type of the morphological feature parameters of the erythrocytes through a feature fusion device established on the basis of fuzzy clustering; performing a statistical analysis on each type of normalized parameters obtained or performing a comprehensive statistical analysis according to a plurality of types of parameters, and expressing the result of the statistical analysis or the comprehensive statistical analysis in the form of graph or numerical table, thereby judging whether the morphology of the erythrocyte is normal. the source and property of the erythrocytes can be identified according to the detection for each type of the erythrocytes with the abnormal morphology. dated 2015-10-27"
9171247,system and method for fast template matching in 3d,"described is a pattern matching system for matching a test image with a 3d template. the system is initiated by generating a library of templates (each individual template is a three-dimensional array, with each pixel in the array representing a value at a particular x, y, and z coordinate). each column of pixels along one axis (e.g., z) is converted into a neural input. each neural input is fed through a neural network to establish a delayed connection between each neural input and output neuron and to generate a template neural network. separately, a test image is converted into neural inputs. the neural inputs of the test image are input the template neural network to generate output neurons. the output neurons are evaluated to identify a location of the template in the test image.",2015-10-27,"The title of the patent is system and method for fast template matching in 3d and its abstract is described is a pattern matching system for matching a test image with a 3d template. the system is initiated by generating a library of templates (each individual template is a three-dimensional array, with each pixel in the array representing a value at a particular x, y, and z coordinate). each column of pixels along one axis (e.g., z) is converted into a neural input. each neural input is fed through a neural network to establish a delayed connection between each neural input and output neuron and to generate a template neural network. separately, a test image is converted into neural inputs. the neural inputs of the test image are input the template neural network to generate output neurons. the output neurons are evaluated to identify a location of the template in the test image. dated 2015-10-27"
9171248,electronic circuit with neuromorphic architecture,"neuromorphic circuits are multi-cell networks configured to imitate the behavior of biological neural networks. a neuromorphic circuit is provided which comprises a network of neurons each identified by a neuron address in the network, each neuron being able to receive and process at least one input signal and then later emit on an output of the neuron a signal representing an event which occurs inside the neuron, and a programmable memory composed of elementary memories each associated with a respective neuron. the elementary memory, which is a memory of post-synaptic addresses and weights, comprises an activation input linked by a conductor to the output of the associated neuron to directly receive an event signal emitted by this neuron without passing through an address encoder or decoder. the post-synaptic addresses extracted from an elementary memory activated by a neuron are applied, with associated synaptic weights, as inputs to the neural network.",2015-10-27,"The title of the patent is electronic circuit with neuromorphic architecture and its abstract is neuromorphic circuits are multi-cell networks configured to imitate the behavior of biological neural networks. a neuromorphic circuit is provided which comprises a network of neurons each identified by a neuron address in the network, each neuron being able to receive and process at least one input signal and then later emit on an output of the neuron a signal representing an event which occurs inside the neuron, and a programmable memory composed of elementary memories each associated with a respective neuron. the elementary memory, which is a memory of post-synaptic addresses and weights, comprises an activation input linked by a conductor to the output of the associated neuron to directly receive an event signal emitted by this neuron without passing through an address encoder or decoder. the post-synaptic addresses extracted from an elementary memory activated by a neuron are applied, with associated synaptic weights, as inputs to the neural network. dated 2015-10-27"
9176104,predicting odor pleasantness with an electronic nose,"apparatus and method for assessing odors, comprises an electronic nose, to be applied to an odor and to output a structure identifying the odor; a neural network which maps an extracted structure to a first location on a pre-learned axis of odor pleasantness; and an output for outputting an assessment of an applied odor based on said first location. the assessment may be a prediction of how pleasant a user will consider the odor.",2015-11-03,"The title of the patent is predicting odor pleasantness with an electronic nose and its abstract is apparatus and method for assessing odors, comprises an electronic nose, to be applied to an odor and to output a structure identifying the odor; a neural network which maps an extracted structure to a first location on a pre-learned axis of odor pleasantness; and an output for outputting an assessment of an applied odor based on said first location. the assessment may be a prediction of how pleasant a user will consider the odor. dated 2015-11-03"
9177246,intelligent modular robotic apparatus and methods,"apparatus and methods for an extensible robotic device with artificial intelligence and receptive to training controls. in one implementation, a modular robotic system that allows a user to fully select the architecture and capability set of their robotic device is disclosed. the user may add/remove modules as their respective functions are required/obviated. in addition, the artificial intelligence is based on a neuronal network (e.g., spiking neural network), and a behavioral control structure that allows a user to train a robotic device in manner conceptually similar to the mode in which one goes about training a domesticated animal such as a dog or cat (e.g., a positive/negative feedback training paradigm) is used. the trainable behavior control structure is based on the artificial neural network, which simulates the neural/synaptic activity of the brain of a living organism.",2015-11-03,"The title of the patent is intelligent modular robotic apparatus and methods and its abstract is apparatus and methods for an extensible robotic device with artificial intelligence and receptive to training controls. in one implementation, a modular robotic system that allows a user to fully select the architecture and capability set of their robotic device is disclosed. the user may add/remove modules as their respective functions are required/obviated. in addition, the artificial intelligence is based on a neuronal network (e.g., spiking neural network), and a behavioral control structure that allows a user to train a robotic device in manner conceptually similar to the mode in which one goes about training a domesticated animal such as a dog or cat (e.g., a positive/negative feedback training paradigm) is used. the trainable behavior control structure is based on the artificial neural network, which simulates the neural/synaptic activity of the brain of a living organism. dated 2015-11-03"
9177550,conservatively adapting a deep neural network in a recognition system,"various technologies described herein pertain to conservatively adapting a deep neural network (dnn) in a recognition system for a particular user or context. a dnn is employed to output a probability distribution over models of context-dependent units responsive to receipt of captured user input. the dnn is adapted for a particular user based upon the captured user input, wherein the adaption is undertaken conservatively such that a deviation between outputs of the adapted dnn and the unadapted dnn is constrained.",2015-11-03,"The title of the patent is conservatively adapting a deep neural network in a recognition system and its abstract is various technologies described herein pertain to conservatively adapting a deep neural network (dnn) in a recognition system for a particular user or context. a dnn is employed to output a probability distribution over models of context-dependent units responsive to receipt of captured user input. the dnn is adapted for a particular user based upon the captured user input, wherein the adaption is undertaken conservatively such that a deviation between outputs of the adapted dnn and the unadapted dnn is constrained. dated 2015-11-03"
9182473,"system, method and product for locating vehicle key using neural networks","a system, method and product for determining a vehicle key fob location. the system may include a control unit for mounting in a vehicle and configured to receive multiple signals, each representing a strength of a wireless signal transmitted between the fob and one of multiple antennas located on a vehicle, and multiple neural networks having a cascade topology. the neural networks may include a first neural network for determining one of a vehicle internal position and a vehicle external position of the fob based on the wireless signal strengths, a second neural network in communication with the first neural network for determining one of multiple vehicle interior positions of the fob based on the wireless signal strengths, and a third neural network in communication with the first neural network for determining one of multiple vehicle exterior positions of the fob based on the wireless signal strengths.",2015-11-10,"The title of the patent is system, method and product for locating vehicle key using neural networks and its abstract is a system, method and product for determining a vehicle key fob location. the system may include a control unit for mounting in a vehicle and configured to receive multiple signals, each representing a strength of a wireless signal transmitted between the fob and one of multiple antennas located on a vehicle, and multiple neural networks having a cascade topology. the neural networks may include a first neural network for determining one of a vehicle internal position and a vehicle external position of the fob based on the wireless signal strengths, a second neural network in communication with the first neural network for determining one of multiple vehicle interior positions of the fob based on the wireless signal strengths, and a third neural network in communication with the first neural network for determining one of multiple vehicle exterior positions of the fob based on the wireless signal strengths. dated 2015-11-10"
9183494,competitive bcm learning rule for identifying features,"disclosed are systems, apparatuses, and methods for implementing a competitive bcm learning rule used in a neural network. such a method includes identifying a maximally responding neuron with respect to a feature of an input signal. the maximally responding neuron is the neuron in a group that has a response with respect to the feature of the input signal that is greater than a response of each other neuron in the group. such a method also includes applying a learning rule to weaken the response of each other neuron with respect to the feature of the input signal. the learning rule may also strengthen the response of the maximally responding neuron with respect to the feature of the input signal.",2015-11-10,"The title of the patent is competitive bcm learning rule for identifying features and its abstract is disclosed are systems, apparatuses, and methods for implementing a competitive bcm learning rule used in a neural network. such a method includes identifying a maximally responding neuron with respect to a feature of an input signal. the maximally responding neuron is the neuron in a group that has a response with respect to the feature of the input signal that is greater than a response of each other neuron in the group. such a method also includes applying a learning rule to weaken the response of each other neuron with respect to the feature of the input signal. the learning rule may also strengthen the response of the maximally responding neuron with respect to the feature of the input signal. dated 2015-11-10"
9183495,structural plasticity in spiking neural networks with symmetric dual of an electronic neuron,"a neural system comprises multiple neurons interconnected via synapse devices. each neuron integrates input signals arriving on its dendrite, generates a spike in response to the integrated input signals exceeding a threshold, and sends the spike to the interconnected neurons via its axon. the system further includes multiple noruens, each noruen is interconnected via the interconnect network with those neurons that the noruen's corresponding neuron sends its axon to. each noruen integrates input spikes from connected spiking neurons and generates a spike in response to the integrated input spikes exceeding a threshold. there can be one noruen for every corresponding neuron. for a first neuron connected via its axon via a synapse to dendrite of a second neuron, a noruen corresponding to the second neuron is connected via its axon through the same synapse to dendrite of the noruen corresponding to the first neuron.",2015-11-10,"The title of the patent is structural plasticity in spiking neural networks with symmetric dual of an electronic neuron and its abstract is a neural system comprises multiple neurons interconnected via synapse devices. each neuron integrates input signals arriving on its dendrite, generates a spike in response to the integrated input signals exceeding a threshold, and sends the spike to the interconnected neurons via its axon. the system further includes multiple noruens, each noruen is interconnected via the interconnect network with those neurons that the noruen's corresponding neuron sends its axon to. each noruen integrates input spikes from connected spiking neurons and generates a spike in response to the integrated input spikes exceeding a threshold. there can be one noruen for every corresponding neuron. for a first neuron connected via its axon via a synapse to dendrite of a second neuron, a noruen corresponding to the second neuron is connected via its axon through the same synapse to dendrite of the noruen corresponding to the first neuron. dated 2015-11-10"
9185126,risk predictive engine,"a method, a device, and a storage medium provide a risk engine that calculates a level of risk stemming from a communication to access a service or an asset. the risk engine operates as a fuzzy logic neural network. the risk engine obtains parameters from the communication and applies rules to calculate the level of risk.",2015-11-10,"The title of the patent is risk predictive engine and its abstract is a method, a device, and a storage medium provide a risk engine that calculates a level of risk stemming from a communication to access a service or an asset. the risk engine operates as a fuzzy logic neural network. the risk engine obtains parameters from the communication and applies rules to calculate the level of risk. dated 2015-11-10"
9189731,structural plasticity in spiking neural networks with symmetric dual of an electronic neuron,"a neural system comprises multiple neurons interconnected via synapse devices. each neuron integrates input signals arriving on its dendrite, generates a spike in response to the integrated input signals exceeding a threshold, and sends the spike to the interconnected neurons via its axon. the system further includes multiple noruens, each noruen is interconnected via the interconnect network with those neurons that the noruen's corresponding neuron sends its axon to. each noruen integrates input spikes from connected spiking neurons and generates a spike in response to the integrated input spikes exceeding a threshold. there can be one noruen for every corresponding neuron. for a first neuron connected via its axon via a synapse to dendrite of a second neuron, a noruen corresponding to the second neuron is connected via its axon through the same synapse to dendrite of the noruen corresponding to the first neuron.",2015-11-17,"The title of the patent is structural plasticity in spiking neural networks with symmetric dual of an electronic neuron and its abstract is a neural system comprises multiple neurons interconnected via synapse devices. each neuron integrates input signals arriving on its dendrite, generates a spike in response to the integrated input signals exceeding a threshold, and sends the spike to the interconnected neurons via its axon. the system further includes multiple noruens, each noruen is interconnected via the interconnect network with those neurons that the noruen's corresponding neuron sends its axon to. each noruen integrates input spikes from connected spiking neurons and generates a spike in response to the integrated input spikes exceeding a threshold. there can be one noruen for every corresponding neuron. for a first neuron connected via its axon via a synapse to dendrite of a second neuron, a noruen corresponding to the second neuron is connected via its axon through the same synapse to dendrite of the noruen corresponding to the first neuron. dated 2015-11-17"
9189732,"method for non-supervised learning in an artificial neural network based on memristive nanodevices, and artificial neural network implementing said method",an unsupervised learning method is provided implemented in an artificial neural network based on memristive devices. it consists notably in producing an increase in the conductance of a synapse when there is temporal overlap between a pre-synaptic pulse and a post-synaptic pulse and in decreasing its conductance on receipt of a post-synaptic pulse when there is no temporal overlap with a pre-synaptic pulse.,2015-11-17,"The title of the patent is method for non-supervised learning in an artificial neural network based on memristive nanodevices, and artificial neural network implementing said method and its abstract is an unsupervised learning method is provided implemented in an artificial neural network based on memristive devices. it consists notably in producing an increase in the conductance of a synapse when there is temporal overlap between a pre-synaptic pulse and a post-synaptic pulse and in decreasing its conductance on receipt of a post-synaptic pulse when there is no temporal overlap with a pre-synaptic pulse. dated 2015-11-17"
9190053,system and method for applying a convolutional neural network to speech recognition,a system and method for applying a convolutional neural network (cnn) to speech recognition. the cnn may provide input to a hidden markov model and has at least one pair of a convolution layer and a pooling layer. the cnn operates along the frequency axis. the cnn has units that operate upon one or more local frequency bands of an acoustic signal. the cnn mitigates acoustic variation.,2015-11-17,The title of the patent is system and method for applying a convolutional neural network to speech recognition and its abstract is a system and method for applying a convolutional neural network (cnn) to speech recognition. the cnn may provide input to a hidden markov model and has at least one pair of a convolution layer and a pooling layer. the cnn operates along the frequency axis. the cnn has units that operate upon one or more local frequency bands of an acoustic signal. the cnn mitigates acoustic variation. dated 2015-11-17
9195656,multilingual prosody generation,"methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for multilingual prosody generation. in some implementations, data indicating a set of linguistic features corresponding to a text is obtained. data indicating the linguistic features and data indicating the language of the text are provided as input to a neural network that has been trained to provide output indicating prosody information for multiple languages. the neural network can be a neural network having been trained using speech in multiple languages. output indicating prosody information for the linguistic features is received from the neural network. audio data representing the text is generated using the output of the neural network.",2015-11-24,"The title of the patent is multilingual prosody generation and its abstract is methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for multilingual prosody generation. in some implementations, data indicating a set of linguistic features corresponding to a text is obtained. data indicating the linguistic features and data indicating the language of the text are provided as input to a neural network that has been trained to provide output indicating prosody information for multiple languages. the neural network can be a neural network having been trained using speech in multiple languages. output indicating prosody information for the linguistic features is received from the neural network. audio data representing the text is generated using the output of the neural network. dated 2015-11-24"
9195935,problem solving by plastic neuronal networks,"more realistic neural networks are disclosed that are able to learn to solve complex problems though a decision making network, modeled as a virtual entity foraging in a digital environment. specifically, the neural networks overcome many of the limitations in prior neural networks by using rewarded stdp bounded with rules to solve a complex problem.",2015-11-24,"The title of the patent is problem solving by plastic neuronal networks and its abstract is more realistic neural networks are disclosed that are able to learn to solve complex problems though a decision making network, modeled as a virtual entity foraging in a digital environment. specifically, the neural networks overcome many of the limitations in prior neural networks by using rewarded stdp bounded with rules to solve a complex problem. dated 2015-11-24"
9202144,regionlets with shift invariant neural patterns for object detection,systems and methods are disclosed for detecting an object in an image by determining convolutional neural network responses on the image; mapping the responses back to their spatial locations in the image; and constructing features densely extract shift invariant activations of a convolutional neural network to produce dense features for the image.,2015-12-01,The title of the patent is regionlets with shift invariant neural patterns for object detection and its abstract is systems and methods are disclosed for detecting an object in an image by determining convolutional neural network responses on the image; mapping the responses back to their spatial locations in the image; and constructing features densely extract shift invariant activations of a convolutional neural network to produce dense features for the image. dated 2015-12-01
9202464,curriculum learning for speech recognition,"methods and apparatus related to training speech recognition devices are presented. a computing device receives training samples for training a neural network to learn an acoustic speech model. a curriculum function for speech modeling can be determined. for each training sample of the training samples, a corresponding curriculum function value for the training sample can be determined using the curriculum function. the training samples can be ordered based on the corresponding curriculum function values. in some embodiments, the neural network can be trained utilizing the ordered training samples. the trained neural network can receive an input of a second plurality of samples corresponding to human speech, where the second plurality of samples differs from the training samples. in response to receiving the second plurality of samples, the trained neural network can generate a plurality of phones corresponding to the captured human speech.",2015-12-01,"The title of the patent is curriculum learning for speech recognition and its abstract is methods and apparatus related to training speech recognition devices are presented. a computing device receives training samples for training a neural network to learn an acoustic speech model. a curriculum function for speech modeling can be determined. for each training sample of the training samples, a corresponding curriculum function value for the training sample can be determined using the curriculum function. the training samples can be ordered based on the corresponding curriculum function values. in some embodiments, the neural network can be trained utilizing the ordered training samples. the trained neural network can receive an input of a second plurality of samples corresponding to human speech, where the second plurality of samples differs from the training samples. in response to receiving the second plurality of samples, the trained neural network can generate a plurality of phones corresponding to the captured human speech. dated 2015-12-01"
9208432,neural network learning and collaboration apparatus and methods,"apparatus and methods for learning and training in neural network-based devices. in one implementation, the devices each comprise multiple spiking neurons, configured to process sensory input. in one approach, alternate heterosynaptic plasticity mechanisms are used to enhance learning and field diversity within the devices. the selection of alternate plasticity rules is based on recent post-synaptic activity of neighboring neurons. apparatus and methods for simplifying training of the devices are also disclosed, including a computer-based application. a data representation of the neural network may be imaged and transferred to another computational environment, effectively copying the brain. techniques and architectures for achieve this training, storing, and distributing these data representations are also disclosed.",2015-12-08,"The title of the patent is neural network learning and collaboration apparatus and methods and its abstract is apparatus and methods for learning and training in neural network-based devices. in one implementation, the devices each comprise multiple spiking neurons, configured to process sensory input. in one approach, alternate heterosynaptic plasticity mechanisms are used to enhance learning and field diversity within the devices. the selection of alternate plasticity rules is based on recent post-synaptic activity of neighboring neurons. apparatus and methods for simplifying training of the devices are also disclosed, including a computer-based application. a data representation of the neural network may be imaged and transferred to another computational environment, effectively copying the brain. techniques and architectures for achieve this training, storing, and distributing these data representations are also disclosed. dated 2015-12-08"
9208475,apparatus and method for email storage,"embodiments of the present invention provide an apparatus for storing emails, comprising a neural network arranged to receive information associated with an email, to determine a storage location of the email according to one or more of the attributes of the email and to output information identifying the determined storage location.",2015-12-08,"The title of the patent is apparatus and method for email storage and its abstract is embodiments of the present invention provide an apparatus for storing emails, comprising a neural network arranged to receive information associated with an email, to determine a storage location of the email according to one or more of the attributes of the email and to output information identifying the determined storage location. dated 2015-12-08"
9213937,apparatus and methods for gating analog and spiking signals in artificial neural networks,"apparatus and methods for universal node design implementing a universal learning rule in a mixed signal spiking neural network. in one implementation, at one instance, the node apparatus, operable according to the parameterized universal learning model, receives a mixture of analog and spiking inputs, and generates a spiking output based on the model parameter for that node that is selected by the parameterized model for that specific mix of inputs. at another instance, the same node receives a different mix of inputs, that also may comprise only analog or only spiking inputs and generates an analog output based on a different value of the node parameter that is selected by the model for the second mix of inputs. in another implementation, the node apparatus may change its output from analog to spiking responsive to a training input for the same inputs.",2015-12-15,"The title of the patent is apparatus and methods for gating analog and spiking signals in artificial neural networks and its abstract is apparatus and methods for universal node design implementing a universal learning rule in a mixed signal spiking neural network. in one implementation, at one instance, the node apparatus, operable according to the parameterized universal learning model, receives a mixture of analog and spiking inputs, and generates a spiking output based on the model parameter for that node that is selected by the parameterized model for that specific mix of inputs. at another instance, the same node receives a different mix of inputs, that also may comprise only analog or only spiking inputs and generates an analog output based on a different value of the node parameter that is selected by the model for the second mix of inputs. in another implementation, the node apparatus may change its output from analog to spiking responsive to a training input for the same inputs. dated 2015-12-15"
9218565,haptic-based artificial neural network training,"in a method for training an artificial neural network based algorithm designed to monitor a first device, a processor receives a first data. a processor determines a first service action recommendation for a first device using the received first data and an artificial neural network (ann) algorithm. a processor causes a second device to provide haptic feedback using the received first data. a processor receives a second service action recommendation for the first device based on the haptic feedback. a processor adjusts at least one parameter of the ann algorithm such that the ann algorithm determines a third service action recommendation for the first device using the received first data, wherein the third service action recommendation is equivalent to the second service action recommendation.",2015-12-22,"The title of the patent is haptic-based artificial neural network training and its abstract is in a method for training an artificial neural network based algorithm designed to monitor a first device, a processor receives a first data. a processor determines a first service action recommendation for a first device using the received first data and an artificial neural network (ann) algorithm. a processor causes a second device to provide haptic feedback using the received first data. a processor receives a second service action recommendation for the first device based on the haptic feedback. a processor adjusts at least one parameter of the ann algorithm such that the ann algorithm determines a third service action recommendation for the first device using the received first data, wherein the third service action recommendation is equivalent to the second service action recommendation. dated 2015-12-22"
9224089,method and apparatus for adaptive bit-allocation in neural systems,"certain aspects of the present disclosure support a technique for adaptive bit-allocation in neural systems. bit-allocation for neural signals and parameters in a neural network described in the present disclosure may comprise for a plurality of synapse circuits in the neural simulator network, dynamically allocating a number of bits to the neural circuit signals based on at least one characteristic of one or more neural potential in the neural simulator network; and for the plurality of synapse circuits in the neural simulator network, dynamically allocating a number of bits to at least one neural processing parameter of the synapse circuit based on at least one condition of the neural simulator network.",2015-12-29,"The title of the patent is method and apparatus for adaptive bit-allocation in neural systems and its abstract is certain aspects of the present disclosure support a technique for adaptive bit-allocation in neural systems. bit-allocation for neural signals and parameters in a neural network described in the present disclosure may comprise for a plurality of synapse circuits in the neural simulator network, dynamically allocating a number of bits to the neural circuit signals based on at least one characteristic of one or more neural potential in the neural simulator network; and for the plurality of synapse circuits in the neural simulator network, dynamically allocating a number of bits to at least one neural processing parameter of the synapse circuit based on at least one condition of the neural simulator network. dated 2015-12-29"
9224090,sensory input processing apparatus in a spiking neural network,"apparatus and methods for feedback in a spiking neural network. in one approach, spiking neurons receive sensory stimulus and context signal that correspond to the same context. when the stimulus provides sufficient excitation, neurons generate response. context connections are adjusted according to inverse spike-timing dependent plasticity. when the context signal precedes the post synaptic spike, context synaptic connections are depressed. conversely, whenever the context signal follows the post synaptic spike, the connections are potentiated. the inverse stdp connection adjustment ensures precise control of feedback-induced firing, eliminates runaway positive feedback loops, enables self-stabilizing network operation. in another aspect of the invention, the connection adjustment methodology facilitates robust context switching when processing visual information. when a context (such an object) becomes intermittently absent, prior context connection potentiation enables firing for a period of time. if the object remains absent, the connection becomes depressed thereby preventing further firing.",2015-12-29,"The title of the patent is sensory input processing apparatus in a spiking neural network and its abstract is apparatus and methods for feedback in a spiking neural network. in one approach, spiking neurons receive sensory stimulus and context signal that correspond to the same context. when the stimulus provides sufficient excitation, neurons generate response. context connections are adjusted according to inverse spike-timing dependent plasticity. when the context signal precedes the post synaptic spike, context synaptic connections are depressed. conversely, whenever the context signal follows the post synaptic spike, the connections are potentiated. the inverse stdp connection adjustment ensures precise control of feedback-induced firing, eliminates runaway positive feedback loops, enables self-stabilizing network operation. in another aspect of the invention, the connection adjustment methodology facilitates robust context switching when processing visual information. when a context (such an object) becomes intermittently absent, prior context connection potentiation enables firing for a period of time. if the object remains absent, the connection becomes depressed thereby preventing further firing. dated 2015-12-29"
9224091,learning artificial neural network using ternary content addressable memory (tcam),"a circuit is provided for that includes one or more tcam arrays including one or more matchlines configured to model a neural network. each of the one or more tcam arrays models a connected group of neurons such that input search data into the one or more matchlines is modeled as neuron dendrite information, and the output from the one or more matchlines is modeled as neuron axon information. the circuit further includes one or more additional bits included within each of the one or more matchlines that are configured to model connectivity strength between each neuron dendrite and axon. the circuit also includes a real-time learning block included within each of the one or more tcam arrays configured to modify the connectivity strength between each neuron dendrite and axon using wild-cards written and stored in the one or more additional bits.",2015-12-29,"The title of the patent is learning artificial neural network using ternary content addressable memory (tcam) and its abstract is a circuit is provided for that includes one or more tcam arrays including one or more matchlines configured to model a neural network. each of the one or more tcam arrays models a connected group of neurons such that input search data into the one or more matchlines is modeled as neuron dendrite information, and the output from the one or more matchlines is modeled as neuron axon information. the circuit further includes one or more additional bits included within each of the one or more matchlines that are configured to model connectivity strength between each neuron dendrite and axon. the circuit also includes a real-time learning block included within each of the one or more tcam arrays configured to modify the connectivity strength between each neuron dendrite and axon using wild-cards written and stored in the one or more additional bits. dated 2015-12-29"
9230208,haptic-based artificial neural network training,"in a method for training an artificial neural network based algorithm designed to monitor a first device, a processor receives a first data. a processor determines a first service action recommendation for a first device using the received first data and an artificial neural network (ann) algorithm. a processor causes a second device to provide haptic feedback using the received first data. a processor receives a second service action recommendation for the first device based on the haptic feedback. a processor adjusts at least one parameter of the ann algorithm such that the ann algorithm determines a third service action recommendation for the first device using the received first data, wherein the third service action recommendation is equivalent to the second service action recommendation.",2016-01-05,"The title of the patent is haptic-based artificial neural network training and its abstract is in a method for training an artificial neural network based algorithm designed to monitor a first device, a processor receives a first data. a processor determines a first service action recommendation for a first device using the received first data and an artificial neural network (ann) algorithm. a processor causes a second device to provide haptic feedback using the received first data. a processor receives a second service action recommendation for the first device based on the haptic feedback. a processor adjusts at least one parameter of the ann algorithm such that the ann algorithm determines a third service action recommendation for the first device using the received first data, wherein the third service action recommendation is equivalent to the second service action recommendation. dated 2016-01-05"
9230550,speaker verification and identification using artificial neural network-based sub-phonetic unit discrimination,"in one embodiment, a computer system stores speech data for a plurality of speakers, where the speech data includes a plurality of feature vectors and, for each feature vector, an associated sub-phonetic class. the computer system then builds, based on the speech data, an artificial neural network (ann) for modeling speech of a target speaker in the plurality of speakers, where the ann is configured to discriminate between instances of sub-phonetic classes uttered by the target speaker and instances of sub-phonetic classes uttered by other speakers in the plurality of speakers.",2016-01-05,"The title of the patent is speaker verification and identification using artificial neural network-based sub-phonetic unit discrimination and its abstract is in one embodiment, a computer system stores speech data for a plurality of speakers, where the speech data includes a plurality of feature vectors and, for each feature vector, an associated sub-phonetic class. the computer system then builds, based on the speech data, an artificial neural network (ann) for modeling speech of a target speaker in the plurality of speakers, where the ann is configured to discriminate between instances of sub-phonetic classes uttered by the target speaker and instances of sub-phonetic classes uttered by other speakers in the plurality of speakers. dated 2016-01-05"
9233242,"non-invasive magnetic or electrical nerve stimulation to treat gastroparesis, functional dyspepsia, and other functional gastrointestinal disorders","devices, systems and methods are disclosed for treating or preventing gastroparesis, functional dyspepsia, and other functional gastrointestinal disorders. the methods comprise transmitting impulses of energy non-invasively to selected nerve fibers, particularly those in a vagus nerve. the methods provide damaged interstitial cells of cajal (icc) with trophic factors via vagal afferent nerve fibers, thereby reversing icc damage, and as a consequence improving gastric motility. the methods also increase levels of inhibitory neurotransmitters in the brain so as to decrease neural activity within the area postrema, or they deactivate a resting state neural network containing parts of the anterior insula and anterior cingulate cortex, which will thereby reduce abnormal interoception and visceral hypersensitivity.",2016-01-12,"The title of the patent is non-invasive magnetic or electrical nerve stimulation to treat gastroparesis, functional dyspepsia, and other functional gastrointestinal disorders and its abstract is devices, systems and methods are disclosed for treating or preventing gastroparesis, functional dyspepsia, and other functional gastrointestinal disorders. the methods comprise transmitting impulses of energy non-invasively to selected nerve fibers, particularly those in a vagus nerve. the methods provide damaged interstitial cells of cajal (icc) with trophic factors via vagal afferent nerve fibers, thereby reversing icc damage, and as a consequence improving gastric motility. the methods also increase levels of inhibitory neurotransmitters in the brain so as to decrease neural activity within the area postrema, or they deactivate a resting state neural network containing parts of the anterior insula and anterior cingulate cortex, which will thereby reduce abnormal interoception and visceral hypersensitivity. dated 2016-01-12"
9235799,discriminative pretraining of deep neural networks,"discriminative pretraining technique embodiments are presented that pretrain the hidden layers of a deep neural network (dnn). in general, a one-hidden-layer neural network is trained first using labels discriminatively with error back-propagation (bp). then, after discarding an output layer in the previous one-hidden-layer neural network, another randomly initialized hidden layer is added on top of the previously trained hidden layer along with a new output layer that represents the targets for classification or recognition. the resulting multiple-hidden-layer dnn is then discriminatively trained using the same strategy, and so on until the desired number of hidden layers is reached. this produces a pretrained dnn. the discriminative pretraining technique embodiments have the advantage of bringing the dnn layer weights close to a good local optimum, while still leaving them in a range with a high gradient so that they can be fine-tuned effectively.",2016-01-12,"The title of the patent is discriminative pretraining of deep neural networks and its abstract is discriminative pretraining technique embodiments are presented that pretrain the hidden layers of a deep neural network (dnn). in general, a one-hidden-layer neural network is trained first using labels discriminatively with error back-propagation (bp). then, after discarding an output layer in the previous one-hidden-layer neural network, another randomly initialized hidden layer is added on top of the previously trained hidden layer along with a new output layer that represents the targets for classification or recognition. the resulting multiple-hidden-layer dnn is then discriminatively trained using the same strategy, and so on until the desired number of hidden layers is reached. this produces a pretrained dnn. the discriminative pretraining technique embodiments have the advantage of bringing the dnn layer weights close to a good local optimum, while still leaving them in a range with a high gradient so that they can be fine-tuned effectively. dated 2016-01-12"
9235800,method for the computer-aided learning of a recurrent neural network for modeling a dynamic system,"a method for the computer-aided learning of a recurrent neural network for modeling a dynamic system which is characterized at respective times by an observable vector with one or more observables as entries is provided. the neural network includes both a causal network with a flow of information that is directed forwards in time and a retro-causal network with a flow of information which is directed backwards in time. the states of the dynamic system are characterized by first state vectors in the causal network and by second state vectors in the retro-causal network, wherein the state vectors each contain observables for the dynamic system and also hidden states of the dynamic system. both networks are linked to one another by a combination of the observables from the relevant first and second state vectors and are learned on the basis of training date including known observables vectors.",2016-01-12,"The title of the patent is method for the computer-aided learning of a recurrent neural network for modeling a dynamic system and its abstract is a method for the computer-aided learning of a recurrent neural network for modeling a dynamic system which is characterized at respective times by an observable vector with one or more observables as entries is provided. the neural network includes both a causal network with a flow of information that is directed forwards in time and a retro-causal network with a flow of information which is directed backwards in time. the states of the dynamic system are characterized by first state vectors in the causal network and by second state vectors in the retro-causal network, wherein the state vectors each contain observables for the dynamic system and also hidden states of the dynamic system. both networks are linked to one another by a combination of the observables from the relevant first and second state vectors and are learned on the basis of training date including known observables vectors. dated 2016-01-12"
9235801,managing computer server capacity,"systems and methods are disclosed for using machine learning (e.g., neural networks and/or combinatorial learning) to solve the non-linear problem of predicting the provisioning of a server farm (e.g., cloud resources). the machine learning may be performed using commercially available products, such as the snns product from the university of stuttgard of germany. the system, which includes a neural network for machine learning, is provided with an identification of inputs and outputs to track, and the system provides correlations between those. rather than static rules, the machine learning provides dynamic provisioning recommendations with corresponding confidence scores. based on the data collected/measured by the neural network, the provisioning recommendations will change as well as the confidence scores.",2016-01-12,"The title of the patent is managing computer server capacity and its abstract is systems and methods are disclosed for using machine learning (e.g., neural networks and/or combinatorial learning) to solve the non-linear problem of predicting the provisioning of a server farm (e.g., cloud resources). the machine learning may be performed using commercially available products, such as the snns product from the university of stuttgard of germany. the system, which includes a neural network for machine learning, is provided with an identification of inputs and outputs to track, and the system provides correlations between those. rather than static rules, the machine learning provides dynamic provisioning recommendations with corresponding confidence scores. based on the data collected/measured by the neural network, the provisioning recommendations will change as well as the confidence scores. dated 2016-01-12"
9239369,method and system for selecting base stations to position mobile device,"a method for selecting a plurality of base stations to position a mobile device is provided. the method includes selecting a plurality of base station sets from the plurality of base stations, wherein each of the plurality of base station sets corresponds to a distance matrix, utilizing a first artificial neural network (ann) unit to select a predefined number of the plurality of base station sets from the plurality of base station sets according to a plurality of distance matrixes corresponding to the plurality of base station sets; and utilizing a second ann unit to position the mobile device according to the predefined number of the plurality of base station sets.",2016-01-19,"The title of the patent is method and system for selecting base stations to position mobile device and its abstract is a method for selecting a plurality of base stations to position a mobile device is provided. the method includes selecting a plurality of base station sets from the plurality of base stations, wherein each of the plurality of base station sets corresponds to a distance matrix, utilizing a first artificial neural network (ann) unit to select a predefined number of the plurality of base station sets from the plurality of base station sets according to a plurality of distance matrixes corresponding to the plurality of base station sets; and utilizing a second ann unit to position the mobile device according to the predefined number of the plurality of base station sets. dated 2016-01-19"
9239828,recurrent conditional random fields,"recurrent conditional random field (r-crf) embodiments are described. in one embodiment, the r-cfr receives feature values corresponding to a sequence of words. semantic labels for words in the sequence of words are then generated and each label is assigned to the appropriate one of the words in the sequence of words. the r-crf used to accomplish these tasks includes a recurrent neural network (rnn) portion and a conditional random field (crf) portion. the rnn portion receives feature values associated with a word in the sequence of words and outputs rnn activation layer activations data that is indicative of a semantic label. the crf portion inputs the rnn activation layer activations data output from the rnn for one or more words in the sequence of words and outputs label data that is indicative of a separate semantic label that is to be assigned to each of the words.",2016-01-19,"The title of the patent is recurrent conditional random fields and its abstract is recurrent conditional random field (r-crf) embodiments are described. in one embodiment, the r-cfr receives feature values corresponding to a sequence of words. semantic labels for words in the sequence of words are then generated and each label is assigned to the appropriate one of the words in the sequence of words. the r-crf used to accomplish these tasks includes a recurrent neural network (rnn) portion and a conditional random field (crf) portion. the rnn portion receives feature values associated with a word in the sequence of words and outputs rnn activation layer activations data that is indicative of a semantic label. the crf portion inputs the rnn activation layer activations data output from the rnn for one or more words in the sequence of words and outputs label data that is indicative of a separate semantic label that is to be assigned to each of the words. dated 2016-01-19"
9239983,quantifying a condition with a neural network,"quantification of a condition of a selected item using a neural network is disclosed. a device defines a good hypertube in a neural state space based on good item state points obtained from one or more items that exhibit desired operating characteristics, and a bad hypertube in the neural state space based on bad item state points obtained from one or more items that exhibit undesirable operating characteristics. a current item state hyperpoint is determined in the neural state space based on a current item state point of the selected item. a condition of the selected item is quantified as a function of a location of the current item state hyperpoint with respect to at least a portion of the good hypertube and with respect to at least a portion of the bad hypertube.",2016-01-19,"The title of the patent is quantifying a condition with a neural network and its abstract is quantification of a condition of a selected item using a neural network is disclosed. a device defines a good hypertube in a neural state space based on good item state points obtained from one or more items that exhibit desired operating characteristics, and a bad hypertube in the neural state space based on bad item state points obtained from one or more items that exhibit undesirable operating characteristics. a current item state hyperpoint is determined in the neural state space based on a current item state point of the selected item. a condition of the selected item is quantified as a function of a location of the current item state hyperpoint with respect to at least a portion of the good hypertube and with respect to at least a portion of the bad hypertube. dated 2016-01-19"
9239984,time-division multiplexed neurosynaptic module with implicit memory addressing for implementing a neural network,"embodiments of the invention relate to a time-division multiplexed neurosynaptic module with implicit memory addressing for implementing a neural network. one embodiment comprises maintaining neuron attributes for multiple neurons and maintaining incoming firing events for different time steps. for each time step, incoming firing events for said time step are integrated in a time-division multiplexing manner. incoming firing events are integrated based on the neuron attributes maintained. for each time step, the neuron attributes maintained are updated in parallel based on the integrated incoming firing events for said time step.",2016-01-19,"The title of the patent is time-division multiplexed neurosynaptic module with implicit memory addressing for implementing a neural network and its abstract is embodiments of the invention relate to a time-division multiplexed neurosynaptic module with implicit memory addressing for implementing a neural network. one embodiment comprises maintaining neuron attributes for multiple neurons and maintaining incoming firing events for different time steps. for each time step, incoming firing events for said time step are integrated in a time-division multiplexing manner. incoming firing events are integrated based on the neuron attributes maintained. for each time step, the neuron attributes maintained are updated in parallel based on the integrated incoming firing events for said time step. dated 2016-01-19"
9239990,hybrid location using pattern recognition of location readings and signal strengths of wireless access points,"a query device scans radio frequencies for visible transmitting devices. the querying device receives at least a signal strength and identifier information associated with each of the transmitting devices. the list of visible devices is used to query a database containing location information for a plurality of visible devices. the list may be sent to a locationing system that may perform a location analysis on the resulting data to return a location to the query device. the weighted average of the locations returned in the database query may be computed to determine the location of the querying device, with the weight for each of the locations being the current signal strength detected by the querying device. neural network analysis may also be used to determine the location of the querying device. learning and seeding operations many also be used to populate the database with location information for transmitting devices.",2016-01-19,"The title of the patent is hybrid location using pattern recognition of location readings and signal strengths of wireless access points and its abstract is a query device scans radio frequencies for visible transmitting devices. the querying device receives at least a signal strength and identifier information associated with each of the transmitting devices. the list of visible devices is used to query a database containing location information for a plurality of visible devices. the list may be sent to a locationing system that may perform a location analysis on the resulting data to return a location to the query device. the weighted average of the locations returned in the database query may be computed to determine the location of the querying device, with the weight for each of the locations being the current signal strength detected by the querying device. neural network analysis may also be used to determine the location of the querying device. learning and seeding operations many also be used to populate the database with location information for transmitting devices. dated 2016-01-19"
9240184,frame-level combination of deep neural network and gaussian mixture models,"a method and system for frame-level merging of hmm state predictions determined by different techniques is disclosed. an audio input signal may be transformed into a first and second sequence of feature vector, the sequences corresponding to each other and to a temporal sequence of frames of the audio input signal on a frame-by-frame basis. the first sequence may be processed by a neural network (nn) to determine nn-based state predictions, and the second sequence may be processed by a gaussian mixture model (gmm) to determine gmm-based state predictions. the nn-based and gmm-based state predictions may be merged as weighted sums for each of a plurality of hmm state on a frame-by-frame basis to determine merged state predictions. the merged state predictions may then be applied to the hmms to speech content of the audio input signal.",2016-01-19,"The title of the patent is frame-level combination of deep neural network and gaussian mixture models and its abstract is a method and system for frame-level merging of hmm state predictions determined by different techniques is disclosed. an audio input signal may be transformed into a first and second sequence of feature vector, the sequences corresponding to each other and to a temporal sequence of frames of the audio input signal on a frame-by-frame basis. the first sequence may be processed by a neural network (nn) to determine nn-based state predictions, and the second sequence may be processed by a gaussian mixture model (gmm) to determine gmm-based state predictions. the nn-based and gmm-based state predictions may be merged as weighted sums for each of a plurality of hmm state on a frame-by-frame basis to determine merged state predictions. the merged state predictions may then be applied to the hmms to speech content of the audio input signal. dated 2016-01-19"
9242370,miniature robot having multiple legs using piezo legs having two degrees of freedom,"the present invention relates to a miniature robot using a plurality of piezo legs capable of deforming in two degrees of freedom by supplying input voltage signals, and an integrated artificial neural network behavior controller capable of modeling complex behavioral patterns and gait patterns by a simple structure. the miniature robot with multiple legs includes: a main body; a plurality of piezo legs constituted by bimorph piezoelectric elements and connected to the main body to generate motion through morphological deformation in two degrees of freedom by applied voltage signals, thereby resulting in movement on the ground; and a control part including an artificial neural network behavior controller, which controls motion patterns by feeding back information with respect to the environment transmitted from external sensors and to electrical signals transmitted when the piezo legs contact the surface of the ground, to control the voltage applied to each of the piezo legs.",2016-01-26,"The title of the patent is miniature robot having multiple legs using piezo legs having two degrees of freedom and its abstract is the present invention relates to a miniature robot using a plurality of piezo legs capable of deforming in two degrees of freedom by supplying input voltage signals, and an integrated artificial neural network behavior controller capable of modeling complex behavioral patterns and gait patterns by a simple structure. the miniature robot with multiple legs includes: a main body; a plurality of piezo legs constituted by bimorph piezoelectric elements and connected to the main body to generate motion through morphological deformation in two degrees of freedom by applied voltage signals, thereby resulting in movement on the ground; and a control part including an artificial neural network behavior controller, which controls motion patterns by feeding back information with respect to the environment transmitted from external sensors and to electrical signals transmitted when the piezo legs contact the surface of the ground, to control the voltage applied to each of the piezo legs. dated 2016-01-26"
9245222,"synaptic, dendritic, somatic, and axonal plasticity in a network of neural cores using a plastic multi-stage crossbar switching","embodiments of the invention provide a neural network comprising multiple functional neural core circuits, and a dynamically reconfigurable switch interconnect between the functional neural core circuits. the interconnect comprises multiple connectivity neural core circuits. each functional neural core circuit comprises a first and a second core module. each core module comprises a plurality of electronic neurons, a plurality of incoming electronic axons, and multiple electronic synapses interconnecting the incoming axons to the neurons. each neuron has a corresponding outgoing electronic axon. in one embodiment, zero or more sets of connectivity neural core circuits interconnect outgoing axons in a functional neural core circuit to incoming axons in the same functional neural core circuit. in another embodiment, zero or more sets of connectivity neural core circuits interconnect outgoing and incoming axons in a functional neural core circuit to incoming and outgoing axons in a different functional neural core circuit, respectively.",2016-01-26,"The title of the patent is synaptic, dendritic, somatic, and axonal plasticity in a network of neural cores using a plastic multi-stage crossbar switching and its abstract is embodiments of the invention provide a neural network comprising multiple functional neural core circuits, and a dynamically reconfigurable switch interconnect between the functional neural core circuits. the interconnect comprises multiple connectivity neural core circuits. each functional neural core circuit comprises a first and a second core module. each core module comprises a plurality of electronic neurons, a plurality of incoming electronic axons, and multiple electronic synapses interconnecting the incoming axons to the neurons. each neuron has a corresponding outgoing electronic axon. in one embodiment, zero or more sets of connectivity neural core circuits interconnect outgoing axons in a functional neural core circuit to incoming axons in the same functional neural core circuit. in another embodiment, zero or more sets of connectivity neural core circuits interconnect outgoing and incoming axons in a functional neural core circuit to incoming and outgoing axons in a different functional neural core circuit, respectively. dated 2016-01-26"
9245223,"unsupervised, supervised and reinforced learning via spiking computation","the present invention relates to unsupervised, supervised and reinforced learning via spiking computation. the neural network comprises a plurality of neural modules. each neural module comprises multiple digital neurons such that each neuron in a neural module has a corresponding neuron in another neural module. an interconnection network comprising a plurality of edges interconnects the plurality of neural modules. each edge interconnects a first neural module to a second neural module, and each edge comprises a weighted synaptic connection between every neuron in the first neural module and a corresponding neuron in the second neural module.",2016-01-26,"The title of the patent is unsupervised, supervised and reinforced learning via spiking computation and its abstract is the present invention relates to unsupervised, supervised and reinforced learning via spiking computation. the neural network comprises a plurality of neural modules. each neural module comprises multiple digital neurons such that each neuron in a neural module has a corresponding neuron in another neural module. an interconnection network comprising a plurality of edges interconnects the plurality of neural modules. each edge interconnects a first neural module to a second neural module, and each edge comprises a weighted synaptic connection between every neuron in the first neural module and a corresponding neuron in the second neural module. dated 2016-01-26"
9245224,optimally configuring an information landscape,"according to an embodiment of the present invention, a system optimizes an information processing environment, and comprises at least one processor. the system collects information pertaining to operational behavior of the information processing environment and including a plurality of parameters. a neural network structure is established to associate the parameters to a desired operational performance characteristic for the information processing environment. the neural network structure is trained with the collected information from the information processing environment to produce a model for the information processing environment. the model is optimized to determine values for the parameters and the information processing environment is adjusted based on the determined parameter values to attain the desired operational performance of the information processing environment. embodiments of the present invention further include a method and computer program product for optimizing an information processing environment in substantially the same manner described above.",2016-01-26,"The title of the patent is optimally configuring an information landscape and its abstract is according to an embodiment of the present invention, a system optimizes an information processing environment, and comprises at least one processor. the system collects information pertaining to operational behavior of the information processing environment and including a plurality of parameters. a neural network structure is established to associate the parameters to a desired operational performance characteristic for the information processing environment. the neural network structure is trained with the collected information from the information processing environment to produce a model for the information processing environment. the model is optimized to determine values for the parameters and the information processing environment is adjusted based on the determined parameter values to attain the desired operational performance of the information processing environment. embodiments of the present invention further include a method and computer program product for optimizing an information processing environment in substantially the same manner described above. dated 2016-01-26"
9248789,method and apparatus for detecting key-on and key-off states using time-to-frequency transforms,a key state detector for a vehicle collects a sequence of battery voltage samples and applies a time-domain to frequency-domain transform (tft) to the collected samples. the results of the tft are then applied to an artificial neural network (ann) to determine if they represent a key-on or key-off state. the ann is trained based on data collected from the vehicle and is periodically retrained so that the detection of key-on and key-off states conforms to the particular vehicle and tracks the aging of vehicle components.,2016-02-02,The title of the patent is method and apparatus for detecting key-on and key-off states using time-to-frequency transforms and its abstract is a key state detector for a vehicle collects a sequence of battery voltage samples and applies a time-domain to frequency-domain transform (tft) to the collected samples. the results of the tft are then applied to an artificial neural network (ann) to determine if they represent a key-on or key-off state. the ann is trained based on data collected from the vehicle and is periodically retrained so that the detection of key-on and key-off states conforms to the particular vehicle and tracks the aging of vehicle components. dated 2016-02-02
9250667,early detection of overheating devices,"a method is provided that monitors the odor within the physical enclosure of a computing device that includes one or more components. the method includes determining whether the odor within the physical enclosure is indicative of an overheating component that is overheating within the physical enclosure of the computing device. determining whether the odor within the physical enclosure may include an artificial neural network (“ann”) to determine whether the odor is indicative of an overheating component. the method includes initiating an overheating protocol in response to determining that the odor within the physical enclosure is indicative of an overheating component. the method may, for example, alert the user and/or applications that a component is overheating.",2016-02-02,"The title of the patent is early detection of overheating devices and its abstract is a method is provided that monitors the odor within the physical enclosure of a computing device that includes one or more components. the method includes determining whether the odor within the physical enclosure is indicative of an overheating component that is overheating within the physical enclosure of the computing device. determining whether the odor within the physical enclosure may include an artificial neural network (“ann”) to determine whether the odor is indicative of an overheating component. the method includes initiating an overheating protocol in response to determining that the odor within the physical enclosure is indicative of an overheating component. the method may, for example, alert the user and/or applications that a component is overheating. dated 2016-02-02"
9251437,system and method for generating training cases for image classification,a system and method for generating training images. an existing training image is associated with a classification. the system includes an image processing module that performs color-space deformation on each pixel of the existing training image and then associates the classification to the color-space deformed training image. the technique may be applied to increase the size of a training set for training a neural network.,2016-02-02,The title of the patent is system and method for generating training cases for image classification and its abstract is a system and method for generating training images. an existing training image is associated with a classification. the system includes an image processing module that performs color-space deformation on each pixel of the existing training image and then associates the classification to the color-space deformed training image. the technique may be applied to increase the size of a training set for training a neural network. dated 2016-02-02
9254824,adaptive anti-collision method for vehicle,"an adaptive anti-collision method for vehicles has steps of creating multiple driving patterns with each driving pattern corresponding to a vehicle speed, a safe distance and a braking distance parameter, such as longer safe distance configured for faster vehicle speed, and higher vehicle speed or shorter safe distance for different road condition, acquiring dynamic information, such as vehicle speed or acceleration, of the vehicle using sensors on the vehicle, combining the dynamic information and drivers' driving behavior to determine a driving pattern through a statistical analysis and a neural network, adjusting control parameters of the vehicle according to the driving pattern for an electronic control unit of the vehicle to issue an alert or activate a braking action according to the driving pattern. accordingly, the anti-collision method can be adapted to different vehicle speed, road condition and drivers' driving habits for adjusting the safe distance and the braking system.",2016-02-09,"The title of the patent is adaptive anti-collision method for vehicle and its abstract is an adaptive anti-collision method for vehicles has steps of creating multiple driving patterns with each driving pattern corresponding to a vehicle speed, a safe distance and a braking distance parameter, such as longer safe distance configured for faster vehicle speed, and higher vehicle speed or shorter safe distance for different road condition, acquiring dynamic information, such as vehicle speed or acceleration, of the vehicle using sensors on the vehicle, combining the dynamic information and drivers' driving behavior to determine a driving pattern through a statistical analysis and a neural network, adjusting control parameters of the vehicle according to the driving pattern for an electronic control unit of the vehicle to issue an alert or activate a braking action according to the driving pattern. accordingly, the anti-collision method can be adapted to different vehicle speed, road condition and drivers' driving habits for adjusting the safe distance and the braking system. dated 2016-02-09"
9255849,"temperature compensation apparatus, methods, and systems","in some embodiments, an apparatus and a system, as well as a method and an article, may operate to receive down hole tool environmental temperature data, axial temperature data, radial temperature data, and log data. further activity may include applying temperature effects compensation associated with the environmental temperature data and the down hole log data using a fitting function model obtained from a trained neural network to transform the down hole log data into corrected log data. additional apparatus, systems, and methods are described.",2016-02-09,"The title of the patent is temperature compensation apparatus, methods, and systems and its abstract is in some embodiments, an apparatus and a system, as well as a method and an article, may operate to receive down hole tool environmental temperature data, axial temperature data, radial temperature data, and log data. further activity may include applying temperature effects compensation associated with the environmental temperature data and the down hole log data using a fitting function model obtained from a trained neural network to transform the down hole log data into corrected log data. additional apparatus, systems, and methods are described. dated 2016-02-09"
9255973,system and method for estimating long term characteristics of battery,"a system for estimating long term characteristics of a battery includes a learning data input unit for receiving initial characteristic learning data and long term characteristic learning data of a battery to be a learning object; a measurement data input unit for receiving initial characteristic measurement data of a battery to be an object for estimation of long term characteristics; and an artificial neural network operation unit for receiving the initial characteristic learning data and the long term characteristic learning data from the learning data input unit to allow learning of an artificial neural network, receiving the initial characteristic measurement data from the measurement data input unit and applying the learned artificial neural network thereto, and thus calculating long term characteristic estimation data from the initial characteristic measurement data of the battery and outputting the long term characteristic estimation data.",2016-02-09,"The title of the patent is system and method for estimating long term characteristics of battery and its abstract is a system for estimating long term characteristics of a battery includes a learning data input unit for receiving initial characteristic learning data and long term characteristic learning data of a battery to be a learning object; a measurement data input unit for receiving initial characteristic measurement data of a battery to be an object for estimation of long term characteristics; and an artificial neural network operation unit for receiving the initial characteristic learning data and the long term characteristic learning data from the learning data input unit to allow learning of an artificial neural network, receiving the initial characteristic measurement data from the measurement data input unit and applying the learned artificial neural network thereto, and thus calculating long term characteristic estimation data from the initial characteristic measurement data of the battery and outputting the long term characteristic estimation data. dated 2016-02-09"
9258607,methods and apparatus to determine locations of audience members,"a disclosed example method involves determining, using a neural network at a stationary unit, a first distance of a first portable audio detector from the stationary unit, the first portable audio detector associated with a first panelist, the stationary unit located in proximity to the media presentation device. the example method also involves determining, using the neural network, a second distance of a second portable audio detector from the stationary unit, the second portable audio detector associated with a second panelist. in response to the first distance being less than a threshold distance, the media is credited as exposed to the first panelist. in response to the second distance being more than the threshold distance, the media is not credited as exposed to the second panelist.",2016-02-09,"The title of the patent is methods and apparatus to determine locations of audience members and its abstract is a disclosed example method involves determining, using a neural network at a stationary unit, a first distance of a first portable audio detector from the stationary unit, the first portable audio detector associated with a first panelist, the stationary unit located in proximity to the media presentation device. the example method also involves determining, using the neural network, a second distance of a second portable audio detector from the stationary unit, the second portable audio detector associated with a second panelist. in response to the first distance being less than a threshold distance, the media is credited as exposed to the first panelist. in response to the second distance being more than the threshold distance, the media is not credited as exposed to the second panelist. dated 2016-02-09"
9263060,artificial neural network based system for classification of the emotional content of digital music,"a system for classification of the emotional content of music is provided. an encoder receives a digital audio recording of a piece of music, and encodes it using musical notes and associated amplitudes. the artificial neural network is configured to take a plurality of encoded time slices and provide output indicative of the emotional content of the music.",2016-02-16,"The title of the patent is artificial neural network based system for classification of the emotional content of digital music and its abstract is a system for classification of the emotional content of music is provided. an encoder receives a digital audio recording of a piece of music, and encodes it using musical notes and associated amplitudes. the artificial neural network is configured to take a plurality of encoded time slices and provide output indicative of the emotional content of the music. dated 2016-02-16"
9268990,apparatus and method for producing an identification device,"an authentication system authenticates an object. the authentication system includes a capture device for capturing at least one biometric output data record (bd) for the object; a reading device for reading configuration data (konf), associated with the object, for an artificial neural network; a processing device designed to produce the artificial neural network and to input the bd into the neural network; a verification device which captures an output from the neural network to authenticate the object, wherein the neural network is a bidirectional associative memory, particularly a hopfield network, having a multiplicity of network states. the verification device is designed to determine the output from the neural network by capturing a final state derived from the input of the bd. the neural network stores a key associated with a particular person. the key is released only when appropriate biometric data are input into the neural network.",2016-02-23,"The title of the patent is apparatus and method for producing an identification device and its abstract is an authentication system authenticates an object. the authentication system includes a capture device for capturing at least one biometric output data record (bd) for the object; a reading device for reading configuration data (konf), associated with the object, for an artificial neural network; a processing device designed to produce the artificial neural network and to input the bd into the neural network; a verification device which captures an output from the neural network to authenticate the object, wherein the neural network is a bidirectional associative memory, particularly a hopfield network, having a multiplicity of network states. the verification device is designed to determine the output from the neural network by capturing a final state derived from the input of the bd. the neural network stores a key associated with a particular person. the key is released only when appropriate biometric data are input into the neural network. dated 2016-02-23"
9269040,event monitoring devices and methods,"a device (10) for processing events (4), including an identifier (8) identifying the event's type, and at least one parameter carrying information about a process, includes an event selector (20) and an event type recognizer (30). the device (10) is configured for receiving an event (4), providing the event (4) to the event selector (20), and providing the identifier (8) to the event type recognizer (30). the event selector (20) stores the provided event (4). the event type recognizer (30) determines, using at least one neural network unit, whether the identifier (8) corresponds to a type for which a subscription exists, and, if so, it causes the event selector (20) to transmit the event (4s) for processing by one or more applications. furthermore, the device (10) is configured for allowing one or more types of event to be subscribed to. the invention also relates to methods for processing events (4).",2016-02-23,"The title of the patent is event monitoring devices and methods and its abstract is a device (10) for processing events (4), including an identifier (8) identifying the event's type, and at least one parameter carrying information about a process, includes an event selector (20) and an event type recognizer (30). the device (10) is configured for receiving an event (4), providing the event (4) to the event selector (20), and providing the identifier (8) to the event type recognizer (30). the event selector (20) stores the provided event (4). the event type recognizer (30) determines, using at least one neural network unit, whether the identifier (8) corresponds to a type for which a subscription exists, and, if so, it causes the event selector (20) to transmit the event (4s) for processing by one or more applications. furthermore, the device (10) is configured for allowing one or more types of event to be subscribed to. the invention also relates to methods for processing events (4). dated 2016-02-23"
9269041,hardware enhancements to radial basis function with restricted coulomb energy learning and/or k-nearest neighbor based neural network classifiers,"this disclosure describes embodiments for a hardware based neural network integrated circuit classifier incorporating natively implemented radial basis functions, restricted coulomb energy function, and/or knn to make it more practical for handling a broader group of parallel algorithms.",2016-02-23,"The title of the patent is hardware enhancements to radial basis function with restricted coulomb energy learning and/or k-nearest neighbor based neural network classifiers and its abstract is this disclosure describes embodiments for a hardware based neural network integrated circuit classifier incorporating natively implemented radial basis functions, restricted coulomb energy function, and/or knn to make it more practical for handling a broader group of parallel algorithms. dated 2016-02-23"
9269044,neuromorphic event-driven neural computing architecture in a scalable neural network,an event-driven neural network includes a plurality of interconnected core circuits is provided. each core circuit includes an electronic synapse array has multiple digital synapses interconnecting a plurality of digital electronic neurons. a synapse interconnects an axon of a pre-synaptic neuron with a dendrite of a post-synaptic neuron. a neuron integrates input spikes and generates a spike event in response to the integrated input spikes exceeding a threshold. each core circuit also has a scheduler that receives a spike event and delivers the spike event to a selected axon in the synapse array based on a schedule for deterministic event delivery.,2016-02-23,The title of the patent is neuromorphic event-driven neural computing architecture in a scalable neural network and its abstract is an event-driven neural network includes a plurality of interconnected core circuits is provided. each core circuit includes an electronic synapse array has multiple digital synapses interconnecting a plurality of digital electronic neurons. a synapse interconnects an axon of a pre-synaptic neuron with a dendrite of a post-synaptic neuron. a neuron integrates input spikes and generates a spike event in response to the integrated input spikes exceeding a threshold. each core circuit also has a scheduler that receives a spike event and delivers the spike event to a selected axon in the synapse array based on a schedule for deterministic event delivery. dated 2016-02-23
9269045,auditory source separation in a spiking neural network,"a method of audio source segregation includes selecting an audio attribute of an audio signal. the method also includes representing a portion of the audio attribute that is dominated by a single source as a source spiking event. in addition, the method includes representing a remaining portion of the audio signal as an audio signal spiking event. the method further includes determining whether the remaining portion coincides with the single source based on coincidence of the source spiking event and audio signal spiking event.",2016-02-23,"The title of the patent is auditory source separation in a spiking neural network and its abstract is a method of audio source segregation includes selecting an audio attribute of an audio signal. the method also includes representing a portion of the audio attribute that is dominated by a single source as a source spiking event. in addition, the method includes representing a remaining portion of the audio signal as an audio signal spiking event. the method further includes determining whether the remaining portion coincides with the single source based on coincidence of the source spiking event and audio signal spiking event. dated 2016-02-23"
9272158,method for non-invasive brain stimulation,magneto-electric nanoparticles in a subject interact with an external magnetic field to cause stimulation of neural networks in the subject. electric signals in the neural network are coupled to magnetic dipoles induced in the nanoparticles to cause changes in electric pulse sequences of the subject's brain.,2016-03-01,The title of the patent is method for non-invasive brain stimulation and its abstract is magneto-electric nanoparticles in a subject interact with an external magnetic field to cause stimulation of neural networks in the subject. electric signals in the neural network are coupled to magnetic dipoles induced in the nanoparticles to cause changes in electric pulse sequences of the subject's brain. dated 2016-03-01
9274036,method and apparatus for characterizing composite materials using an artificial neural network,"this invention relates to a method and apparatus for characterizing composite materials, and in particular, to utilizing an artificial neural network for predicting an impact resistance of a composite material. a method for predicting an impact resistance of a composite material in accordance with the present invention includes the steps of designing an artificial neural network including a plurality of neurons, training the artificial neural network to predict the impact resistance by adjusting an output of the plurality of neurons according to sample data and known results of the sample data, inputting data of the composite material into the artificial neural network, and utilizing the artificial neural network to predict the impact resistance of the composite material.",2016-03-01,"The title of the patent is method and apparatus for characterizing composite materials using an artificial neural network and its abstract is this invention relates to a method and apparatus for characterizing composite materials, and in particular, to utilizing an artificial neural network for predicting an impact resistance of a composite material. a method for predicting an impact resistance of a composite material in accordance with the present invention includes the steps of designing an artificial neural network including a plurality of neurons, training the artificial neural network to predict the impact resistance by adjusting an output of the plurality of neurons according to sample data and known results of the sample data, inputting data of the composite material into the artificial neural network, and utilizing the artificial neural network to predict the impact resistance of the composite material. dated 2016-03-01"
9275308,object detection using deep neural networks,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for detecting objects in images. one of the methods includes receiving an input image. a full object mask is generated by providing the input image to a first deep neural network object detector that produces a full object mask for an object of a particular object type depicted in the input image. a partial object mask is generated by providing the input image to a second deep neural network object detector that produces a partial object mask for a portion of the object of the particular object type depicted in the input image. a bounding box is determined for the object in the image using the full object mask and the partial object mask.",2016-03-01,"The title of the patent is object detection using deep neural networks and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for detecting objects in images. one of the methods includes receiving an input image. a full object mask is generated by providing the input image to a first deep neural network object detector that produces a full object mask for an object of a particular object type depicted in the input image. a partial object mask is generated by providing the input image to a second deep neural network object detector that produces a partial object mask for a portion of the object of the particular object type depicted in the input image. a bounding box is determined for the object in the image using the full object mask and the partial object mask. dated 2016-03-01"
9275327,ai for relating herbal ingredients to illnesses classified in traditional chinese medicine/tcm using probabilities and a relevance index,described herein are systems and methods for identifying herbal ingredients effective in treating illnesses in traditional chinese medicine (tcm) using an artificial neural network.,2016-03-01,The title of the patent is ai for relating herbal ingredients to illnesses classified in traditional chinese medicine/tcm using probabilities and a relevance index and its abstract is described herein are systems and methods for identifying herbal ingredients effective in treating illnesses in traditional chinese medicine (tcm) using an artificial neural network. dated 2016-03-01
9275328,neuromorphic compiler,a neuromorphic compiler includes a placement module to provide analytic placement of neurons in a neural network description. the analytic placement is to produce placed neurons. the neuromorphic compiler further includes a smoothing module to perform diffusion-based smoothing of the placed neurons; a legalization module to adjust locations of the placed neurons to correspond to legal locations of neuromorphic neurons within a neural fabric; and a simulated annealing module to refine locations of the placed neurons within the neural fabric using simulated annealing following location adjustment by the legalization module. the neural fabric is to implement a synaptic time-multiplexed (stm) neuromorphic network.,2016-03-01,The title of the patent is neuromorphic compiler and its abstract is a neuromorphic compiler includes a placement module to provide analytic placement of neurons in a neural network description. the analytic placement is to produce placed neurons. the neuromorphic compiler further includes a smoothing module to perform diffusion-based smoothing of the placed neurons; a legalization module to adjust locations of the placed neurons to correspond to legal locations of neuromorphic neurons within a neural fabric; and a simulated annealing module to refine locations of the placed neurons within the neural fabric using simulated annealing following location adjustment by the legalization module. the neural fabric is to implement a synaptic time-multiplexed (stm) neuromorphic network. dated 2016-03-01
9275330,multi-compartment neurons with neural cores,embodiments of the invention provide a neural core circuit comprising a synaptic interconnect network including plural electronic synapses for interconnecting one or more source electronic neurons with one or more target electronic neurons. the interconnect network further includes multiple axon paths and multiple dendrite paths. each synapse is at a cross-point junction of the interconnect network between a dendrite path and an axon path. the core circuit further comprises a routing module maintaining routing information. the routing module routes output from a source electronic neuron to one or more selected axon paths. each synapse provides a configurable level of signal conduction from an axon path of a source electronic neuron to a dendrite path of a target electronic neuron.,2016-03-01,The title of the patent is multi-compartment neurons with neural cores and its abstract is embodiments of the invention provide a neural core circuit comprising a synaptic interconnect network including plural electronic synapses for interconnecting one or more source electronic neurons with one or more target electronic neurons. the interconnect network further includes multiple axon paths and multiple dendrite paths. each synapse is at a cross-point junction of the interconnect network between a dendrite path and an axon path. the core circuit further comprises a routing module maintaining routing information. the routing module routes output from a source electronic neuron to one or more selected axon paths. each synapse provides a configurable level of signal conduction from an axon path of a source electronic neuron to a dendrite path of a target electronic neuron. dated 2016-03-01
9277208,system and method for estimating quality of video with frame freezing artifacts,"a method and system that assesses video quality of transmitted video packet signals suffering from packet loss and delay. this packet loss and delay can create freeze events, which are observed as a jerkiness while viewing the video. the system and method compares the frames in a video transmission to determine freeze events; extracts a set of features from the locations of the freeze events and decoded video frames; and maps the set of features into a video quality score using a neural network. the video quality score provides an assessment of the effects of irregular frame freezes due to packet loss or delay on the quality of the video.",2016-03-01,"The title of the patent is system and method for estimating quality of video with frame freezing artifacts and its abstract is a method and system that assesses video quality of transmitted video packet signals suffering from packet loss and delay. this packet loss and delay can create freeze events, which are observed as a jerkiness while viewing the video. the system and method compares the frames in a video transmission to determine freeze events; extracts a set of features from the locations of the freeze events and decoded video frames; and maps the set of features into a video quality score using a neural network. the video quality score provides an assessment of the effects of irregular frame freezes due to packet loss or delay on the quality of the video. dated 2016-03-01"
9286524,multi-task deep convolutional neural networks for efficient and robust traffic lane detection,"disclosed herein are devices, systems, and methods for detecting the presence and orientation of traffic lane markings. deep convolutional neural networks are used with convolutional layers and max-pooling layers to generate fully connected nodes. after the convolutional and max-pooling layers, two sublayers are applied, one to determine presence and one to determine geometry. the presence of a lane marking segment as detected by the first sublayer can serve as a gate for the second sublayer by regulating the credit assignment for training the network. only when the first sublayer predicts actual presence will the geometric layout of the lane marking segment contribute to the training of the overall network. this achieves advantages with respect to accuracy and efficiency and contributes to efficient robust model selection.",2016-03-15,"The title of the patent is multi-task deep convolutional neural networks for efficient and robust traffic lane detection and its abstract is disclosed herein are devices, systems, and methods for detecting the presence and orientation of traffic lane markings. deep convolutional neural networks are used with convolutional layers and max-pooling layers to generate fully connected nodes. after the convolutional and max-pooling layers, two sublayers are applied, one to determine presence and one to determine geometry. the presence of a lane marking segment as detected by the first sublayer can serve as a gate for the second sublayer by regulating the credit assignment for training the network. only when the first sublayer predicts actual presence will the geometric layout of the lane marking segment contribute to the training of the overall network. this achieves advantages with respect to accuracy and efficiency and contributes to efficient robust model selection. dated 2016-03-15"
9289810,apparatus and method for measuring the bending angle of a sheet,"an apparatus for measuring a bending angle of a sheet, comprising a processing unit and at least one sensor comprising a light source which projects a light pattern on at least one side of the sheet, and recording means adapted to record an image of the projection of said light pattern on the at least one side of the sheet. the processing unit is adapted to control the recording means for recording the image in at least one time instant (treg1; treg1, treg2 . . . tregn) during an operation of bending the sheet; a control unit is capable of transforming the recorded image into a point cloud and comprises a neural network adapted to associate a bending angle value with the point cloud.",2016-03-22,"The title of the patent is apparatus and method for measuring the bending angle of a sheet and its abstract is an apparatus for measuring a bending angle of a sheet, comprising a processing unit and at least one sensor comprising a light source which projects a light pattern on at least one side of the sheet, and recording means adapted to record an image of the projection of said light pattern on the at least one side of the sheet. the processing unit is adapted to control the recording means for recording the image in at least one time instant (treg1; treg1, treg2 . . . tregn) during an operation of bending the sheet; a control unit is capable of transforming the recorded image into a point cloud and comprises a neural network adapted to associate a bending angle value with the point cloud. dated 2016-03-22"
9290756,apparatus and methods for high throughput network electrophysiology and cellular analysis,"provided herein are apparatus and methods relating to the development of instrumentation for high throughput network electrophysiology and cellular analysis. more specifically, provided herein are multiwell microelectrode arrays (meas) and methods for the development of such an apparatus in an inexpensive fashion with a flexible, ansi/sbs-compliant (american national standards institute/society for biomolecular screening) format. microelectrode arrays are a grid of tightly spaced microelectrodes useful for stimulating and sensing electrically active cells, networks and tissue. the techniques described herein relate to the use of microfabrication in combination with certain large-area processes that have been employed to achieve multiwell meas in ansi/sbs-compliant culture well formats, which are also transparent for inverted/backside microscopy compatibility. these multiwell meas can be used to investigate two and three-dimensional networks of electrically active cells and tissue such as cardiac, neural, and muscular in a high throughput fashion. also being ansi/sbs-compliant, they are compatible with machinery and robotics developed for the pharmaceutical industry for drug screening applications.",2016-03-22,"The title of the patent is apparatus and methods for high throughput network electrophysiology and cellular analysis and its abstract is provided herein are apparatus and methods relating to the development of instrumentation for high throughput network electrophysiology and cellular analysis. more specifically, provided herein are multiwell microelectrode arrays (meas) and methods for the development of such an apparatus in an inexpensive fashion with a flexible, ansi/sbs-compliant (american national standards institute/society for biomolecular screening) format. microelectrode arrays are a grid of tightly spaced microelectrodes useful for stimulating and sensing electrically active cells, networks and tissue. the techniques described herein relate to the use of microfabrication in combination with certain large-area processes that have been employed to achieve multiwell meas in ansi/sbs-compliant culture well formats, which are also transparent for inverted/backside microscopy compatibility. these multiwell meas can be used to investigate two and three-dimensional networks of electrically active cells and tissue such as cardiac, neural, and muscular in a high throughput fashion. also being ansi/sbs-compliant, they are compatible with machinery and robotics developed for the pharmaceutical industry for drug screening applications. dated 2016-03-22"
9291486,method and system for measuring fluid flow in bell nipples using pressure measurement,the volumetric rate of flow out of a flow line (4) of a bell nipple (2) is a function of the pressure of the head of liquid (7) above the threshold of the intersection of the vertical pipe (3) and the flow line (4). a pressure which is related to that pressure is measured by a pressure transducer (10) which is mounted below the threshold. density of the fluid (7) is inferred by using the measured pressure as an input to a trained neural network.,2016-03-22,The title of the patent is method and system for measuring fluid flow in bell nipples using pressure measurement and its abstract is the volumetric rate of flow out of a flow line (4) of a bell nipple (2) is a function of the pressure of the head of liquid (7) above the threshold of the intersection of the vertical pipe (3) and the flow line (4). a pressure which is related to that pressure is measured by a pressure transducer (10) which is mounted below the threshold. density of the fluid (7) is inferred by using the measured pressure as an input to a trained neural network. dated 2016-03-22
9292787,computer-implemented deep tensor neural network,"a deep tensor neural network (dtnn) is described herein, wherein the dtnn is suitable for employment in a computer-implemented recognition/classification system. hidden layers in the dtnn comprise at least one projection layer, which includes a first subspace of hidden units and a second subspace of hidden units. the first subspace of hidden units receives a first nonlinear projection of input data to a projection layer and generates the first set of output data based at least in part thereon, and the second subspace of hidden units receives a second nonlinear projection of the input data to the projection layer and generates the second set of output data based at least in part thereon. a tensor layer, which can converted into a conventional layer of a dnn, generates the third set of output data based upon the first set of output data and the second set of output data.",2016-03-22,"The title of the patent is computer-implemented deep tensor neural network and its abstract is a deep tensor neural network (dtnn) is described herein, wherein the dtnn is suitable for employment in a computer-implemented recognition/classification system. hidden layers in the dtnn comprise at least one projection layer, which includes a first subspace of hidden units and a second subspace of hidden units. the first subspace of hidden units receives a first nonlinear projection of input data to a projection layer and generates the first set of output data based at least in part thereon, and the second subspace of hidden units receives a second nonlinear projection of the input data to the projection layer and generates the second set of output data based at least in part thereon. a tensor layer, which can converted into a conventional layer of a dnn, generates the third set of output data based upon the first set of output data and the second set of output data. dated 2016-03-22"
9292788,event-driven universal neural network circuit,the present invention provides an event-driven universal neural network circuit. the circuit comprises a plurality of neural modules. each neural module comprises multiple digital neurons such that each neuron in a neural module has a corresponding neuron in another neural module. an interconnection network comprising a plurality of digital synapses interconnects the neural modules. each synapse interconnects a first neural module to a second neural module by interconnecting a neuron in the first neural module to a corresponding neuron in the second neural module. corresponding neurons in the first neural module and the second neural module communicate via the synapses. each synapse comprises a learning rule associating a neuron in the first neural module with a corresponding neuron in the second neural module. a control module generates signals which define a set of time steps for event-driven operation of the neurons and event communication via the interconnection network.,2016-03-22,The title of the patent is event-driven universal neural network circuit and its abstract is the present invention provides an event-driven universal neural network circuit. the circuit comprises a plurality of neural modules. each neural module comprises multiple digital neurons such that each neuron in a neural module has a corresponding neuron in another neural module. an interconnection network comprising a plurality of digital synapses interconnects the neural modules. each synapse interconnects a first neural module to a second neural module by interconnecting a neuron in the first neural module to a corresponding neuron in the second neural module. corresponding neurons in the first neural module and the second neural module communicate via the synapses. each synapse comprises a learning rule associating a neuron in the first neural module with a corresponding neuron in the second neural module. a control module generates signals which define a set of time steps for event-driven operation of the neurons and event communication via the interconnection network. dated 2016-03-22
9292789,continuous-weight neural networks,a computer-based multi-layer artificial network named continuous-weight neural network (cwnn) configured to receive an input feature set wherein the input feature set comprises a variable number of features is disclosed. a method for classifying input sets based on a trained cwnn is also disclosed. various implementation examples are also provided.,2016-03-22,The title of the patent is continuous-weight neural networks and its abstract is a computer-based multi-layer artificial network named continuous-weight neural network (cwnn) configured to receive an input feature set wherein the input feature set comprises a variable number of features is disclosed. a method for classifying input sets based on a trained cwnn is also disclosed. various implementation examples are also provided. dated 2016-03-22
9295429,predicting acute cardiopulmonary events and survivability of a patient,"a method of predicting survivability of a patient. the method includes storing in an electronic database patient health data comprising a plurality of sets of data, each set having a first parameter relating to heart rate variability data including at least one of st segment elevation and depression, a second parameter relating to vital sign data, and a third parameter relating to patient survivability; providing a network of nodes interconnected to form an artificial neural network, the nodes comprising a plurality of neurons, each having at least one input with an associated weight; and training the neural network using the patient health data such that the associated weight of the at least one input of each neuron is adjusted in response to respective first, second and third parameters of different sets of data from the patient health data, such that the neural network is trained to produce a prediction on the survivability of a patient within the next 72 hours.",2016-03-29,"The title of the patent is predicting acute cardiopulmonary events and survivability of a patient and its abstract is a method of predicting survivability of a patient. the method includes storing in an electronic database patient health data comprising a plurality of sets of data, each set having a first parameter relating to heart rate variability data including at least one of st segment elevation and depression, a second parameter relating to vital sign data, and a third parameter relating to patient survivability; providing a network of nodes interconnected to form an artificial neural network, the nodes comprising a plurality of neurons, each having at least one input with an associated weight; and training the neural network using the patient health data such that the associated weight of the at least one input of each neuron is adjusted in response to respective first, second and third parameters of different sets of data from the patient health data, such that the neural network is trained to produce a prediction on the survivability of a patient within the next 72 hours. dated 2016-03-29"
9299010,data fusion analysis for maritime automatic target recognition,"a system and method for performing automatic target recognition by combining the outputs of several classifiers. in one embodiment, feature vectors are extracted from radar images and fed to three classifiers. the classifiers include a gaussian mixture model neural network, a radial basis function neural network, and a vector quantization classifier. the class designations generated by the classifiers are combined in a weighted voting system, i.e., the mode of the weighted classification decisions is selected as the overall class designation of the target. a confidence metric may be formed from the extent to which the class designations of the several classifiers are the same. this system is also designed to handle unknown target types and subsequent re-integration at a later time, effectively, artificially and automatically increasing the training database size.",2016-03-29,"The title of the patent is data fusion analysis for maritime automatic target recognition and its abstract is a system and method for performing automatic target recognition by combining the outputs of several classifiers. in one embodiment, feature vectors are extracted from radar images and fed to three classifiers. the classifiers include a gaussian mixture model neural network, a radial basis function neural network, and a vector quantization classifier. the class designations generated by the classifiers are combined in a weighted voting system, i.e., the mode of the weighted classification decisions is selected as the overall class designation of the target. a confidence metric may be formed from the extent to which the class designations of the several classifiers are the same. this system is also designed to handle unknown target types and subsequent re-integration at a later time, effectively, artificially and automatically increasing the training database size. dated 2016-03-29"
9299022,intelligent modular robotic apparatus and methods,"apparatus and methods for an extensible robotic device with artificial intelligence and receptive to training controls. in one implementation, a modular robotic system that allows a user to fully select the architecture and capability set of their robotic device is disclosed. the user may add/remove modules as their respective functions are required/obviated. in addition, the artificial intelligence is based on a neuronal network (e.g., spiking neural network), and a behavioral control structure that allows a user to train a robotic device in manner conceptually similar to the mode in which one goes about training a domesticated animal such as a dog or cat (e.g., a positive/negative feedback training paradigm) is used. the trainable behavior control structure is based on the artificial neural network, which simulates the neural/synaptic activity of the brain of a living organism.",2016-03-29,"The title of the patent is intelligent modular robotic apparatus and methods and its abstract is apparatus and methods for an extensible robotic device with artificial intelligence and receptive to training controls. in one implementation, a modular robotic system that allows a user to fully select the architecture and capability set of their robotic device is disclosed. the user may add/remove modules as their respective functions are required/obviated. in addition, the artificial intelligence is based on a neuronal network (e.g., spiking neural network), and a behavioral control structure that allows a user to train a robotic device in manner conceptually similar to the mode in which one goes about training a domesticated animal such as a dog or cat (e.g., a positive/negative feedback training paradigm) is used. the trainable behavior control structure is based on the artificial neural network, which simulates the neural/synaptic activity of the brain of a living organism. dated 2016-03-29"
9305050,"aggregator, filter and delivery system for online context dependent interaction, systems and methods","a method of providing information to a user is provided. the method includes; establishing an user system interface between a client device and an information system; processing informal queries input from the client device with at least one neural network that converts the informal queries from the client device into formal queries; storing interface context in a browser of the client device, the interface context created in forming formal queries from informal queries, wherein the client device contains unique interface context in the client device's browser that is secure to the client device, the interface context aiding in the determination of future formal queries from future informal queries; searching at least one database in response to the formal queries; and providing responses to the informal queries processed by the neural network to a user through the client device.",2016-04-05,"The title of the patent is aggregator, filter and delivery system for online context dependent interaction, systems and methods and its abstract is a method of providing information to a user is provided. the method includes; establishing an user system interface between a client device and an information system; processing informal queries input from the client device with at least one neural network that converts the informal queries from the client device into formal queries; storing interface context in a browser of the client device, the interface context created in forming formal queries from informal queries, wherein the client device contains unique interface context in the client device's browser that is secure to the client device, the interface context aiding in the determination of future formal queries from future informal queries; searching at least one database in response to the formal queries; and providing responses to the informal queries processed by the neural network to a user through the client device. dated 2016-04-05"
9311546,biometric identity verification for access control using a trained statistical classifier,"a method and apparatus for providing biometric authentication of a user uses a registration process in which a reference data sample representative of a biometric attribute of a reference user is used to train a statistical classifier such as a neural network to achieve a target output. the set of parameters of the statistical classifier, e.g. the weights that achieve this in the neural network, are stored on a user's device as a first data set. for subsequent authentication of a user to be tested at an access point, the first data set is retrieved from the user device and a second data set representative of the biometric attribute of the test user is generated directly from the test user. the first data set is used as a set of parameters in a statistical classifier, e.g. as weights in an artificial neural network, to generate a trained classifier or neural network and the second data set is then used as input to the trained classifier or neural network. the output of the trained classifier or neural network is then used to determine a degree of correlation between the biometric attribute of the reference user and the biometric attribute of the test user to be authenticated.",2016-04-12,"The title of the patent is biometric identity verification for access control using a trained statistical classifier and its abstract is a method and apparatus for providing biometric authentication of a user uses a registration process in which a reference data sample representative of a biometric attribute of a reference user is used to train a statistical classifier such as a neural network to achieve a target output. the set of parameters of the statistical classifier, e.g. the weights that achieve this in the neural network, are stored on a user's device as a first data set. for subsequent authentication of a user to be tested at an access point, the first data set is retrieved from the user device and a second data set representative of the biometric attribute of the test user is generated directly from the test user. the first data set is used as a set of parameters in a statistical classifier, e.g. as weights in an artificial neural network, to generate a trained classifier or neural network and the second data set is then used as input to the trained classifier or neural network. the output of the trained classifier or neural network is then used to determine a degree of correlation between the biometric attribute of the reference user and the biometric attribute of the test user to be authenticated. dated 2016-04-12"
9311595,neural network device with engineered delays for pattern storage and matching,"described is a system for searching a continuous data stream for exact matches with a priori stored data sequences. the system includes a neural network with an input and an output layer. the input layer has one neuron for each possible character or number in the data stream, and the output layer has one neuron for each stored pattern. importantly, the delays of the connections from input to output layer are engineered to match the temporal occurrence of an input character within a stored sequence. thus, if an input sequence has the proper time gaps between characters, matching a stored pattern, then the delayed neural signals result in a simultaneous activation at the receiving neuron, which indicates a detected pattern. for storing a pattern, only one connection for each pair of input character and output neuron has to be specified resulting in sparse coding and quick storage.",2016-04-12,"The title of the patent is neural network device with engineered delays for pattern storage and matching and its abstract is described is a system for searching a continuous data stream for exact matches with a priori stored data sequences. the system includes a neural network with an input and an output layer. the input layer has one neuron for each possible character or number in the data stream, and the output layer has one neuron for each stored pattern. importantly, the delays of the connections from input to output layer are engineered to match the temporal occurrence of an input character within a stored sequence. thus, if an input sequence has the proper time gaps between characters, matching a stored pattern, then the delayed neural signals result in a simultaneous activation at the receiving neuron, which indicates a detected pattern. for storing a pattern, only one connection for each pair of input character and output neuron has to be specified resulting in sparse coding and quick storage. dated 2016-04-12"
9311915,context-based speech recognition,"a processing system receives an audio signal encoding a portion of an utterance. the processing system receives context information associated with the utterance, wherein the context information is not derived from the audio signal or any other audio signal. the processing system provides, as input to a neural network, data corresponding to the audio signal and the context information, and generates a transcription for the utterance based on at least an output of the neural network.",2016-04-12,"The title of the patent is context-based speech recognition and its abstract is a processing system receives an audio signal encoding a portion of an utterance. the processing system receives context information associated with the utterance, wherein the context information is not derived from the audio signal or any other audio signal. the processing system provides, as input to a neural network, data corresponding to the audio signal and the context information, and generates a transcription for the utterance based on at least an output of the neural network. dated 2016-04-12"
9317779,training an image processing neural network without human selection of features,"a method for training an image processing neural network without human selection of features may include providing a training set of images labeled with two or more classifications, providing an image processing toolbox with image transforms that can be applied to the training set, generating a random set of feature extraction pipelines, where each feature extraction pipeline includes a sequence of image transforms randomly selected from the image processing toolbox and randomly selected control parameters associated with the sequence of image transforms. the method may also include coupling a first stage classifier to an output of each feature extraction pipeline and executing a genetic algorithm to conduct genetic modification of each feature extraction pipeline and train each first stage classifier on the training set, and coupling a second stage classifier to each of the first stage classifiers in order to increase classification accuracy.",2016-04-19,"The title of the patent is training an image processing neural network without human selection of features and its abstract is a method for training an image processing neural network without human selection of features may include providing a training set of images labeled with two or more classifications, providing an image processing toolbox with image transforms that can be applied to the training set, generating a random set of feature extraction pipelines, where each feature extraction pipeline includes a sequence of image transforms randomly selected from the image processing toolbox and randomly selected control parameters associated with the sequence of image transforms. the method may also include coupling a first stage classifier to an output of each feature extraction pipeline and executing a genetic algorithm to conduct genetic modification of each feature extraction pipeline and train each first stage classifier on the training set, and coupling a second stage classifier to each of the first stage classifiers in order to increase classification accuracy. dated 2016-04-19"
9322768,sensor system and method for the panoramic standoff detection of chemical contaminants on terrestrial surfaces,"a pseudo-active chemical imaging sensor including irradiative transient heating, temperature nonequilibrium thermal luminescence spectroscopy, differential hyperspectral imaging, and artificial neural network technologies integrated together. the sensor may be applied to the terrestrial chemical contamination problem, where the interstitial contaminant compounds of detection interest (analytes) comprise liquid chemical warfare agents, their various derivative condensed phase compounds, and other material of a life-threatening nature. the sensor measures and processes a dynamic pattern of absorptive-emissive middle infrared molecular signature spectra of subject analytes to perform its chemical imaging and standoff detection functions successfully.",2016-04-26,"The title of the patent is sensor system and method for the panoramic standoff detection of chemical contaminants on terrestrial surfaces and its abstract is a pseudo-active chemical imaging sensor including irradiative transient heating, temperature nonequilibrium thermal luminescence spectroscopy, differential hyperspectral imaging, and artificial neural network technologies integrated together. the sensor may be applied to the terrestrial chemical contamination problem, where the interstitial contaminant compounds of detection interest (analytes) comprise liquid chemical warfare agents, their various derivative condensed phase compounds, and other material of a life-threatening nature. the sensor measures and processes a dynamic pattern of absorptive-emissive middle infrared molecular signature spectra of subject analytes to perform its chemical imaging and standoff detection functions successfully. dated 2016-04-26"
9324320,neural network-based speech processing,"pairs of feature vectors are obtained that represent speech. some pairs represent two samples of speech from the same speakers, and other pairs represent two samples of speech from different speakers. a neural network feeds each feature vector in a sample pair into a separate bottleneck layer, with a weight matrix on the input of both vectors tied to one another. the neural network is trained using the feature vectors and an objective function that induces the network to classify whether the speech samples come from the same speaker. the weights from the tied weight matrix are extracted for use in generating derived features for a speech processing system that can benefit from features that are thus transformed to better reflect speaker identity.",2016-04-26,"The title of the patent is neural network-based speech processing and its abstract is pairs of feature vectors are obtained that represent speech. some pairs represent two samples of speech from the same speakers, and other pairs represent two samples of speech from different speakers. a neural network feeds each feature vector in a sample pair into a separate bottleneck layer, with a weight matrix on the input of both vectors tied to one another. the neural network is trained using the feature vectors and an objective function that induces the network to classify whether the speech samples come from the same speaker. the weights from the tied weight matrix are extracted for use in generating derived features for a speech processing system that can benefit from features that are thus transformed to better reflect speaker identity. dated 2016-04-26"
9324321,low-footprint adaptation and personalization for a deep neural network,"the adaptation and personalization of a deep neural network (dnn) model for automatic speech recognition is provided. an utterance which includes speech features for one or more speakers may be received in asr tasks such as voice search or short message dictation. a decomposition approach may then be applied to an original matrix in the dnn model. in response to applying the decomposition approach, the original matrix may be converted into multiple new matrices which are smaller than the original matrix. a square matrix may then be added to the new matrices. speaker-specific parameters may then be stored in the square matrix. the dnn model may then be adapted by updating the square matrix. this process may be applied to all of a number of original matrices in the dnn model. the adapted dnn model may include a reduced number of parameters than those received in the original dnn model.",2016-04-26,"The title of the patent is low-footprint adaptation and personalization for a deep neural network and its abstract is the adaptation and personalization of a deep neural network (dnn) model for automatic speech recognition is provided. an utterance which includes speech features for one or more speakers may be received in asr tasks such as voice search or short message dictation. a decomposition approach may then be applied to an original matrix in the dnn model. in response to applying the decomposition approach, the original matrix may be converted into multiple new matrices which are smaller than the original matrix. a square matrix may then be added to the new matrices. speaker-specific parameters may then be stored in the square matrix. the dnn model may then be adapted by updating the square matrix. this process may be applied to all of a number of original matrices in the dnn model. the adapted dnn model may include a reduced number of parameters than those received in the original dnn model. dated 2016-04-26"
9328644,exhaust system and method of estimating diesel particulate filter soot loading for same using two-tier neural network,"a method of estimating soot loading in a diesel particulate filter (dpf) in a vehicle exhaust system includes estimating an engine-out soot rate using a first neural network that has a first set of vehicle operating conditions as inputs. the method further includes estimating dpf soot loading using a second neural network that has the estimated engine-out soot rate from the first neural network and a second set of vehicle operating conditions as inputs. estimating the engine-out soot rate and estimating the dpf soot loading are performed by an electronic controller that executes the first and the second neural networks. the method also provides for training the first and second neural networks both offline (for initial settings of the neural networks in the vehicle), and online (when the vehicle is being used by a vehicle operator). an exhaust system has a controller that implements the method.",2016-05-03,"The title of the patent is exhaust system and method of estimating diesel particulate filter soot loading for same using two-tier neural network and its abstract is a method of estimating soot loading in a diesel particulate filter (dpf) in a vehicle exhaust system includes estimating an engine-out soot rate using a first neural network that has a first set of vehicle operating conditions as inputs. the method further includes estimating dpf soot loading using a second neural network that has the estimated engine-out soot rate from the first neural network and a second set of vehicle operating conditions as inputs. estimating the engine-out soot rate and estimating the dpf soot loading are performed by an electronic controller that executes the first and the second neural networks. the method also provides for training the first and second neural networks both offline (for initial settings of the neural networks in the vehicle), and online (when the vehicle is being used by a vehicle operator). an exhaust system has a controller that implements the method. dated 2016-05-03"
9332028,"system, method, and apparatus for providing network security","methods, systems, and apparatuses for proactively protecting a computing network are disclosed. a proactive security mechanism is disclosed, among other things, with the ability to monitor a protected domain in real-time and safely identify inoculation procedures for responding to threats introduced to the protected domain via malware. the proactive security mechanism includes an artificial neural network interface (anni) configured to execute at least some features of the proactive security mechanism.",2016-05-03,"The title of the patent is system, method, and apparatus for providing network security and its abstract is methods, systems, and apparatuses for proactively protecting a computing network are disclosed. a proactive security mechanism is disclosed, among other things, with the ability to monitor a protected domain in real-time and safely identify inoculation procedures for responding to threats introduced to the protected domain via malware. the proactive security mechanism includes an artificial neural network interface (anni) configured to execute at least some features of the proactive security mechanism. dated 2016-05-03"
9336239,system and method for deep packet inspection and intrusion detection,"the present invention relates to a system for deep packet inspection and intrusion detection. the system uses a pattern matching module receiving as an input a data stream in a neural network. neurons are activated such that when active, the neuron fires to all connecting output neurons to form a neuron spike, each neuron spike from the assigned neuron to a connecting output neuron having a delay. a delay is associated with each input character in the pattern, such that a position of each input character relative to an end of the pattern is stored in an alphabet-pattern-delay matrix (apdfm). an activation matrix (am) is used to match each input character with a stored pattern to generate a similarity match and determine if the string of characters is the stored pattern.",2016-05-10,"The title of the patent is system and method for deep packet inspection and intrusion detection and its abstract is the present invention relates to a system for deep packet inspection and intrusion detection. the system uses a pattern matching module receiving as an input a data stream in a neural network. neurons are activated such that when active, the neuron fires to all connecting output neurons to form a neuron spike, each neuron spike from the assigned neuron to a connecting output neuron having a delay. a delay is associated with each input character in the pattern, such that a position of each input character relative to an end of the pattern is stored in an alphabet-pattern-delay matrix (apdfm). an activation matrix (am) is used to match each input character with a stored pattern to generate a similarity match and determine if the string of characters is the stored pattern. dated 2016-05-10"
9336482,predicting likelihoods of conditions being satisfied using recurrent neural networks,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for predicting likelihoods of conditions being satisfied using recurrent neural networks. one of the systems is configured to process a temporal sequence comprising a respective input at each of a plurality of time steps and comprises: one or more recurrent neural network layers; one or more logistic regression nodes, wherein each of the logistic regression nodes corresponds to a respective condition from a predetermined set of conditions, and wherein each of the logistic regression nodes is configured to, for each of the plurality of time steps: receive the network internal state for the time step; and process the network internal state for the time step in accordance with current values of a set of parameters of the logistic regression node to generate a future condition score for the corresponding condition for the time step.",2016-05-10,"The title of the patent is predicting likelihoods of conditions being satisfied using recurrent neural networks and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for predicting likelihoods of conditions being satisfied using recurrent neural networks. one of the systems is configured to process a temporal sequence comprising a respective input at each of a plurality of time steps and comprises: one or more recurrent neural network layers; one or more logistic regression nodes, wherein each of the logistic regression nodes corresponds to a respective condition from a predetermined set of conditions, and wherein each of the logistic regression nodes is configured to, for each of the plurality of time steps: receive the network internal state for the time step; and process the network internal state for the time step in accordance with current values of a set of parameters of the logistic regression node to generate a future condition score for the corresponding condition for the time step. dated 2016-05-10"
9336483,dynamically updated neural network structures for content distribution networks,"dynamically updating neural network systems may be implemented to generate, train, evaluate and update artificial neural network data structures used by content distribution networks. such systems and methods described herein may include generating and training neural networks, using neural networks to perform predictive analysis and other decision-making processes within content distribution networks, evaluating the performance of neural networks, and generating and training pluralities of replacement candidate neural networks within cloud computing architectures and/or other computing environments.",2016-05-10,"The title of the patent is dynamically updated neural network structures for content distribution networks and its abstract is dynamically updating neural network systems may be implemented to generate, train, evaluate and update artificial neural network data structures used by content distribution networks. such systems and methods described herein may include generating and training neural networks, using neural networks to perform predictive analysis and other decision-making processes within content distribution networks, evaluating the performance of neural networks, and generating and training pluralities of replacement candidate neural networks within cloud computing architectures and/or other computing environments. dated 2016-05-10"
9342782,stochastic delay plasticity,a method of operating a spiking neural network having neurons coupled together with a synapse includes monitoring a timing of a presynaptic spike and monitoring a timing of a postsynaptic spike. the method also includes determining a time difference between the postsynaptic spike and the presynaptic spike. the method further includes calculating a stochastic update of a delay for the synapse based on the time difference between the postsynaptic spike and the presynaptic spike.,2016-05-17,The title of the patent is stochastic delay plasticity and its abstract is a method of operating a spiking neural network having neurons coupled together with a synapse includes monitoring a timing of a presynaptic spike and monitoring a timing of a postsynaptic spike. the method also includes determining a time difference between the postsynaptic spike and the presynaptic spike. the method further includes calculating a stochastic update of a delay for the synapse based on the time difference between the postsynaptic spike and the presynaptic spike. dated 2016-05-17
9342793,training a self-learning network using interpolated input sets based on a target output,"embodiments relate to systems and methods for training a self-learning network using interpolated input sets based on a target output. a database management system can store sets of operational data, such as financial, medical, climate or other information. a user can input or access a set of target data, representing an output which a user wishes to be generated from an interpolated set of input data. the interpolation engine can generate a conformal interpolation function and input sets that map to the set of target output data. after interpolation, the interpolation engine can transmit the interpolated inputs, along with the set of target output data and other information, to a self-learning network such as a neural or fuzzy logic network. the self-learning network can be trained to converge to the target output based on the interpolated input results as generated by the interpolation engine, thus reproducing the desired interpolation function.",2016-05-17,"The title of the patent is training a self-learning network using interpolated input sets based on a target output and its abstract is embodiments relate to systems and methods for training a self-learning network using interpolated input sets based on a target output. a database management system can store sets of operational data, such as financial, medical, climate or other information. a user can input or access a set of target data, representing an output which a user wishes to be generated from an interpolated set of input data. the interpolation engine can generate a conformal interpolation function and input sets that map to the set of target output data. after interpolation, the interpolation engine can transmit the interpolated inputs, along with the set of target output data and other information, to a self-learning network such as a neural or fuzzy logic network. the self-learning network can be trained to converge to the target output based on the interpolated input results as generated by the interpolation engine, thus reproducing the desired interpolation function. dated 2016-05-17"
9345413,heart rate extraction using neural wavelet adaptive gain control and neural pattern processing,"an improved heart rate monitor is provided that can detect and distinguish a heartbeat from an otherwise contaminated system with noise components potentially larger than the signal of interest. embodiments of the inventive monitor have an amplification system that eliminates large noise components so as not to saturate the system during detection of a desired low amplitude signal. in embodiments the elimination of noise components is accomplished through wavelet decomposition, and the removal of undesired components including interference components during adaptive gain control (agc), in addition to hunting algorithms which minimize the error with techniques such as neural network least mean squares type back propagation algorithms.",2016-05-24,"The title of the patent is heart rate extraction using neural wavelet adaptive gain control and neural pattern processing and its abstract is an improved heart rate monitor is provided that can detect and distinguish a heartbeat from an otherwise contaminated system with noise components potentially larger than the signal of interest. embodiments of the inventive monitor have an amplification system that eliminates large noise components so as not to saturate the system during detection of a desired low amplitude signal. in embodiments the elimination of noise components is accomplished through wavelet decomposition, and the removal of undesired components including interference components during adaptive gain control (agc), in addition to hunting algorithms which minimize the error with techniques such as neural network least mean squares type back propagation algorithms. dated 2016-05-24"
9347430,adaptive pitch control system for wind generators,"the adaptive pitch control system for wind generators is utilized in variable speed doubly fed induction generator (dfig) systems. an adaptive neural network generates optimized controller gains for pitch control. the pitch controller parameters are generated using intelligent differential evolution, a type of genetic algorithm. a back propagation neural network is trained using the generated pitch controller parameters, thereby tuning the weights of the network according to the system states in a variable wind speed environment.",2016-05-24,"The title of the patent is adaptive pitch control system for wind generators and its abstract is the adaptive pitch control system for wind generators is utilized in variable speed doubly fed induction generator (dfig) systems. an adaptive neural network generates optimized controller gains for pitch control. the pitch controller parameters are generated using intelligent differential evolution, a type of genetic algorithm. a back propagation neural network is trained using the generated pitch controller parameters, thereby tuning the weights of the network according to the system states in a variable wind speed environment. dated 2016-05-24"
9349092,neural network for reinforcement learning,"a neural model for reinforcement-learning and for action-selection includes a plurality of channels, a population of input neurons in each of the channels, a population of output neurons in each of the channels, each population of input neurons in each of the channels coupled to each population of output neurons in each of the channels, and a population of reward neurons in each of the channels. each channel of a population of reward neurons receives input from an environmental input, and is coupled only to output neurons in a channel that the reward neuron is part of. if the environmental input for a channel is positive, the corresponding channel of a population of output neurons are rewarded and have their responses reinforced, otherwise the corresponding channel of a population of output neurons are punished and have their responses attenuated.",2016-05-24,"The title of the patent is neural network for reinforcement learning and its abstract is a neural model for reinforcement-learning and for action-selection includes a plurality of channels, a population of input neurons in each of the channels, a population of output neurons in each of the channels, each population of input neurons in each of the channels coupled to each population of output neurons in each of the channels, and a population of reward neurons in each of the channels. each channel of a population of reward neurons receives input from an environmental input, and is coupled only to output neurons in a channel that the reward neuron is part of. if the environmental input for a channel is positive, the corresponding channel of a population of output neurons are rewarded and have their responses reinforced, otherwise the corresponding channel of a population of output neurons are punished and have their responses attenuated. dated 2016-05-24"
9361534,image recognition apparatus using neural network processing,"an image recognition apparatus determines whether an image of a pedestrian is captured in a frame of video data captured by a vehicle mounted camera. a pre-processing unit determines a detection block from within a frame, and cuts out block image data corresponding to the detection block from the frame. block data with a predetermined size that is smaller than the size of the detection block is created from the block image data. a neuro calculation unit executes neuro calculation on the block data, and calculates an output synapse. a post-processing unit determines whether a pedestrian exists within the detection block on the basis of the output synapse. when a pedestrian is detected, the post-processing unit creates result data, which is obtained by superimposing the detection block within which the pedestrian was detected onto the frame.",2016-06-07,"The title of the patent is image recognition apparatus using neural network processing and its abstract is an image recognition apparatus determines whether an image of a pedestrian is captured in a frame of video data captured by a vehicle mounted camera. a pre-processing unit determines a detection block from within a frame, and cuts out block image data corresponding to the detection block from the frame. block data with a predetermined size that is smaller than the size of the detection block is created from the block image data. a neuro calculation unit executes neuro calculation on the block data, and calculates an output synapse. a post-processing unit determines whether a pedestrian exists within the detection block on the basis of the output synapse. when a pedestrian is detected, the post-processing unit creates result data, which is obtained by superimposing the detection block within which the pedestrian was detected onto the frame. dated 2016-06-07"
9361575,method of programming a neural network computer,"a method is disclosed for programming a target neural network computer for use in cognitive computing systems in automotive safety applications. an observer neural network computer is integrated into active safety systems of a plurality of vehicles to observe signals. each respective observer neural network computer is arranged to observe signals from a forward facing camera and signals from a driver action monitor of its respective vehicle, to process the observed signals from the forward facing camera of its respective vehicle and correlate them with the observed signals from the driver action monitor of its respective vehicle. the correlated signals from the plurality of observer neural network computers are combined, and the target neural network computer is programmed for use in cognitive computing systems in automotive safety applications based on said combined correlated signals.",2016-06-07,"The title of the patent is method of programming a neural network computer and its abstract is a method is disclosed for programming a target neural network computer for use in cognitive computing systems in automotive safety applications. an observer neural network computer is integrated into active safety systems of a plurality of vehicles to observe signals. each respective observer neural network computer is arranged to observe signals from a forward facing camera and signals from a driver action monitor of its respective vehicle, to process the observed signals from the forward facing camera of its respective vehicle and correlate them with the observed signals from the driver action monitor of its respective vehicle. the correlated signals from the plurality of observer neural network computers are combined, and the target neural network computer is programmed for use in cognitive computing systems in automotive safety applications based on said combined correlated signals. dated 2016-06-07"
9367288,device and method responsive to influences of mind,"an anomalous effect detector responsive to an influence of mind comprises a source of non-deterministic random numbers, sndrn, a phase-sensitive filter and a results interface. in some embodiments, the phase-sensitive filter comprises a complex filter. an artificial sensory neuron comprises a sndrn. preferably, several artificial sensory neurons are grouped in a small volume. an analog artificial sensory detector comprises a plurality of analog artificial sensory neurons, an abstracting processor and a control or feedback unit. some embodiments include an artificial neural network. an artificial consciousness network contains a plurality of artificial neural networks. one of the artificial neural networks comprises an activation pattern meta-analyzer. an artificial consciousness device comprises a cluster of artificial consciousness networks, a sensory input device to provide sensory input signals to the input of one or more anns in acd, and an output device.",2016-06-14,"The title of the patent is device and method responsive to influences of mind and its abstract is an anomalous effect detector responsive to an influence of mind comprises a source of non-deterministic random numbers, sndrn, a phase-sensitive filter and a results interface. in some embodiments, the phase-sensitive filter comprises a complex filter. an artificial sensory neuron comprises a sndrn. preferably, several artificial sensory neurons are grouped in a small volume. an analog artificial sensory detector comprises a plurality of analog artificial sensory neurons, an abstracting processor and a control or feedback unit. some embodiments include an artificial neural network. an artificial consciousness network contains a plurality of artificial neural networks. one of the artificial neural networks comprises an activation pattern meta-analyzer. an artificial consciousness device comprises a cluster of artificial consciousness networks, a sensory input device to provide sensory input signals to the input of one or more anns in acd, and an output device. dated 2016-06-14"
9367796,method for predicting the properties of crude oils by the application of neural networks,"a method for predicting the properties of crude oils by the application of neural networks articulated in phases and characterized by determining the t2 nmr relaxation curve of an unknown crude oil and converting it to a logarithmic relaxation curve; selecting the values of the logarithmic relaxation curve lying on a characterization grid; entering the selected values as input data for a multilayer neural network of the back propagation type, trained and optimized by means of genetic algorithms; predicting, by means of the trained and optimized neural network, the physico-chemical factors of the unknown crude oil.the method comprises a training and optimization process of the multilayer neural network of the back propagation type.the method thus defined allows the most representative physico-chemical factors of crude oils to be predicted rapidly and without onerous laboratory structures, or alternatively the distillation curve of crude oils with an acceptable approximation degree.",2016-06-14,"The title of the patent is method for predicting the properties of crude oils by the application of neural networks and its abstract is a method for predicting the properties of crude oils by the application of neural networks articulated in phases and characterized by determining the t2 nmr relaxation curve of an unknown crude oil and converting it to a logarithmic relaxation curve; selecting the values of the logarithmic relaxation curve lying on a characterization grid; entering the selected values as input data for a multilayer neural network of the back propagation type, trained and optimized by means of genetic algorithms; predicting, by means of the trained and optimized neural network, the physico-chemical factors of the unknown crude oil.the method comprises a training and optimization process of the multilayer neural network of the back propagation type.the method thus defined allows the most representative physico-chemical factors of crude oils to be predicted rapidly and without onerous laboratory structures, or alternatively the distillation curve of crude oils with an acceptable approximation degree. dated 2016-06-14"
9367799,"neural network based cluster visualization that computes pairwise distances between centroid locations, and determines a projected centroid location in a multidimensional space","a computing device presents a cluster visualization based on a neural network computation. first centroid locations are computed for first clusters. second centroid locations are computed for second clusters. each centroid location includes a plurality of coordinate values where each coordinate value relates to a single variable of a plurality of variables. distances are computed pairwise between each centroid location. an optimum pairing is selected based on a minimum distance of the computed pairwise distances where each pair is associated with a different cluster of a set of composite clusters. noised centroid location data is created. a multi-layer neural network is trained with the noised centroid location data. a projected centroid location is determined in a multidimensional space for each centroid location as values of hidden units of a middle layer of the multi-layer neural network. a graph is presented for display that indicates the determined, projected centroid locations.",2016-06-14,"The title of the patent is neural network based cluster visualization that computes pairwise distances between centroid locations, and determines a projected centroid location in a multidimensional space and its abstract is a computing device presents a cluster visualization based on a neural network computation. first centroid locations are computed for first clusters. second centroid locations are computed for second clusters. each centroid location includes a plurality of coordinate values where each coordinate value relates to a single variable of a plurality of variables. distances are computed pairwise between each centroid location. an optimum pairing is selected based on a minimum distance of the computed pairwise distances where each pair is associated with a different cluster of a set of composite clusters. noised centroid location data is created. a multi-layer neural network is trained with the noised centroid location data. a projected centroid location is determined in a multidimensional space for each centroid location as values of hidden units of a middle layer of the multi-layer neural network. a graph is presented for display that indicates the determined, projected centroid locations. dated 2016-06-14"
9368110,method for distinguishing components of an acoustic signal,"a method distinguishes components of an acoustic signal by processing the signal to estimate a set of analysis features, wherein each analysis feature defines an element of the signal and has feature values that represent parts of the signal, processing the signal to estimate input features of the signal, and processing the input features using a deep neural network to assign an associative descriptor to each element of the signal, wherein a degree of similarity between the associative descriptors of different elements is related to a degree to which the parts of the signal represented by the elements belong to a single component of the signal. the the similarities between associative descriptors are processed to estimate correspondences between the elements of the signal and the components in the signal. then, the signal is processed using the correspondences to distinguish component parts of the signal.",2016-06-14,"The title of the patent is method for distinguishing components of an acoustic signal and its abstract is a method distinguishes components of an acoustic signal by processing the signal to estimate a set of analysis features, wherein each analysis feature defines an element of the signal and has feature values that represent parts of the signal, processing the signal to estimate input features of the signal, and processing the input features using a deep neural network to assign an associative descriptor to each element of the signal, wherein a degree of similarity between the associative descriptors of different elements is related to a degree to which the parts of the signal represented by the elements belong to a single component of the signal. the the similarities between associative descriptors are processed to estimate correspondences between the elements of the signal and the components in the signal. then, the signal is processed using the correspondences to distinguish component parts of the signal. dated 2016-06-14"
9370316,mri estimation of contrast agent concentration using a neural network approach,"the invention comprises systems, methods, and apparatus to correlate changes in mri data with ca concentration, using an adaptive neural network in mri techniques, cas are used to estimate vascular properties such as blood flow, blood volume, and transfer constant of tissue microvessels however, the relationship between the contrast in the mri image and the contrast agent concentration is not linear, instead depending on factors such as the nature of the sequence, the nature of the tissue, and the tissue concentration of contrast agent, and thus limiting the reliability of vascular properties using mri.",2016-06-21,"The title of the patent is mri estimation of contrast agent concentration using a neural network approach and its abstract is the invention comprises systems, methods, and apparatus to correlate changes in mri data with ca concentration, using an adaptive neural network in mri techniques, cas are used to estimate vascular properties such as blood flow, blood volume, and transfer constant of tissue microvessels however, the relationship between the contrast in the mri image and the contrast agent concentration is not linear, instead depending on factors such as the nature of the sequence, the nature of the tissue, and the tissue concentration of contrast agent, and thus limiting the reliability of vascular properties using mri. dated 2016-06-21"
9373057,training a neural network to detect objects in images,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network to detect object in images. one of the methods includes receiving a training image and object location data for the training image; providing the training image to a neural network and obtaining bounding box data for the training image from the neural network, wherein the bounding box data comprises data defining a plurality of candidate bounding boxes in the training image and a respective confidence score for each candidate bounding box in the training image; determining an optimal set of assignments using the object location data for the training image and the bounding box data for the training image, wherein the optimal set of assignments assigns a respective candidate bounding box to each of the object locations; and training the neural network on the training image using the optimal set of assignments.",2016-06-21,"The title of the patent is training a neural network to detect objects in images and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network to detect object in images. one of the methods includes receiving a training image and object location data for the training image; providing the training image to a neural network and obtaining bounding box data for the training image from the neural network, wherein the bounding box data comprises data defining a plurality of candidate bounding boxes in the training image and a respective confidence score for each candidate bounding box in the training image; determining an optimal set of assignments using the object location data for the training image and the bounding box data for the training image, wherein the optimal set of assignments assigns a respective candidate bounding box to each of the object locations; and training the neural network on the training image using the optimal set of assignments. dated 2016-06-21"
9373059,systems and methods for applying a convolutional network to spatial data,"systems and methods for test object classification are provided in which the test object is docked with a target object in a plurality of different poses to form voxel maps. the maps are vectorized and fed into a convolutional neural network comprising an input layer, a plurality of individually weighted convolutional layers, and an output scorer. the convolutional layers include initial and final layers. responsive to vectorized input, the input layer feeds values into the initial convolutional layer. each respective convolutional layer, other than the final convolutional layer, feeds intermediate values as a function of the weights and input values of the respective layer into another of the convolutional layers. the final convolutional layer feeds values into one or more fully connected layers as a function of the final layer weights and input values. the one or more full connected layers feed values into the scorer which scores each input vector to thereby classify the test object.",2016-06-21,"The title of the patent is systems and methods for applying a convolutional network to spatial data and its abstract is systems and methods for test object classification are provided in which the test object is docked with a target object in a plurality of different poses to form voxel maps. the maps are vectorized and fed into a convolutional neural network comprising an input layer, a plurality of individually weighted convolutional layers, and an output scorer. the convolutional layers include initial and final layers. responsive to vectorized input, the input layer feeds values into the initial convolutional layer. each respective convolutional layer, other than the final convolutional layer, feeds intermediate values as a function of the weights and input values of the respective layer into another of the convolutional layers. the final convolutional layer feeds values into one or more fully connected layers as a function of the final layer weights and input values. the one or more full connected layers feed values into the scorer which scores each input vector to thereby classify the test object. dated 2016-06-21"
9378435,image segmentation in optical character recognition using neural networks,neural-network-based image segmentation techniques are provided herein. an input image that includes a plurality of characters can be received. boundaries between the characters can be identified using a trained neural network. the input image can be segmented along the boundaries identified between the characters. the neural network can be trained using a training image and a training target vector. the training target vector can indicate one or more boundaries between characters in the training image. neural-network-based segmentation can be used alone or in conjunction with other segmentation techniques to improve overall segmentation accuracy.,2016-06-28,The title of the patent is image segmentation in optical character recognition using neural networks and its abstract is neural-network-based image segmentation techniques are provided herein. an input image that includes a plurality of characters can be received. boundaries between the characters can be identified using a trained neural network. the input image can be segmented along the boundaries identified between the characters. the neural network can be trained using a training image and a training target vector. the training target vector can indicate one or more boundaries between characters in the training image. neural-network-based segmentation can be used alone or in conjunction with other segmentation techniques to improve overall segmentation accuracy. dated 2016-06-28
9378538,image interpolation method and image interpolation device and image apparatus using the same,"an image interpolation method and an image interpolation device and an image apparatus using the image interpolation method are provided. the image interpolation method uses a probabilistic neural network model to perform an adaptive interpolation on an image. the image interpolation method includes the following steps. firstly, plural reference points neighboring an interpolation point are selected. then, an anisotropic gaussian function value of each reference point of the plural reference points is obtained according to an edge direction angle, a horizontal smoothing parameter, a vertical smoothing parameter and a distance between each reference point and the interpolation point. afterwards, a statistics method is performed to integrate and compute the anisotropic gaussian function values of the plural reference points, thereby obtaining an interpolation value of the interpolation point.",2016-06-28,"The title of the patent is image interpolation method and image interpolation device and image apparatus using the same and its abstract is an image interpolation method and an image interpolation device and an image apparatus using the image interpolation method are provided. the image interpolation method uses a probabilistic neural network model to perform an adaptive interpolation on an image. the image interpolation method includes the following steps. firstly, plural reference points neighboring an interpolation point are selected. then, an anisotropic gaussian function value of each reference point of the plural reference points is obtained according to an edge direction angle, a horizontal smoothing parameter, a vertical smoothing parameter and a distance between each reference point and the interpolation point. afterwards, a statistics method is performed to integrate and compute the anisotropic gaussian function values of the plural reference points, thereby obtaining an interpolation value of the interpolation point. dated 2016-06-28"
9378733,keyword detection without decoding,"embodiments pertain to automatic speech recognition in mobile devices to establish the presence of a keyword. an audio waveform is received at a mobile device. front-end feature extraction is performed on the audio waveform, followed by acoustic modeling, high level feature extraction, and output classification to detect the keyword. acoustic modeling may use a neural network or a vector quantization dictionary and high level feature extraction may use pooling.",2016-06-28,"The title of the patent is keyword detection without decoding and its abstract is embodiments pertain to automatic speech recognition in mobile devices to establish the presence of a keyword. an audio waveform is received at a mobile device. front-end feature extraction is performed on the audio waveform, followed by acoustic modeling, high level feature extraction, and output classification to detect the keyword. acoustic modeling may use a neural network or a vector quantization dictionary and high level feature extraction may use pooling. dated 2016-06-28"
9378735,estimating speaker-specific affine transforms for neural network based speech recognition systems,"features are disclosed for estimating affine transforms in log filter-bank energy space (“lfbe” space) in order to adapt artificial neural network-based acoustic models to a new speaker or environment. neural network-based acoustic models may be trained using concatenated lfbes as input features. the affine transform may be estimated by minimizing the least squares error between corresponding linear and bias transform parts for the resultant neural network feature vector and some standard speaker-specific feature vector obtained for a gmm-based acoustic model using constrained maximum likelihood linear regression (“cmllr”) techniques. alternatively, the affine transform may be estimated by minimizing the least squares error between the resultant transformed neural network feature and some standard speaker-specific feature obtained for a gmm-based acoustic model.",2016-06-28,"The title of the patent is estimating speaker-specific affine transforms for neural network based speech recognition systems and its abstract is features are disclosed for estimating affine transforms in log filter-bank energy space (“lfbe” space) in order to adapt artificial neural network-based acoustic models to a new speaker or environment. neural network-based acoustic models may be trained using concatenated lfbes as input features. the affine transform may be estimated by minimizing the least squares error between corresponding linear and bias transform parts for the resultant neural network feature vector and some standard speaker-specific feature vector obtained for a gmm-based acoustic model using constrained maximum likelihood linear regression (“cmllr”) techniques. alternatively, the affine transform may be estimated by minimizing the least squares error between the resultant transformed neural network feature and some standard speaker-specific feature obtained for a gmm-based acoustic model. dated 2016-06-28"
9379546,vector control of grid-connected power electronic converter using artificial neural networks,"systems and methods for performing vector control of grid-connected power electronic converters using a neural network are described herein. optionally, the grid-connected power electronic converters can be used in renewable and electric power system applications. in order to improve performance and stability under disturbance conditions, integrals of error signals can be introduced as inputs to the neural network. alternatively or additionally, grid disturbance voltage can be introduced to an output of a trained neural network.",2016-06-28,"The title of the patent is vector control of grid-connected power electronic converter using artificial neural networks and its abstract is systems and methods for performing vector control of grid-connected power electronic converters using a neural network are described herein. optionally, the grid-connected power electronic converters can be used in renewable and electric power system applications. in order to improve performance and stability under disturbance conditions, integrals of error signals can be introduced as inputs to the neural network. alternatively or additionally, grid disturbance voltage can be introduced to an output of a trained neural network. dated 2016-06-28"
9384062,artificial neural network for balancing workload by migrating computing tasks across hosts,"methods and apparatuses for balancing computing workload via migrating computing tasks are disclosed. an artificial neural network (ann) is trained based on the workload distribution over time for a host. the ann predicts the workload for the host, and an indication may be sent to migrate at least one computing task away from the host. the indication is sent when the method is operating in a proactive mode and when the predicted workload is outside of a desired operating range. some embodiments monitor the workload; and automatically switch the method to the proactive mode, when a difference between the monitored workload and the predicted workload is small. other embodiments monitor the workload; and automatically switch the method to a reactive mode, when the monitored workload is outside of a failsafe operating range for the particular host.",2016-07-05,"The title of the patent is artificial neural network for balancing workload by migrating computing tasks across hosts and its abstract is methods and apparatuses for balancing computing workload via migrating computing tasks are disclosed. an artificial neural network (ann) is trained based on the workload distribution over time for a host. the ann predicts the workload for the host, and an indication may be sent to migrate at least one computing task away from the host. the indication is sent when the method is operating in a proactive mode and when the predicted workload is outside of a desired operating range. some embodiments monitor the workload; and automatically switch the method to the proactive mode, when a difference between the monitored workload and the predicted workload is small. other embodiments monitor the workload; and automatically switch the method to a reactive mode, when the monitored workload is outside of a failsafe operating range for the particular host. dated 2016-07-05"
9384444,web analytics neural network modeling prediction,"a system and method are disclosed for optimizing website effectiveness. original input data associated with a plurality of website effectiveness variables is processed using a website effectiveness model to generate a first website effectiveness value, which in turn is processed to generate a dependent variable. input data corresponding to an individual website effectiveness variable is then processed to generate changed input data, which in turn is processed by the website effectiveness model with the original input data and the dependent variable to generate a second website effectiveness value. the first and second website effectiveness values are then processed to determine the effect of the changed data on the first website effectiveness value.",2016-07-05,"The title of the patent is web analytics neural network modeling prediction and its abstract is a system and method are disclosed for optimizing website effectiveness. original input data associated with a plurality of website effectiveness variables is processed using a website effectiveness model to generate a first website effectiveness value, which in turn is processed to generate a dependent variable. input data corresponding to an individual website effectiveness variable is then processed to generate changed input data, which in turn is processed by the website effectiveness model with the original input data and the dependent variable to generate a second website effectiveness value. the first and second website effectiveness values are then processed to determine the effect of the changed data on the first website effectiveness value. dated 2016-07-05"
9384560,contamination level estimation method for high voltage insulators,"the contamination level estimation method for high voltage insulators collects samples of naturally contaminated insulators and builds an image data set for the collected insulators. flashover voltages of several insulators samples are measured. esdd levels of the collected insulators are estimated. images are input to image processing algorithms to extract representative features. the images are segmented. transforming the image from rgb color space into grayscale model excludes the background from the image. subsequently, the segmented images are transferred back to rgb color space model using matrix manipulation. since contaminants on the insulator surface affect the color of the insulator, the segmented image is transformed from rgb to hsv color space which is used for extracting statistical and linear algebraic features from the hue image. a trained artificial neural network correlates the extracted features to the contamination levels enabling testing of other contaminated insulators.",2016-07-05,"The title of the patent is contamination level estimation method for high voltage insulators and its abstract is the contamination level estimation method for high voltage insulators collects samples of naturally contaminated insulators and builds an image data set for the collected insulators. flashover voltages of several insulators samples are measured. esdd levels of the collected insulators are estimated. images are input to image processing algorithms to extract representative features. the images are segmented. transforming the image from rgb color space into grayscale model excludes the background from the image. subsequently, the segmented images are transferred back to rgb color space model using matrix manipulation. since contaminants on the insulator surface affect the color of the insulator, the segmented image is transformed from rgb to hsv color space which is used for extracting statistical and linear algebraic features from the hue image. a trained artificial neural network correlates the extracted features to the contamination levels enabling testing of other contaminated insulators. dated 2016-07-05"
9390370,training deep neural network acoustic models using distributed hessian-free optimization,"a method for training a neural network includes receiving labeled training data at a master node, generating, by the master node, partitioned training data from the labeled training data and a held-out set of the labeled training data, determining a plurality of gradients for the partitioned training data, wherein the determination of the gradients is distributed across a plurality of worker nodes, determining a plurality of curvature matrix-vector products over the plurality of samples of the partitioned training data, wherein the determination of the plurality of curvature matrix-vector products is distributed across the plurality of worker nodes, and determining, by the master node, a second-order optimization of the plurality of gradients and the plurality of curvature matrix-vector products, producing a trained neural network configured to perform a structured classification task using a sequence-discriminative criterion.",2016-07-12,"The title of the patent is training deep neural network acoustic models using distributed hessian-free optimization and its abstract is a method for training a neural network includes receiving labeled training data at a master node, generating, by the master node, partitioned training data from the labeled training data and a held-out set of the labeled training data, determining a plurality of gradients for the partitioned training data, wherein the determination of the gradients is distributed across a plurality of worker nodes, determining a plurality of curvature matrix-vector products over the plurality of samples of the partitioned training data, wherein the determination of the plurality of curvature matrix-vector products is distributed across the plurality of worker nodes, and determining, by the master node, a second-order optimization of the plurality of gradients and the plurality of curvature matrix-vector products, producing a trained neural network configured to perform a structured classification task using a sequence-discriminative criterion. dated 2016-07-12"
9390372,"unsupervised, supervised, and reinforced learning via spiking computation","the present invention relates to unsupervised, supervised and reinforced learning via spiking computation. the neural network comprises a plurality of neural modules. each neural module comprises multiple digital neurons such that each neuron in a neural module has a corresponding neuron in another neural module. an interconnection network comprising a plurality of edges interconnects the plurality of neural modules. each edge interconnects a first neural module to a second neural module, and each edge comprises a weighted synaptic connection between every neuron in the first neural module and a corresponding neuron in the second neural module.",2016-07-12,"The title of the patent is unsupervised, supervised, and reinforced learning via spiking computation and its abstract is the present invention relates to unsupervised, supervised and reinforced learning via spiking computation. the neural network comprises a plurality of neural modules. each neural module comprises multiple digital neurons such that each neuron in a neural module has a corresponding neuron in another neural module. an interconnection network comprising a plurality of edges interconnects the plurality of neural modules. each edge interconnects a first neural module to a second neural module, and each edge comprises a weighted synaptic connection between every neuron in the first neural module and a corresponding neuron in the second neural module. dated 2016-07-12"
9390373,neural network and method of neural network training,"a neural network includes a plurality of inputs for receiving input signals, and synapses connected to the inputs and having corrective weights. the network additionally includes distributors. each distributor is connected to one of the inputs for receiving the respective input signal and selects one or more corrective weights in correlation with the input value. the network also includes neurons. each neuron has an output connected with at least one of the inputs via one synapse and generates a neuron sum by summing corrective weights selected from each synapse connected to the respective neuron. furthermore, the network includes a weight correction calculator that receives a desired output signal, determines a deviation of the neuron sum from the desired output signal value, and modifies respective corrective weights using the determined deviation. adding up the modified corrective weights to determine the neuron sum minimizes the subject deviation for training the neural network.",2016-07-12,"The title of the patent is neural network and method of neural network training and its abstract is a neural network includes a plurality of inputs for receiving input signals, and synapses connected to the inputs and having corrective weights. the network additionally includes distributors. each distributor is connected to one of the inputs for receiving the respective input signal and selects one or more corrective weights in correlation with the input value. the network also includes neurons. each neuron has an output connected with at least one of the inputs via one synapse and generates a neuron sum by summing corrective weights selected from each synapse connected to the respective neuron. furthermore, the network includes a weight correction calculator that receives a desired output signal, determines a deviation of the neuron sum from the desired output signal value, and modifies respective corrective weights using the determined deviation. adding up the modified corrective weights to determine the neuron sum minimizes the subject deviation for training the neural network. dated 2016-07-12"
9390712,mixed speech recognition,"the claimed subject matter includes a system and method for recognizing mixed speech from a source. the method includes training a first neural network to recognize the speech signal spoken by the speaker with a higher level of a speech characteristic from a mixed speech sample. the method also includes training a second neural network to recognize the speech signal spoken by the speaker with a lower level of the speech characteristic from the mixed speech sample. additionally, the method includes decoding the mixed speech sample with the first neural network and the second neural network by optimizing the joint likelihood of observing the two speech signals considering the probability that a specific frame is a switching point of the speech characteristic.",2016-07-12,"The title of the patent is mixed speech recognition and its abstract is the claimed subject matter includes a system and method for recognizing mixed speech from a source. the method includes training a first neural network to recognize the speech signal spoken by the speaker with a higher level of a speech characteristic from a mixed speech sample. the method also includes training a second neural network to recognize the speech signal spoken by the speaker with a lower level of the speech characteristic from the mixed speech sample. additionally, the method includes decoding the mixed speech sample with the first neural network and the second neural network by optimizing the joint likelihood of observing the two speech signals considering the probability that a specific frame is a switching point of the speech characteristic. dated 2016-07-12"
9396415,neural network image representation,"a method for representing an input image includes the steps of applying a trained neural network on the input image, selecting a plurality of feature maps, determining a location of each of the plurality of feature maps in an image space of the input image, defining a plurality of interest points of the input image, and employing the plurality of interest points for representing the input image for performing a visual task. the plurality of feature maps are selected of an output of at least one selected layer of the trained neural network according to values attributed to the plurality of feature maps by the trained neural network. the plurality of interest points of the input image are defined based on the locations corresponding to the plurality of feature maps.",2016-07-19,"The title of the patent is neural network image representation and its abstract is a method for representing an input image includes the steps of applying a trained neural network on the input image, selecting a plurality of feature maps, determining a location of each of the plurality of feature maps in an image space of the input image, defining a plurality of interest points of the input image, and employing the plurality of interest points for representing the input image for performing a visual task. the plurality of feature maps are selected of an output of at least one selected layer of the trained neural network according to values attributed to the plurality of feature maps by the trained neural network. the plurality of interest points of the input image are defined based on the locations corresponding to the plurality of feature maps. dated 2016-07-19"
9396431,network of artificial neurons based on complementary memristive devices,"a neural network comprises a plurality of artificial neurons and a plurality of artificial synapses each input neuron being connected to each output neuron by way of an artificial synapse, the network being characterized in that each synapse consists of a first memristive device connected to a first input of an output neuron and of a second memristive device, mounted in opposition to said first device and connected to a second, complemented, input of said output neuron so that said output neuron integrates the difference between the currents originating from the first and second devices.",2016-07-19,"The title of the patent is network of artificial neurons based on complementary memristive devices and its abstract is a neural network comprises a plurality of artificial neurons and a plurality of artificial synapses each input neuron being connected to each output neuron by way of an artificial synapse, the network being characterized in that each synapse consists of a first memristive device connected to a first input of an output neuron and of a second memristive device, mounted in opposition to said first device and connected to a second, complemented, input of said output neuron so that said output neuron integrates the difference between the currents originating from the first and second devices. dated 2016-07-19"
9396738,methods and apparatus for signal quality analysis,"a non-intrusive objective speech quality assessment is performed on a degraded speech signal. the methods are well suited for systems where random and bursty packet losses may occur and/or packet stream regeneration may also occur prior to speech signal quality assessment. in one embodiment received packetized speech is analyzed to determine to an overall final signal quality score. a limited set of trained neural networks, e.g., 5, corresponding to different signal features, each determine a signal feature quality score. a trained joint quality score determination module determines a joint quality score based on the signal feature quality scores. packet loss is estimated based on received packet header information and/or detected gap durations. the determined joint quality score is adjusted, based on estimated packet loss information obtained from examining the speech signal, network level statistics and/or codec parameters to generate the final quality score.",2016-07-19,"The title of the patent is methods and apparatus for signal quality analysis and its abstract is a non-intrusive objective speech quality assessment is performed on a degraded speech signal. the methods are well suited for systems where random and bursty packet losses may occur and/or packet stream regeneration may also occur prior to speech signal quality assessment. in one embodiment received packetized speech is analyzed to determine to an overall final signal quality score. a limited set of trained neural networks, e.g., 5, corresponding to different signal features, each determine a signal feature quality score. a trained joint quality score determination module determines a joint quality score based on the signal feature quality scores. packet loss is estimated based on received packet header information and/or detected gap durations. the determined joint quality score is adjusted, based on estimated packet loss information obtained from examining the speech signal, network level statistics and/or codec parameters to generate the final quality score. dated 2016-07-19"
9400490,method and a system for an automatic recovery from a fault situation in a production plant,"a method automatically recovers from a fault situation in a production plant and provides production resources and a manufacturing execution system having a production modeler for modeling the production resources into a plant model and a production scheduler to schedule operations of the modeled production resources. a production controller executes the production process and a fault manager detects fault situations and automatically decides a corrective action. a production resource runs an application for the operation of the production resource and a fault analysis agent provides categorized error situations and checks operational data representing the operation of the production resource against the categorized error situations and when, an error situation occurs, forwards an error event to the fault manager. the error events are collected and then analyzed by a neural network system to assign the error event to an error category. a corrective action is executed on the production resource.",2016-07-26,"The title of the patent is method and a system for an automatic recovery from a fault situation in a production plant and its abstract is a method automatically recovers from a fault situation in a production plant and provides production resources and a manufacturing execution system having a production modeler for modeling the production resources into a plant model and a production scheduler to schedule operations of the modeled production resources. a production controller executes the production process and a fault manager detects fault situations and automatically decides a corrective action. a production resource runs an application for the operation of the production resource and a fault analysis agent provides categorized error situations and checks operational data representing the operation of the production resource against the categorized error situations and when, an error situation occurs, forwards an error event to the fault manager. the error events are collected and then analyzed by a neural network system to assign the error event to an error category. a corrective action is executed on the production resource. dated 2016-07-26"
9400919,learning deep face representation,"face representation is a crucial step of face recognition systems. an optimal face representation should be discriminative, robust, compact, and very easy to implement. while numerous hand-crafted and learning-based representations have been proposed, considerable room for improvement is still present. a very easy-to-implement deep learning framework for face representation is presented. the framework bases on pyramid convolutional neural network (cnn). the pyramid cnn adopts a greedy-filter-and-down-sample operation, which enables the training procedure to be very fast and computation efficient. in addition, the structure of pyramid cnn can naturally incorporate feature sharing across multi-scale face representations, increasing the discriminative ability of resulting representation.",2016-07-26,"The title of the patent is learning deep face representation and its abstract is face representation is a crucial step of face recognition systems. an optimal face representation should be discriminative, robust, compact, and very easy to implement. while numerous hand-crafted and learning-based representations have been proposed, considerable room for improvement is still present. a very easy-to-implement deep learning framework for face representation is presented. the framework bases on pyramid convolutional neural network (cnn). the pyramid cnn adopts a greedy-filter-and-down-sample operation, which enables the training procedure to be very fast and computation efficient. in addition, the structure of pyramid cnn can naturally incorporate feature sharing across multi-scale face representations, increasing the discriminative ability of resulting representation. dated 2016-07-26"
9400922,facial landmark localization using coarse-to-fine cascaded neural networks,"the present invention overcomes the limitations of the prior art by performing facial landmark localization in a coarse-to-fine manner with a cascade of neural network levels, and enforcing geometric constraints for each of the neural network levels. in one approach, the neural network levels may be implemented with deep convolutional neural network. one aspect concerns a system for localizing landmarks on face images. the system includes an input for receiving a face image, and an output for presenting landmarks identified by the system. neural network levels are coupled in a cascade from the input to the output for the system. each neural network level produces an estimate of landmarks. the estimate of landmarks is more refined than an estimate of landmark of a previous neural network level.",2016-07-26,"The title of the patent is facial landmark localization using coarse-to-fine cascaded neural networks and its abstract is the present invention overcomes the limitations of the prior art by performing facial landmark localization in a coarse-to-fine manner with a cascade of neural network levels, and enforcing geometric constraints for each of the neural network levels. in one approach, the neural network levels may be implemented with deep convolutional neural network. one aspect concerns a system for localizing landmarks on face images. the system includes an input for receiving a face image, and an output for presenting landmarks identified by the system. neural network levels are coupled in a cascade from the input to the output for the system. each neural network level produces an estimate of landmarks. the estimate of landmarks is more refined than an estimate of landmark of a previous neural network level. dated 2016-07-26"
9400925,pose-aligned networks for deep attribute modeling,"technology is disclosed for inferring human attributes from images of people. the attributes can include, for example, gender, age, hair, and/or clothing. the technology uses part-based models, e.g., poselets, to locate multiple normalized part patches from an image. the normalized part patches are provided into trained convolutional neural networks to generate feature data. each convolution neural network applies multiple stages of convolution operations to one part patch to generate a set of fully connected feature data. the feature data for all part patches are concatenated and then provided into multiple trained classifiers (e.g., linear support vector machines) to predict attributes of the image.",2016-07-26,"The title of the patent is pose-aligned networks for deep attribute modeling and its abstract is technology is disclosed for inferring human attributes from images of people. the attributes can include, for example, gender, age, hair, and/or clothing. the technology uses part-based models, e.g., poselets, to locate multiple normalized part patches from an image. the normalized part patches are provided into trained convolutional neural networks to generate feature data. each convolution neural network applies multiple stages of convolution operations to one part patch to generate a set of fully connected feature data. the feature data for all part patches are concatenated and then provided into multiple trained classifiers (e.g., linear support vector machines) to predict attributes of the image. dated 2016-07-26"
9400954,multi-scale spatio-temporal neural network system,"embodiments of the invention relate to a multi-scale spatio-temporal neural network system. one embodiment comprises a neural network including multiple heterogeneous neuron populations that operate at different time scales. each neuron population comprises at least one digital neuron. each neuron population further comprises a time scale generation circuit that controls timing for operation of said neuron population, wherein each neuron of said neuron population integrates neuronal firing events at a time scale corresponding to said neuron population. the neural network further comprises a plurality of synapses interconnecting the neurons, wherein each synapse interconnects a neuron with another neuron. at least one neuron receives neuronal firing events from an interconnected neuron that operates at a different time scale.",2016-07-26,"The title of the patent is multi-scale spatio-temporal neural network system and its abstract is embodiments of the invention relate to a multi-scale spatio-temporal neural network system. one embodiment comprises a neural network including multiple heterogeneous neuron populations that operate at different time scales. each neuron population comprises at least one digital neuron. each neuron population further comprises a time scale generation circuit that controls timing for operation of said neuron population, wherein each neuron of said neuron population integrates neuronal firing events at a time scale corresponding to said neuron population. the neural network further comprises a plurality of synapses interconnecting the neurons, wherein each synapse interconnects a neuron with another neuron. at least one neuron receives neuronal firing events from an interconnected neuron that operates at a different time scale. dated 2016-07-26"
9400955,reducing dynamic range of low-rank decomposition matrices,"features are disclosed for reducing the dynamic range of an approximated trained artificial neural network weight matrix in an automatic speech recognition system. the weight matrix may be approximated as two low-rank matrices using a decomposition technique. this approximation technique may insert an additional layer between the two original layers connected by the weight matrix. the dynamic range of the low-rank decomposition may be reduced by applying the square root of singular values, combining them with both low-rank matrices, and utilizing a random rotation matrix to further compress the low-rank matrices. reduction of dynamic range may make fixed point scoring more effective due to smaller quantization error, as well as make the neural network system more favorable for retraining after approximating a neural network weight matrix. features are also disclosed for adjusting the learning rate during retraining to account for the low-rank approximations.",2016-07-26,"The title of the patent is reducing dynamic range of low-rank decomposition matrices and its abstract is features are disclosed for reducing the dynamic range of an approximated trained artificial neural network weight matrix in an automatic speech recognition system. the weight matrix may be approximated as two low-rank matrices using a decomposition technique. this approximation technique may insert an additional layer between the two original layers connected by the weight matrix. the dynamic range of the low-rank decomposition may be reduced by applying the square root of singular values, combining them with both low-rank matrices, and utilizing a random rotation matrix to further compress the low-rank matrices. reduction of dynamic range may make fixed point scoring more effective due to smaller quantization error, as well as make the neural network system more favorable for retraining after approximating a neural network weight matrix. features are also disclosed for adjusting the learning rate during retraining to account for the low-rank approximations. dated 2016-07-26"
9401143,cluster specific speech model,"methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for receiving data representing acoustic characteristics of a user's voice; selecting a cluster for the data from among a plurality of clusters, where each cluster includes a plurality of vectors, and where each cluster is associated with a speech model trained by a neural network using at least one or more vectors of the plurality of vectors in the respective cluster; and in response to receiving one or more utterances of the user, providing the speech model associated with the cluster for transcribing the one or more utterances.",2016-07-26,"The title of the patent is cluster specific speech model and its abstract is methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for receiving data representing acoustic characteristics of a user's voice; selecting a cluster for the data from among a plurality of clusters, where each cluster includes a plurality of vectors, and where each cluster is associated with a speech model trained by a neural network using at least one or more vectors of the plurality of vectors in the respective cluster; and in response to receiving one or more utterances of the user, providing the speech model associated with the cluster for transcribing the one or more utterances. dated 2016-07-26"
9401148,speaker verification using neural networks,"methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for inputting speech data that corresponds to a particular utterance to a neural network; determining an evaluation vector based on output at a hidden layer of the neural network; comparing the evaluation vector with a reference vector that corresponds to a past utterance of a particular speaker; and based on comparing the evaluation vector and the reference vector, determining whether the particular utterance was likely spoken by the particular speaker.",2016-07-26,"The title of the patent is speaker verification using neural networks and its abstract is methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for inputting speech data that corresponds to a particular utterance to a neural network; determining an evaluation vector based on output at a hidden layer of the neural network; comparing the evaluation vector with a reference vector that corresponds to a past utterance of a particular speaker; and based on comparing the evaluation vector and the reference vector, determining whether the particular utterance was likely spoken by the particular speaker. dated 2016-07-26"
9403001,"non-invasive magnetic or electrical nerve stimulation to treat gastroparesis, functional dyspepsia, and other functional gastrointestinal disorders","devices, systems and methods are disclosed for treating or preventing gastroparesis, functional dyspepsia, and other functional gastrointestinal disorders. the methods comprise transmitting impulses of energy non-invasively to selected nerve fibers, particularly those in a vagus nerve. the methods provide damaged interstitial cells of cajal (icc) with trophic factors via vagal afferent nerve fibers, thereby reversing icc damage, and as a consequence improving gastric motility. the methods also increase levels of inhibitory neurotransmitters in the brain so as to decrease neural activity within the area postrema, or they deactivate a resting state neural network containing parts of the anterior insula and anterior cingulate cortex, which will thereby reduce abnormal interoception and visceral hypersensitivity.",2016-08-02,"The title of the patent is non-invasive magnetic or electrical nerve stimulation to treat gastroparesis, functional dyspepsia, and other functional gastrointestinal disorders and its abstract is devices, systems and methods are disclosed for treating or preventing gastroparesis, functional dyspepsia, and other functional gastrointestinal disorders. the methods comprise transmitting impulses of energy non-invasively to selected nerve fibers, particularly those in a vagus nerve. the methods provide damaged interstitial cells of cajal (icc) with trophic factors via vagal afferent nerve fibers, thereby reversing icc damage, and as a consequence improving gastric motility. the methods also increase levels of inhibitory neurotransmitters in the brain so as to decrease neural activity within the area postrema, or they deactivate a resting state neural network containing parts of the anterior insula and anterior cingulate cortex, which will thereby reduce abnormal interoception and visceral hypersensitivity. dated 2016-08-02"
9405960,face hallucination using convolutional neural networks,"face hallucination using a bi-channel deep convolutional neural network (bcnn), which can adaptively fuse two channels of information. in one example, the bcnn is implemented to extract high level features from an input image. the extracted high level features are combined with low level details in the input image to produce the higher resolution image. preferably, a proper coefficient is obtained to adaptively combine the high level features and the low level details.",2016-08-02,"The title of the patent is face hallucination using convolutional neural networks and its abstract is face hallucination using a bi-channel deep convolutional neural network (bcnn), which can adaptively fuse two channels of information. in one example, the bcnn is implemented to extract high level features from an input image. the extracted high level features are combined with low level details in the input image to produce the higher resolution image. preferably, a proper coefficient is obtained to adaptively combine the high level features and the low level details. dated 2016-08-02"
9406016,method and apparatus for monitoring network traffic,"a system that collects data from monitored network traffic. the system inputs, in parallel, the data through inputs of a neural network. the system compares an output of the neural network, generated in response to the inputted data, to at least one predetermined output. if the output of the neural network corresponds to the at least one predetermined output, the system provides a notification relating to the data.",2016-08-02,"The title of the patent is method and apparatus for monitoring network traffic and its abstract is a system that collects data from monitored network traffic. the system inputs, in parallel, the data through inputs of a neural network. the system compares an output of the neural network, generated in response to the inputted data, to at least one predetermined output. if the output of the neural network corresponds to the at least one predetermined output, the system provides a notification relating to the data. dated 2016-08-02"
9406017,system and method for addressing overfitting in a neural network,"a system for training a neural network. a switch is linked to feature detectors in at least some of the layers of the neural network. for each training case, the switch randomly selectively disables each of the feature detectors in accordance with a preconfigured probability. the weights from each training case are then normalized for applying the neural network to test data.",2016-08-02,"The title of the patent is system and method for addressing overfitting in a neural network and its abstract is a system for training a neural network. a switch is linked to feature detectors in at least some of the layers of the neural network. for each training case, the switch randomly selectively disables each of the feature detectors in accordance with a preconfigured probability. the weights from each training case are then normalized for applying the neural network to test data. dated 2016-08-02"
9408570,physiological feature extraction and fusion to assist in the diagnosis of post-traumatic stress disorder,"post-traumatic stress disorder (ptsd), and other anxiety disorders, are diagnosed via clinical interviews in which subjective self-reports of traumatic events and associated experiences are discussed with a mental health professional. the system and methods described herein classify and diagnose patients as suffering from anxiety disorders by measuring objective physiological measures, such as inter-heartbeat interval and skin conductance. the system measures various physiological measures and then extracts features from the physiological measures. a diagnosis is then made by classifying the extracted features using one of a neural network, bayesian network, or a support vector machine.",2016-08-09,"The title of the patent is physiological feature extraction and fusion to assist in the diagnosis of post-traumatic stress disorder and its abstract is post-traumatic stress disorder (ptsd), and other anxiety disorders, are diagnosed via clinical interviews in which subjective self-reports of traumatic events and associated experiences are discussed with a mental health professional. the system and methods described herein classify and diagnose patients as suffering from anxiety disorders by measuring objective physiological measures, such as inter-heartbeat interval and skin conductance. the system measures various physiological measures and then extracts features from the physiological measures. a diagnosis is then made by classifying the extracted features using one of a neural network, bayesian network, or a support vector machine. dated 2016-08-09"
9415076,muscle tissue regeneration using muscle fiber fragments,"the invention is directed to methods and compositions for obtaining uniform sized muscle fiber fragments for transplantation. these muscle fiber fragments are able to reconstitute into long fibers that are oriented along native muscle. the implanted muscle cells integrate with native vascular and neural network, as confirmed by histology and immunohistochemistry. this invention is particularly advantageous because autologous muscle can be harvested from a donor site, processed and injected into target sites in the operating room. the fragmented muscle fibers can be readily integrated within the host.",2016-08-16,"The title of the patent is muscle tissue regeneration using muscle fiber fragments and its abstract is the invention is directed to methods and compositions for obtaining uniform sized muscle fiber fragments for transplantation. these muscle fiber fragments are able to reconstitute into long fibers that are oriented along native muscle. the implanted muscle cells integrate with native vascular and neural network, as confirmed by histology and immunohistochemistry. this invention is particularly advantageous because autologous muscle can be harvested from a donor site, processed and injected into target sites in the operating room. the fragmented muscle fibers can be readily integrated within the host. dated 2016-08-16"
9418319,object detection using cascaded convolutional neural networks,"different candidate windows in an image are identified, such as by sliding a rectangular or other geometric shape of different sizes over an image to identify portions of the image (groups of pixels in the image). the candidate windows are analyzed by a set of convolutional neural networks, which are cascaded so that the input of one convolutional neural network layer is based on the input of another convolutional neural network layer. each convolutional neural network layer drops or rejects one or more candidate windows that the convolutional neural network layer determines does not include an object (e.g., a face). the candidate windows that are identified as including an object (e.g., a face) are analyzed by another one of the convolutional neural network layers. the candidate windows identified by the last of the convolutional neural network layers are the indications of the objects (e.g., faces) in the image.",2016-08-16,"The title of the patent is object detection using cascaded convolutional neural networks and its abstract is different candidate windows in an image are identified, such as by sliding a rectangular or other geometric shape of different sizes over an image to identify portions of the image (groups of pixels in the image). the candidate windows are analyzed by a set of convolutional neural networks, which are cascaded so that the input of one convolutional neural network layer is based on the input of another convolutional neural network layer. each convolutional neural network layer drops or rejects one or more candidate windows that the convolutional neural network layer determines does not include an object (e.g., a face). the candidate windows that are identified as including an object (e.g., a face) are analyzed by another one of the convolutional neural network layers. the candidate windows identified by the last of the convolutional neural network layers are the indications of the objects (e.g., faces) in the image. dated 2016-08-16"
9418331,methods and apparatus for tagging classes using supervised learning,certain aspects of the present disclosure provide methods and apparatus for creating tags (static or dynamic) for input/output classes of a neural network model using supervised learning. the method includes augmenting a neural network model with a plurality of neurons and training the augmented network using spike timing dependent plasticity (stdp) to determine one or more tags.,2016-08-16,The title of the patent is methods and apparatus for tagging classes using supervised learning and its abstract is certain aspects of the present disclosure provide methods and apparatus for creating tags (static or dynamic) for input/output classes of a neural network model using supervised learning. the method includes augmenting a neural network model with a plurality of neurons and training the augmented network using spike timing dependent plasticity (stdp) to determine one or more tags. dated 2016-08-16
9418458,graph image representation from convolutional neural networks,"a method for producing a graph representation of an input image, the method including the procedures of applying convolutional layers of a trained convolutional neural network on the input image, defining a receptive field of a last convolutional layer of the trained convolutional neural network as a vertex of the graph representation, defining a vector of a three dimensional output matrix of the last convolutional layer that is mapped to the receptive field as a descriptor for the vertex and determining an edge between a pair of vertices of the graph representation by applying an operator on a pair of descriptors respective of the pair of vertices.",2016-08-16,"The title of the patent is graph image representation from convolutional neural networks and its abstract is a method for producing a graph representation of an input image, the method including the procedures of applying convolutional layers of a trained convolutional neural network on the input image, defining a receptive field of a last convolutional layer of the trained convolutional neural network as a vertex of the graph representation, defining a vector of a three dimensional output matrix of the last convolutional layer that is mapped to the receptive field as a descriptor for the vertex and determining an edge between a pair of vertices of the graph representation by applying an operator on a pair of descriptors respective of the pair of vertices. dated 2016-08-16"
9420957,system and method for predicting acute cardiopulmonary events and survivability of a patient,"a method of producing an artificial neural network capable of predicting the survivability of a patient, including: storing in an electronic database patient health data comprising a plurality of sets of data, each set having at least one of a first parameter relating to heart rate variability data and a second parameter relating to vital sign data, each set further having a third parameter relating to patient survivability; providing a network of nodes interconnected to form an artificial neural network, the nodes comprising a plurality of artificial neurons, each artificial neuron having at least one input with an associated weight; and training the artificial neural network using the patient health data such that the associated weight of the at least one input of each artificial neuron is adjusted in response to respective first, second and third parameters of different sets of data from the patient health data.",2016-08-23,"The title of the patent is system and method for predicting acute cardiopulmonary events and survivability of a patient and its abstract is a method of producing an artificial neural network capable of predicting the survivability of a patient, including: storing in an electronic database patient health data comprising a plurality of sets of data, each set having at least one of a first parameter relating to heart rate variability data and a second parameter relating to vital sign data, each set further having a third parameter relating to patient survivability; providing a network of nodes interconnected to form an artificial neural network, the nodes comprising a plurality of artificial neurons, each artificial neuron having at least one input with an associated weight; and training the artificial neural network using the patient health data such that the associated weight of the at least one input of each artificial neuron is adjusted in response to respective first, second and third parameters of different sets of data from the patient health data. dated 2016-08-23"
9424493,generic object detection in images,"neural networks for object detection in images are used with a spatial pyramid pooling (spp) layer. using the spp network structure, a fixed-length representation is generated regardless of image size and scale. the feature maps are computed from the entire image once, and the features are pooled in arbitrary regions (sub-images) to generate fixed-length representations for training the detectors. thus, repeated computation of the convolutional features is avoided while accuracy is enhanced.",2016-08-23,"The title of the patent is generic object detection in images and its abstract is neural networks for object detection in images are used with a spatial pyramid pooling (spp) layer. using the spp network structure, a fixed-length representation is generated regardless of image size and scale. the feature maps are computed from the entire image once, and the features are pooled in arbitrary regions (sub-images) to generate fixed-length representations for training the detectors. thus, repeated computation of the convolutional features is avoided while accuracy is enhanced. dated 2016-08-23"
9424494,pure convolutional neural network localization,"an approach is provided in which a knowledge manager processes an image using a convolutional neural network. the knowledge manager generates a pixel-level heat map of the image that includes multiple decision points corresponding to multiple pixels of the image. the knowledge manager analyzes the pixel-level heat map and detects sets of decision points that correspond to target objects. in turn, the knowledge manager marks regions of the heat map corresponding to the detected sets of per-pixel decision points, each of the regions indicating a location of the target objects.",2016-08-23,"The title of the patent is pure convolutional neural network localization and its abstract is an approach is provided in which a knowledge manager processes an image using a convolutional neural network. the knowledge manager generates a pixel-level heat map of the image that includes multiple decision points corresponding to multiple pixels of the image. the knowledge manager analyzes the pixel-level heat map and detects sets of decision points that correspond to target objects. in turn, the knowledge manager marks regions of the heat map corresponding to the detected sets of per-pixel decision points, each of the regions indicating a location of the target objects. dated 2016-08-23"
9424509,system for application personalization for a mobile device,a system for controlling applications of a wireless mobile device includes a server for receiving data related to an adaptive user profile and for controlling operations of applications within the wireless mobile device. an adaptive neural/fuzzy logic control application implemented within the network server generates the adaptive user profile responsive to the received data. the adaptive user profile controls operations of the applications within the wireless mobile device and changes in real time responsive to the received data.,2016-08-23,The title of the patent is system for application personalization for a mobile device and its abstract is a system for controlling applications of a wireless mobile device includes a server for receiving data related to an adaptive user profile and for controlling operations of applications within the wireless mobile device. an adaptive neural/fuzzy logic control application implemented within the network server generates the adaptive user profile responsive to the received data. the adaptive user profile controls operations of the applications within the wireless mobile device and changes in real time responsive to the received data. dated 2016-08-23
9424514,synapse maintenance in the developmental networks,"the developmental neural network is trained using a synaptic maintenance process. synaptogenic trimming is first performed on the neuron inputs using a synaptogenic factor for each neuron based on standard deviation of a measured match between the input and synaptic weight value. a top-k competition among all neurons then selects a subset of said neurons as winning neurons. neuronal learning is applied only to these winning neurons, updating their synaptic weights and updating their synaptogenic factors.",2016-08-23,"The title of the patent is synapse maintenance in the developmental networks and its abstract is the developmental neural network is trained using a synaptic maintenance process. synaptogenic trimming is first performed on the neuron inputs using a synaptogenic factor for each neuron based on standard deviation of a measured match between the input and synaptic weight value. a top-k competition among all neurons then selects a subset of said neurons as winning neurons. neuronal learning is applied only to these winning neurons, updating their synaptic weights and updating their synaptogenic factors. dated 2016-08-23"
9427539,system and method for diagnosis and treatment of a breathing pattern of a patient,"described is a system including a sensor and a processing arrangement. the sensor measures data corresponding to a patient's breathing patterns. the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of a rem sleep state. in another embodiment, the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of one of the following states: (i) a wake state and (ii) a sleep state. in another embodiment, a neural network analyzes the data to determine whether the breathing patterns are indicative of one of the following states: (i) a rem sleep state, (ii) a wake state and (iii) a sleep state. in another embodiment, the processing arrangement analyzes the data to determine whether the breathing pattern is indicative of an arousal.",2016-08-30,"The title of the patent is system and method for diagnosis and treatment of a breathing pattern of a patient and its abstract is described is a system including a sensor and a processing arrangement. the sensor measures data corresponding to a patient's breathing patterns. the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of a rem sleep state. in another embodiment, the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of one of the following states: (i) a wake state and (ii) a sleep state. in another embodiment, a neural network analyzes the data to determine whether the breathing patterns are indicative of one of the following states: (i) a rem sleep state, (ii) a wake state and (iii) a sleep state. in another embodiment, the processing arrangement analyzes the data to determine whether the breathing pattern is indicative of an arousal. dated 2016-08-30"
9430734,method and system for validating energy measurement in a high pressure gas distribution network,"a method and system for validating energy measurement in a high pressure gas distribution network. the method comprises the steps of calculating a validation energy value using an artificial neural network (ann) engine based on measured parameters associated with a gas flow in the gas distribution network; measuring an actual energy value of the gas flow; and comparing the validation energy value and the actual energy value, wherein the actual energy value is validated if the validation energy value and the actual energy value are substantially equal.",2016-08-30,"The title of the patent is method and system for validating energy measurement in a high pressure gas distribution network and its abstract is a method and system for validating energy measurement in a high pressure gas distribution network. the method comprises the steps of calculating a validation energy value using an artificial neural network (ann) engine based on measured parameters associated with a gas flow in the gas distribution network; measuring an actual energy value of the gas flow; and comparing the validation energy value and the actual energy value, wherein the actual energy value is validated if the validation energy value and the actual energy value are substantially equal. dated 2016-08-30"
9430735,neural network in a memory device,"devices, systems and methods for operating a memory device facilitating a neural network in a memory device are disclosed. in at least one embodiment, the memory device is operated having a feed-ward neural network operating scheme. in at least one other embodiment, memory cells are operated to emulate a number of neural models to facilitate one or more neural network operating characteristics in the memory device.",2016-08-30,"The title of the patent is neural network in a memory device and its abstract is devices, systems and methods for operating a memory device facilitating a neural network in a memory device are disclosed. in at least one embodiment, the memory device is operated having a feed-ward neural network operating scheme. in at least one other embodiment, memory cells are operated to emulate a number of neural models to facilitate one or more neural network operating characteristics in the memory device. dated 2016-08-30"
9430736,firing rate independent spike message passing in large scale neural network modeling,"""a neural network portion comprising n pre-synaptic neurons capable each of firing an action potential, wherein the number n can be encoded in a word of n bits;    """,2016-08-30,"The title of the patent is firing rate independent spike message passing in large scale neural network modeling and its abstract is ""a neural network portion comprising n pre-synaptic neurons capable each of firing an action potential, wherein the number n can be encoded in a word of n bits;    "" dated 2016-08-30"
9430737,spiking model to learn arbitrary multiple transformations for a             self-realizing network,"a neural network, wherein a portion of the neural network comprises: a first array having a first number of neurons, wherein the dendrite of each neuron of the first array is provided for receiving an input signal indicating that a measured parameter gets closer to a predetermined value assigned to said neuron; and a second array having a second number of neurons, wherein the second number is smaller than the first number, the dendrite of each neuron of the second array forming an excitatory stdp synapse with the axon of a plurality of neurons of the first array; the dendrite of each neuron of the second array forming an excitatory stdp synapse with the axon of neighboring neurons of the second array.",2016-08-30,"The title of the patent is spiking model to learn arbitrary multiple transformations for a             self-realizing network and its abstract is a neural network, wherein a portion of the neural network comprises: a first array having a first number of neurons, wherein the dendrite of each neuron of the first array is provided for receiving an input signal indicating that a measured parameter gets closer to a predetermined value assigned to said neuron; and a second array having a second number of neurons, wherein the second number is smaller than the first number, the dendrite of each neuron of the second array forming an excitatory stdp synapse with the axon of a plurality of neurons of the first array; the dendrite of each neuron of the second array forming an excitatory stdp synapse with the axon of neighboring neurons of the second array. dated 2016-08-30"
9430829,automatic detection of mitosis using handcrafted and convolutional neural network features,"one example apparatus associated with detecting mitosis in breast cancer pathology images by combining handcrafted (hc) and convolutional neural network (cnn) features in a cascaded architecture includes a set of logics that acquires an image of a region of tissue, partitions the image into candidate patches, generates a first probability that the patch is mitotic using an hc feature set and a second probability that the patch is mitotic using a cnn-learned feature set, and classifies the patch based on the first probability and the second probability. if the first and second probabilities do not agree, the apparatus trains a cascaded classifier on the cnn-learned feature set and the hc feature set, generates a third probability that the patch is mitotic, and classifies the patch based on a weighted average of the first probability, the second probability, and the third probability.",2016-08-30,"The title of the patent is automatic detection of mitosis using handcrafted and convolutional neural network features and its abstract is one example apparatus associated with detecting mitosis in breast cancer pathology images by combining handcrafted (hc) and convolutional neural network (cnn) features in a cascaded architecture includes a set of logics that acquires an image of a region of tissue, partitions the image into candidate patches, generates a first probability that the patch is mitotic using an hc feature set and a second probability that the patch is mitotic using a cnn-learned feature set, and classifies the patch based on the first probability and the second probability. if the first and second probabilities do not agree, the apparatus trains a cascaded classifier on the cnn-learned feature set and the hc feature set, generates a third probability that the patch is mitotic, and classifies the patch based on a weighted average of the first probability, the second probability, and the third probability. dated 2016-08-30"
9433703,neural graft,"a neural graft includes a biological substrate, a carbon nanotube structure and a neural network. the carbon nanotube structure is located on the biological substrate. the carbon nanotube structure includes a number of carbon nanotube wires crossed with each other to define a number of pores. the neural network includes a number of neural cell bodies and a number of neurites branched from the neural cell bodies. an effective diameter of each pore is larger than or equal to a diameter of the neural cell body, the neurites substantially extend along the carbon nanotube wires such that the neurites are patterned.",2016-09-06,"The title of the patent is neural graft and its abstract is a neural graft includes a biological substrate, a carbon nanotube structure and a neural network. the carbon nanotube structure is located on the biological substrate. the carbon nanotube structure includes a number of carbon nanotube wires crossed with each other to define a number of pores. the neural network includes a number of neural cell bodies and a number of neurites branched from the neural cell bodies. an effective diameter of each pore is larger than or equal to a diameter of the neural cell body, the neurites substantially extend along the carbon nanotube wires such that the neurites are patterned. dated 2016-09-06"
9433843,method and apparatus for creating cost data for use in generating a route across an electronic map,"a method is disclosed involving receiving gps data from persona portable training devices of users when traversing an off-road segment of an electronic map together with associated data indicative of a heart rate of a user during the movements. the position and heart rate data for each user traversing the segment are processed using data indicative of a fitness profile for the user. the resulting data is used to determine a normalized cost to be associated with the segment, indicative of the difficulty in traversing the segment. the cost data is generated using a neural network. the resulting cost data for different segments in a network of segments is used to generate route suggestions for users based upon desired workout intensity, fitness levels, etc.",2016-09-06,"The title of the patent is method and apparatus for creating cost data for use in generating a route across an electronic map and its abstract is a method is disclosed involving receiving gps data from persona portable training devices of users when traversing an off-road segment of an electronic map together with associated data indicative of a heart rate of a user during the movements. the position and heart rate data for each user traversing the segment are processed using data indicative of a fitness profile for the user. the resulting data is used to determine a normalized cost to be associated with the segment, indicative of the difficulty in traversing the segment. the cost data is generated using a neural network. the resulting cost data for different segments in a network of segments is used to generate route suggestions for users based upon desired workout intensity, fitness levels, etc. dated 2016-09-06"
9435732,hyperspectral identification of egg fertility and gender,a hyperspectral method for detecting the present condition of an avian egg is disclosed in which a neural network algorithm is used to compare the spectrum of a test egg against a spectral library. the method can detect fertility with greater than 90% reliability on the day of laying and the gender of the chick with greater than 75% reliability on the 12th day after laying.,2016-09-06,The title of the patent is hyperspectral identification of egg fertility and gender and its abstract is a hyperspectral method for detecting the present condition of an avian egg is disclosed in which a neural network algorithm is used to compare the spectrum of a test egg against a spectral library. the method can detect fertility with greater than 90% reliability on the day of laying and the gender of the chick with greater than 75% reliability on the 12th day after laying. dated 2016-09-06
9436895,method for determining similarity of objects represented in images,"a method re-identifies objects in a pair of images by applying a convolutional neural network (cnn). each layer in the network operates on an output of a previous layer. the layers include a first convolutional layer and a first max pooling layer to determine a feature map, a cross-input neighborhood differences layer to produce neighborhood difference maps, a patch summary layer to produce patch summary feature maps, a first fully connected layer to produce a feature vector representing higher order relationships in the patch summary feature maps, a second fully connected layer to produce two scores representing positive pair and negative pair classes, and a softmax layer to produce positive pair and negative pair probabilities. then, the positive pair probability is output to signal whether the two images represent the same object or not.",2016-09-06,"The title of the patent is method for determining similarity of objects represented in images and its abstract is a method re-identifies objects in a pair of images by applying a convolutional neural network (cnn). each layer in the network operates on an output of a previous layer. the layers include a first convolutional layer and a first max pooling layer to determine a feature map, a cross-input neighborhood differences layer to produce neighborhood difference maps, a patch summary layer to produce patch summary feature maps, a first fully connected layer to produce a feature vector representing higher order relationships in the patch summary feature maps, a second fully connected layer to produce two scores representing positive pair and negative pair classes, and a softmax layer to produce positive pair and negative pair probabilities. then, the positive pair probability is output to signal whether the two images represent the same object or not. dated 2016-09-06"
9436907,method and system for calculating value of website visitor,"calculating a value of a website visitor includes initializing a calculation model for calculating the value of the website visitor, the calculation model being a neural network model with visitor information as an input and the visitor's value as an output; training the calculation model by using a data sample and determining the calculation model; and obtaining the visitor information, and calculating the value of the visitor by using the determined calculation model.",2016-09-06,"The title of the patent is method and system for calculating value of website visitor and its abstract is calculating a value of a website visitor includes initializing a calculation model for calculating the value of the website visitor, the calculation model being a neural network model with visitor information as an input and the visitor's value as an output; training the calculation model by using a data sample and determining the calculation model; and obtaining the visitor information, and calculating the value of the visitor by using the determined calculation model. dated 2016-09-06"
9436911,neural networking system and methods,a method/apparatus/system for generating a request for improvement of a data object in a neural network is described herein. the neural network contains a plurality of data objects each made of an aggregation of content. the data objects of the neural network are interconnected based on one or several skill levels embodied in the content of the data objects via a plurality of connecting vectors. these connecting vectors can be generated and/or modified based on data collected from the iterative transversal of the connecting vectors by one or several users of the neural network.,2016-09-06,The title of the patent is neural networking system and methods and its abstract is a method/apparatus/system for generating a request for improvement of a data object in a neural network is described herein. the neural network contains a plurality of data objects each made of an aggregation of content. the data objects of the neural network are interconnected based on one or several skill levels embodied in the content of the data objects via a plurality of connecting vectors. these connecting vectors can be generated and/or modified based on data collected from the iterative transversal of the connecting vectors by one or several users of the neural network. dated 2016-09-06
9443192,universal artificial intelligence engine for autonomous computing devices and software applications,"aspects of the disclosure generally relate to computing devices and may be generally directed to devices, systems, methods, and/or applications for learning the operation of a computing device or software application, storing this knowledge in a knowledgebase, neural network, or other repository, and enabling autonomous operation of the computing device or software application with partial, minimal, or no user input.",2016-09-13,"The title of the patent is universal artificial intelligence engine for autonomous computing devices and software applications and its abstract is aspects of the disclosure generally relate to computing devices and may be generally directed to devices, systems, methods, and/or applications for learning the operation of a computing device or software application, storing this knowledge in a knowledgebase, neural network, or other repository, and enabling autonomous operation of the computing device or software application with partial, minimal, or no user input. dated 2016-09-13"
9443517,generating sounds for detectability by neural networks,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network. one of the methods includes accessing a first neural network that was trained to recognize a given keyword or keyphrase using a set of hotword training data, wherein the hotword training data includes positive hotword training data that correspond to utterances of the keyword or keyphrase, and negative hotword training data that corresponds to utterances of words or phrases that are other than the keyword or keyphrase, selecting a seed hotsound, mapping, to a feature space, (i) the positive hotword training data, (ii) the negative hotword training data, and (iii) the seed hotsound, performing an optimization of a position of the seed hotsound within the feature space to generate a modified seed hotsound, generating a set of hotsound training data using the modified seed hotsound, training a second neural network to recognize the modified seed hotsound using the generated set of hotsound training data, and using the trained second neural network to recognize the modified hotsound.",2016-09-13,"The title of the patent is generating sounds for detectability by neural networks and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network. one of the methods includes accessing a first neural network that was trained to recognize a given keyword or keyphrase using a set of hotword training data, wherein the hotword training data includes positive hotword training data that correspond to utterances of the keyword or keyphrase, and negative hotword training data that corresponds to utterances of words or phrases that are other than the keyword or keyphrase, selecting a seed hotsound, mapping, to a feature space, (i) the positive hotword training data, (ii) the negative hotword training data, and (iii) the seed hotsound, performing an optimization of a position of the seed hotsound within the feature space to generate a modified seed hotsound, generating a set of hotsound training data using the modified seed hotsound, training a second neural network to recognize the modified seed hotsound using the generated set of hotsound training data, and using the trained second neural network to recognize the modified hotsound. dated 2016-09-13"
9449257,dynamically reconstructable multistage parallel single instruction multiple data array processing system,"the present invention proposes a dynamically reconfigurable multi-level parallel single instruction multiple data array processing system which has a pixel level parallel image processing element array and a row-parallel array processor. the pe array mainly implements a linear operation which is adapted to be executed in parallel in the low and middle levels of image processing and the rp array implements an operation which is adapted to execute in row-parallel in the low and middle levels of image processing or more complex nonlinear operations. in particularly, such a system can be dynamically reconfigured as an som neural network at a low cost of area, and the neural network supports high level of image processing such as a high speed online neural network training and image feature recognition, and completely overcomes a defect that a high level of image processing can't be done by pixel-level parallel processing array in the existing programmable vision chips and parallel vision processors, and facilitates an intelligent and portable real time on-chip vision image system with a complete function at low device cost and low power consumption.",2016-09-20,"The title of the patent is dynamically reconstructable multistage parallel single instruction multiple data array processing system and its abstract is the present invention proposes a dynamically reconfigurable multi-level parallel single instruction multiple data array processing system which has a pixel level parallel image processing element array and a row-parallel array processor. the pe array mainly implements a linear operation which is adapted to be executed in parallel in the low and middle levels of image processing and the rp array implements an operation which is adapted to execute in row-parallel in the low and middle levels of image processing or more complex nonlinear operations. in particularly, such a system can be dynamically reconfigured as an som neural network at a low cost of area, and the neural network supports high level of image processing such as a high speed online neural network training and image feature recognition, and completely overcomes a defect that a high level of image processing can't be done by pixel-level parallel processing array in the existing programmable vision chips and parallel vision processors, and facilitates an intelligent and portable real time on-chip vision image system with a complete function at low device cost and low power consumption. dated 2016-09-20"
9449271,classifying resources using a deep network,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for scoring concept terms using a deep network. one of the methods includes receiving an input comprising a plurality of features of a resource, wherein each feature is a value of a respective attribute of the resource; processing each of the features using a respective embedding function to generate one or more numeric values; processing the numeric values using one or more neural network layers to generate an alternative representation of the features, wherein processing the floating point values comprises applying one or more non-linear transformations to the floating point values; and processing the alternative representation of the input using a classifier to generate a respective category score for each category in a pre-determined set of categories, wherein each of the respective category scores measure a predicted likelihood that the resource belongs to the corresponding category.",2016-09-20,"The title of the patent is classifying resources using a deep network and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for scoring concept terms using a deep network. one of the methods includes receiving an input comprising a plurality of features of a resource, wherein each feature is a value of a respective attribute of the resource; processing each of the features using a respective embedding function to generate one or more numeric values; processing the numeric values using one or more neural network layers to generate an alternative representation of the features, wherein processing the floating point values comprises applying one or more non-linear transformations to the floating point values; and processing the alternative representation of the input using a classifier to generate a respective category score for each category in a pre-determined set of categories, wherein each of the respective category scores measure a predicted likelihood that the resource belongs to the corresponding category. dated 2016-09-20"
9449272,doppler effect processing in a neural network model,a method of frequency discrimination associated with the doppler effect is presented. the method includes mapping a first signal to a first plurality of frequency bins and a second signal to a second plurality of frequency bins. the first signal and the second signal corresponding to different times. the method also includes firing a first plurality of neurons based on contents of the first plurality of frequency bins and firing a second plurality of neurons based on contents of the second plurality of frequency bins.,2016-09-20,The title of the patent is doppler effect processing in a neural network model and its abstract is a method of frequency discrimination associated with the doppler effect is presented. the method includes mapping a first signal to a first plurality of frequency bins and a second signal to a second plurality of frequency bins. the first signal and the second signal corresponding to different times. the method also includes firing a first plurality of neurons based on contents of the first plurality of frequency bins and firing a second plurality of neurons based on contents of the second plurality of frequency bins. dated 2016-09-20
9449285,system and method for using pattern recognition to monitor and maintain status quo,"a system for prospectively identifying media characteristics for inclusion in media content is disclosed. a neural network database including media characteristic information and feature information may associate relationships among the media characteristic information and feature information. personal characteristic information associated with target media consumers may be used to select a subset of the neural network database. a first set of nodes, representing selected feature information, may be activated. the node interactions may be calculated to detect the activation of a second set of nodes, the second set of nodes representing media characteristic information. generally, a node is activated when an activation value of the node exceeds a threshold value. media characteristic information may be identified for inclusion in media content based on the second set of nodes.",2016-09-20,"The title of the patent is system and method for using pattern recognition to monitor and maintain status quo and its abstract is a system for prospectively identifying media characteristics for inclusion in media content is disclosed. a neural network database including media characteristic information and feature information may associate relationships among the media characteristic information and feature information. personal characteristic information associated with target media consumers may be used to select a subset of the neural network database. a first set of nodes, representing selected feature information, may be activated. the node interactions may be calculated to detect the activation of a second set of nodes, the second set of nodes representing media characteristic information. generally, a node is activated when an activation value of the node exceeds a threshold value. media characteristic information may be identified for inclusion in media content based on the second set of nodes. dated 2016-09-20"
9454714,sequence transcription with deep neural networks,"systems and methods for sequence transcription with neural networks are provided. more particularly, a neural network can be implemented to map a plurality of training images received by the neural network into a probabilistic model of sequences comprising p(s|x) by maximizing log p(s|x) on the plurality of training images. x represents an input image and s represents an output sequence of characters for the input image. the trained neural network can process a received image containing characters associated with building numbers. the trained neural network can generate a predicted sequence of characters by processing the received image.",2016-09-27,"The title of the patent is sequence transcription with deep neural networks and its abstract is systems and methods for sequence transcription with neural networks are provided. more particularly, a neural network can be implemented to map a plurality of training images received by the neural network into a probabilistic model of sequences comprising p(s|x) by maximizing log p(s|x) on the plurality of training images. x represents an input image and s represents an output sequence of characters for the input image. the trained neural network can process a received image containing characters associated with building numbers. the trained neural network can generate a predicted sequence of characters by processing the received image. dated 2016-09-27"
9454725,passage justification scoring for question answering,"according to an aspect, passage justification scoring includes creating a multi-layered neural network from domain knowledge and training the multi-layered neural network with labeled data and unlabeled data. a further aspect includes inputting at least one of an existing passage justification component and raw input data for a question and passage to the multi-layered neural network, extracting concepts determined to have passage justification with respect to a candidate answer contained in a respective passage, and creating a passage justification model from the extracted concepts and from passage justification ground truth.",2016-09-27,"The title of the patent is passage justification scoring for question answering and its abstract is according to an aspect, passage justification scoring includes creating a multi-layered neural network from domain knowledge and training the multi-layered neural network with labeled data and unlabeled data. a further aspect includes inputting at least one of an existing passage justification component and raw input data for a question and passage to the multi-layered neural network, extracting concepts determined to have passage justification with respect to a candidate answer contained in a respective passage, and creating a passage justification model from the extracted concepts and from passage justification ground truth. dated 2016-09-27"
9454958,exploiting heterogeneous data in deep neural network-based speech recognition systems,"technologies pertaining to training a deep neural network (dnn) for use in a recognition system are described herein. the dnn is trained using heterogeneous data, the heterogeneous data including narrowband signals and wideband signals. the dnn, subsequent to being trained, receives an input signal that can be either a wideband signal or narrowband signal. the dnn estimates the class posterior probability of the input signal regardless of whether the input signal is the wideband signal or the narrowband signal.",2016-09-27,"The title of the patent is exploiting heterogeneous data in deep neural network-based speech recognition systems and its abstract is technologies pertaining to training a deep neural network (dnn) for use in a recognition system are described herein. the dnn is trained using heterogeneous data, the heterogeneous data including narrowband signals and wideband signals. the dnn, subsequent to being trained, receives an input signal that can be either a wideband signal or narrowband signal. the dnn estimates the class posterior probability of the input signal regardless of whether the input signal is the wideband signal or the narrowband signal. dated 2016-09-27"
9456174,neural network for video editing,an automated video editing system uses user inputs and metadata combined with machine learning technology to gradually improve editing techniques as more footage is edited. the system is designed to work primarily with a network of automated video recording systems that use cooperative tracking methods. the system is also designed to improve tracking algorithms used in cooperative tracking and to enable systems to begin using image recognition based tracking when the results of machine learning are utilized.,2016-09-27,The title of the patent is neural network for video editing and its abstract is an automated video editing system uses user inputs and metadata combined with machine learning technology to gradually improve editing techniques as more footage is edited. the system is designed to work primarily with a network of automated video recording systems that use cooperative tracking methods. the system is also designed to improve tracking algorithms used in cooperative tracking and to enable systems to begin using image recognition based tracking when the results of machine learning are utilized. dated 2016-09-27
9459142,flame detectors and testing methods,"exemplary embodiments of a flame detector and operating method. optical energy is received at one or more optical sensors, and the detector processes the energy to determine whether the received energy is from a known remote test source. if so, the flame detector is operated in a test mode. if the processing indicates that the received optical energy is not a test signal, the flame detector is operated in a flame detection operating mode. the detector processing uses an artificial neural network in an exemplary embodiment in the flame detection operation mode.",2016-10-04,"The title of the patent is flame detectors and testing methods and its abstract is exemplary embodiments of a flame detector and operating method. optical energy is received at one or more optical sensors, and the detector processes the energy to determine whether the received energy is from a known remote test source. if so, the flame detector is operated in a test mode. if the processing indicates that the received optical energy is not a test signal, the flame detector is operated in a flame detection operating mode. the detector processing uses an artificial neural network in an exemplary embodiment in the flame detection operation mode. dated 2016-10-04"
9460382,neural watchdog,a method of monitoring a neural network includes monitoring activity of the neural network. the method also includes detecting a condition based on the activity. the method further includes performing an exception event based on the detected condition.,2016-10-04,The title of the patent is neural watchdog and its abstract is a method of monitoring a neural network includes monitoring activity of the neural network. the method also includes detecting a condition based on the activity. the method further includes performing an exception event based on the detected condition. dated 2016-10-04
9460383,reconfigurable and customizable general-purpose circuits for neural networks,"a reconfigurable neural network circuit is provided. the reconfigurable neural network circuit comprises an electronic synapse array including multiple synapses interconnecting a plurality of digital electronic neurons. each neuron comprises an integrator that integrates input spikes and generates a signal when the integrated inputs exceed a threshold. the circuit further comprises a control module for reconfiguring the synapse array. the control module comprises a global final state machine that controls timing for operation of the circuit, and a priority encoder that allows spiking neurons to sequentially access the synapse array.",2016-10-04,"The title of the patent is reconfigurable and customizable general-purpose circuits for neural networks and its abstract is a reconfigurable neural network circuit is provided. the reconfigurable neural network circuit comprises an electronic synapse array including multiple synapses interconnecting a plurality of digital electronic neurons. each neuron comprises an integrator that integrates input spikes and generates a signal when the integrated inputs exceed a threshold. the circuit further comprises a control module for reconfiguring the synapse array. the control module comprises a global final state machine that controls timing for operation of the circuit, and a priority encoder that allows spiking neurons to sequentially access the synapse array. dated 2016-10-04"
9460384,effecting modulation by global scalar values in a spiking neural network,methods and apparatus are provided for effecting modulation using global scalar values in a spiking neural network. one example method for operating an artificial nervous system generally includes determining one or more updated values for artificial neuromodulators to be used by a plurality of entities in a neuron model and providing the updated values to the plurality of entities.,2016-10-04,The title of the patent is effecting modulation by global scalar values in a spiking neural network and its abstract is methods and apparatus are provided for effecting modulation using global scalar values in a spiking neural network. one example method for operating an artificial nervous system generally includes determining one or more updated values for artificial neuromodulators to be used by a plurality of entities in a neuron model and providing the updated values to the plurality of entities. dated 2016-10-04
9460386,passage justification scoring for question answering,"according to an aspect, passage justification scoring is implemented by a processor executing computer readable instructions. the computer readable instructions include creating a multi-layered neural network from domain knowledge and training the multi-layered neural network with labeled data and unlabeled data. the computer readable instructions further include inputting at least one of an existing passage justification component and raw input data for a question and passage to the multi-layered neural network, extracting concepts determined to have passage justification with respect to a candidate answer contained in a respective passage, and creating a passage justification model from the extracted concepts and from passage justification ground truth.",2016-10-04,"The title of the patent is passage justification scoring for question answering and its abstract is according to an aspect, passage justification scoring is implemented by a processor executing computer readable instructions. the computer readable instructions include creating a multi-layered neural network from domain knowledge and training the multi-layered neural network with labeled data and unlabeled data. the computer readable instructions further include inputting at least one of an existing passage justification component and raw input data for a question and passage to the multi-layered neural network, extracting concepts determined to have passage justification with respect to a candidate answer contained in a respective passage, and creating a passage justification model from the extracted concepts and from passage justification ground truth. dated 2016-10-04"
9460704,deep networks for unit selection speech synthesis,"methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for providing a representation based on structured data in resources. the methods, systems, and apparatus include actions of receiving target acoustic features output from a neural network that has been trained to predict acoustic features given linguistic features. additional actions include determining a distance between the target acoustic features and acoustic features of a stored acoustic sample. further actions include selecting the acoustic sample to be used in speech synthesis based at least on the determined distance and synthesizing speech based on the selected acoustic sample.",2016-10-04,"The title of the patent is deep networks for unit selection speech synthesis and its abstract is methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for providing a representation based on structured data in resources. the methods, systems, and apparatus include actions of receiving target acoustic features output from a neural network that has been trained to predict acoustic features given linguistic features. additional actions include determining a distance between the target acoustic features and acoustic features of a stored acoustic sample. further actions include selecting the acoustic sample to be used in speech synthesis based at least on the determined distance and synthesizing speech based on the selected acoustic sample. dated 2016-10-04"
9460711,"multilingual, acoustic deep neural networks","methods and systems for processing multilingual dnn acoustic models are described. an example method may include receiving training data that includes a respective training data set for each of two or more or languages. a multilingual deep neural network (dnn) acoustic model may be processed based on the training data. the multilingual dnn acoustic model may include a feedforward neural network having multiple layers of one or more nodes. each node of a given layer may connect with a respective weight to each node of a subsequent layer, and the multiple layers of one or more nodes may include one or more shared hidden layers of nodes and a language-specific output layer of nodes corresponding to each of the two or more languages. additionally, weights associated with the multiple layers of one or more nodes of the processed multilingual dnn acoustic model may be stored in a database.",2016-10-04,"The title of the patent is multilingual, acoustic deep neural networks and its abstract is methods and systems for processing multilingual dnn acoustic models are described. an example method may include receiving training data that includes a respective training data set for each of two or more or languages. a multilingual deep neural network (dnn) acoustic model may be processed based on the training data. the multilingual dnn acoustic model may include a feedforward neural network having multiple layers of one or more nodes. each node of a given layer may connect with a respective weight to each node of a subsequent layer, and the multiple layers of one or more nodes may include one or more shared hidden layers of nodes and a language-specific output layer of nodes corresponding to each of the two or more languages. additionally, weights associated with the multiple layers of one or more nodes of the processed multilingual dnn acoustic model may be stored in a database. dated 2016-10-04"
9466022,hardware architecture for simulating a neural network of neurons,"embodiments of the invention relate to a neural network system for simulating neurons of a neural model. one embodiment comprises a memory device that maintains neuronal states for multiple neurons, a lookup table that maintains state transition information for multiple neuronal states, and a controller unit that manages the memory device. the controller unit updates a neuronal state for each neuron based on incoming spike events targeting said neuron and state transition information corresponding to said neuronal state.",2016-10-11,"The title of the patent is hardware architecture for simulating a neural network of neurons and its abstract is embodiments of the invention relate to a neural network system for simulating neurons of a neural model. one embodiment comprises a memory device that maintains neuronal states for multiple neurons, a lookup table that maintains state transition information for multiple neuronal states, and a controller unit that manages the memory device. the controller unit updates a neuronal state for each neuron based on incoming spike events targeting said neuron and state transition information corresponding to said neuronal state. dated 2016-10-11"
9466292,online incremental adaptation of deep neural networks using auxiliary gaussian mixture models in speech recognition,"methods and systems for online incremental adaptation of neural networks using gaussian mixture models in speech recognition are described. in an example, a computing device may be configured to receive an audio signal and a subsequent audio signal, both signals having speech content. the computing device may be configured to apply a speaker-specific feature transform to the audio signal to obtain a transformed audio signal. the speaker-specific feature transform may be configured to include speaker-specific speech characteristics of a speaker-profile relating to the speech content. further, the computing device may be configured to process the transformed audio signal using a neural network trained to estimate a respective speech content of the audio signal. based on outputs of the neural network, the computing device may be configured to modify the speaker-specific feature transform, and apply the modified speaker-specific feature transform to a subsequent audio signal.",2016-10-11,"The title of the patent is online incremental adaptation of deep neural networks using auxiliary gaussian mixture models in speech recognition and its abstract is methods and systems for online incremental adaptation of neural networks using gaussian mixture models in speech recognition are described. in an example, a computing device may be configured to receive an audio signal and a subsequent audio signal, both signals having speech content. the computing device may be configured to apply a speaker-specific feature transform to the audio signal to obtain a transformed audio signal. the speaker-specific feature transform may be configured to include speaker-specific speech characteristics of a speaker-profile relating to the speech content. further, the computing device may be configured to process the transformed audio signal using a neural network trained to estimate a respective speech content of the audio signal. based on outputs of the neural network, the computing device may be configured to modify the speaker-specific feature transform, and apply the modified speaker-specific feature transform to a subsequent audio signal. dated 2016-10-11"
9471852,user-configurable settings for content obfuscation,"an aspect of providing user-configurable settings for content obfuscation includes, for each media segment in a media file, inputting the media segment to a neural network, applying a classifier to features output by the neural network, and determining from results of the classifier images in the media segment that contain the sensitive characteristics. the classifier specifies images that are predetermined to include sensitive characteristics. an aspect further includes assigning a tag to each of the images in the media segment that contain the sensitive characteristics. the tag indicates a type of sensitivity. an aspect also includes receiving at least one user-defined sensitivity, the user-defined sensitivity indicating an action or condition that is considered objectionable to a user, identifying a subset of the tagged images that correlate to the user-defined sensitivity, and visually modifying, during playback of the media file, an appearance of the subset of the tagged images.",2016-10-18,"The title of the patent is user-configurable settings for content obfuscation and its abstract is an aspect of providing user-configurable settings for content obfuscation includes, for each media segment in a media file, inputting the media segment to a neural network, applying a classifier to features output by the neural network, and determining from results of the classifier images in the media segment that contain the sensitive characteristics. the classifier specifies images that are predetermined to include sensitive characteristics. an aspect further includes assigning a tag to each of the images in the media segment that contain the sensitive characteristics. the tag indicates a type of sensitivity. an aspect also includes receiving at least one user-defined sensitivity, the user-defined sensitivity indicating an action or condition that is considered objectionable to a user, identifying a subset of the tagged images that correlate to the user-defined sensitivity, and visually modifying, during playback of the media file, an appearance of the subset of the tagged images. dated 2016-10-18"
9483727,reduction of computation complexity of neural network sensitivity analysis,"as part of neural network sensitivity analysis, base outputs of hidden layer nodes of a neural network model for non-perturbed variables can be reused when perturbing the variables. such an arrangement greatly reduces complexity of the calculations required to generate outputs of the model. related apparatus, systems, techniques and articles are also described.",2016-11-01,"The title of the patent is reduction of computation complexity of neural network sensitivity analysis and its abstract is as part of neural network sensitivity analysis, base outputs of hidden layer nodes of a neural network model for non-perturbed variables can be reused when perturbing the variables. such an arrangement greatly reduces complexity of the calculations required to generate outputs of the model. related apparatus, systems, techniques and articles are also described. dated 2016-11-01"
9483728,systems and methods for combining stochastic average gradient and hessian-free optimization for sequence training of deep neural networks,"a method for training a deep neural network (dnn), comprises receiving and formatting speech data for the training, performing hessian-free sequence training (hfst) on a first subset of a plurality of subsets of the speech data, and iteratively performing the hfst on successive subsets of the plurality of subsets of the speech data, wherein iteratively performing the hfst comprises reusing information from at least one previous iteration.",2016-11-01,"The title of the patent is systems and methods for combining stochastic average gradient and hessian-free optimization for sequence training of deep neural networks and its abstract is a method for training a deep neural network (dnn), comprises receiving and formatting speech data for the training, performing hessian-free sequence training (hfst) on a first subset of a plurality of subsets of the speech data, and iteratively performing the hfst on successive subsets of the plurality of subsets of the speech data, wherein iteratively performing the hfst comprises reusing information from at least one previous iteration. dated 2016-11-01"
9484022,training multiple neural networks with different accuracy,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a deep neural network. one of the methods includes generating a plurality of feature vectors that each model a different portion of an audio waveform, generating a first posterior probability vector for a first feature vector using a first neural network, determining whether one of the scores in the first posterior probability vector satisfies a first threshold value, generating a second posterior probability vector for each subsequent feature vector using a second neural network, wherein the second neural network is trained to identify the same key words and key phrases and includes more inner layer nodes than the first neural network, and determining whether one of the scores in the second posterior probability vector satisfies a second threshold value.",2016-11-01,"The title of the patent is training multiple neural networks with different accuracy and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a deep neural network. one of the methods includes generating a plurality of feature vectors that each model a different portion of an audio waveform, generating a first posterior probability vector for a first feature vector using a first neural network, determining whether one of the scores in the first posterior probability vector satisfies a first threshold value, generating a second posterior probability vector for each subsequent feature vector using a second neural network, wherein the second neural network is trained to identify the same key words and key phrases and includes more inner layer nodes than the first neural network, and determining whether one of the scores in the second posterior probability vector satisfies a second threshold value. dated 2016-11-01"
9484974,methods and systems for multi-layer perceptron based non-linear interference management in multi-technology communication devices,"the various embodiments include methods and apparatuses for canceling nonlinear interference during concurrent communication of multi-technology wireless communication devices. nonlinear interference may be estimated using a multilayer perceptron neural network with hammerstein structure by dividing an aggressor signal into real and imaginary components, augmenting the components by weight factors, executing a linear combination of the augmented components, and executing a nonlinear sigmoid function for the combined components at a hidden layer of multilayer perceptron neural network to produce a hidden layer output signal. at an output layer, hidden layer output signals may be augmented by weight factors, and the augmented hidden layer output signals may be linearly combined to produce real and imaginary components of an estimated jammer signal. a linear filter function may be executed for the components of the jammer signal, and to produce a nonlinear interference estimate used to cancel the nonlinear interference of a victim signal.",2016-11-01,"The title of the patent is methods and systems for multi-layer perceptron based non-linear interference management in multi-technology communication devices and its abstract is the various embodiments include methods and apparatuses for canceling nonlinear interference during concurrent communication of multi-technology wireless communication devices. nonlinear interference may be estimated using a multilayer perceptron neural network with hammerstein structure by dividing an aggressor signal into real and imaginary components, augmenting the components by weight factors, executing a linear combination of the augmented components, and executing a nonlinear sigmoid function for the combined components at a hidden layer of multilayer perceptron neural network to produce a hidden layer output signal. at an output layer, hidden layer output signals may be augmented by weight factors, and the augmented hidden layer output signals may be linearly combined to produce real and imaginary components of an estimated jammer signal. a linear filter function may be executed for the components of the jammer signal, and to produce a nonlinear interference estimate used to cancel the nonlinear interference of a victim signal. dated 2016-11-01"
9489618,electronic comparison systems,"an electronic comparison system includes input stages that successively provide bits of code words. one-shots connected to respective stages successively provide a first bit value until receiving a bit having a non-preferred value concurrently with an enable signal, and then provide a second, different bit value. an enable circuit provides the enable signal if at least one of the one-shots is providing the first bit value. a neural network system includes a crossbar with row and column electrodes and resistive memory elements at their intersections. a writing circuit stores weights in the elements. a signal source applies signals to the row electrodes. comparators compare signals on the column electrodes to corresponding references using domain-wall neurons and store bit values in cmos latches by comparison with a threshold.",2016-11-08,"The title of the patent is electronic comparison systems and its abstract is an electronic comparison system includes input stages that successively provide bits of code words. one-shots connected to respective stages successively provide a first bit value until receiving a bit having a non-preferred value concurrently with an enable signal, and then provide a second, different bit value. an enable circuit provides the enable signal if at least one of the one-shots is providing the first bit value. a neural network system includes a crossbar with row and column electrodes and resistive memory elements at their intersections. a writing circuit stores weights in the elements. a signal source applies signals to the row electrodes. comparators compare signals on the column electrodes to corresponding references using domain-wall neurons and store bit values in cmos latches by comparison with a threshold. dated 2016-11-08"
9489619,method for the computer-assisted modeling of a technical system,"a method for computer-assisted modeling of a technical system is disclosed. at multiple different operating points, the technical system is described by a first state vector with first state variable(s) and by a second state vector with second state variable(s). a neural network comprising a special form of a feed-forward network is used for the computer-assisted modeling of said system. the feed-forward network includes at least one bridging connector that connects a neural layer with an output layer, thereby bridging at least one hidden layer, which allows the training of networks with multiple hidden layers in a simple manner with known learning methods, e.g., the gradient descent method. the method may be used for modeling a gas turbine system, in which a neural network trained using the method may be used to estimate or predict nitrogen oxide or carbon monoxide emissions or parameters relating to combustion chamber vibrations.",2016-11-08,"The title of the patent is method for the computer-assisted modeling of a technical system and its abstract is a method for computer-assisted modeling of a technical system is disclosed. at multiple different operating points, the technical system is described by a first state vector with first state variable(s) and by a second state vector with second state variable(s). a neural network comprising a special form of a feed-forward network is used for the computer-assisted modeling of said system. the feed-forward network includes at least one bridging connector that connects a neural layer with an output layer, thereby bridging at least one hidden layer, which allows the training of networks with multiple hidden layers in a simple manner with known learning methods, e.g., the gradient descent method. the method may be used for modeling a gas turbine system, in which a neural network trained using the method may be used to estimate or predict nitrogen oxide or carbon monoxide emissions or parameters relating to combustion chamber vibrations. dated 2016-11-08"
9489620,quick analysis of residual stress and distortion in cast aluminum components,"a computer-implemented system and method of rapidly predicting at least one of residual stress and distortion of a quenched aluminum casting. input data corresponding to at least one of topological features, geometrical features and quenching process parameters associated with the casting is operated upon by the computer that is configured as a neural network to determine output data corresponding to at least one of the residual stress and distortion based on the input data. the neural network is trained to determine the validity of at least one of the input data and output data and to retrain the network when an error threshold is exceeded. thereby, residual stresses and distortion in the quenched aluminum castings can be predicted using the embodiments in a tiny fraction of the time required by conventional finite-element based approaches.",2016-11-08,"The title of the patent is quick analysis of residual stress and distortion in cast aluminum components and its abstract is a computer-implemented system and method of rapidly predicting at least one of residual stress and distortion of a quenched aluminum casting. input data corresponding to at least one of topological features, geometrical features and quenching process parameters associated with the casting is operated upon by the computer that is configured as a neural network to determine output data corresponding to at least one of the residual stress and distortion based on the input data. the neural network is trained to determine the validity of at least one of the input data and output data and to retrain the network when an error threshold is exceeded. thereby, residual stresses and distortion in the quenched aluminum castings can be predicted using the embodiments in a tiny fraction of the time required by conventional finite-element based approaches. dated 2016-11-08"
9489622,event-driven universal neural network circuit,the present invention provides an event-driven universal neural network circuit. the circuit comprises a plurality of neural modules. each neural module comprises multiple digital neurons such that each neuron in a neural module has a corresponding neuron in another neural module. an interconnection network comprising a plurality of digital synapses interconnects the neural modules. each synapse interconnects a first neural module to a second neural module by interconnecting a neuron in the first neural module to a corresponding neuron in the second neural module. corresponding neurons in the first neural module and the second neural module communicate via the synapses. each synapse comprises a learning rule associating a neuron in the first neural module with a corresponding neuron in the second neural module. a control module generates signals which define a set of time steps for event-driven operation of the neurons and event communication via the interconnection network.,2016-11-08,The title of the patent is event-driven universal neural network circuit and its abstract is the present invention provides an event-driven universal neural network circuit. the circuit comprises a plurality of neural modules. each neural module comprises multiple digital neurons such that each neuron in a neural module has a corresponding neuron in another neural module. an interconnection network comprising a plurality of digital synapses interconnects the neural modules. each synapse interconnects a first neural module to a second neural module by interconnecting a neuron in the first neural module to a corresponding neuron in the second neural module. corresponding neurons in the first neural module and the second neural module communicate via the synapses. each synapse comprises a learning rule associating a neuron in the first neural module with a corresponding neuron in the second neural module. a control module generates signals which define a set of time steps for event-driven operation of the neurons and event communication via the interconnection network. dated 2016-11-08
9495619,systems and methods for image object recognition based on location information and object categories,"systems, methods, and non-transitory computer-readable media can identify a set of regions corresponding to a geographical area. a collection of training images can be acquired. each training image in the collection can be associated with one or more respective recognized objects and with a respective region in the set of regions. histogram metrics for a plurality of object categories within each region in the set of regions can be determined based at least in part on the collection of training images. a neural network can be developed based at least in part on the histogram metrics for the plurality of object categories within each region in the set of regions and on the collection of training images.",2016-11-15,"The title of the patent is systems and methods for image object recognition based on location information and object categories and its abstract is systems, methods, and non-transitory computer-readable media can identify a set of regions corresponding to a geographical area. a collection of training images can be acquired. each training image in the collection can be associated with one or more respective recognized objects and with a respective region in the set of regions. histogram metrics for a plurality of object categories within each region in the set of regions can be determined based at least in part on the collection of training images. a neural network can be developed based at least in part on the histogram metrics for the plurality of object categories within each region in the set of regions and on the collection of training images. dated 2016-11-15"
9495633,recurrent neural networks for malware analysis,"using a recurrent neural network (rnn) that has been trained to a satisfactory level of performance, highly discriminative features can be extracted by running a sample through the rnn, and then extracting a final hidden state hi, where i is the number of instructions of the sample. this resulting feature vector may then be concatenated with the other hand-engineered features, and a larger classifier may then be trained on hand-engineered as well as automatically determined features. related apparatus, systems, techniques and articles are also described.",2016-11-15,"The title of the patent is recurrent neural networks for malware analysis and its abstract is using a recurrent neural network (rnn) that has been trained to a satisfactory level of performance, highly discriminative features can be extracted by running a sample through the rnn, and then extracting a final hidden state hi, where i is the number of instructions of the sample. this resulting feature vector may then be concatenated with the other hand-engineered features, and a larger classifier may then be trained on hand-engineered as well as automatically determined features. related apparatus, systems, techniques and articles are also described. dated 2016-11-15"
9501724,font recognition and font similarity learning using a deep neural network,"a convolutional neural network (cnn) is trained for font recognition and font similarity learning. in a training phase, text images with font labels are synthesized by introducing variances to minimize the gap between the training images and real-world text images. training images are generated and input into the cnn. the output is fed into an n-way softmax function dependent on the number of fonts the cnn is being trained on, producing a distribution of classified text images over n class labels. in a testing phase, each test image is normalized in height and squeezed in aspect ratio resulting in a plurality of test patches. the cnn averages the probabilities of each test patch belonging to a set of fonts to obtain a classification. feature representations may be extracted and utilized to define font similarity between fonts, which may be utilized in font suggestion, font browsing, or font recognition applications.",2016-11-22,"The title of the patent is font recognition and font similarity learning using a deep neural network and its abstract is a convolutional neural network (cnn) is trained for font recognition and font similarity learning. in a training phase, text images with font labels are synthesized by introducing variances to minimize the gap between the training images and real-world text images. training images are generated and input into the cnn. the output is fed into an n-way softmax function dependent on the number of fonts the cnn is being trained on, producing a distribution of classified text images over n class labels. in a testing phase, each test image is normalized in height and squeezed in aspect ratio resulting in a plurality of test patches. the cnn averages the probabilities of each test patch belonging to a set of fonts to obtain a classification. feature representations may be extracted and utilized to define font similarity between fonts, which may be utilized in font suggestion, font browsing, or font recognition applications. dated 2016-11-22"
9501740,predicting well markers from artificial neural-network-predicted lithostratigraphic facies,"this disclosure generally describes methods and systems, including computer-implemented methods, computer-program products, and computer systems, for predicting well markers. one computer-implemented method includes separating neural-network (nn)-predicted facies output associated with a plurality of wells into two sets, a first set of nn-predicted facies output of training wells and a second set of nn-predicted facies output of target wells, calculating, for each training well of the plurality of wells, a sameness score between zones of nn-predicted facies output and human-identified lithostratigraphic units (finer zones), calculating a mean sameness score for the finer zones for all training wells, identifying finer zones with a mean sameness score greater than a threshold value as dominant facies zones, and iterating over each target well to calculate a top and depth position of each dominant facies zone determined based upon the nn-predicted facies output of the target well.",2016-11-22,"The title of the patent is predicting well markers from artificial neural-network-predicted lithostratigraphic facies and its abstract is this disclosure generally describes methods and systems, including computer-implemented methods, computer-program products, and computer systems, for predicting well markers. one computer-implemented method includes separating neural-network (nn)-predicted facies output associated with a plurality of wells into two sets, a first set of nn-predicted facies output of training wells and a second set of nn-predicted facies output of target wells, calculating, for each training well of the plurality of wells, a sameness score between zones of nn-predicted facies output and human-identified lithostratigraphic units (finer zones), calculating a mean sameness score for the finer zones for all training wells, identifying finer zones with a mean sameness score greater than a threshold value as dominant facies zones, and iterating over each target well to calculate a top and depth position of each dominant facies zone determined based upon the nn-predicted facies output of the target well. dated 2016-11-22"
9502038,method and device for voiceprint recognition,"a method and device for voiceprint recognition, include: establishing a first-level deep neural network (dnn) model based on unlabeled speech data, the unlabeled speech data containing no speaker labels and the first-level dnn model specifying a plurality of basic voiceprint features for the unlabeled speech data; obtaining a plurality of high-level voiceprint features by tuning the first-level dnn model based on labeled speech data, the labeled speech data containing speech samples with respective speaker labels, and the tuning producing a second-level dnn model specifying the plurality of high-level voiceprint features; based on the second-level dnn model, registering a respective high-level voiceprint feature sequence for a user based on a registration speech sample received from the user; and performing speaker verification for the user based on the respective high-level voiceprint feature sequence registered for the user.",2016-11-22,"The title of the patent is method and device for voiceprint recognition and its abstract is a method and device for voiceprint recognition, include: establishing a first-level deep neural network (dnn) model based on unlabeled speech data, the unlabeled speech data containing no speaker labels and the first-level dnn model specifying a plurality of basic voiceprint features for the unlabeled speech data; obtaining a plurality of high-level voiceprint features by tuning the first-level dnn model based on labeled speech data, the labeled speech data containing speech samples with respective speaker labels, and the tuning producing a second-level dnn model specifying the plurality of high-level voiceprint features; based on the second-level dnn model, registering a respective high-level voiceprint feature sequence for a user based on a registration speech sample received from the user; and performing speaker verification for the user based on the respective high-level voiceprint feature sequence registered for the user. dated 2016-11-22"
9508340,user specified keyword spotting using long short term memory neural network feature extractor,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for recognizing keywords using a long short term memory neural network. one of the methods includes receiving, by a device for each of multiple variable length enrollment audio signals, a respective plurality of enrollment feature vectors that represent features of the respective variable length enrollment audio signal, processing each of the plurality of enrollment feature vectors using a long short term memory (lstm) neural network to generate a respective enrollment lstm output vector for each enrollment feature vector, and generating, for the respective variable length enrollment audio signal, a template fixed length representation for use in determining whether another audio signal encodes another spoken utterance of the enrollment phrase by combining at most a quantity k of the enrollment lstm output vectors for the enrollment audio signal.",2016-11-29,"The title of the patent is user specified keyword spotting using long short term memory neural network feature extractor and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for recognizing keywords using a long short term memory neural network. one of the methods includes receiving, by a device for each of multiple variable length enrollment audio signals, a respective plurality of enrollment feature vectors that represent features of the respective variable length enrollment audio signal, processing each of the plurality of enrollment feature vectors using a long short term memory (lstm) neural network to generate a respective enrollment lstm output vector for each enrollment feature vector, and generating, for the respective variable length enrollment audio signal, a template fixed length representation for use in determining whether another audio signal encodes another spoken utterance of the enrollment phrase by combining at most a quantity k of the enrollment lstm output vectors for the enrollment audio signal. dated 2016-11-29"
9511366,microfluidic device and its use for positioning of cells or organisms,"a micro fluidic device comprises one microstructure layer (5) and one cover layer (1), wherein the cover layer (1) is connected to the microstructure layer (5). the microstructure layer (5) comprises one bottom layer and a plurality of microstructures on it to position samples. the cover layer (1) comprises one top layer, one positioning well (6) and at least one inlet pool (4). the positioning well (6) is right above the microstructures and connected with each other. the inlet pools (4) and the positioning well (6) are connected by microchannels (3) which are formed between the microstructure layer (5) and the cover layer (1). the micro fluidic device can be applied in vitro fertilization, in determining how glial cells affect neurons, in constructing neural network and in detecting cell growth conditions.",2016-12-06,"The title of the patent is microfluidic device and its use for positioning of cells or organisms and its abstract is a micro fluidic device comprises one microstructure layer (5) and one cover layer (1), wherein the cover layer (1) is connected to the microstructure layer (5). the microstructure layer (5) comprises one bottom layer and a plurality of microstructures on it to position samples. the cover layer (1) comprises one top layer, one positioning well (6) and at least one inlet pool (4). the positioning well (6) is right above the microstructures and connected with each other. the inlet pools (4) and the positioning well (6) are connected by microchannels (3) which are formed between the microstructure layer (5) and the cover layer (1). the micro fluidic device can be applied in vitro fertilization, in determining how glial cells affect neurons, in constructing neural network and in detecting cell growth conditions. dated 2016-12-06"
9514389,training a neural network to detect objects in images,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network to detect object in images. one of the methods includes receiving a training image and object location data for the training image; providing the training image to a neural network and obtaining bounding box data for the training image from the neural network, wherein the bounding box data comprises data defining a plurality of candidate bounding boxes in the training image and a respective confidence score for each candidate bounding box in the training image; determining an optimal set of assignments using the object location data for the training image and the bounding box data for the training image, wherein the optimal set of assignments assigns a respective candidate bounding box to each of the object locations; and training the neural network on the training image using the optimal set of assignments.",2016-12-06,"The title of the patent is training a neural network to detect objects in images and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network to detect object in images. one of the methods includes receiving a training image and object location data for the training image; providing the training image to a neural network and obtaining bounding box data for the training image from the neural network, wherein the bounding box data comprises data defining a plurality of candidate bounding boxes in the training image and a respective confidence score for each candidate bounding box in the training image; determining an optimal set of assignments using the object location data for the training image and the bounding box data for the training image, wherein the optimal set of assignments assigns a respective candidate bounding box to each of the object locations; and training the neural network on the training image using the optimal set of assignments. dated 2016-12-06"
9514390,systems and methods for identifying users in media content based on poselets and neural networks,"systems, methods, and non-transitory computer-readable media can receive a first image including a representation of a first user. a second image including a representation of a second user can be received. a first set of poselets associated with the first user can be detected in the first image. a second set of poselets associated with the second user can be detected in the second image. the first image including the first set of poselets can be inputted into a first instance of a neural network to generate a first multi-dimensional vector. the second image including the second set of poselets can be inputted into a second instance of the neural network to generate a second multi-dimensional vector. a first distance metric between the first multi-dimensional vector and the second multi-dimensional vector can be determined.",2016-12-06,"The title of the patent is systems and methods for identifying users in media content based on poselets and neural networks and its abstract is systems, methods, and non-transitory computer-readable media can receive a first image including a representation of a first user. a second image including a representation of a second user can be received. a first set of poselets associated with the first user can be detected in the first image. a second set of poselets associated with the second user can be detected in the second image. the first image including the first set of poselets can be inputted into a first instance of a neural network to generate a first multi-dimensional vector. the second image including the second set of poselets can be inputted into a second instance of the neural network to generate a second multi-dimensional vector. a first distance metric between the first multi-dimensional vector and the second multi-dimensional vector can be determined. dated 2016-12-06"
9514391,fisher vectors meet neural networks: a hybrid visual classification architecture,"in an image classification method, a feature vector representing an input image is generated by unsupervised operations including extracting local descriptors from patches distributed over the input image, and a classification value for the input image is generated by applying a neural network (nn) to the feature vector. extracting the feature vector may include encoding the local descriptors extracted from each patch using a generative model, such as fisher vector encoding, aggregating the encoded local descriptors to form a vector, projecting the vector into a space of lower dimensionality, for example using principal component analysis (pca), and normalizing the feature vector of lower dimensionality to produce the feature vector representing the input image. a set of mid-level features representing the input image may be generated as the output of an intermediate layer of the nn.",2016-12-06,"The title of the patent is fisher vectors meet neural networks: a hybrid visual classification architecture and its abstract is in an image classification method, a feature vector representing an input image is generated by unsupervised operations including extracting local descriptors from patches distributed over the input image, and a classification value for the input image is generated by applying a neural network (nn) to the feature vector. extracting the feature vector may include encoding the local descriptors extracted from each patch using a generative model, such as fisher vector encoding, aggregating the encoded local descriptors to form a vector, projecting the vector into a space of lower dimensionality, for example using principal component analysis (pca), and normalizing the feature vector of lower dimensionality to produce the feature vector representing the input image. a set of mid-level features representing the input image may be generated as the output of an intermediate layer of the nn. dated 2016-12-06"
9519857,apparatus and method for sensing characterizing features of a deformable structure,"an apparatus includes a deformable structure in which a neural network comprising a plurality of deformation sensors, e.g. nanowire sensors, and distributed in-situ processing circuits. the circuits generate a signal characterizing features of the local deformation of the structure and/or a command signal corresponding to the detected deformation. the structure may be a wearable sleeve that conforms to deformations of a user's skin, part of an electronic device, such as a touch sensitive screen, or an object in itself. the apparatus can provide a user interface, wherein a command corresponding to a current shape of the structure is generated and acted upon by a integrated or remote device, or a device for monitoring a user's position or movement e.g. for replication by a robotic device. the apparatus may have machine learning capability to improve the matching of commands with determined shapes of the deformable structure.",2016-12-13,"The title of the patent is apparatus and method for sensing characterizing features of a deformable structure and its abstract is an apparatus includes a deformable structure in which a neural network comprising a plurality of deformation sensors, e.g. nanowire sensors, and distributed in-situ processing circuits. the circuits generate a signal characterizing features of the local deformation of the structure and/or a command signal corresponding to the detected deformation. the structure may be a wearable sleeve that conforms to deformations of a user's skin, part of an electronic device, such as a touch sensitive screen, or an object in itself. the apparatus can provide a user interface, wherein a command corresponding to a current shape of the structure is generated and acted upon by a integrated or remote device, or a device for monitoring a user's position or movement e.g. for replication by a robotic device. the apparatus may have machine learning capability to improve the matching of commands with determined shapes of the deformable structure. dated 2016-12-13"
9519858,feature-augmented neural networks and applications of same,"a system is described herein which uses a neural network having an input layer that accepts an input vector and a feature vector. the input vector represents at least part of input information, such as, but not limited to, a word or phrase in a sequence of input words. the feature vector provides supplemental information pertaining to the input information. the neural network produces an output vector based on the input vector and the feature vector. in one implementation, the neural network is a recurrent neural network. also described herein are various applications of the system, including a machine translation application.",2016-12-13,"The title of the patent is feature-augmented neural networks and applications of same and its abstract is a system is described herein which uses a neural network having an input layer that accepts an input vector and a feature vector. the input vector represents at least part of input information, such as, but not limited to, a word or phrase in a sequence of input words. the feature vector provides supplemental information pertaining to the input information. the neural network produces an output vector based on the input vector and the feature vector. in one implementation, the neural network is a recurrent neural network. also described herein are various applications of the system, including a machine translation application. dated 2016-12-13"
9520127,shared hidden layer combination for speech recognition systems,providing a framework for merging automatic speech recognition (asr) systems having a shared deep neural network (dnn) feature transformation is provided. a received utterance may be evaluated to generate a dnn-derived feature from the top hidden layer of a dnn. the top hidden layer output may then be utilized to generate a network including a bottleneck layer and an output layer. weights representing a feature dimension reduction may then be extracted between the top hidden layer and the bottleneck layer. scores may then be generated and combined to merge the asr systems which share the dnn feature transformation.,2016-12-13,The title of the patent is shared hidden layer combination for speech recognition systems and its abstract is providing a framework for merging automatic speech recognition (asr) systems having a shared deep neural network (dnn) feature transformation is provided. a received utterance may be evaluated to generate a dnn-derived feature from the top hidden layer of a dnn. the top hidden layer output may then be utilized to generate a network including a bottleneck layer and an output layer. weights representing a feature dimension reduction may then be extracted between the top hidden layer and the bottleneck layer. scores may then be generated and combined to merge the asr systems which share the dnn feature transformation. dated 2016-12-13
9520128,frame skipping with extrapolation and outputs on demand neural network for automatic speech recognition,techniques related to implementing neural networks for speech recognition systems are discussed. such techniques may include implementing frame skipping with approximated skip frames and/or distances on demand such that only those outputs needed by a speech decoder are provided via the neural network or approximation techniques.,2016-12-13,The title of the patent is frame skipping with extrapolation and outputs on demand neural network for automatic speech recognition and its abstract is techniques related to implementing neural networks for speech recognition systems are discussed. such techniques may include implementing frame skipping with approximated skip frames and/or distances on demand such that only those outputs needed by a speech decoder are provided via the neural network or approximation techniques. dated 2016-12-13
9524716,systems and methods for providing unnormalized language models,"some embodiments relate to using an unnormalized neural network language model in connection with a speech processing application. the techniques include obtaining a language segment sequence comprising one or more language segments in a vocabulary; accessing an unnormalized neural network language model having a normalizer node and an output layer comprising a plurality of output nodes, each of the plurality of output nodes associated with a respective language segment in the vocabulary; and determining, using the unnormalized neural network language model, a first likelihood that a first language segment in the vocabulary follows the language segment sequence.",2016-12-20,"The title of the patent is systems and methods for providing unnormalized language models and its abstract is some embodiments relate to using an unnormalized neural network language model in connection with a speech processing application. the techniques include obtaining a language segment sequence comprising one or more language segments in a vocabulary; accessing an unnormalized neural network language model having a normalizer node and an output layer comprising a plurality of output nodes, each of the plurality of output nodes associated with a respective language segment in the vocabulary; and determining, using the unnormalized neural network language model, a first likelihood that a first language segment in the vocabulary follows the language segment sequence. dated 2016-12-20"
9524730,monaural speech filter,"a system receives monaural sound which includes speech and background noises. the received sound is divided by frequency and time into time-frequency units (tfus). each tfu is classified as speech or non-speech by a processing unit. the processing unit for each frequency range includes at least one of a deep neural network (dnn) or a linear support vector machine (lsvm). the dnn extracts and classifies the features of the tfu and includes a pre-trained stack of restricted boltzmann machines (rbm), and each rbm includes a visible and a hidden layer. the lsvm classifies each tfu based on extracted features from the dnn, including those from the visible layer of the first rbm, and those from the hidden layer of the last rbm in the stack. the lsvm and dnn include training with a plurality of training noises. each tfu classified as speech is output.",2016-12-20,"The title of the patent is monaural speech filter and its abstract is a system receives monaural sound which includes speech and background noises. the received sound is divided by frequency and time into time-frequency units (tfus). each tfu is classified as speech or non-speech by a processing unit. the processing unit for each frequency range includes at least one of a deep neural network (dnn) or a linear support vector machine (lsvm). the dnn extracts and classifies the features of the tfu and includes a pre-trained stack of restricted boltzmann machines (rbm), and each rbm includes a visible and a hidden layer. the lsvm classifies each tfu based on extracted features from the dnn, including those from the visible layer of the first rbm, and those from the hidden layer of the last rbm in the stack. the lsvm and dnn include training with a plurality of training noises. each tfu classified as speech is output. dated 2016-12-20"
9526436,amplifiers including tunable tunnel field effect transistor pseudo resistors and related devices,neural signal amplifiers include an operational amplifier and a feedback network coupled between an output and an input thereof. the feedback network includes a tunnel field effect transistor (“tfet”) pseudo resistor that exhibits bi-directional conductivity. a drain region of the tfet may be electrically connected to the gate electrode thereof to provide a bi-directional resistor having good symmetry in terms of resistance as a function of voltage polarity.,2016-12-27,The title of the patent is amplifiers including tunable tunnel field effect transistor pseudo resistors and related devices and its abstract is neural signal amplifiers include an operational amplifier and a feedback network coupled between an output and an input thereof. the feedback network includes a tunnel field effect transistor (“tfet”) pseudo resistor that exhibits bi-directional conductivity. a drain region of the tfet may be electrically connected to the gate electrode thereof to provide a bi-directional resistor having good symmetry in terms of resistance as a function of voltage polarity. dated 2016-12-27
9530042,method for fingerprint classification,"the method for fingerprint classification uses a local gradient directional binary pattern (lgdbp) descriptor. the method acquires digital images of fingerprints from a scanner or the like, and the lgdbp descriptors corresponding to the directional ridge patterns are calculated. using the lgdbp descriptors as a fingerprint representation, an extreme learning machine neural network with a radial basis function kernel may is used to reduce substantially the search space to a predefined number of classes of known fingerprints to be searched to identify the fingerprint.",2016-12-27,"The title of the patent is method for fingerprint classification and its abstract is the method for fingerprint classification uses a local gradient directional binary pattern (lgdbp) descriptor. the method acquires digital images of fingerprints from a scanner or the like, and the lgdbp descriptors corresponding to the directional ridge patterns are calculated. using the lgdbp descriptors as a fingerprint representation, an extreme learning machine neural network with a radial basis function kernel may is used to reduce substantially the search space to a predefined number of classes of known fingerprints to be searched to identify the fingerprint. dated 2016-12-27"
9530047,method and system for face image recognition,"a method for face image recognition is disclosed. the method comprises generating one or more face region pairs of face images to be compared and recognized; forming a plurality of feature modes by exchanging the two face regions of each face region pair and horizontally flipping each face region of each face region pair; receiving, by one or more convolutional neural networks, the plurality of feature modes, each of which forms a plurality of input maps in the convolutional neural network; extracting, by the one or more convolutional neural networks, relational features from the input maps, which reflect identity similarities of the face images; and recognizing whether the compared face images belong to the same identity based on the extracted relational features of the face images. in addition, a system for face image recognition is also disclosed.",2016-12-27,"The title of the patent is method and system for face image recognition and its abstract is a method for face image recognition is disclosed. the method comprises generating one or more face region pairs of face images to be compared and recognized; forming a plurality of feature modes by exchanging the two face regions of each face region pair and horizontally flipping each face region of each face region pair; receiving, by one or more convolutional neural networks, the plurality of feature modes, each of which forms a plurality of input maps in the convolutional neural network; extracting, by the one or more convolutional neural networks, relational features from the input maps, which reflect identity similarities of the face images; and recognizing whether the compared face images belong to the same identity based on the extracted relational features of the face images. in addition, a system for face image recognition is also disclosed. dated 2016-12-27"
9530071,hierarchical interlinked multi-scale convolutional network for image parsing,"a disclosed facial recognition system (and method) includes face parsing. in one approach, the face parsing is based on hierarchical interlinked multiscale convolutional neural network (him) to identify locations and/or footprints of components of a face image. the him generates multiple levels of image patches from different resolution images of the face image, where image patches for different levels have different resolutions. moreover, the him integrates the image patches for different levels to generate interlinked image patches for different levels, where interlinked image patches for different levels have different resolutions. furthermore, the him combines the interlinked image patches to identify refined locations and/or footprints of components.",2016-12-27,"The title of the patent is hierarchical interlinked multi-scale convolutional network for image parsing and its abstract is a disclosed facial recognition system (and method) includes face parsing. in one approach, the face parsing is based on hierarchical interlinked multiscale convolutional neural network (him) to identify locations and/or footprints of components of a face image. the him generates multiple levels of image patches from different resolution images of the face image, where image patches for different levels have different resolutions. moreover, the him integrates the image patches for different levels to generate interlinked image patches for different levels, where interlinked image patches for different levels have different resolutions. furthermore, the him combines the interlinked image patches to identify refined locations and/or footprints of components. dated 2016-12-27"
9530092,haptic-based artificial neural network training,"in a method for training an artificial neural network based algorithm designed to monitor a first device, a processor receives a first data. a processor determines a first service action recommendation for a first device using the received first data and an artificial neural network (ann) algorithm. a processor causes a second device to provide haptic feedback using the received first data. a processor receives a second service action recommendation for the first device based on the haptic feedback. a processor adjusts at least one parameter of the ann algorithm such that the ann algorithm determines a third service action recommendation for the first device using the received first data, wherein the third service action recommendation is equivalent to the second service action recommendation.",2016-12-27,"The title of the patent is haptic-based artificial neural network training and its abstract is in a method for training an artificial neural network based algorithm designed to monitor a first device, a processor receives a first data. a processor determines a first service action recommendation for a first device using the received first data and an artificial neural network (ann) algorithm. a processor causes a second device to provide haptic feedback using the received first data. a processor receives a second service action recommendation for the first device based on the haptic feedback. a processor adjusts at least one parameter of the ann algorithm such that the ann algorithm determines a third service action recommendation for the first device using the received first data, wherein the third service action recommendation is equivalent to the second service action recommendation. dated 2016-12-27"
9530400,system and method for compressed domain language identification,"embodiments included herein are directed towards a system and method for compressed domain language identification. embodiments may include receiving a bitstream of a sequence of packets at one or more computing devices and classifying each packet into speech or non-speech based upon, at least in part, compressed domain voice activity detection (vad). embodiments may further include extracting a pseudo-cepstral representation from the speech detected packets and partially decoding without extracting a pcm format and generating a sequence of multi-frames, based upon, at least in part, the pseudo-cepstral representation. embodiments may also include providing in real time the sequence of multi-frames to a deep neural network (dnn), wherein the dnn has been trained off-line for one or more desired target languages.",2016-12-27,"The title of the patent is system and method for compressed domain language identification and its abstract is embodiments included herein are directed towards a system and method for compressed domain language identification. embodiments may include receiving a bitstream of a sequence of packets at one or more computing devices and classifying each packet into speech or non-speech based upon, at least in part, compressed domain voice activity detection (vad). embodiments may further include extracting a pseudo-cepstral representation from the speech detected packets and partially decoding without extracting a pcm format and generating a sequence of multi-frames, based upon, at least in part, the pseudo-cepstral representation. embodiments may also include providing in real time the sequence of multi-frames to a deep neural network (dnn), wherein the dnn has been trained off-line for one or more desired target languages. dated 2016-12-27"
9533115,system and method for diagnosis and treatment of a breathing pattern of a patient,"described is a system including a sensor and a processing arrangement. the sensor measures data corresponding to a patient's breathing patterns. the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of a rem sleep state. in another embodiment, the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of one of the following states: (i) a wake state and (ii) a sleep state. in another embodiment, a neural network analyzes the data to determine whether the breathing patterns are indicative of one of the following states: (i) a rem sleep state, (ii) a wake state and (iii) a sleep state. in another embodiment, the processing arrangement analyzes the data to determine whether the breathing pattern is indicative of an arousal.",2017-01-03,"The title of the patent is system and method for diagnosis and treatment of a breathing pattern of a patient and its abstract is described is a system including a sensor and a processing arrangement. the sensor measures data corresponding to a patient's breathing patterns. the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of a rem sleep state. in another embodiment, the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of one of the following states: (i) a wake state and (ii) a sleep state. in another embodiment, a neural network analyzes the data to determine whether the breathing patterns are indicative of one of the following states: (i) a rem sleep state, (ii) a wake state and (iii) a sleep state. in another embodiment, the processing arrangement analyzes the data to determine whether the breathing pattern is indicative of an arousal. dated 2017-01-03"
9535897,content recommendation system using a neural network language model,"the present disclosure relates to applying techniques similar to those used in neural network language modeling systems to a content recommendation system. for example, by associating consumed media content to words of a language model, the system may provide content predictions based on an ordering. thus, the systems and techniques described herein may produce enhanced prediction results for recommending content (e.g. word) in a given sequence of consumed content. in addition, the system may account for additional user actions by representing particular actions as punctuation in the language model.",2017-01-03,"The title of the patent is content recommendation system using a neural network language model and its abstract is the present disclosure relates to applying techniques similar to those used in neural network language modeling systems to a content recommendation system. for example, by associating consumed media content to words of a language model, the system may provide content predictions based on an ordering. thus, the systems and techniques described herein may produce enhanced prediction results for recommending content (e.g. word) in a given sequence of consumed content. in addition, the system may account for additional user actions by representing particular actions as punctuation in the language model. dated 2017-01-03"
9535960,context-sensitive search using a deep learning model,"a search engine is described herein for providing search results based on a context in which a query has been submitted, as expressed by context information. the search engine operates by ranking a plurality of documents based on a consideration of the query, and based, in part, on a context concept vector and a plurality of document concept vectors, both generated using a deep learning model (such as a deep neural network). the context concept vector is formed by a projection of the context information into a semantic space using the deep learning model. each document concept vector is formed by a projection of document information, associated with a particular document, into the same semantic space using the deep learning model. the ranking operates by favoring documents that are relevant to the context within the semantic space, and disfavoring documents that are not relevant to the context.",2017-01-03,"The title of the patent is context-sensitive search using a deep learning model and its abstract is a search engine is described herein for providing search results based on a context in which a query has been submitted, as expressed by context information. the search engine operates by ranking a plurality of documents based on a consideration of the query, and based, in part, on a context concept vector and a plurality of document concept vectors, both generated using a deep learning model (such as a deep neural network). the context concept vector is formed by a projection of the context information into a semantic space using the deep learning model. each document concept vector is formed by a projection of document information, associated with a particular document, into the same semantic space using the deep learning model. the ranking operates by favoring documents that are relevant to the context within the semantic space, and disfavoring documents that are not relevant to the context. dated 2017-01-03"
9536189,phase-coding for coordinate transformation,a method for coordinate transformation in a spiking neural network includes encoding a first positional representation as phase information in the spiking neural network. the method also includes shifting the phase information to modify the first positional representation into a second positional representation.,2017-01-03,The title of the patent is phase-coding for coordinate transformation and its abstract is a method for coordinate transformation in a spiking neural network includes encoding a first positional representation as phase information in the spiking neural network. the method also includes shifting the phase information to modify the first positional representation into a second positional representation. dated 2017-01-03
9536190,dynamically assigning and examining synaptic delay,a method for dynamically modifying synaptic delays in a neural network includes initializing a delay parameter and operating the neural network. the method further includes dynamically updating the delay parameter based on a program which is based on a statement including the delay parameter.,2017-01-03,The title of the patent is dynamically assigning and examining synaptic delay and its abstract is a method for dynamically modifying synaptic delays in a neural network includes initializing a delay parameter and operating the neural network. the method further includes dynamically updating the delay parameter based on a program which is based on a statement including the delay parameter. dated 2017-01-03
9536293,image assessment using deep convolutional neural networks,"deep convolutional neural networks receive local and global representations of images as inputs and learn the best representation for a particular feature through multiple convolutional and fully connected layers. a double-column neural network structure receives each of the local and global representations as two heterogeneous parallel inputs to the two columns. after some layers of transformations, the two columns are merged to form the final classifier. additionally, features may be learned in one of the fully connected layers. the features of the images may be leveraged to boost classification accuracy of other features by learning a regularized double-column neural network.",2017-01-03,"The title of the patent is image assessment using deep convolutional neural networks and its abstract is deep convolutional neural networks receive local and global representations of images as inputs and learn the best representation for a particular feature through multiple convolutional and fully connected layers. a double-column neural network structure receives each of the local and global representations as two heterogeneous parallel inputs to the two columns. after some layers of transformations, the two columns are merged to form the final classifier. additionally, features may be learned in one of the fully connected layers. the features of the images may be leveraged to boost classification accuracy of other features by learning a regularized double-column neural network. dated 2017-01-03"
9538925,method and system for machine learning based assessment of fractional flow reserve,"a method and system for determining fractional flow reserve (ffr) for a coronary artery stenosis of a patient is disclosed. in one embodiment, medical image data of the patient including the stenosis is received, a set of features for the stenosis is extracted from the medical image data of the patient, and an ffr value for the stenosis is determined based on the extracted set of features using a trained machine-learning based mapping. in another embodiment, a medical image of the patient including the stenosis of interest is received, image patches corresponding to the stenosis of interest and a coronary tree of the patient are detected, an ffr value for the stenosis of interest is determined using a trained deep neural network regressor applied directly to the detected image patches.",2017-01-10,"The title of the patent is method and system for machine learning based assessment of fractional flow reserve and its abstract is a method and system for determining fractional flow reserve (ffr) for a coronary artery stenosis of a patient is disclosed. in one embodiment, medical image data of the patient including the stenosis is received, a set of features for the stenosis is extracted from the medical image data of the patient, and an ffr value for the stenosis is determined based on the extracted set of features using a trained machine-learning based mapping. in another embodiment, a medical image of the patient including the stenosis of interest is received, image patches corresponding to the stenosis of interest and a coronary tree of the patient are detected, an ffr value for the stenosis of interest is determined using a trained deep neural network regressor applied directly to the detected image patches. dated 2017-01-10"
9542626,augmenting layer-based object detection with deep convolutional neural networks,"by way of example, the technology disclosed by this document receives image data; extracts a depth image and a color image from the image data; creates a mask image by segmenting the depth image; determines a first likelihood score from the depth image and the mask image using a layered classifier; determines a second likelihood score from the color image and the mask image using a deep convolutional neural network; and determines a class of at least a portion of the image data based on the first likelihood score and the second likelihood score. further, the technology can pre-filter the mask image using the layered classifier and then use the pre-filtered mask image and the color image to calculate a second likelihood score using the deep convolutional neural network to speed up processing.",2017-01-10,"The title of the patent is augmenting layer-based object detection with deep convolutional neural networks and its abstract is by way of example, the technology disclosed by this document receives image data; extracts a depth image and a color image from the image data; creates a mask image by segmenting the depth image; determines a first likelihood score from the depth image and the mask image using a layered classifier; determines a second likelihood score from the color image and the mask image using a deep convolutional neural network; and determines a class of at least a portion of the image data based on the first likelihood score and the second likelihood score. further, the technology can pre-filter the mask image using the layered classifier and then use the pre-filtered mask image and the color image to calculate a second likelihood score using the deep convolutional neural network to speed up processing. dated 2017-01-10"
9542642,packet data neural network system and method,"this application discloses a neural network that also functions as a connection oriented packet data network using an mpls-type label switching technology. the neural network uses its intelligence to build and manage label switched paths (lsps) to transport user packets and solve complex mathematical problems. however, the methods taught here can be applied to other data networks including ad-hoc, mobile, and traditional packet networks, cell or frame-switched networks, time-slot networks and the like.",2017-01-10,"The title of the patent is packet data neural network system and method and its abstract is this application discloses a neural network that also functions as a connection oriented packet data network using an mpls-type label switching technology. the neural network uses its intelligence to build and manage label switched paths (lsps) to transport user packets and solve complex mathematical problems. however, the methods taught here can be applied to other data networks including ad-hoc, mobile, and traditional packet networks, cell or frame-switched networks, time-slot networks and the like. dated 2017-01-10"
9542645,plastic synapse management,a method for managing synapse plasticity in a neural network includes converting a first set of synapses from a plastic synapse type to a fixed synapse type. the method may also include converting a second set of synapses from the fixed synapse type to the plastic synapse type.,2017-01-10,The title of the patent is plastic synapse management and its abstract is a method for managing synapse plasticity in a neural network includes converting a first set of synapses from a plastic synapse type to a fixed synapse type. the method may also include converting a second set of synapses from the fixed synapse type to the plastic synapse type. dated 2017-01-10
9542948,text-dependent speaker identification,"methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for speaker verification. the methods, systems, and apparatus include actions of inputting speech data that corresponds to a particular utterance to a first neural network and determining an evaluation vector based on output at a hidden layer of the first neural network. additional actions include obtaining a reference vector that corresponds to a past utterance of a particular speaker. further actions include inputting the evaluation vector and the reference vector to a second neural network that is trained on a set of labeled pairs of feature vectors to identify whether speakers associated with the labeled pairs of feature vectors are the same speaker. more actions include determining, based on an output of the second neural network, whether the particular utterance was likely spoken by the particular speaker.",2017-01-10,"The title of the patent is text-dependent speaker identification and its abstract is methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for speaker verification. the methods, systems, and apparatus include actions of inputting speech data that corresponds to a particular utterance to a first neural network and determining an evaluation vector based on output at a hidden layer of the first neural network. additional actions include obtaining a reference vector that corresponds to a past utterance of a particular speaker. further actions include inputting the evaluation vector and the reference vector to a second neural network that is trained on a set of labeled pairs of feature vectors to identify whether speakers associated with the labeled pairs of feature vectors are the same speaker. more actions include determining, based on an output of the second neural network, whether the particular utterance was likely spoken by the particular speaker. dated 2017-01-10"
9547316,thermostat classification method and system,"the disclosure provides a computer-implemented method and system of reducing commodity usage by providing tailored consumer information to the consumer. the method utilizes neural network and machine learning techniques to calculate and cluster statistical data to classify the premises for desired observable condition, including the presence of a programmed thermostat. a score is determined that corresponds to at least one of: (i) a present state of an observable condition, (ii) a non-present state of the observable condition, and (iii) a degree of a condition of the observable condition, to provide tailored consumer information associated to the consumer's usage of the commodity.",2017-01-17,"The title of the patent is thermostat classification method and system and its abstract is the disclosure provides a computer-implemented method and system of reducing commodity usage by providing tailored consumer information to the consumer. the method utilizes neural network and machine learning techniques to calculate and cluster statistical data to classify the premises for desired observable condition, including the presence of a programmed thermostat. a score is determined that corresponds to at least one of: (i) a present state of an observable condition, (ii) a non-present state of the observable condition, and (iii) a degree of a condition of the observable condition, to provide tailored consumer information associated to the consumer's usage of the commodity. dated 2017-01-17"
9547820,method of classifying input pattern and pattern classification apparatus,"a method of classifying an input pattern and a pattern classification apparatus are provided. the method includes enabling an artificial neural network to learn based on learning input data received by an input layer of the artificial neural network, determining classification of an input pattern received by the input layer of the enabled artificial neural network according to an output value obtained from an output layer of the artificial neural network, the obtained output value being based on the input pattern, updating connection intensities of a plurality of connection lines of the enabled artificial neural network to output a result value indicating the determined classification from the output layer when the input pattern, and determining updated classification of the input pattern according to an updated output value obtained from an output layer of the updated artificial neural network, the obtained updated output value being based on the input pattern.",2017-01-17,"The title of the patent is method of classifying input pattern and pattern classification apparatus and its abstract is a method of classifying an input pattern and a pattern classification apparatus are provided. the method includes enabling an artificial neural network to learn based on learning input data received by an input layer of the artificial neural network, determining classification of an input pattern received by the input layer of the enabled artificial neural network according to an output value obtained from an output layer of the artificial neural network, the obtained output value being based on the input pattern, updating connection intensities of a plurality of connection lines of the enabled artificial neural network to output a result value indicating the determined classification from the output layer when the input pattern, and determining updated classification of the input pattern according to an updated output value obtained from an output layer of the updated artificial neural network, the obtained updated output value being based on the input pattern. dated 2017-01-17"
9547821,deep learning for algorithm portfolios,"automated feature construction for algorithm portfolios in machine learning is provided. a gray scale image is generated from a text representing a problem instance. the gray scale image is rescaled or reshaped to a predefined size that is smaller than an initial size of the gray scale image. the rescaled gray scale image represents features of the problem instance. the rescaled gray scale image is input as features to a machine learning-based convolutional neural network. based on the rescaled gray scale image, the machine learning-based convolutional neural network is automatically trained to learn to automatically determine one or more problem solvers from a portfolio of problem solvers suited for solving the problem instance.",2017-01-17,"The title of the patent is deep learning for algorithm portfolios and its abstract is automated feature construction for algorithm portfolios in machine learning is provided. a gray scale image is generated from a text representing a problem instance. the gray scale image is rescaled or reshaped to a predefined size that is smaller than an initial size of the gray scale image. the rescaled gray scale image represents features of the problem instance. the rescaled gray scale image is input as features to a machine learning-based convolutional neural network. based on the rescaled gray scale image, the machine learning-based convolutional neural network is automatically trained to learn to automatically determine one or more problem solvers from a portfolio of problem solvers suited for solving the problem instance. dated 2017-01-17"
9552547,normalizing electronic communications using a neural-network normalizer and a neural-network flagger,"electronic communications can be normalized using neural networks. for example, an electronic representation of a noncanonical communication can be received. a normalized version of the noncanonical communication can be determined using a normalizer including a neural network. the neural network can receive a single vector at an input layer of the neural network and transform an output of a hidden layer of the neural network into multiple values that sum to a total value of one. each value of the multiple values can be a number between zero and one and represent a probability of a particular character being in a particular position in the normalized version of the noncanonical communication. the neural network can determine the normalized version of the noncanonical communication based on the multiple values. whether the normalized version should be output can be determined based on a result from a flagger including another neural network.",2017-01-24,"The title of the patent is normalizing electronic communications using a neural-network normalizer and a neural-network flagger and its abstract is electronic communications can be normalized using neural networks. for example, an electronic representation of a noncanonical communication can be received. a normalized version of the noncanonical communication can be determined using a normalizer including a neural network. the neural network can receive a single vector at an input layer of the neural network and transform an output of a hidden layer of the neural network into multiple values that sum to a total value of one. each value of the multiple values can be a number between zero and one and represent a probability of a particular character being in a particular position in the normalized version of the noncanonical communication. the neural network can determine the normalized version of the noncanonical communication based on the multiple values. whether the normalized version should be output can be determined based on a result from a flagger including another neural network. dated 2017-01-24"
9552549,ranking approach to train deep neural nets for multilabel image annotation,"systems and techniques are provided for a ranking approach to train deep neural nets for multilabel image annotation. label scores may be received for labels determined by a neural network for training examples. each label may be a positive label or a negative label for the training example. an error of the neural network may be determined based on a comparison, for each of the training examples, of the label scores for positive labels and negative labels for the training example and a semantic distance between each positive label and each negative label for the training example. updated weights may be determined for the neural network based on a gradient of the determined error of the neural network. the updated weights may be applied to the neural network to train the neural network.",2017-01-24,"The title of the patent is ranking approach to train deep neural nets for multilabel image annotation and its abstract is systems and techniques are provided for a ranking approach to train deep neural nets for multilabel image annotation. label scores may be received for labels determined by a neural network for training examples. each label may be a positive label or a negative label for the training example. an error of the neural network may be determined based on a comparison, for each of the training examples, of the label scores for positive labels and negative labels for the training example and a semantic distance between each positive label and each negative label for the training example. updated weights may be determined for the neural network based on a gradient of the determined error of the neural network. the updated weights may be applied to the neural network to train the neural network. dated 2017-01-24"
9553741,adaptive demodulation method and apparatus using an artificial neural network to improve data recovery in high speed channels,a neural network demodulator is used within a receiver to provide inter symbol interference (isi) channel equalization and to correct for i/q/phase imbalance. the neural network is trained with a single integrated training step to simultaneously handle the channel impairments of isi equalization and i/q phase imbalance as opposed to prior art methods of separately addressing each channel impairment in sequence.,2017-01-24,The title of the patent is adaptive demodulation method and apparatus using an artificial neural network to improve data recovery in high speed channels and its abstract is a neural network demodulator is used within a receiver to provide inter symbol interference (isi) channel equalization and to correct for i/q/phase imbalance. the neural network is trained with a single integrated training step to simultaneously handle the channel impairments of isi equalization and i/q phase imbalance as opposed to prior art methods of separately addressing each channel impairment in sequence. dated 2017-01-24
9558442,monitoring neural networks with shadow networks,a method for generating an event includes monitoring a first neural network with a second neural network. the method also includes generating an event based on the monitoring. the event is generated at the second neural network. the event may be generated based on a spike received at the second network during the monitoring.,2017-01-31,The title of the patent is monitoring neural networks with shadow networks and its abstract is a method for generating an event includes monitoring a first neural network with a second neural network. the method also includes generating an event based on the monitoring. the event is generated at the second neural network. the event may be generated based on a spike received at the second network during the monitoring. dated 2017-01-31
9558742,mixed speech recognition,"the claimed subject matter includes a system and method for recognizing mixed speech from a source. the method includes training a first neural network to recognize the speech signal spoken by the speaker with a higher level of a speech characteristic from a mixed speech sample. the method also includes training a second neural network to recognize the speech signal spoken by the speaker with a lower level of the speech characteristic from the mixed speech sample. additionally, the method includes decoding the mixed speech sample with the first neural network and the second neural network by optimizing the joint likelihood of observing the two speech signals considering the probability that a specific frame is a switching point of the speech characteristic.",2017-01-31,"The title of the patent is mixed speech recognition and its abstract is the claimed subject matter includes a system and method for recognizing mixed speech from a source. the method includes training a first neural network to recognize the speech signal spoken by the speaker with a higher level of a speech characteristic from a mixed speech sample. the method also includes training a second neural network to recognize the speech signal spoken by the speaker with a lower level of the speech characteristic from the mixed speech sample. additionally, the method includes decoding the mixed speech sample with the first neural network and the second neural network by optimizing the joint likelihood of observing the two speech signals considering the probability that a specific frame is a switching point of the speech characteristic. dated 2017-01-31"
9558743,integration of semantic context information,"in one implementation, a computer-implemented method includes receiving, at a computer system, a request to predict a next word in a dialog being uttered by a speaker; accessing, by the computer system, a neural network comprising i) an input layer, ii) one or more hidden layers, and iii) an output layer; identifying the local context for the dialog of the speaker; selecting, by the computer system and using a semantic model, at least one vector that represents the semantic context for the dialog; applying input to the input layer of the neural network, the input comprising i) the local context of the dialog and ii) the values for the at least one vector; generating probability values for at least a portion of the candidate words; and providing, by the computer system and based on the probability values, information that identifies one or more of the candidate words.",2017-01-31,"The title of the patent is integration of semantic context information and its abstract is in one implementation, a computer-implemented method includes receiving, at a computer system, a request to predict a next word in a dialog being uttered by a speaker; accessing, by the computer system, a neural network comprising i) an input layer, ii) one or more hidden layers, and iii) an output layer; identifying the local context for the dialog of the speaker; selecting, by the computer system and using a semantic model, at least one vector that represents the semantic context for the dialog; applying input to the input layer of the neural network, the input comprising i) the local context of the dialog and ii) the values for the at least one vector; generating probability values for at least a portion of the candidate words; and providing, by the computer system and based on the probability values, information that identifies one or more of the candidate words. dated 2017-01-31"
9563825,convolutional neural network using a binarized convolution layer,"a convolutional neural network is trained to analyze input data in various different manners. the convolutional neural network includes multiple layers, one of which is a convolution layer that performs a convolution, for each of one or more filters in the convolution layer, of the filter over the input data. the convolution includes generation of an inner product based on the filter and the input data. both the filter of the convolution layer and the input data are binarized, allowing the inner product to be computed using particular operations that are typically faster than multiplication of floating point values. the possible results for the convolution layer can optionally be pre-computed and stored in a look-up table. thus, during operation of the convolutional neural network, rather than performing the convolution on the input data, the pre-computed result can be obtained from the look-up table.",2017-02-07,"The title of the patent is convolutional neural network using a binarized convolution layer and its abstract is a convolutional neural network is trained to analyze input data in various different manners. the convolutional neural network includes multiple layers, one of which is a convolution layer that performs a convolution, for each of one or more filters in the convolution layer, of the filter over the input data. the convolution includes generation of an inner product based on the filter and the input data. both the filter of the convolution layer and the input data are binarized, allowing the inner product to be computed using particular operations that are typically faster than multiplication of floating point values. the possible results for the convolution layer can optionally be pre-computed and stored in a look-up table. thus, during operation of the convolutional neural network, rather than performing the convolution on the input data, the pre-computed result can be obtained from the look-up table. dated 2017-02-07"
9563839,artificial neural network system,"a non-biological asynchronous neural network system comprising multiple neurons to receive respective input signals representing an input stimulus for the network, supply an output signal representing a spatio-temporal sequence of rhythmic electric pulses to an external system, wherein respective ones of the multiple neurons are connected using multiple mutually inhibitory links.",2017-02-07,"The title of the patent is artificial neural network system and its abstract is a non-biological asynchronous neural network system comprising multiple neurons to receive respective input signals representing an input stimulus for the network, supply an output signal representing a spatio-temporal sequence of rhythmic electric pulses to an external system, wherein respective ones of the multiple neurons are connected using multiple mutually inhibitory links. dated 2017-02-07"
9563840,system and method for parallelizing convolutional neural networks,a parallel convolutional neural network is provided. the cnn is implemented by a plurality of convolutional neural networks each on a respective processing node. each cnn has a plurality of layers. a subset of the layers are interconnected between processing nodes such that activations are fed forward across nodes. the remaining subset is not so interconnected.,2017-02-07,The title of the patent is system and method for parallelizing convolutional neural networks and its abstract is a parallel convolutional neural network is provided. the cnn is implemented by a plurality of convolutional neural networks each on a respective processing node. each cnn has a plurality of layers. a subset of the layers are interconnected between processing nodes such that activations are fed forward across nodes. the remaining subset is not so interconnected. dated 2017-02-07
9563841,globally asynchronous and locally synchronous (gals) neuromorphic network,"embodiments of the invention relate to a globally asynchronous and locally synchronous neuromorphic network. one embodiment comprises generating a synchronization signal that is distributed to a plurality of neural core circuits. in response to the synchronization signal, in at least one core circuit, incoming spike events maintained by said at least one core circuit are processed to generate an outgoing spike event. spike events are asynchronously communicated between the core circuits via a routing fabric comprising multiple asynchronous routers.",2017-02-07,"The title of the patent is globally asynchronous and locally synchronous (gals) neuromorphic network and its abstract is embodiments of the invention relate to a globally asynchronous and locally synchronous neuromorphic network. one embodiment comprises generating a synchronization signal that is distributed to a plurality of neural core circuits. in response to the synchronization signal, in at least one core circuit, incoming spike events maintained by said at least one core circuit are processed to generate an outgoing spike event. spike events are asynchronously communicated between the core circuits via a routing fabric comprising multiple asynchronous routers. dated 2017-02-07"
9563842,structural plasticity in spiking neural networks with symmetric dual of an electronic neuron,"a neural system comprises multiple neurons interconnected via synapse devices. each neuron integrates input signals arriving on its dendrite, generates a spike in response to the integrated input signals exceeding a threshold, and sends the spike to the interconnected neurons via its axon. the system further includes multiple noruens, each noruen is interconnected via the interconnect network with those neurons that the noruen's corresponding neuron sends its axon to. each noruen integrates input spikes from connected spiking neurons and generates a spike in response to the integrated input spikes exceeding a threshold. there can be one noruen for every corresponding neuron. for a first neuron connected via its axon via a synapse to dendrite of a second neuron, a noruen corresponding to the second neuron is connected via its axon through the same synapse to dendrite of the noruen corresponding to the first neuron.",2017-02-07,"The title of the patent is structural plasticity in spiking neural networks with symmetric dual of an electronic neuron and its abstract is a neural system comprises multiple neurons interconnected via synapse devices. each neuron integrates input signals arriving on its dendrite, generates a spike in response to the integrated input signals exceeding a threshold, and sends the spike to the interconnected neurons via its axon. the system further includes multiple noruens, each noruen is interconnected via the interconnect network with those neurons that the noruen's corresponding neuron sends its axon to. each noruen integrates input spikes from connected spiking neurons and generates a spike in response to the integrated input spikes exceeding a threshold. there can be one noruen for every corresponding neuron. for a first neuron connected via its axon via a synapse to dendrite of a second neuron, a noruen corresponding to the second neuron is connected via its axon through the same synapse to dendrite of the noruen corresponding to the first neuron. dated 2017-02-07"
9569722,optimal persistence of a business process,"aspects of the invention provide for automatically selecting optimal fetch settings for business processes as a function of database query load and relational context by monitoring usage of a data retrieval point with respect to a defined unit of work. a multilayer feed-forward neural network is used to predict, as a function of training sets composed of historical data generated by the monitored usage of the data retrieval point, a future value of a data size of results from an eager fetch setting for the data retrieval point. the eager fetch is automatically revised to a lazy fetch setting in response to determining that the future data size value of the eager fetch setting results is larger than a permissible memory resource threshold.",2017-02-14,"The title of the patent is optimal persistence of a business process and its abstract is aspects of the invention provide for automatically selecting optimal fetch settings for business processes as a function of database query load and relational context by monitoring usage of a data retrieval point with respect to a defined unit of work. a multilayer feed-forward neural network is used to predict, as a function of training sets composed of historical data generated by the monitored usage of the data retrieval point, a future value of a data size of results from an eager fetch setting for the data retrieval point. the eager fetch is automatically revised to a lazy fetch setting in response to determining that the future data size value of the eager fetch setting results is larger than a permissible memory resource threshold. dated 2017-02-14"
9570069,sectioned memory networks for online word-spotting in continuous speech,"systems, methods, and computer program products to detect a keyword in speech, by generating, from a sequence of spectral feature vectors generated from the speech, a plurality of blocked feature vector sequences, and analyzing, by a neural network, each of the plurality of blocked feature vector sequences to detect the presence of the keyword in the speech.",2017-02-14,"The title of the patent is sectioned memory networks for online word-spotting in continuous speech and its abstract is systems, methods, and computer program products to detect a keyword in speech, by generating, from a sequence of spectral feature vectors generated from the speech, a plurality of blocked feature vector sequences, and analyzing, by a neural network, each of the plurality of blocked feature vector sequences to detect the presence of the keyword in the speech. dated 2017-02-14"
9582753,neural networks for transforming signals,"a method for transforms input signals, by first defining a model for transforming the input signals, wherein the model is specified by constraints and a set of model parameters. an iterative inference procedure is derived from the model and the set of model parameters and unfolded into a set of layers, wherein there is one layer for each iteration of the procedure, and wherein a same set of network parameters is used by all layers. a neural network is formed by untying the set of network parameters such that there is one set of network parameters for each layer and each set of network parameters is separately maintainable and separately applicable to the corresponding layer. the neural network is trined to obtain a trained neural network, and then input signals are transformed using the trained neural network to obtain output signals.",2017-02-28,"The title of the patent is neural networks for transforming signals and its abstract is a method for transforms input signals, by first defining a model for transforming the input signals, wherein the model is specified by constraints and a set of model parameters. an iterative inference procedure is derived from the model and the set of model parameters and unfolded into a set of layers, wherein there is one layer for each iteration of the procedure, and wherein a same set of network parameters is used by all layers. a neural network is formed by untying the set of network parameters such that there is one set of network parameters for each layer and each set of network parameters is separately maintainable and separately applicable to the corresponding layer. the neural network is trined to obtain a trained neural network, and then input signals are transformed using the trained neural network to obtain output signals. dated 2017-02-28"
9582762,"devices, systems, and methods for learning and using artificially intelligent interactive memories","aspects of the disclosure generally relate to computing devices and may be generally directed to devices, systems, methods, and/or applications for learning conversations among two or more conversation participants, storing this knowledge in a knowledgebase (i.e. neural network, graph, sequences, etc.), and enabling a user to simulate a conversation with an artificially intelligent conversation participant.",2017-02-28,"The title of the patent is devices, systems, and methods for learning and using artificially intelligent interactive memories and its abstract is aspects of the disclosure generally relate to computing devices and may be generally directed to devices, systems, methods, and/or applications for learning conversations among two or more conversation participants, storing this knowledge in a knowledgebase (i.e. neural network, graph, sequences, etc.), and enabling a user to simulate a conversation with an artificially intelligent conversation participant. dated 2017-02-28"
9588580,system and method for single domain and multi-domain decision aid for product on the web,"a system and method for problem solving in multiple domains on the web is provided. two facets of preference are applied regardless of domain: first, criteria selected by the user which indicates which elements relate to the user, and second, level of importance to the user. for each decision aid that the user saves to his/her member account, a series of methods applied thereto assist the user in making decisions through intelligent agent expertise, as well as through related ecommerce, social networking, guided content search and delivery of context-rich content. relevancy of results is also calculated. depending on characteristics inherent in a particular domain, one of two primary methods is employed. the multi-product method uses ontology and a neural network engine to reveal the subset of relevant results based on any combination of user inputs, implicitly and explicitly derived. the single-product method maps inputs to results using sub-category analysis of fit and then applies user-centric filters and discounting rules to return meaningful coaching and relevancy of results.",2017-03-07,"The title of the patent is system and method for single domain and multi-domain decision aid for product on the web and its abstract is a system and method for problem solving in multiple domains on the web is provided. two facets of preference are applied regardless of domain: first, criteria selected by the user which indicates which elements relate to the user, and second, level of importance to the user. for each decision aid that the user saves to his/her member account, a series of methods applied thereto assist the user in making decisions through intelligent agent expertise, as well as through related ecommerce, social networking, guided content search and delivery of context-rich content. relevancy of results is also calculated. depending on characteristics inherent in a particular domain, one of two primary methods is employed. the multi-product method uses ontology and a neural network engine to reveal the subset of relevant results based on any combination of user inputs, implicitly and explicitly derived. the single-product method maps inputs to results using sub-category analysis of fit and then applies user-centric filters and discounting rules to return meaningful coaching and relevancy of results. dated 2017-03-07"
9589210,broad area geospatial object detection using autogenerated deep learning models,"a system for automated geospatial image analysis comprising a deep learning model module and a convolutional neural network serving as an automated image analysis software module. the deep learning module receives a plurality of orthorectified geospatial images, pre-labeled to demarcate objects of interest, and optimized for the purpose of training the neural network of the image analysis software module. the module presents marked geospatial images and a second set of unmarked, optimized, training geospatial images to the convolutional neural network. this process may be repeated so that an image analysis software module can detect multiple object types or categories. the image analysis software module receives a plurality of orthorectified geospatial images from one or more geospatial image caches. using multi-scale sliding window submodule, image analysis modules scan geospatial images, detect objects present and locate them on the geographical latitude-longitude system. the system reports the results in the requestor's preferred format.",2017-03-07,"The title of the patent is broad area geospatial object detection using autogenerated deep learning models and its abstract is a system for automated geospatial image analysis comprising a deep learning model module and a convolutional neural network serving as an automated image analysis software module. the deep learning module receives a plurality of orthorectified geospatial images, pre-labeled to demarcate objects of interest, and optimized for the purpose of training the neural network of the image analysis software module. the module presents marked geospatial images and a second set of unmarked, optimized, training geospatial images to the convolutional neural network. this process may be repeated so that an image analysis software module can detect multiple object types or categories. the image analysis software module receives a plurality of orthorectified geospatial images from one or more geospatial image caches. using multi-scale sliding window submodule, image analysis modules scan geospatial images, detect objects present and locate them on the geographical latitude-longitude system. the system reports the results in the requestor's preferred format. dated 2017-03-07"
9589237,"systems, methods and computer products for recommending media suitable for a designated activity","media content is recommended based on suitability for a designated activity. a vector engine is trained using a plurality of lists, each of the lists containing metadata associated with a plurality of media objects. the vector engine includes a neural network trained with corpus data including (i) the plurality of lists (ii) a plurality of titles, each one of the titles associated with one of the lists, and (iii) the metadata associated with the plurality of media objects. training the vector engine involves initializing, using the vector engine, a plurality of feature vectors representing each of the lists, each of the media objects, and each of a plurality of words in the titles of the lists. the training then further involves nudging, using the vector engine, the feature vectors based on a plurality of co-occurrences of the lists, the media objects, the words in the titles of the lists, or a combination thereof. a feature vector corresponding to an activity is identified among the feature vectors. at least one of the media objects, (ii) at least one of lists or (iii) a combination thereof suitable for the activity is selected based on cosine similarities between the feature vector corresponding to the activity and others of the feature vectors.",2017-03-07,"The title of the patent is systems, methods and computer products for recommending media suitable for a designated activity and its abstract is media content is recommended based on suitability for a designated activity. a vector engine is trained using a plurality of lists, each of the lists containing metadata associated with a plurality of media objects. the vector engine includes a neural network trained with corpus data including (i) the plurality of lists (ii) a plurality of titles, each one of the titles associated with one of the lists, and (iii) the metadata associated with the plurality of media objects. training the vector engine involves initializing, using the vector engine, a plurality of feature vectors representing each of the lists, each of the media objects, and each of a plurality of words in the titles of the lists. the training then further involves nudging, using the vector engine, the feature vectors based on a plurality of co-occurrences of the lists, the media objects, the words in the titles of the lists, or a combination thereof. a feature vector corresponding to an activity is identified among the feature vectors. at least one of the media objects, (ii) at least one of lists or (iii) a combination thereof suitable for the activity is selected based on cosine similarities between the feature vector corresponding to the activity and others of the feature vectors. dated 2017-03-07"
9594984,business discovery from imagery,"aspects of the present disclosure relate to a method includes training a deep neural network using training images and data identifying one or more business storefront locations in the training images. the deep neural network outputs tight bounding boxes on each image. at the deep neural network, a first image may be received. the first image may be evaluated using the deep neural network. bounding boxes may then be generated identifying business storefront locations in the first image.",2017-03-14,"The title of the patent is business discovery from imagery and its abstract is aspects of the present disclosure relate to a method includes training a deep neural network using training images and data identifying one or more business storefront locations in the training images. the deep neural network outputs tight bounding boxes on each image. at the deep neural network, a first image may be received. the first image may be evaluated using the deep neural network. bounding boxes may then be generated identifying business storefront locations in the first image. dated 2017-03-14"
9595002,normalizing electronic communications using a vector having a repeating substring as input for a neural network,"electronic communications can be normalized using a neural network. for example, a noncanonical communication that includes multiple terms can be received. the noncanonical communication can be preprocessed by (i) generating a vector including multiple characters from a term of the multiple terms; and (ii) repeating a substring of the term in the vector such that a last character of the substring is positioned in a last position in the vector. the vector can be transmitted to a neural network configured to receive the vector and generate multiple probabilities based on the vector. a normalized version of the noncanonical communication can be determined using one or more of the multiple probabilities generated by the neural network. whether the normalized version of the noncanonical communication should be outputted can also be determined using at least one of the multiple probabilities generated by the neural network.",2017-03-14,"The title of the patent is normalizing electronic communications using a vector having a repeating substring as input for a neural network and its abstract is electronic communications can be normalized using a neural network. for example, a noncanonical communication that includes multiple terms can be received. the noncanonical communication can be preprocessed by (i) generating a vector including multiple characters from a term of the multiple terms; and (ii) repeating a substring of the term in the vector such that a last character of the substring is positioned in a last position in the vector. the vector can be transmitted to a neural network configured to receive the vector and generate multiple probabilities based on the vector. a normalized version of the noncanonical communication can be determined using one or more of the multiple probabilities generated by the neural network. whether the normalized version of the noncanonical communication should be outputted can also be determined using at least one of the multiple probabilities generated by the neural network. dated 2017-03-14"
9595257,downsampling schemes in a hierarchical neural network structure for phoneme recognition,an approach for phoneme recognition is described. a sequence of intermediate output posterior vectors is generated from an input sequence of cepstral features using a first layer perceptron. the intermediate output posterior vectors are then downsampled to form a reduced input set of intermediate posterior vectors for a second layer perceptron. a sequence of final posterior vectors is generated from the reduced input set of intermediate posterior vectors using the second layer perceptron. then the final posterior vectors are decoded to determine an output recognized phoneme sequence representative of the input sequence of cepstral features.,2017-03-14,The title of the patent is downsampling schemes in a hierarchical neural network structure for phoneme recognition and its abstract is an approach for phoneme recognition is described. a sequence of intermediate output posterior vectors is generated from an input sequence of cepstral features using a first layer perceptron. the intermediate output posterior vectors are then downsampled to form a reduced input set of intermediate posterior vectors for a second layer perceptron. a sequence of final posterior vectors is generated from the reduced input set of intermediate posterior vectors using the second layer perceptron. then the final posterior vectors are decoded to determine an output recognized phoneme sequence representative of the input sequence of cepstral features. dated 2017-03-14
9600763,"information processing method, information processing device, and non-transitory recording medium for storing program","an computer-implemented information processing method for a convolutional neural network processing input data, includes: identifying, by a computer, for each of elements of a kernel used in convolution operation, input values to be multiplied by the respective elements in the convolution operation from among input values included in the input data; calculating a sum total of identified input values; calculating, for each of the elements of the kernel, a product of the sum total and the element; calculating an average of calculated products; and performing the convolution operation within the convolutional neural network based on the average of the calculated products.",2017-03-21,"The title of the patent is information processing method, information processing device, and non-transitory recording medium for storing program and its abstract is an computer-implemented information processing method for a convolutional neural network processing input data, includes: identifying, by a computer, for each of elements of a kernel used in convolution operation, input values to be multiplied by the respective elements in the convolution operation from among input values included in the input data; calculating a sum total of identified input values; calculating, for each of the elements of the kernel, a product of the sum total and the element; calculating an average of calculated products; and performing the convolution operation within the convolutional neural network based on the average of the calculated products. dated 2017-03-21"
9600764,markov-based sequence tagging using neural networks,"features are disclosed for using a neural network to tag sequential input without using an internal representation of the neural network generated when scoring previous positions in the sequence. a predicted or determined label (e.g., the highest scoring or otherwise most probable label) for input at a given position in the sequence can be used when scoring input corresponding to the next position the sequence. additional features are disclosed for training a neural network for use in tagging sequential input without using an internal representation of the neural network generated when scoring previous positions the sequence.",2017-03-21,"The title of the patent is markov-based sequence tagging using neural networks and its abstract is features are disclosed for using a neural network to tag sequential input without using an internal representation of the neural network generated when scoring previous positions in the sequence. a predicted or determined label (e.g., the highest scoring or otherwise most probable label) for input at a given position in the sequence can be used when scoring input corresponding to the next position the sequence. additional features are disclosed for training a neural network for use in tagging sequential input without using an internal representation of the neural network generated when scoring previous positions the sequence. dated 2017-03-21"
9601109,systems and methods for accelerating hessian-free optimization for deep neural networks by implicit preconditioning and sampling,"a method for training a deep neural network, comprises receiving and formatting speech data for the training, preconditioning a system of equations to be used for analyzing the speech data in connection with the training by using a non-fixed point quasi-newton preconditioning scheme, and employing flexible krylov subspace solvers in response to variations in the preconditioning scheme for different iterations of the training.",2017-03-21,"The title of the patent is systems and methods for accelerating hessian-free optimization for deep neural networks by implicit preconditioning and sampling and its abstract is a method for training a deep neural network, comprises receiving and formatting speech data for the training, preconditioning a system of equations to be used for analyzing the speech data in connection with the training by using a non-fixed point quasi-newton preconditioning scheme, and employing flexible krylov subspace solvers in response to variations in the preconditioning scheme for different iterations of the training. dated 2017-03-21"
9606530,decision support system for order prioritization,"a method for order prioritization includes calculating a cycle time for a product order of a plurality of product orders using an artificial neural network, determining a first order priority of the product order based on a priority index using an analytic hierarchy process, determining a second order priority of the product order based on event based simulation model, and determining a shipping date for the product order based on the second order priority. the artificial neural network calculates the cycle time based upon product order type and a plurality of component counts. the analytic hierarchy process determines a first order priority based upon a plurality of product order attributes. the simulation model determines a second order priority and completion time based upon the first order priority, product model, product type, a plurality of component counts, manufacturing capacity and inventory data, and production time data for historical product orders.",2017-03-28,"The title of the patent is decision support system for order prioritization and its abstract is a method for order prioritization includes calculating a cycle time for a product order of a plurality of product orders using an artificial neural network, determining a first order priority of the product order based on a priority index using an analytic hierarchy process, determining a second order priority of the product order based on event based simulation model, and determining a shipping date for the product order based on the second order priority. the artificial neural network calculates the cycle time based upon product order type and a plurality of component counts. the analytic hierarchy process determines a first order priority based upon a plurality of product order attributes. the simulation model determines a second order priority and completion time based upon the first order priority, product model, product type, a plurality of component counts, manufacturing capacity and inventory data, and production time data for historical product orders. dated 2017-03-28"
9607217,generating preference indices for image content,"briefly, embodiments of methods and/or systems of generating preference indices for contiguous portions of digital images are disclosed. for one embodiment, as an example, parameters of a neural network may be developed to generate object labels for digital images. the developed parameters may be transferred to a neural network utilized to generate signal sample value levels corresponding to preference indices for contiguous portions of digital images.",2017-03-28,"The title of the patent is generating preference indices for image content and its abstract is briefly, embodiments of methods and/or systems of generating preference indices for contiguous portions of digital images are disclosed. for one embodiment, as an example, parameters of a neural network may be developed to generate object labels for digital images. the developed parameters may be transferred to a neural network utilized to generate signal sample value levels corresponding to preference indices for contiguous portions of digital images. dated 2017-03-28"
9607265,accurate and fast neural network training for library-based critical dimension (cd) metrology,"embodiments are generally directed to neural network training for library-based critical dimension metrology. an embodiment of a method includes optimizing a threshold for a principal component analysis of a spectrum data set to provide a principal component value, estimating a training target for one or more neural networks, training the one or more neural networks based both on the training target and on the principal component value provided from optimizing the threshold for the principal component analysis, and providing a spectral library based on the one or more trained neural networks.",2017-03-28,"The title of the patent is accurate and fast neural network training for library-based critical dimension (cd) metrology and its abstract is embodiments are generally directed to neural network training for library-based critical dimension metrology. an embodiment of a method includes optimizing a threshold for a principal component analysis of a spectrum data set to provide a principal component value, estimating a training target for one or more neural networks, training the one or more neural networks based both on the training target and on the principal component value provided from optimizing the threshold for the principal component analysis, and providing a spectral library based on the one or more trained neural networks. dated 2017-03-28"
9607616,method for using a multi-scale recurrent neural network with pretraining for spoken language understanding tasks,"a spoken language understanding (slu) system receives a sequence of words corresponding to one or more spoken utterances of a user, which is passed through a spoken language understanding module to produce a sequence of intentions. the sequence of words are passed through a first subnetwork of a multi-scale recurrent neural network (msrnn), and the sequence of intentions are passed through a second subnetwork of the multi-scale recurrent neural network (msrnn). then, the outputs of the first subnetwork and the second subnetwork are combined to predict a goal of the user.",2017-03-28,"The title of the patent is method for using a multi-scale recurrent neural network with pretraining for spoken language understanding tasks and its abstract is a spoken language understanding (slu) system receives a sequence of words corresponding to one or more spoken utterances of a user, which is passed through a spoken language understanding module to produce a sequence of intentions. the sequence of words are passed through a first subnetwork of a multi-scale recurrent neural network (msrnn), and the sequence of intentions are passed through a second subnetwork of the multi-scale recurrent neural network (msrnn). then, the outputs of the first subnetwork and the second subnetwork are combined to predict a goal of the user. dated 2017-03-28"
9613058,neural network image curation control,"neural network image curation techniques are described. in one or more implementations, curation is controlled of images that represent a repository of images. a plurality of images of the repository are curated by one or more computing devices to select representative images of the repository. the curation includes calculating a score based on image and face aesthetics, jointly, for each of the plurality of images through processing by a neural network, ranking the plurality of images based on respective said scores, and selecting one or more of the plurality of images as one of the representative images of the repository based on the ranking and a determination that the one or more said images are not visually similar to images that have already been selected as one of the representative images of the repository.",2017-04-04,"The title of the patent is neural network image curation control and its abstract is neural network image curation techniques are described. in one or more implementations, curation is controlled of images that represent a repository of images. a plurality of images of the repository are curated by one or more computing devices to select representative images of the repository. the curation includes calculating a score based on image and face aesthetics, jointly, for each of the plurality of images through processing by a neural network, ranking the plurality of images based on respective said scores, and selecting one or more of the plurality of images as one of the representative images of the repository based on the ranking and a determination that the one or more said images are not visually similar to images that have already been selected as one of the representative images of the repository. dated 2017-04-04"
9613310,neural network learning and collaboration apparatus and methods,"apparatus and methods for learning and training in neural network-based devices. in one implementation, the devices each comprise multiple spiking neurons, configured to process sensory input. in one approach, alternate heterosynaptic plasticity mechanisms are used to enhance learning and field diversity within the devices. the selection of alternate plasticity rules is based on recent post-synaptic activity of neighboring neurons. apparatus and methods for simplifying training of the devices are also disclosed, including a computer-based application. a data representation of the neural network may be imaged and transferred to another computational environment, effectively copying the brain. techniques and architectures for achieve this training, storing, and distributing these data representations are also disclosed.",2017-04-04,"The title of the patent is neural network learning and collaboration apparatus and methods and its abstract is apparatus and methods for learning and training in neural network-based devices. in one implementation, the devices each comprise multiple spiking neurons, configured to process sensory input. in one approach, alternate heterosynaptic plasticity mechanisms are used to enhance learning and field diversity within the devices. the selection of alternate plasticity rules is based on recent post-synaptic activity of neighboring neurons. apparatus and methods for simplifying training of the devices are also disclosed, including a computer-based application. a data representation of the neural network may be imaged and transferred to another computational environment, effectively copying the brain. techniques and architectures for achieve this training, storing, and distributing these data representations are also disclosed. dated 2017-04-04"
9619735,pure convolutional neural network localization,"an approach is provided in which a knowledge manager processes an image using a convolutional neural network. the knowledge manager generates a pixel-level heat map of the image that includes multiple decision points corresponding to multiple pixels of the image. the knowledge manager analyzes the pixel-level heat map and detects sets of decision points that correspond to target objects. in turn, the knowledge manager marks regions of the heat map corresponding to the detected sets of per-pixel decision points, each of the regions indicating a location of the target objects.",2017-04-11,"The title of the patent is pure convolutional neural network localization and its abstract is an approach is provided in which a knowledge manager processes an image using a convolutional neural network. the knowledge manager generates a pixel-level heat map of the image that includes multiple decision points corresponding to multiple pixels of the image. the knowledge manager analyzes the pixel-level heat map and detects sets of decision points that correspond to target objects. in turn, the knowledge manager marks regions of the heat map corresponding to the detected sets of per-pixel decision points, each of the regions indicating a location of the target objects. dated 2017-04-11"
9619747,prospective media content generation using neural network modeling,"a system for prospectively identifying media characteristics for inclusion in media content is disclosed. a neural network database including media characteristic information and feature information may associate relationships among the media characteristic information and feature information. personal characteristic information associated with target media consumers may be used to select a subset of the neural network database. a first set of nodes, representing selected feature information, may be activated. the node interactions may be calculated to detect the activation of a second set of nodes, the second set of nodes representing media characteristic information. generally, a node is activated when an activation value of the node exceeds a threshold value. media characteristic information may be identified for inclusion in media content based on the second set of nodes.",2017-04-11,"The title of the patent is prospective media content generation using neural network modeling and its abstract is a system for prospectively identifying media characteristics for inclusion in media content is disclosed. a neural network database including media characteristic information and feature information may associate relationships among the media characteristic information and feature information. personal characteristic information associated with target media consumers may be used to select a subset of the neural network database. a first set of nodes, representing selected feature information, may be activated. the node interactions may be calculated to detect the activation of a second set of nodes, the second set of nodes representing media characteristic information. generally, a node is activated when an activation value of the node exceeds a threshold value. media characteristic information may be identified for inclusion in media content based on the second set of nodes. dated 2017-04-11"
9619748,intelligent control with hierarchical stacked neural networks,"an intelligent control system based on an explicit model of cognitive development (table 1) performs high-level functions. it comprises up to o hierarchically stacked neural networks, nm, . . . , nm+(o−1), where m denotes the stage/order tasks performed in the first neural network, nm, and o denotes the highest stage/order tasks performed in the highest-level neural network. the type of processing actions performed in a network, nm, corresponds to the complexity for stage/order m. thus n1 performs tasks at the level corresponding to stage/order 1. n5 processes information at the level corresponding to stage/order 5. stacked neural networks begin and end at any stage/order, but information must be processed by each stage in ascending order sequence. stages/orders cannot be skipped. each neural network in a stack may use different architectures, interconnections, algorithms, and training methods, depending on the stage/order of the neural network and the type of intelligent control system implemented.",2017-04-11,"The title of the patent is intelligent control with hierarchical stacked neural networks and its abstract is an intelligent control system based on an explicit model of cognitive development (table 1) performs high-level functions. it comprises up to o hierarchically stacked neural networks, nm, . . . , nm+(o−1), where m denotes the stage/order tasks performed in the first neural network, nm, and o denotes the highest stage/order tasks performed in the highest-level neural network. the type of processing actions performed in a network, nm, corresponds to the complexity for stage/order m. thus n1 performs tasks at the level corresponding to stage/order 1. n5 processes information at the level corresponding to stage/order 5. stacked neural networks begin and end at any stage/order, but information must be processed by each stage in ascending order sequence. stages/orders cannot be skipped. each neural network in a stack may use different architectures, interconnections, algorithms, and training methods, depending on the stage/order of the neural network and the type of intelligent control system implemented. dated 2017-04-11"
9619749,neural network and method of neural network training,"a neural network includes a plurality of inputs for receiving input signals, and synapses connected to the inputs and having corrective weights established by a memory element that retains a respective weight value. the network additionally includes distributors. each distributor is connected to one of the inputs for receiving the respective input signal and selects one or more corrective weights in correlation with the input value. the network also includes neurons. each neuron has an output connected with at least one of the inputs via one synapse and generates a neuron sum by summing corrective weights selected from each synapse connected to the respective neuron. the output of each neuron provides the respective neuron sum to establish operational output signal of the network. a method of operating a neural network includes processing data thereby and using modified corrective weight values established by a separate analogous neural network during training thereof.",2017-04-11,"The title of the patent is neural network and method of neural network training and its abstract is a neural network includes a plurality of inputs for receiving input signals, and synapses connected to the inputs and having corrective weights established by a memory element that retains a respective weight value. the network additionally includes distributors. each distributor is connected to one of the inputs for receiving the respective input signal and selects one or more corrective weights in correlation with the input value. the network also includes neurons. each neuron has an output connected with at least one of the inputs via one synapse and generates a neuron sum by summing corrective weights selected from each synapse connected to the respective neuron. the output of each neuron provides the respective neuron sum to establish operational output signal of the network. a method of operating a neural network includes processing data thereby and using modified corrective weight values established by a separate analogous neural network during training thereof. dated 2017-04-11"
9620145,context-dependent state tying using a neural network,"the technology described herein can be embodied in a method that includes receiving an audio signal encoding a portion of an utterance, and providing, to a first neural network, data corresponding to the audio signal. the method also includes generating, by a processor, data representing a transcription for the utterance based on an output of the first neural network. the first neural network is trained using features of multiple context-dependent states, the context-dependent states being derived from a plurality of context-independent states provided by a second neural network.",2017-04-11,"The title of the patent is context-dependent state tying using a neural network and its abstract is the technology described herein can be embodied in a method that includes receiving an audio signal encoding a portion of an utterance, and providing, to a first neural network, data corresponding to the audio signal. the method also includes generating, by a processor, data representing a transcription for the utterance based on an output of the first neural network. the first neural network is trained using features of multiple context-dependent states, the context-dependent states being derived from a plurality of context-independent states provided by a second neural network. dated 2017-04-11"
9626621,systems and methods for combining stochastic average gradient and hessian-free optimization for sequence training of deep neural networks,"a method for training a deep neural network (dnn), comprises receiving and formatting speech data for the training, performing hessian-free sequence training (hfst) on a first subset of a plurality of subsets of the speech data, and iteratively performing the hfst on successive subsets of the plurality of subsets of the speech data, wherein iteratively performing the hfst comprises reusing information from at least one previous iteration.",2017-04-18,"The title of the patent is systems and methods for combining stochastic average gradient and hessian-free optimization for sequence training of deep neural networks and its abstract is a method for training a deep neural network (dnn), comprises receiving and formatting speech data for the training, performing hessian-free sequence training (hfst) on a first subset of a plurality of subsets of the speech data, and iteratively performing the hfst on successive subsets of the plurality of subsets of the speech data, wherein iteratively performing the hfst comprises reusing information from at least one previous iteration. dated 2017-04-18"
9627532,methods and apparatus for training an artificial neural network for use in speech recognition,"methods and apparatus for training a multi-layer artificial neural network for use in speech recognition. the method comprises determining for a first speech pattern of the plurality of speech patterns, using a first processing pipeline, network activations for a plurality of nodes of the artificial neural network in response to providing the first speech pattern as input to the artificial neural network, determining based, at least in part, on the network activations and a selection criterion, whether the artificial neural network should be trained on the first speech pattern, and updating, using a second processing pipeline, network weights between nodes of the artificial neural network based, at least in part, on the network activations when it is determined that the artificial neural network should be trained on the first speech pattern.",2017-04-18,"The title of the patent is methods and apparatus for training an artificial neural network for use in speech recognition and its abstract is methods and apparatus for training a multi-layer artificial neural network for use in speech recognition. the method comprises determining for a first speech pattern of the plurality of speech patterns, using a first processing pipeline, network activations for a plurality of nodes of the artificial neural network in response to providing the first speech pattern as input to the artificial neural network, determining based, at least in part, on the network activations and a selection criterion, whether the artificial neural network should be trained on the first speech pattern, and updating, using a second processing pipeline, network weights between nodes of the artificial neural network based, at least in part, on the network activations when it is determined that the artificial neural network should be trained on the first speech pattern. dated 2017-04-18"
9628926,modeling loudspeakers based on cascading lumped parameter models with neural networks,"in one embodiment of the present invention, a loudspeaker modeling subsystem configures a neural lumped parameter loudspeaker (nelp) model to represent the behavior of a loudspeaker. the nelp model is implemented as a cascaded combination of a lumped parameter model (lpm) and a neural network (nn) model. to configure the model, the loudspeaker modeling subsystem first estimates values for the parameters used in the lpm. the loudspeaker modeling subsystem then “fixes” these parameters and trains the nn model to act on a predicted output pressure that is generated via the lpm. more specifically, the loudspeaker modeling subsystem configures the nn to modify the predicted output pressure to minimize the error between the predicted output pressure and a measured loudspeaker output pressure. notably, by strategically fusing the lpm and the nn model, the nelp model leverages the strengths and mitigates the weaknesses typically associated with conventional loudspeaker modeling techniques.",2017-04-18,"The title of the patent is modeling loudspeakers based on cascading lumped parameter models with neural networks and its abstract is in one embodiment of the present invention, a loudspeaker modeling subsystem configures a neural lumped parameter loudspeaker (nelp) model to represent the behavior of a loudspeaker. the nelp model is implemented as a cascaded combination of a lumped parameter model (lpm) and a neural network (nn) model. to configure the model, the loudspeaker modeling subsystem first estimates values for the parameters used in the lpm. the loudspeaker modeling subsystem then “fixes” these parameters and trains the nn model to act on a predicted output pressure that is generated via the lpm. more specifically, the loudspeaker modeling subsystem configures the nn to modify the predicted output pressure to minimize the error between the predicted output pressure and a measured loudspeaker output pressure. notably, by strategically fusing the lpm and the nn model, the nelp model leverages the strengths and mitigates the weaknesses typically associated with conventional loudspeaker modeling techniques. dated 2017-04-18"
9633268,method and device for gait recognition,"disclose is a gait recognition method, firstly, extracting an initial gait feature of a gait video of a person to be recognized; secondly obtaining a corresponding optimized gait feature according to a trained sub neural network and the initial gait feature; then determining corresponding degrees of similarity according to the optimized gait feature of the person to be recognized and the optimized gait feature of each known person in a matching library, and determining information of the person to be recognized according to information of the known person in the matching library corresponding to the optimized gait feature which has the highest degree of similarity with the optimized gait feature of the person to be recognized.",2017-04-25,"The title of the patent is method and device for gait recognition and its abstract is disclose is a gait recognition method, firstly, extracting an initial gait feature of a gait video of a person to be recognized; secondly obtaining a corresponding optimized gait feature according to a trained sub neural network and the initial gait feature; then determining corresponding degrees of similarity according to the optimized gait feature of the person to be recognized and the optimized gait feature of each known person in a matching library, and determining information of the person to be recognized according to information of the known person in the matching library corresponding to the optimized gait feature which has the highest degree of similarity with the optimized gait feature of the person to be recognized. dated 2017-04-25"
9633282,cross-trained convolutional neural networks using multimodal images,"embodiments of a computer-implemented method for training a convolutional neural network (cnn) that is pre-trained using a set of color images are disclosed. the method comprises receiving a training dataset including multiple multidimensional images, each multidimensional image including a color image and a depth image; performing a fine-tuning of the pre-trained cnn using the depth image for each of the plurality of multidimensional images; obtaining a depth cnn based on the pre-trained cnn, wherein the depth cnn is associated with a first set of parameters; replicating the depth cnn to obtain a duplicate depth cnn being initialized with the first set of parameters; and obtaining a depth-enhanced color cnn based on the duplicate depth cnn being fine-tuned using the color image for each of the plurality of multidimensional images, wherein the depth-enhanced color cnn is associated with a second set of parameters.",2017-04-25,"The title of the patent is cross-trained convolutional neural networks using multimodal images and its abstract is embodiments of a computer-implemented method for training a convolutional neural network (cnn) that is pre-trained using a set of color images are disclosed. the method comprises receiving a training dataset including multiple multidimensional images, each multidimensional image including a color image and a depth image; performing a fine-tuning of the pre-trained cnn using the depth image for each of the plurality of multidimensional images; obtaining a depth cnn based on the pre-trained cnn, wherein the depth cnn is associated with a first set of parameters; replicating the depth cnn to obtain a duplicate depth cnn being initialized with the first set of parameters; and obtaining a depth-enhanced color cnn based on the duplicate depth cnn being fine-tuned using the color image for each of the plurality of multidimensional images, wherein the depth-enhanced color cnn is associated with a second set of parameters. dated 2017-04-25"
9633306,method and system for approximating deep neural networks for anatomical object detection,a method and system for approximating a deep neural network for anatomical object detection is discloses. a deep neural network is trained to detect an anatomical object in medical images. an approximation of the trained deep neural network is calculated that reduces the computational complexity of the trained deep neural network. the anatomical object is detected in an input medical image of a patient using the approximation of the trained deep neural network.,2017-04-25,The title of the patent is method and system for approximating deep neural networks for anatomical object detection and its abstract is a method and system for approximating a deep neural network for anatomical object detection is discloses. a deep neural network is trained to detect an anatomical object in medical images. an approximation of the trained deep neural network is calculated that reduces the computational complexity of the trained deep neural network. the anatomical object is detected in an input medical image of a patient using the approximation of the trained deep neural network. dated 2017-04-25
9633307,method to predict the effluent ammonia-nitrogen concentration based on a recurrent self-organizing neural network,"an intelligent method is designed for predicting the effluent ammonia-nitrogen concentration in the urban wastewater treatment process (wwtp). the technology of this invention is part of advanced manufacturing technology, belongs to both the field of control engineering and environment engineering. in order to improve the predicting efficiency, a recurrent self-organizing neural network, which can adjust the structure and parameters concurrently to train the parameters, is developed to design this intelligent method. this intelligent method can predict the effluent ammonia-nitrogen concentration with acceptable accuracy and solve the problem that the effluent ammonia-nitrogen concentration is difficult to be measured online. moreover, the online information of effluent ammonia-nitrogen concentration, predicted by this intelligent method, can enhance the quality monitoring level and alleviate the current situation of wastewater to strengthen the whole management of wwtp.",2017-04-25,"The title of the patent is method to predict the effluent ammonia-nitrogen concentration based on a recurrent self-organizing neural network and its abstract is an intelligent method is designed for predicting the effluent ammonia-nitrogen concentration in the urban wastewater treatment process (wwtp). the technology of this invention is part of advanced manufacturing technology, belongs to both the field of control engineering and environment engineering. in order to improve the predicting efficiency, a recurrent self-organizing neural network, which can adjust the structure and parameters concurrently to train the parameters, is developed to design this intelligent method. this intelligent method can predict the effluent ammonia-nitrogen concentration with acceptable accuracy and solve the problem that the effluent ammonia-nitrogen concentration is difficult to be measured online. moreover, the online information of effluent ammonia-nitrogen concentration, predicted by this intelligent method, can enhance the quality monitoring level and alleviate the current situation of wastewater to strengthen the whole management of wwtp. dated 2017-04-25"
9639070,controlling a turbine with a recurrent neural network,"a method for controlling a turbine is proposed, which is characterized at any point in the control by a hidden state. the dynamic behavior of the turbine is modeled with a recurrent neural network comprising a recurrent hidden layer. in this case, the recurrent hidden layer is formed from vectors of neurons, which describe the hidden state of the turbine at the time points of the regulation, wherein two vectors are chronologically linked for each time point with a first connection bridging a time and second connection bridging at least two points in time. short-term effects can be controlled by means of the first connections and long-term effects can be adjusted by means of the second connections. secondly, emissions and also occurring dynamics in the turbine can be minimized. furthermore, a regulating device and a turbine with such a regulating device are proposed.",2017-05-02,"The title of the patent is controlling a turbine with a recurrent neural network and its abstract is a method for controlling a turbine is proposed, which is characterized at any point in the control by a hidden state. the dynamic behavior of the turbine is modeled with a recurrent neural network comprising a recurrent hidden layer. in this case, the recurrent hidden layer is formed from vectors of neurons, which describe the hidden state of the turbine at the time points of the regulation, wherein two vectors are chronologically linked for each time point with a first connection bridging a time and second connection bridging at least two points in time. short-term effects can be controlled by means of the first connections and long-term effects can be adjusted by means of the second connections. secondly, emissions and also occurring dynamics in the turbine can be minimized. furthermore, a regulating device and a turbine with such a regulating device are proposed. dated 2017-05-02"
9639802,"multi-modal neural network for universal, online learning","in one embodiment, the present invention provides a neural network comprising multiple modalities. each modality comprises multiple neurons. the neural network further comprises an interconnection lattice for cross-associating signaling between the neurons in different modalities. the interconnection lattice includes a plurality of perception neuron populations along a number of bottom-up signaling pathways, and a plurality of action neuron populations along a number of top-down signaling pathways. each perception neuron along a bottom-up signaling pathway has a corresponding action neuron along a reciprocal top-down signaling pathway. an input neuron population configured to receive sensory input drives perception neurons along a number of bottom-up signaling pathways. a first set of perception neurons along bottom-up signaling pathways drive a first set of action neurons along top-down signaling pathways. action neurons along a number of top-down signaling pathways drive an output neuron population configured to generate motor output.",2017-05-02,"The title of the patent is multi-modal neural network for universal, online learning and its abstract is in one embodiment, the present invention provides a neural network comprising multiple modalities. each modality comprises multiple neurons. the neural network further comprises an interconnection lattice for cross-associating signaling between the neurons in different modalities. the interconnection lattice includes a plurality of perception neuron populations along a number of bottom-up signaling pathways, and a plurality of action neuron populations along a number of top-down signaling pathways. each perception neuron along a bottom-up signaling pathway has a corresponding action neuron along a reciprocal top-down signaling pathway. an input neuron population configured to receive sensory input drives perception neurons along a number of bottom-up signaling pathways. a first set of perception neurons along bottom-up signaling pathways drive a first set of action neurons along top-down signaling pathways. action neurons along a number of top-down signaling pathways drive an output neuron population configured to generate motor output. dated 2017-05-02"
9646230,image segmentation in optical character recognition using neural networks,neural-network-based image segmentation techniques are provided herein. an input image that includes a plurality of characters can be received. boundaries between the characters can be identified using a trained neural network. the input image can be segmented along the boundaries identified between the characters. the neural network can be trained using a training image and a training target vector. the training target vector can indicate one or more boundaries between characters in the training image. neural-network-based segmentation can be used alone or in conjunction with other segmentation techniques to improve overall segmentation accuracy.,2017-05-09,The title of the patent is image segmentation in optical character recognition using neural networks and its abstract is neural-network-based image segmentation techniques are provided herein. an input image that includes a plurality of characters can be received. boundaries between the characters can be identified using a trained neural network. the input image can be segmented along the boundaries identified between the characters. the neural network can be trained using a training image and a training target vector. the training target vector can indicate one or more boundaries between characters in the training image. neural-network-based segmentation can be used alone or in conjunction with other segmentation techniques to improve overall segmentation accuracy. dated 2017-05-09
9646243,convolutional neural networks using resistive processing unit array,"technical solutions are described for implementing a convolutional neural network (cnn) using resistive processing unit (rpu) array. an example method includes configuring an rpu array corresponding to a convolution layer in the cnn based on convolution kernels of the layer. the method further includes performing forward pass computations via the rpu array by transmitting voltage pulses corresponding to input data to the rpu array, and storing values corresponding to output currents from the rpu arrays as output maps. the method further includes performing backward pass computations via the rpu array by transmitting voltage pulses corresponding to error of the output maps, and storing the output currents from the rpu arrays as backward error maps. the method further includes performing update pass computations via the rpu array by transmitting voltage pulses corresponding to the input data of the convolution layer and the error of the output maps to the rpu array.",2017-05-09,"The title of the patent is convolutional neural networks using resistive processing unit array and its abstract is technical solutions are described for implementing a convolutional neural network (cnn) using resistive processing unit (rpu) array. an example method includes configuring an rpu array corresponding to a convolution layer in the cnn based on convolution kernels of the layer. the method further includes performing forward pass computations via the rpu array by transmitting voltage pulses corresponding to input data to the rpu array, and storing values corresponding to output currents from the rpu arrays as output maps. the method further includes performing backward pass computations via the rpu array by transmitting voltage pulses corresponding to error of the output maps, and storing the output currents from the rpu arrays as backward error maps. the method further includes performing update pass computations via the rpu array by transmitting voltage pulses corresponding to the input data of the convolution layer and the error of the output maps to the rpu array. dated 2017-05-09"
9646244,predicting likelihoods of conditions being satisfied using recurrent neural networks,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for predicting likelihoods of conditions being satisfied using recurrent neural networks. one of the systems is configured to process a temporal sequence comprising a respective input at each of a plurality of time steps and comprises: one or more recurrent neural network layers; one or more logistic regression nodes, wherein each of the logistic regression nodes corresponds to a respective condition from a predetermined set of conditions, and wherein each of the logistic regression nodes is configured to, for each of the plurality of time steps: receive the network internal state for the time step; and process the network internal state for the time step in accordance with current values of a set of parameters of the logistic regression node to generate a future condition score for the corresponding condition for the time step.",2017-05-09,"The title of the patent is predicting likelihoods of conditions being satisfied using recurrent neural networks and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for predicting likelihoods of conditions being satisfied using recurrent neural networks. one of the systems is configured to process a temporal sequence comprising a respective input at each of a plurality of time steps and comprises: one or more recurrent neural network layers; one or more logistic regression nodes, wherein each of the logistic regression nodes corresponds to a respective condition from a predetermined set of conditions, and wherein each of the logistic regression nodes is configured to, for each of the plurality of time steps: receive the network internal state for the time step; and process the network internal state for the time step in accordance with current values of a set of parameters of the logistic regression node to generate a future condition score for the corresponding condition for the time step. dated 2017-05-09"
9646634,low-rank hidden input layer for speech recognition neural network,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a deep neural network. one of the methods for training a deep neural network that includes a low rank hidden input layer and an adjoining hidden layer, the low rank hidden input layer including a first matrix a and a second matrix b with dimensions i×m and m×o, respectively, to identify a keyword includes receiving a feature vector including i values that represent features of an audio signal encoding an utterance, determining, using the low rank hidden input layer, an output vector including o values using the feature vector, determining, using the adjoining hidden layer, another vector using the output vector, determining a confidence score that indicates whether the utterance includes the keyword using the other vector, and adjusting weights for the low rank hidden input layer using the confidence score.",2017-05-09,"The title of the patent is low-rank hidden input layer for speech recognition neural network and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a deep neural network. one of the methods for training a deep neural network that includes a low rank hidden input layer and an adjoining hidden layer, the low rank hidden input layer including a first matrix a and a second matrix b with dimensions i×m and m×o, respectively, to identify a keyword includes receiving a feature vector including i values that represent features of an audio signal encoding an utterance, determining, using the low rank hidden input layer, an output vector including o values using the feature vector, determining, using the adjoining hidden layer, another vector using the output vector, determining a confidence score that indicates whether the utterance includes the keyword using the other vector, and adjusting weights for the low rank hidden input layer using the confidence score. dated 2017-05-09"
9651529,artificial olfactory system and an application thereof,"the present invention relates to an artificial olfactory system (100), comprising of an inlet (101); a gas chamber (110) having a detector means, connected to a data acquisition system (104); a heater (112) and a plurality of fans (115); a humidity absorber (111); an outlet (102); a vacuum pump (103); characterized by the detector means having a plurality of sensors (121) in each of a plurality of clusters (120), wherein the plurality of sensors (121) in each of the plurality of clusters (120) comprises identical sensors capable of responding to a particular gas or vapor. the present invention also relates to a method for detecting a gas or a vapor from the artificial olfactory system (100), comprising the step of exposing the gas or vapor to the plurality of sensors (121) to produce a plurality of output signals from the plurality of sensors (121); transferring the plurality of output signals to the data acquisition system (104); extracting median data from the plurality of output signals; applying a principal component analysis (pca), neural network, and least square regression analysis on the median data from all of the plurality of clusters (120).",2017-05-16,"The title of the patent is artificial olfactory system and an application thereof and its abstract is the present invention relates to an artificial olfactory system (100), comprising of an inlet (101); a gas chamber (110) having a detector means, connected to a data acquisition system (104); a heater (112) and a plurality of fans (115); a humidity absorber (111); an outlet (102); a vacuum pump (103); characterized by the detector means having a plurality of sensors (121) in each of a plurality of clusters (120), wherein the plurality of sensors (121) in each of the plurality of clusters (120) comprises identical sensors capable of responding to a particular gas or vapor. the present invention also relates to a method for detecting a gas or a vapor from the artificial olfactory system (100), comprising the step of exposing the gas or vapor to the plurality of sensors (121) to produce a plurality of output signals from the plurality of sensors (121); transferring the plurality of output signals to the data acquisition system (104); extracting median data from the plurality of output signals; applying a principal component analysis (pca), neural network, and least square regression analysis on the median data from all of the plurality of clusters (120). dated 2017-05-16"
9652696,apparatus and method for surface and subsurface tactile sensation imaging,"a tactile sensor, computer readable medium, methods of using and manufacturing the tactile sensor, and methods and apparatuses for processing the information generated by the tactile sensor. the tactile sensor includes a planar optical waveguide comprised of a flexible and transparent layer; a light configured to direct light into the optical waveguide; a light sensor or an imager facing the optical waveguide and configured to generate signals from light scattered out of the optical waveguide; and a controller which may be configured to generate an image of the object and characteristics of the object. the waveguide may be configured so that some of the light directed into the optical waveguide is scattered out of the waveguide if the waveguide is deformed by being pressed against the object. a finite element and a neural network are used to estimate mechanical characteristics of the objects.",2017-05-16,"The title of the patent is apparatus and method for surface and subsurface tactile sensation imaging and its abstract is a tactile sensor, computer readable medium, methods of using and manufacturing the tactile sensor, and methods and apparatuses for processing the information generated by the tactile sensor. the tactile sensor includes a planar optical waveguide comprised of a flexible and transparent layer; a light configured to direct light into the optical waveguide; a light sensor or an imager facing the optical waveguide and configured to generate signals from light scattered out of the optical waveguide; and a controller which may be configured to generate an image of the object and characteristics of the object. the waveguide may be configured so that some of the light directed into the optical waveguide is scattered out of the waveguide if the waveguide is deformed by being pressed against the object. a finite element and a neural network are used to estimate mechanical characteristics of the objects. dated 2017-05-16"
9652712,analyzing health events using recurrent neural networks,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for using recurrent neural networks to analyze health events. one of the methods includes: processing each of a plurality of initial temporal sequences of health events to generate, for each of the initial temporal sequences, a respective network internal state of a recurrent neural network for each time step in the initial temporal sequence; storing, for each of the initial temporal sequences, one or more of the network internal states for the time steps in the temporal sequence in a repository; obtaining a first temporal sequence; processing the first temporal sequence using the recurrent neural network to generate a sequence internal state for the first temporal sequence; and selecting one or more initial temporal sequences that are likely to include health events that are predictive of future health events in the first temporal sequence.",2017-05-16,"The title of the patent is analyzing health events using recurrent neural networks and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for using recurrent neural networks to analyze health events. one of the methods includes: processing each of a plurality of initial temporal sequences of health events to generate, for each of the initial temporal sequences, a respective network internal state of a recurrent neural network for each time step in the initial temporal sequence; storing, for each of the initial temporal sequences, one or more of the network internal states for the time steps in the temporal sequence in a repository; obtaining a first temporal sequence; processing the first temporal sequence using the recurrent neural network to generate a sequence internal state for the first temporal sequence; and selecting one or more initial temporal sequences that are likely to include health events that are predictive of future health events in the first temporal sequence. dated 2017-05-16"
9653093,generative modeling of speech using neural networks,"features are disclosed for using an artificial neural network to generate customized speech recognition models during the speech recognition process. by dynamically generating the speech recognition models during the speech recognition process, the models can be customized based on the specific context of individual frames within the audio data currently being processed. in this way, dependencies between frames in the current sequence can form the basis of the models used to score individual frames of the current sequence. thus, each frame of the current sequence (or some subset thereof) may be scored using one or more models customized for the particular frame in context.",2017-05-16,"The title of the patent is generative modeling of speech using neural networks and its abstract is features are disclosed for using an artificial neural network to generate customized speech recognition models during the speech recognition process. by dynamically generating the speech recognition models during the speech recognition process, the models can be customized based on the specific context of individual frames within the audio data currently being processed. in this way, dependencies between frames in the current sequence can form the basis of the models used to score individual frames of the current sequence. thus, each frame of the current sequence (or some subset thereof) may be scored using one or more models customized for the particular frame in context. dated 2017-05-16"
9659247,system and method for employing the use of neural networks for the purpose of real-time business intelligence and automation control,"a system and integration infrastructure to provide a distributed matrix or neural network of connected real-time decision support modules designed to perform business intelligence evaluations in real time. the system and integration infrastructure provide a network of intelligence superimposed upon any company's existing it data centers, and cloud computing connections. the system is highly customizable to the unique business model deployed by the client company within the best practices of the client company's industry. whether or not the client company has integrated their diverse enterprise systems, the elements of the matrix are annealed to the various data sources, transaction logs and client software installations currently deployed. these matrix elements or neurons are designed to house critical operational data, determined by the operational model of the client company to be of critical importance. when combined with monitor neurons, they automatically assess the gap between the desired state of a critical element and the current condition in real time. trigger conditions are pre-established, but modified by an executive controller in real-time, and the system is pre programmed to automatically respond in a prescribed manner to critical conditions having been met even when these conditions come from otherwise stove-piped enterprise applications.",2017-05-23,"The title of the patent is system and method for employing the use of neural networks for the purpose of real-time business intelligence and automation control and its abstract is a system and integration infrastructure to provide a distributed matrix or neural network of connected real-time decision support modules designed to perform business intelligence evaluations in real time. the system and integration infrastructure provide a network of intelligence superimposed upon any company's existing it data centers, and cloud computing connections. the system is highly customizable to the unique business model deployed by the client company within the best practices of the client company's industry. whether or not the client company has integrated their diverse enterprise systems, the elements of the matrix are annealed to the various data sources, transaction logs and client software installations currently deployed. these matrix elements or neurons are designed to house critical operational data, determined by the operational model of the client company to be of critical importance. when combined with monitor neurons, they automatically assess the gap between the desired state of a critical element and the current condition in real time. trigger conditions are pre-established, but modified by an executive controller in real-time, and the system is pre programmed to automatically respond in a prescribed manner to critical conditions having been met even when these conditions come from otherwise stove-piped enterprise applications. dated 2017-05-23"
9659248,machine learning and training a computer-implemented neural network to retrieve semantically equivalent questions using hybrid in-memory representations,"determining semantically equivalent text or questions using hybrid representations based on neural network learning. weighted bag-of-words and convolutional neural networks (cnn) based distributed vector representations of questions or text may be generated to compute the semantic similarity between questions or text. weighted bag-of-words and cnn based distributed vector representations may be jointly used to compute the semantic similarity. a pair-wise ranking loss function trains neural network. in one embodiment, the parameters of the system are trained by minimizing a pair-wise ranking loss function over a training set using stochastic gradient descent (sgd).",2017-05-23,"The title of the patent is machine learning and training a computer-implemented neural network to retrieve semantically equivalent questions using hybrid in-memory representations and its abstract is determining semantically equivalent text or questions using hybrid representations based on neural network learning. weighted bag-of-words and convolutional neural networks (cnn) based distributed vector representations of questions or text may be generated to compute the semantic similarity between questions or text. weighted bag-of-words and cnn based distributed vector representations may be jointly used to compute the semantic similarity. a pair-wise ranking loss function trains neural network. in one embodiment, the parameters of the system are trained by minimizing a pair-wise ranking loss function over a training set using stochastic gradient descent (sgd). dated 2017-05-23"
9659249,pre-programmed resistive cross-point array for neural network,"technical solutions are described for forming a semiconductor device for a crosspoint array that implements a pre-programmed neural network. an example method includes sequentially depositing a semiconducting layer, a top insulating layer, and a shunting layer onto a base insulating layer. the method further includes etching selective portions of the top insulating layer corresponding to resistance values associated with weights of the crossbar that implements the neural network.",2017-05-23,"The title of the patent is pre-programmed resistive cross-point array for neural network and its abstract is technical solutions are described for forming a semiconductor device for a crosspoint array that implements a pre-programmed neural network. an example method includes sequentially depositing a semiconducting layer, a top insulating layer, and a shunting layer onto a base insulating layer. the method further includes etching selective portions of the top insulating layer corresponding to resistance values associated with weights of the crossbar that implements the neural network. dated 2017-05-23"
9659384,"systems, methods, and computer program products for searching and sorting images by aesthetic quality","a system, method, and computer program product for assigning an aesthetic score to an image. a method of the present invention includes receiving an image. the method further includes executing a neural network on the image to generate learned features. the method further includes applying a machine-learned model to assign an aesthetic score to the image, where a more aesthetically-pleasing image is given a higher aesthetic score and a less aesthetically-pleasing image is given a lower aesthetic score. the learned features are inputs to the machine-learned model.",2017-05-23,"The title of the patent is systems, methods, and computer program products for searching and sorting images by aesthetic quality and its abstract is a system, method, and computer program product for assigning an aesthetic score to an image. a method of the present invention includes receiving an image. the method further includes executing a neural network on the image to generate learned features. the method further includes applying a machine-learned model to assign an aesthetic score to the image, where a more aesthetically-pleasing image is given a higher aesthetic score and a less aesthetically-pleasing image is given a lower aesthetic score. the learned features are inputs to the machine-learned model. dated 2017-05-23"
9659560,semi-supervised learning of word embeddings,"software that trains an artificial neural network for generating vector representations for natural language text, by performing the following steps: (i) receiving, by one or more processors, a set of natural language text; (ii) generating, by one or more processors, a set of first metadata for the set of natural language text, where the first metadata is generated using supervised learning method(s); (iii) generating, by one or more processors, a set of second metadata for the set of natural language text, where the second metadata is generated using unsupervised learning method(s); and (iv) training, by one or more processors, an artificial neural network adapted to generate vector representations for natural language text, where the training is based, at least in part, on the received natural language text, the generated set of first metadata, and the generated set of second metadata.",2017-05-23,"The title of the patent is semi-supervised learning of word embeddings and its abstract is software that trains an artificial neural network for generating vector representations for natural language text, by performing the following steps: (i) receiving, by one or more processors, a set of natural language text; (ii) generating, by one or more processors, a set of first metadata for the set of natural language text, where the first metadata is generated using supervised learning method(s); (iii) generating, by one or more processors, a set of second metadata for the set of natural language text, where the second metadata is generated using unsupervised learning method(s); and (iv) training, by one or more processors, an artificial neural network adapted to generate vector representations for natural language text, where the training is based, at least in part, on the received natural language text, the generated set of first metadata, and the generated set of second metadata. dated 2017-05-23"
9660571,"method for hybrid solar tracking, and apparatus for hybrid solar tracking and photovoltaic blind system using same","a method for hybrid solar tracking, and an apparatus for hybrid solar tracking and a photovoltaic blind system using the same are disclosed. the method includes generating first predicted power output data by analyzing first measured power output data of the past; generating a lagged error; constructing a regression analysis (ra) model, and deriving second predicted power output data; constructing an artificial neural network (ann) model, and deriving third predicted power output data; selecting either a method for solar tracking based on photovoltaic power output or a method for solar tracking based on location and time depending on whether the second measured power output data of the present time falls within a filtering range based on a first error range and a second error range; and determining the directions of photovoltaic panels according to the selected method for solar tracking.",2017-05-23,"The title of the patent is method for hybrid solar tracking, and apparatus for hybrid solar tracking and photovoltaic blind system using same and its abstract is a method for hybrid solar tracking, and an apparatus for hybrid solar tracking and a photovoltaic blind system using the same are disclosed. the method includes generating first predicted power output data by analyzing first measured power output data of the past; generating a lagged error; constructing a regression analysis (ra) model, and deriving second predicted power output data; constructing an artificial neural network (ann) model, and deriving third predicted power output data; selecting either a method for solar tracking based on photovoltaic power output or a method for solar tracking based on location and time depending on whether the second measured power output data of the present time falls within a filtering range based on a first error range and a second error range; and determining the directions of photovoltaic panels according to the selected method for solar tracking. dated 2017-05-23"
9665799,convolutional neural network,"a convolutional neural network (cnn) for an image processing system comprises an image cache responsive to a request to read a block of n×m pixels extending from a specified location within an input map to provide a block of n×m pixels at an output port. a convolution engine reads blocks of pixels from the output port, combines blocks of pixels with a corresponding set of weights to provide a product, and subjects the product to an activation function to provide an output pixel value. the image cache comprises a plurality of interleaved memories capable of simultaneously providing the n×m pixels at the output port in a single clock cycle. a controller provides a set of weights to the convolution engine before processing an input map, causes the convolution engine to scan across the input map by incrementing a specified location for successive blocks of pixels and generates an output map within the image cache by writing output pixel values to successive locations within the image cache.",2017-05-30,"The title of the patent is convolutional neural network and its abstract is a convolutional neural network (cnn) for an image processing system comprises an image cache responsive to a request to read a block of n×m pixels extending from a specified location within an input map to provide a block of n×m pixels at an output port. a convolution engine reads blocks of pixels from the output port, combines blocks of pixels with a corresponding set of weights to provide a product, and subjects the product to an activation function to provide an output pixel value. the image cache comprises a plurality of interleaved memories capable of simultaneously providing the n×m pixels at the output port in a single clock cycle. a controller provides a set of weights to the convolution engine before processing an input map, causes the convolution engine to scan across the input map by incrementing a specified location for successive blocks of pixels and generates an output map within the image cache by writing output pixel values to successive locations within the image cache. dated 2017-05-30"
9665802,object-centric fine-grained image classification,"systems and methods are disclosed for classifying vehicles by performing scale aware detection; performing detection assisted sampling for convolutional neural network (cnn) training, and performing deep cnn fine-grained image classification to classify the vehicle type.",2017-05-30,"The title of the patent is object-centric fine-grained image classification and its abstract is systems and methods are disclosed for classifying vehicles by performing scale aware detection; performing detection assisted sampling for convolutional neural network (cnn) training, and performing deep cnn fine-grained image classification to classify the vehicle type. dated 2017-05-30"
9665823,method and system for joint training of hybrid neural networks for acoustic modeling in automatic speech recognition,"systems and methods for training networks are provided. a method for training networks comprises receiving an input from each of a plurality of neural networks differing from each other in at least one of architecture, input modality, and feature type, connecting the plurality of neural networks through a common output layer, or through one or more common hidden layers and a common output layer to result in a joint network, and training the joint network.",2017-05-30,"The title of the patent is method and system for joint training of hybrid neural networks for acoustic modeling in automatic speech recognition and its abstract is systems and methods for training networks are provided. a method for training networks comprises receiving an input from each of a plurality of neural networks differing from each other in at least one of architecture, input modality, and feature type, connecting the plurality of neural networks through a common output layer, or through one or more common hidden layers and a common output layer to result in a joint network, and training the joint network. dated 2017-05-30"
9666184,method and apparatus for training language model and recognizing speech,"a method and apparatus for training a neural network language model, and a method and apparatus for recognizing speech data based on a trained language model are provided. the method of training a language model involves converting, using a processor, training data into error-containing training data, and training a neural network language model using the error-containing training data.",2017-05-30,"The title of the patent is method and apparatus for training language model and recognizing speech and its abstract is a method and apparatus for training a neural network language model, and a method and apparatus for recognizing speech data based on a trained language model are provided. the method of training a language model involves converting, using a processor, training data into error-containing training data, and training a neural network language model using the error-containing training data. dated 2017-05-30"
9668075,estimating parameter values for a lumped parameter model of a loudspeaker,"in one embodiment of the present invention, a loudspeaker parameter estimation subsystem efficiently and accurately estimates parameter values for a lumped parameter model (lpm) of a loudspeaker. in operation, the loudspeaker parameter estimation subsystem trains a neural network model based on responses generated via the lumped parameter model and the corresponding sets of parameter values. subsequently, based on the relationship between the measured output response of a loudspeaker to an input stimulus, the loudspeaker parameter estimation subsystem estimates parameter values for the lpm of the loudspeaker. advantageously, by sagaciously estimating parameter values for the lpm of loudspeakers, these nn-based techniques enable designers to leverage the lpm to reliably improve the design of loudspeakers, perform nonlinear correction of loudspeakers, and the like.",2017-05-30,"The title of the patent is estimating parameter values for a lumped parameter model of a loudspeaker and its abstract is in one embodiment of the present invention, a loudspeaker parameter estimation subsystem efficiently and accurately estimates parameter values for a lumped parameter model (lpm) of a loudspeaker. in operation, the loudspeaker parameter estimation subsystem trains a neural network model based on responses generated via the lumped parameter model and the corresponding sets of parameter values. subsequently, based on the relationship between the measured output response of a loudspeaker to an input stimulus, the loudspeaker parameter estimation subsystem estimates parameter values for the lpm of the loudspeaker. advantageously, by sagaciously estimating parameter values for the lpm of loudspeakers, these nn-based techniques enable designers to leverage the lpm to reliably improve the design of loudspeakers, perform nonlinear correction of loudspeakers, and the like. dated 2017-05-30"
9668699,method and system for anatomical object detection using marginal space deep neural networks,"a method and system for anatomical object detection using marginal space deep neural networks is disclosed. the pose parameter space for an anatomical object is divided into a series of marginal search spaces with increasing dimensionality. a respective deep neural network is trained for each of the marginal search spaces, resulting in a series of trained deep neural networks. each of the trained deep neural networks can evaluate hypotheses in a current parameter space using discriminative classification or a regression function. an anatomical object is detected in a medical image by sequentially applying the series of trained deep neural networks to the medical image.",2017-06-06,"The title of the patent is method and system for anatomical object detection using marginal space deep neural networks and its abstract is a method and system for anatomical object detection using marginal space deep neural networks is disclosed. the pose parameter space for an anatomical object is divided into a series of marginal search spaces with increasing dimensionality. a respective deep neural network is trained for each of the marginal search spaces, resulting in a series of trained deep neural networks. each of the trained deep neural networks can evaluate hypotheses in a current parameter space using discriminative classification or a regression function. an anatomical object is detected in a medical image by sequentially applying the series of trained deep neural networks to the medical image. dated 2017-06-06"
9669232,method for non-invasive brain stimulation,magneto-electric nanoparticles in a subject interact with an external magnetic field to cause stimulation of neural networks in the subject. electric signals in the neural network are coupled to magnetic dipoles induced in the nanoparticles to cause changes in electric pulse sequences of the subject's brain.,2017-06-06,The title of the patent is method for non-invasive brain stimulation and its abstract is magneto-electric nanoparticles in a subject interact with an external magnetic field to cause stimulation of neural networks in the subject. electric signals in the neural network are coupled to magnetic dipoles induced in the nanoparticles to cause changes in electric pulse sequences of the subject's brain. dated 2017-06-06
9672760,personalized eeg-based encryptor,"a user-specific, electroencephalogram data-based secure encryption generator maps artificial neural network neuron elements to electroencephalogram data signals generated from scanning neural activity of a user while the user executes a mental activity. weighting factors are trained to transform the electroencephalogram data signals into a first set of weighted signals that are different from weighted signals generated from scanning neural activity of the user while the user executes another, different activity, and from weighted signals generated from scanning neural activity of another user while executing a similar mental activity. the trained weighting factors are associated with the first set of electroencephalogram data signals and the current mental activity. thus, a reproducible electroencephalogram encryption key is defined that is unique to the user as a function of one or both of the trained weighting factors and the first weighted set of electroencephalogram data signals.",2017-06-06,"The title of the patent is personalized eeg-based encryptor and its abstract is a user-specific, electroencephalogram data-based secure encryption generator maps artificial neural network neuron elements to electroencephalogram data signals generated from scanning neural activity of a user while the user executes a mental activity. weighting factors are trained to transform the electroencephalogram data signals into a first set of weighted signals that are different from weighted signals generated from scanning neural activity of the user while the user executes another, different activity, and from weighted signals generated from scanning neural activity of another user while executing a similar mental activity. the trained weighting factors are associated with the first set of electroencephalogram data signals and the current mental activity. thus, a reproducible electroencephalogram encryption key is defined that is unique to the user as a function of one or both of the trained weighting factors and the first weighted set of electroencephalogram data signals. dated 2017-06-06"
9672814,semi-supervised learning of word embeddings,"software that trains an artificial neural network for generating vector representations for natural language text, by performing the following steps: (i) receiving, by one or more processors, a set of natural language text; (ii) generating, by one or more processors, a set of first metadata for the set of natural language text, where the first metadata is generated using supervised learning method(s); (iii) generating, by one or more processors, a set of second metadata for the set of natural language text, where the second metadata is generated using unsupervised learning method(s); and (iv) training, by one or more processors, an artificial neural network adapted to generate vector representations for natural language text, where the training is based, at least in part, on the received natural language text, the generated set of first metadata, and the generated set of second metadata.",2017-06-06,"The title of the patent is semi-supervised learning of word embeddings and its abstract is software that trains an artificial neural network for generating vector representations for natural language text, by performing the following steps: (i) receiving, by one or more processors, a set of natural language text; (ii) generating, by one or more processors, a set of first metadata for the set of natural language text, where the first metadata is generated using supervised learning method(s); (iii) generating, by one or more processors, a set of second metadata for the set of natural language text, where the second metadata is generated using unsupervised learning method(s); and (iv) training, by one or more processors, an artificial neural network adapted to generate vector representations for natural language text, where the training is based, at least in part, on the received natural language text, the generated set of first metadata, and the generated set of second metadata. dated 2017-06-06"
9678664,neural network for keyboard input decoding,"in some examples, a computing device includes at least one processor; and at least one module, operable by the at least one processor to: output, for display at an output device, a graphical keyboard; receive an indication of a gesture detected at a location of a presence-sensitive input device, wherein the location of the presence-sensitive input device corresponds to a location of the output device that outputs the graphical keyboard; determine, based on at least one spatial feature of the gesture that is processed by the computing device using a neural network, at least one character string, wherein the at least one spatial feature indicates at least one physical property of the gesture; and output, for display at the output device, based at least in part on the processing of the at least one spatial feature of the gesture using the neural network, the at least one character string.",2017-06-13,"The title of the patent is neural network for keyboard input decoding and its abstract is in some examples, a computing device includes at least one processor; and at least one module, operable by the at least one processor to: output, for display at an output device, a graphical keyboard; receive an indication of a gesture detected at a location of a presence-sensitive input device, wherein the location of the presence-sensitive input device corresponds to a location of the output device that outputs the graphical keyboard; determine, based on at least one spatial feature of the gesture that is processed by the computing device using a neural network, at least one character string, wherein the at least one spatial feature indicates at least one physical property of the gesture; and output, for display at the output device, based at least in part on the processing of the at least one spatial feature of the gesture using the neural network, the at least one character string. dated 2017-06-13"
9679243,system and method for detecting platform anomalies through neural networks,"a system and method for detecting behavior of a computing platform that includes obtaining platform data; for each data motif identifiers in a set data motif identifiers, performing data motif detection on data in an associated timescale, wherein a first data motif identifier operates on data in a first timescale, wherein a second data motif identifier operates on data in a second timescale, wherein the first timescale and second timescale are different; in a neural network model, synthesizing platform data anomaly detection with at least a set of features inputs from data motif detection of the set of motif identifiers; and signaling if a platform data anomaly is detected through the neural network model.",2017-06-13,"The title of the patent is system and method for detecting platform anomalies through neural networks and its abstract is a system and method for detecting behavior of a computing platform that includes obtaining platform data; for each data motif identifiers in a set data motif identifiers, performing data motif detection on data in an associated timescale, wherein a first data motif identifier operates on data in a first timescale, wherein a second data motif identifier operates on data in a second timescale, wherein the first timescale and second timescale are different; in a neural network model, synthesizing platform data anomaly detection with at least a set of features inputs from data motif detection of the set of motif identifiers; and signaling if a platform data anomaly is detected through the neural network model. dated 2017-06-13"
9679244,method for predicting quality or manufacturing condition of cement,"provided is a method capable of predicting the quality of cement in a short time period and with high accuracy. the method of predicting the quality or manufacturing conditions of cement through use of a neural network including an input layer and an output layer includes: performing learning of the neural network for a sufficiently large number of times of learning such that σl<σm is obtained, using learning data and monitor data; then repeating the learning of the neural network until σl≧σm is obtained while the number of times of learning is decreased; inputting specific observation data to the input layer of the neural network in which a judgment value for analysis degree obtained from the neural network after the learning is less than a preset value; and outputting an estimated value of specific evaluation data from the output layer of the neural network.",2017-06-13,"The title of the patent is method for predicting quality or manufacturing condition of cement and its abstract is provided is a method capable of predicting the quality of cement in a short time period and with high accuracy. the method of predicting the quality or manufacturing conditions of cement through use of a neural network including an input layer and an output layer includes: performing learning of the neural network for a sufficiently large number of times of learning such that σl<σm is obtained, using learning data and monitor data; then repeating the learning of the neural network until σl≧σm is obtained while the number of times of learning is decreased; inputting specific observation data to the input layer of the neural network in which a judgment value for analysis degree obtained from the neural network after the learning is less than a preset value; and outputting an estimated value of specific evaluation data from the output layer of the neural network. dated 2017-06-13"
9679258,methods and apparatus for reinforcement learning,we describe a method of reinforcement learning for a subject system having multiple states and actions to move from one state to the next. training data is generated by operating on the system with a succession of actions and used to train a second neural network. target values for training the second neural network are derived from a first neural network which is generated by copying weights of the second neural network at intervals.,2017-06-13,The title of the patent is methods and apparatus for reinforcement learning and its abstract is we describe a method of reinforcement learning for a subject system having multiple states and actions to move from one state to the next. training data is generated by operating on the system with a succession of actions and used to train a second neural network. target values for training the second neural network are derived from a first neural network which is generated by copying weights of the second neural network at intervals. dated 2017-06-13
9679259,systems and methods for training and employing a machine learning system in evaluating entity pairs,"a matching or pairing system and method for matching first and second entities having a greater likelihood of forming a successful pairing includes a trained machine learning system to provide heuristic values useful in determining a compatibility score for the pairing. during training of the machine learning system, a training example selection device can provide attribute values logically associated with entities engaged in historically successful pairings and a number of hypothetically successful pairings. the hypothetically successful pairings may be based at least in part on historically successful pairings where at least one attribute value logically associated with at least one entity in the pairing is varied, adjusted, or subjected to a loosened constraint. during run-time operation a screening device can screen unsuccessful pairings and forward potentially successful pairings that meet a threshold value to the neural network. the system can then determine a compatibility score for the pairing.",2017-06-13,"The title of the patent is systems and methods for training and employing a machine learning system in evaluating entity pairs and its abstract is a matching or pairing system and method for matching first and second entities having a greater likelihood of forming a successful pairing includes a trained machine learning system to provide heuristic values useful in determining a compatibility score for the pairing. during training of the machine learning system, a training example selection device can provide attribute values logically associated with entities engaged in historically successful pairings and a number of hypothetically successful pairings. the hypothetically successful pairings may be based at least in part on historically successful pairings where at least one attribute value logically associated with at least one entity in the pairing is varied, adjusted, or subjected to a loosened constraint. during run-time operation a screening device can screen unsuccessful pairings and forward potentially successful pairings that meet a threshold value to the neural network. the system can then determine a compatibility score for the pairing. dated 2017-06-13"
9681815,reconstruction of a surface electrocardiogram from an endocardial electrogram using non-linear filtering,"the present invention relates to an active medical device that uses non-linear filtering for reconstructing a surface electrocardiogram from an endocardial electrogram. at least one endocardial egm electrogram signal is collected from of samples collected from at least one endocardial or epicardial derivation (71′, 72′, 73′), and at least one of a reconstructed surface electrocardiogram (ecg) signal through the processing of collected egm samples by a transfer function (tf) of a neural network (60′). the neural network (60′) is a time-delay-type network that simultaneously processes said at least one endocardial egm electrogram signal, formed by a first sequence of collected samples, and at least one delayed version of this egm signal, formed by a second sequence of collected samples distinct from the first sequence collected samples. the neural network (60′) provides said reconstructed ecg signal from the egm signal and its delayed version.",2017-06-20,"The title of the patent is reconstruction of a surface electrocardiogram from an endocardial electrogram using non-linear filtering and its abstract is the present invention relates to an active medical device that uses non-linear filtering for reconstructing a surface electrocardiogram from an endocardial electrogram. at least one endocardial egm electrogram signal is collected from of samples collected from at least one endocardial or epicardial derivation (71′, 72′, 73′), and at least one of a reconstructed surface electrocardiogram (ecg) signal through the processing of collected egm samples by a transfer function (tf) of a neural network (60′). the neural network (60′) is a time-delay-type network that simultaneously processes said at least one endocardial egm electrogram signal, formed by a first sequence of collected samples, and at least one delayed version of this egm signal, formed by a second sequence of collected samples distinct from the first sequence collected samples. the neural network (60′) provides said reconstructed ecg signal from the egm signal and its delayed version. dated 2017-06-20"
9683989,optically sensitive cell network,a neural network is disclosed. the neural network comprises a plurality of optogenetically modified neural cells being three-dimensionally distributed in a hydrogel medium and being disconnected from any solid support having a shear modulus above 1 gpa.,2017-06-20,The title of the patent is optically sensitive cell network and its abstract is a neural network is disclosed. the neural network comprises a plurality of optogenetically modified neural cells being three-dimensionally distributed in a hydrogel medium and being disconnected from any solid support having a shear modulus above 1 gpa. dated 2017-06-20
9685155,method for distinguishing components of signal of environment,"a method distinguishes components of a signal by processing the signal to estimate a set of analysis features, wherein each analysis feature defines an element of the signal and has feature values that represent parts of the signal, processing the signal to estimate input features of the signal, and processing the input features using a deep neural network to assign an associative descriptor to each element of the signal, wherein a degree of similarity between the associative descriptors of different elements is related to a degree to which the parts of the signal represented by the elements belong to a single component of the signal. the similarities between associative descriptors are processed to estimate correspondences between the elements of the signal and the components in the signal. then, the signal is processed using the correspondences to distinguish component parts of the signal.",2017-06-20,"The title of the patent is method for distinguishing components of signal of environment and its abstract is a method distinguishes components of a signal by processing the signal to estimate a set of analysis features, wherein each analysis feature defines an element of the signal and has feature values that represent parts of the signal, processing the signal to estimate input features of the signal, and processing the input features using a deep neural network to assign an associative descriptor to each element of the signal, wherein a degree of similarity between the associative descriptors of different elements is related to a degree to which the parts of the signal represented by the elements belong to a single component of the signal. the similarities between associative descriptors are processed to estimate correspondences between the elements of the signal and the components in the signal. then, the signal is processed using the correspondences to distinguish component parts of the signal. dated 2017-06-20"
9686126,automotive neural network,network node modules within a vehicle are arranged to form a reconfigurable automotive neural network. each network node module includes one or more subsystems for performing one or more operations and a local processing module for communicating with the one or more subsystems. a management system enables traffic from the one or more subsystems of a particular network node module to be re-routed to an external processing module upon failure of the local processing module of that particular network node module.,2017-06-20,The title of the patent is automotive neural network and its abstract is network node modules within a vehicle are arranged to form a reconfigurable automotive neural network. each network node module includes one or more subsystems for performing one or more operations and a local processing module for communicating with the one or more subsystems. a management system enables traffic from the one or more subsystems of a particular network node module to be re-routed to an external processing module upon failure of the local processing module of that particular network node module. dated 2017-06-20
9691019,depth concatenation using a matrix computation unit,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for depth concatenation using a matrix computation unit. one of the methods includes: receiving a request to process network inputs to a neural network using an integrated circuit, the neural network comprising a depth concatenation neural network layer; and generating instructions that, when executed by the integrated circuit, cause the integrated circuit to performing operations comprising: for each spatial location in a first input tensor to the depth concatenation layer and a second input tensor to the depth concatenation layer: multiplying, using the matrix computation unit, a second depth vector for the spatial location by a shift weight matrix for the depth concatenation layer to generate a shifted second depth vector; and adding the shifted second depth vector and a first input depth vector for the spatial location to generate a concatenated depth vector.",2017-06-27,"The title of the patent is depth concatenation using a matrix computation unit and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for depth concatenation using a matrix computation unit. one of the methods includes: receiving a request to process network inputs to a neural network using an integrated circuit, the neural network comprising a depth concatenation neural network layer; and generating instructions that, when executed by the integrated circuit, cause the integrated circuit to performing operations comprising: for each spatial location in a first input tensor to the depth concatenation layer and a second input tensor to the depth concatenation layer: multiplying, using the matrix computation unit, a second depth vector for the spatial location by a shift weight matrix for the depth concatenation layer to generate a shifted second depth vector; and adding the shifted second depth vector and a first input depth vector for the spatial location to generate a concatenated depth vector. dated 2017-06-27"
9691020,"deep neural network learning method and apparatus, and category-independent sub-network learning apparatus","provided is a dnn learning method that can reduce dnn learning time using data belonging to a plurality of categories. the method includes the steps of training a language-independent sub-network 120 and language-dependent sub-networks 122 and 124 with training data of japanese and english. this step includes: a first step of training a dnn obtained by connecting neurons in an output layer of the sub-network 120 with neurons in an input layer of sub-network 122 with japanese training data; a step of forming a dnn by connecting the sub-network 124 in place of the sub-network 122 to the sub-network 120 and training it with english data; repeating these steps alternately until all training data ends; and after completion, separating the first sub-network 120 from other sub-networks and storing it as a category-independent sub-network in a storage medium.",2017-06-27,"The title of the patent is deep neural network learning method and apparatus, and category-independent sub-network learning apparatus and its abstract is provided is a dnn learning method that can reduce dnn learning time using data belonging to a plurality of categories. the method includes the steps of training a language-independent sub-network 120 and language-dependent sub-networks 122 and 124 with training data of japanese and english. this step includes: a first step of training a dnn obtained by connecting neurons in an output layer of the sub-network 120 with neurons in an input layer of sub-network 122 with japanese training data; a step of forming a dnn by connecting the sub-network 124 in place of the sub-network 122 to the sub-network 120 and training it with english data; repeating these steps alternately until all training data ends; and after completion, separating the first sub-network 120 from other sub-networks and storing it as a category-independent sub-network in a storage medium. dated 2017-06-27"
9691289,monotonous game-like task to promote effortless automatic recognition of sight words,"system and methods are provided to promote effortless automatic recognition of common sight words. a subject performs a game-like task that generates novel non-verbal visual stimuli that triggers visual attention shifts that enhance foveal and parafoveal recognition of non-verbal and verbal stimuli laterally presented in the right or left visual field. the present invention engages a shared motor-perceptual-cognitive neural network involving oculomotor, visuo-motor and selective executive cognitive behaviors on both brain hemispheres. the present invention has applications to a wide range of non-verbal pre-orthographic visual processes and early lexical processes, not only contributing to enabling reading fluency to dyslexics, reluctant and slow readers, but also to beginning readers. the present invention has wide applications in learning disabilities and normative individuals learning to read.",2017-06-27,"The title of the patent is monotonous game-like task to promote effortless automatic recognition of sight words and its abstract is system and methods are provided to promote effortless automatic recognition of common sight words. a subject performs a game-like task that generates novel non-verbal visual stimuli that triggers visual attention shifts that enhance foveal and parafoveal recognition of non-verbal and verbal stimuli laterally presented in the right or left visual field. the present invention engages a shared motor-perceptual-cognitive neural network involving oculomotor, visuo-motor and selective executive cognitive behaviors on both brain hemispheres. the present invention has applications to a wide range of non-verbal pre-orthographic visual processes and early lexical processes, not only contributing to enabling reading fluency to dyslexics, reluctant and slow readers, but also to beginning readers. the present invention has wide applications in learning disabilities and normative individuals learning to read. dated 2017-06-27"
9696699,self-organizing sensing and actuation for automatic control,"a self-organizing process control architecture is introduced with a sensing layer, control layer, actuation layer, process layer, as well as self-organizing sensors (sos) and self-organizing actuators (soa). a self-organizing sensor for a process variable with one or multiple input variables is disclosed. an artificial neural network (ann) based dynamic modeling mechanism as part of the self-organizing sensor is described. as a case example, a self-organizing soft-sensor for cfb boiler bed height is presented. also provided is a method to develop a self-organizing sensor.",2017-07-04,"The title of the patent is self-organizing sensing and actuation for automatic control and its abstract is a self-organizing process control architecture is introduced with a sensing layer, control layer, actuation layer, process layer, as well as self-organizing sensors (sos) and self-organizing actuators (soa). a self-organizing sensor for a process variable with one or multiple input variables is disclosed. an artificial neural network (ann) based dynamic modeling mechanism as part of the self-organizing sensor is described. as a case example, a self-organizing soft-sensor for cfb boiler bed height is presented. also provided is a method to develop a self-organizing sensor. dated 2017-07-04"
9697416,object detection using cascaded convolutional neural networks,"different candidate windows in an image are identified, such as by sliding a rectangular or other geometric shape of different sizes over an image to identify portions of the image (groups of pixels in the image). the candidate windows are analyzed by a set of convolutional neural networks, which are cascaded so that the input of one convolutional neural network layer is based on the input of another convolutional neural network layer. each convolutional neural network layer drops or rejects one or more candidate windows that the convolutional neural network layer determines does not include an object (e.g., a face). the candidate windows that are identified as including an object (e.g., a face) are analyzed by another one of the convolutional neural network layers. the candidate windows identified by the last of the convolutional neural network layers are the indications of the objects (e.g., faces) in the image.",2017-07-04,"The title of the patent is object detection using cascaded convolutional neural networks and its abstract is different candidate windows in an image are identified, such as by sliding a rectangular or other geometric shape of different sizes over an image to identify portions of the image (groups of pixels in the image). the candidate windows are analyzed by a set of convolutional neural networks, which are cascaded so that the input of one convolutional neural network layer is based on the input of another convolutional neural network layer. each convolutional neural network layer drops or rejects one or more candidate windows that the convolutional neural network layer determines does not include an object (e.g., a face). the candidate windows that are identified as including an object (e.g., a face) are analyzed by another one of the convolutional neural network layers. the candidate windows identified by the last of the convolutional neural network layers are the indications of the objects (e.g., faces) in the image. dated 2017-07-04"
9697444,convolutional-neural-network-based classifier and classifying method and training methods for the same,"the present invention relates to a convolutional-neural-network-based classifier, a classifying method by using a convolutional-neural-network-based classifier and a method for training the convolutional-neural-network-based classifier. the convolutional-neural-network-based classifier comprises: a plurality of feature map layers, at least one feature map in at least one of the plurality of feature map layers being divided into a plurality of regions; and a plurality of convolutional templates corresponding to the plurality of regions respectively, each of the convolutional templates being used for obtaining a response value of a neuron in the corresponding region.",2017-07-04,"The title of the patent is convolutional-neural-network-based classifier and classifying method and training methods for the same and its abstract is the present invention relates to a convolutional-neural-network-based classifier, a classifying method by using a convolutional-neural-network-based classifier and a method for training the convolutional-neural-network-based classifier. the convolutional-neural-network-based classifier comprises: a plurality of feature map layers, at least one feature map in at least one of the plurality of feature map layers being divided into a plurality of regions; and a plurality of convolutional templates corresponding to the plurality of regions respectively, each of the convolutional templates being used for obtaining a response value of a neuron in the corresponding region. dated 2017-07-04"
9697461,"universal, online learning in multi-modal perception-action semilattices","in one embodiment, the present invention provides a method for interconnecting neurons in a neural network. at least one node among a first set of nodes is interconnected to at least one node among a second set of nodes, and nodes of the first and second set are arranged in a lattice. at least one node of the first set represents a sensory-motor modality of the neural network. at least one node of the second set is a union of at least two nodes of the first set. each node in the lattice has an acyclic digraph comprising multiple vertices and directed edges. each vertex represents a neuron population. each directed edge comprises multiple synaptic connections. vertices in different acyclic digraphs are interconnected using an acyclic bottom-up digraph. the bottom-up digraph has a corresponding acyclic top-down digraph. vertices in the bottom-up digraph are interconnected to vertices in the top-down digraph.",2017-07-04,"The title of the patent is universal, online learning in multi-modal perception-action semilattices and its abstract is in one embodiment, the present invention provides a method for interconnecting neurons in a neural network. at least one node among a first set of nodes is interconnected to at least one node among a second set of nodes, and nodes of the first and second set are arranged in a lattice. at least one node of the first set represents a sensory-motor modality of the neural network. at least one node of the second set is a union of at least two nodes of the first set. each node in the lattice has an acyclic digraph comprising multiple vertices and directed edges. each vertex represents a neuron population. each directed edge comprises multiple synaptic connections. vertices in different acyclic digraphs are interconnected using an acyclic bottom-up digraph. the bottom-up digraph has a corresponding acyclic top-down digraph. vertices in the bottom-up digraph are interconnected to vertices in the top-down digraph. dated 2017-07-04"
9697462,synaptic time multiplexing,"a synaptic time-multiplexed (stm) neuromorphic network includes a neural fabric that includes nodes and switches to define inter-nodal connections between selected nodes of the neural fabric. the stm neuromorphic network further includes a neuromorphic controller to form subsets of a set of the inter-nodal connections representing a fully connected neural network. each subset is formed during a different time slot of a plurality of time slots of a time multiplexing cycle of the stm neuromorphic network. in combination, the inter-nodal connection subsets implement the fully connected neural network. a method of synaptic time multiplexing a neuromorphic network includes providing the neural fabric and forming the subsets of the set of inter-nodal connections.",2017-07-04,"The title of the patent is synaptic time multiplexing and its abstract is a synaptic time-multiplexed (stm) neuromorphic network includes a neural fabric that includes nodes and switches to define inter-nodal connections between selected nodes of the neural fabric. the stm neuromorphic network further includes a neuromorphic controller to form subsets of a set of the inter-nodal connections representing a fully connected neural network. each subset is formed during a different time slot of a plurality of time slots of a time multiplexing cycle of the stm neuromorphic network. in combination, the inter-nodal connection subsets implement the fully connected neural network. a method of synaptic time multiplexing a neuromorphic network includes providing the neural fabric and forming the subsets of the set of inter-nodal connections. dated 2017-07-04"
9697463,computing convolutions using a neural network processor,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for computing a layer output for a convolutional neural network layer, the method comprising: receiving the layer input, the layer input comprising a plurality of activation inputs, the plurality of activation inputs represented as a multi-dimensional matrix comprising a plurality of depth levels, each depth level being a respective matrix of distinct activation inputs from the plurality of activation inputs; sending each respective kernel matrix structure to a distinct cell along a first dimension of the systolic array; for each depth level, sending the respective matrix of distinct activation inputs to a distinct cell along a second dimension of the systolic array; causing the systolic array to generate an accumulated output from the respective matrices sent to the cells; and generating the layer output from the accumulated output.",2017-07-04,"The title of the patent is computing convolutions using a neural network processor and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for computing a layer output for a convolutional neural network layer, the method comprising: receiving the layer input, the layer input comprising a plurality of activation inputs, the plurality of activation inputs represented as a multi-dimensional matrix comprising a plurality of depth levels, each depth level being a respective matrix of distinct activation inputs from the plurality of activation inputs; sending each respective kernel matrix structure to a distinct cell along a first dimension of the systolic array; for each depth level, sending the respective matrix of distinct activation inputs to a distinct cell along a second dimension of the systolic array; causing the systolic array to generate an accumulated output from the respective matrices sent to the cells; and generating the layer output from the accumulated output. dated 2017-07-04"
9697826,processing multi-channel audio waveforms,"methods, including computer programs encoded on a computer storage medium, for enhancing the processing of audio waveforms for speech recognition using various neural network processing techniques. in one aspect, a method includes: receiving multiple channels of audio data corresponding to an utterance; convolving each of multiple filters, in a time domain, with each of the multiple channels of audio waveform data to generate convolution outputs, wherein the multiple filters have parameters that have been learned during a training process that jointly trains the multiple filters and trains a deep neural network as an acoustic model; combining, for each of the multiple filters, the convolution outputs for the filter for the multiple channels of audio waveform data; inputting the combined convolution outputs to the deep neural network trained jointly with the multiple filters; and providing a transcription for the utterance that is determined.",2017-07-04,"The title of the patent is processing multi-channel audio waveforms and its abstract is methods, including computer programs encoded on a computer storage medium, for enhancing the processing of audio waveforms for speech recognition using various neural network processing techniques. in one aspect, a method includes: receiving multiple channels of audio data corresponding to an utterance; convolving each of multiple filters, in a time domain, with each of the multiple channels of audio waveform data to generate convolution outputs, wherein the multiple filters have parameters that have been learned during a training process that jointly trains the multiple filters and trains a deep neural network as an acoustic model; combining, for each of the multiple filters, the convolution outputs for the filter for the multiple channels of audio waveform data; inputting the combined convolution outputs to the deep neural network trained jointly with the multiple filters; and providing a transcription for the utterance that is determined. dated 2017-07-04"
9697833,audio-visual speech recognition with scattering operators,"aspects described herein are directed towards methods, computing devices, systems, and computer-readable media that apply scattering operations to extracted visual features of audiovisual input to generate predictions regarding the speech status of a subject. visual scattering coefficients generated according to one or more aspects described herein may be used as input to a neural network operative to generate the predictions regarding the speech status of the subject. predictions generated based on the visual features may be combined with predictions based on audio input associated with the visual features. in some embodiments, the extracted visual features may be combined with the audio input to generate a combined feature vector for use in generating predictions.",2017-07-04,"The title of the patent is audio-visual speech recognition with scattering operators and its abstract is aspects described herein are directed towards methods, computing devices, systems, and computer-readable media that apply scattering operations to extracted visual features of audiovisual input to generate predictions regarding the speech status of a subject. visual scattering coefficients generated according to one or more aspects described herein may be used as input to a neural network operative to generate the predictions regarding the speech status of the subject. predictions generated based on the visual features may be combined with predictions based on audio input associated with the visual features. in some embodiments, the extracted visual features may be combined with the audio input to generate a combined feature vector for use in generating predictions. dated 2017-07-04"
9700219,method and system for machine learning based assessment of fractional flow reserve,"a method and system for determining fractional flow reserve (ffr) for a coronary artery stenosis of a patient is disclosed. in one embodiment, medical image data of the patient including the stenosis is received, a set of features for the stenosis is extracted from the medical image data of the patient, and an ffr value for the stenosis is determined based on the extracted set of features using a trained machine-learning based mapping. in another embodiment, a medical image of the patient including the stenosis of interest is received, image patches corresponding to the stenosis of interest and a coronary tree of the patient are detected, an ffr value for the stenosis of interest is determined using a trained deep neural network regressor applied directly to the detected image patches.",2017-07-11,"The title of the patent is method and system for machine learning based assessment of fractional flow reserve and its abstract is a method and system for determining fractional flow reserve (ffr) for a coronary artery stenosis of a patient is disclosed. in one embodiment, medical image data of the patient including the stenosis is received, a set of features for the stenosis is extracted from the medical image data of the patient, and an ffr value for the stenosis is determined based on the extracted set of features using a trained machine-learning based mapping. in another embodiment, a medical image of the patient including the stenosis of interest is received, image patches corresponding to the stenosis of interest and a coronary tree of the patient are detected, an ffr value for the stenosis of interest is determined using a trained deep neural network regressor applied directly to the detected image patches. dated 2017-07-11"
9704029,systems and methods for identifying users in media content based on poselets and neural networks,"systems, methods, and non-transitory computer-readable media can receive a first image including a representation of a first user. a second image including a representation of a second user can be received. a first set of poselets associated with the first user can be detected in the first image. a second set of poselets associated with the second user can be detected in the second image. the first image including the first set of poselets can be inputted into a first instance of a neural network to generate a first multi-dimensional vector. the second image including the second set of poselets can be inputted into a second instance of the neural network to generate a second multi-dimensional vector. a first distance metric between the first multi-dimensional vector and the second multi-dimensional vector can be determined.",2017-07-11,"The title of the patent is systems and methods for identifying users in media content based on poselets and neural networks and its abstract is systems, methods, and non-transitory computer-readable media can receive a first image including a representation of a first user. a second image including a representation of a second user can be received. a first set of poselets associated with the first user can be detected in the first image. a second set of poselets associated with the second user can be detected in the second image. the first image including the first set of poselets can be inputted into a first instance of a neural network to generate a first multi-dimensional vector. the second image including the second set of poselets can be inputted into a second instance of the neural network to generate a second multi-dimensional vector. a first distance metric between the first multi-dimensional vector and the second multi-dimensional vector can be determined. dated 2017-07-11"
9704068,system and method for labelling aerial images,a system and method for labelling aerial images. a neural network generates predicted map data. the parameters of the neural network are trained by optimizing an objective function which compensates for noise in the map images. the function compensates both omission noise and registration noise.,2017-07-11,The title of the patent is system and method for labelling aerial images and its abstract is a system and method for labelling aerial images. a neural network generates predicted map data. the parameters of the neural network are trained by optimizing an objective function which compensates for noise in the map images. the function compensates both omission noise and registration noise. dated 2017-07-11
9704097,automatically constructing training sets for electronic sentiment analysis,"training data for training a neural network usable for electronic sentiment analysis can be automatically constructed. for example, an electronic communication usable for training the neural network and including multiple characters can be received. a sentiment dictionary including multiple expressions mapped to multiple sentiment values representing different sentiments can be received. each expression in the sentiment dictionary can be mapped to a corresponding sentiment value. an overall sentiment for the electronic communication can be determined using the sentiment dictionary. training data usable for training the neural network can be automatically constructed based on the overall sentiment of the electronic communication. the neural network can be trained using the training data. a second electronic communication including an unknown sentiment can be received. at least one sentiment associated with the second electronic communication can be determined using the neural network.",2017-07-11,"The title of the patent is automatically constructing training sets for electronic sentiment analysis and its abstract is training data for training a neural network usable for electronic sentiment analysis can be automatically constructed. for example, an electronic communication usable for training the neural network and including multiple characters can be received. a sentiment dictionary including multiple expressions mapped to multiple sentiment values representing different sentiments can be received. each expression in the sentiment dictionary can be mapped to a corresponding sentiment value. an overall sentiment for the electronic communication can be determined using the sentiment dictionary. training data usable for training the neural network can be automatically constructed based on the overall sentiment of the electronic communication. the neural network can be trained using the training data. a second electronic communication including an unknown sentiment can be received. at least one sentiment associated with the second electronic communication can be determined using the neural network. dated 2017-07-11"
9704257,system and method for semantic segmentation using gaussian random field network,"a computer-implemented method for semantic segmentation of an image determines unary energy of each pixel in an image using a first subnetwork, determines pairwise energy of at least some pairs of pixels of the image using a second subnetwork, and determines, using a third subnetwork, an inference on a gaussian random field (grf) minimizing an energy function including a combination of the unary energy and the pairwise energy. the grf inference defining probabilities of semantic labels for each pixel in the image, and the method converts the image into a semantically segmented image by assigning to a pixel in the semantically segmented image a semantic label having the highest probability for a corresponding pixel in the image among the probabilities determined by the third subnetwork. the first subnetwork, the second subnetwork, and the third subnetwork are parts of a neural network.",2017-07-11,"The title of the patent is system and method for semantic segmentation using gaussian random field network and its abstract is a computer-implemented method for semantic segmentation of an image determines unary energy of each pixel in an image using a first subnetwork, determines pairwise energy of at least some pairs of pixels of the image using a second subnetwork, and determines, using a third subnetwork, an inference on a gaussian random field (grf) minimizing an energy function including a combination of the unary energy and the pairwise energy. the grf inference defining probabilities of semantic labels for each pixel in the image, and the method converts the image into a semantically segmented image by assigning to a pixel in the semantically segmented image a semantic label having the highest probability for a corresponding pixel in the image among the probabilities determined by the third subnetwork. the first subnetwork, the second subnetwork, and the third subnetwork are parts of a neural network. dated 2017-07-11"
9705904,neural attention mechanisms for malware analysis,"as part of an analysis of the likelihood that a given input (e.g. a file, etc.) includes malicious code, a convolutional neural network can be used to review a sequence of chunks into which an input is divided to assess how best to navigate through the input and to classify parts of the input in a most optimal manner. at least some of the sequence of chunks can be further examined using a recurrent neural network in series with the convolutional neural network to determine how to progress through the sequence of chunks. a state of the at least some of the chunks examined using the recurrent neural network summarized to form an output indicative of the likelihood that the input includes malicious code. methods, systems, and articles of manufacture are also described.",2017-07-11,"The title of the patent is neural attention mechanisms for malware analysis and its abstract is as part of an analysis of the likelihood that a given input (e.g. a file, etc.) includes malicious code, a convolutional neural network can be used to review a sequence of chunks into which an input is divided to assess how best to navigate through the input and to classify parts of the input in a most optimal manner. at least some of the sequence of chunks can be further examined using a recurrent neural network in series with the convolutional neural network to determine how to progress through the sequence of chunks. a state of the at least some of the chunks examined using the recurrent neural network summarized to form an output indicative of the likelihood that the input includes malicious code. methods, systems, and articles of manufacture are also described. dated 2017-07-11"
9710265,neural network compute tile,"a computing unit is disclosed, comprising a first memory bank for storing input activations and a second memory bank for storing parameters used in performing computations. the computing unit includes at least one cell comprising at least one multiply accumulate (“mac”) operator that receives parameters from the second memory bank and performs computations. the computing unit further includes a first traversal unit that provides a control signal to the first memory bank to cause an input activation to be provided to a data bus accessible by the mac operator. the computing unit performs one or more computations associated with at least one element of a data array, the one or more computations being performed by the mac operator and comprising, in part, a multiply operation of the input activation received from the data bus and a parameter received from the second memory bank.",2017-07-18,"The title of the patent is neural network compute tile and its abstract is a computing unit is disclosed, comprising a first memory bank for storing input activations and a second memory bank for storing parameters used in performing computations. the computing unit includes at least one cell comprising at least one multiply accumulate (“mac”) operator that receives parameters from the second memory bank and performs computations. the computing unit further includes a first traversal unit that provides a control signal to the first memory bank to cause an input activation to be provided to a data bus accessible by the mac operator. the computing unit performs one or more computations associated with at least one element of a data array, the one or more computations being performed by the mac operator and comprising, in part, a multiply operation of the input activation received from the data bus and a parameter received from the second memory bank. dated 2017-07-18"
9710695,characterizing pathology images with statistical analysis of local neural network responses,"for digital pathology imaging, intelligent processing, such as automatic recognition or content-based retrieval, is one significant benefit that drives the wide application of this technology. before any intelligent processing on pathology images, every image is converted into a feature vector which quantitatively capture its visual characteristics. an algorithm characterizing pathology images with statistical analysis of local responses of neural networks is described herein. the algorithm framework enables extracting sophisticated textural features that are well adapted to the image data of interest.",2017-07-18,"The title of the patent is characterizing pathology images with statistical analysis of local neural network responses and its abstract is for digital pathology imaging, intelligent processing, such as automatic recognition or content-based retrieval, is one significant benefit that drives the wide application of this technology. before any intelligent processing on pathology images, every image is converted into a feature vector which quantitatively capture its visual characteristics. an algorithm characterizing pathology images with statistical analysis of local responses of neural networks is described herein. the algorithm framework enables extracting sophisticated textural features that are well adapted to the image data of interest. dated 2017-07-18"
9710748,neural network processor,"a circuit for performing neural network computations for a neural network comprising a plurality of neural network layers, the circuit comprising: a matrix computation unit configured to, for each of the plurality of neural network layers: receive a plurality of weight inputs and a plurality of activation inputs for the neural network layer, and generate a plurality of accumulated values based on the plurality of weight inputs and the plurality of activation inputs; and a vector computation unit communicatively coupled to the matrix computation unit and configured to, for each of the plurality of neural network layers: apply an activation function to each accumulated value generated by the matrix computation unit to generate a plurality of activated values for the neural network layer.",2017-07-18,"The title of the patent is neural network processor and its abstract is a circuit for performing neural network computations for a neural network comprising a plurality of neural network layers, the circuit comprising: a matrix computation unit configured to, for each of the plurality of neural network layers: receive a plurality of weight inputs and a plurality of activation inputs for the neural network layer, and generate a plurality of accumulated values based on the plurality of weight inputs and the plurality of activation inputs; and a vector computation unit communicatively coupled to the matrix computation unit and configured to, for each of the plurality of neural network layers: apply an activation function to each accumulated value generated by the matrix computation unit to generate a plurality of activated values for the neural network layer. dated 2017-07-18"
9710910,"image registration device, image registration method, and ultrasonic diagnosis apparatus having image registration device","there is provided an image registration device and an image registration method. the device includes: a feature extractor configured to extract, from a first image, a first feature group and to extract, from a second image, a second feature group; a feature converter configured to convert, using a converted neural network in which a correlation between features is learned, the extracted second feature group to correspond to the extracted first feature group, to obtain a converted group; and a register configured to register the first image and the second image based on the converted group and the extracted first feature group.",2017-07-18,"The title of the patent is image registration device, image registration method, and ultrasonic diagnosis apparatus having image registration device and its abstract is there is provided an image registration device and an image registration method. the device includes: a feature extractor configured to extract, from a first image, a first feature group and to extract, from a second image, a second feature group; a feature converter configured to convert, using a converted neural network in which a correlation between features is learned, the extracted second feature group to correspond to the extracted first feature group, to obtain a converted group; and a register configured to register the first image and the second image based on the converted group and the extracted first feature group. dated 2017-07-18"
9712146,mixed signal processors,"various processor architectures for mixed signal computation exploit the unique characteristics of advanced cmos technologies, such as fin-based, multi-gate field effect transistors, and/or emerging technologies such as tunnel field effect transistors (tfets). the example processors disclosed herein are cellular neural network (cnn)-inspired and eliminate the need for voltage controlled current sources (vccss), which have previously been utilized to realize feedback and feed-forward templates in cnns and are the dominant source of power consumption in a cnn array. the example processors replace vccss with comparators, which can be efficiently realized with tfets given their high intrinsic gain. power efficiencies are in the order of 10,000 giga-operations per second per watt (gops/w), which represents an improvement of more than ten times over state-of-the-art architectures seeking to accomplish similar information processing tasks.",2017-07-18,"The title of the patent is mixed signal processors and its abstract is various processor architectures for mixed signal computation exploit the unique characteristics of advanced cmos technologies, such as fin-based, multi-gate field effect transistors, and/or emerging technologies such as tunnel field effect transistors (tfets). the example processors disclosed herein are cellular neural network (cnn)-inspired and eliminate the need for voltage controlled current sources (vccss), which have previously been utilized to realize feedback and feed-forward templates in cnns and are the dominant source of power consumption in a cnn array. the example processors replace vccss with comparators, which can be efficiently realized with tfets given their high intrinsic gain. power efficiencies are in the order of 10,000 giga-operations per second per watt (gops/w), which represents an improvement of more than ten times over state-of-the-art architectures seeking to accomplish similar information processing tasks. dated 2017-07-18"
9714885,fault prediction and condition-based repair method of urban rail train bogie,"the present invention provides a fault prediction and condition-based repair method of an urban rail train bogie. an optimum service life distribution model of a framework, a spring device, a connecting device, a wheel set and axle box, a driving mechanism, and a basic brake device of a bogie is determined by adopting a method based on survival analysis; a reliability characteristic function of each subsystem is obtained; then, a failure rate of each subsystem of the bogie is calculated by adopting a neural network model optimized by an evolutionary algorithm; and finally, proportional risk modelling is conducted by taking the failure rate and safe operation days of each subsystem of the bogie as concomitant variables; and on the basis of cost optimization, thresholds and control limits for condition-based repair of a bogie system are obtained.",2017-07-25,"The title of the patent is fault prediction and condition-based repair method of urban rail train bogie and its abstract is the present invention provides a fault prediction and condition-based repair method of an urban rail train bogie. an optimum service life distribution model of a framework, a spring device, a connecting device, a wheel set and axle box, a driving mechanism, and a basic brake device of a bogie is determined by adopting a method based on survival analysis; a reliability characteristic function of each subsystem is obtained; then, a failure rate of each subsystem of the bogie is calculated by adopting a neural network model optimized by an evolutionary algorithm; and finally, proportional risk modelling is conducted by taking the failure rate and safe operation days of each subsystem of the bogie as concomitant variables; and on the basis of cost optimization, thresholds and control limits for condition-based repair of a bogie system are obtained. dated 2017-07-25"
9715642,processing images using deep neural networks,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for image processing using deep neural networks. one of the methods includes receiving data characterizing an input image; processing the data characterizing the input image using a deep neural network to generate an alternative representation of the input image, wherein the deep neural network comprises a plurality of subnetworks, wherein the subnetworks are arranged in a sequence from lowest to highest, and wherein processing the data characterizing the input image using the deep neural network comprises processing the data through each of the subnetworks in the sequence; and processing the alternative representation of the input image through an output layer to generate an output from the input image.",2017-07-25,"The title of the patent is processing images using deep neural networks and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for image processing using deep neural networks. one of the methods includes receiving data characterizing an input image; processing the data characterizing the input image using a deep neural network to generate an alternative representation of the input image, wherein the deep neural network comprises a plurality of subnetworks, wherein the subnetworks are arranged in a sequence from lowest to highest, and wherein processing the data characterizing the input image using the deep neural network comprises processing the data through each of the subnetworks in the sequence; and processing the alternative representation of the input image through an output layer to generate an output from the input image. dated 2017-07-25"
9715653,multi-scale spatio-temporal neural network system,"embodiments of the invention relate to a multi-scale spatio-temporal neural network system. one embodiment comprises a neural network including multiple heterogeneous neuron populations that operate at different time scales. each neuron population comprises at least one digital neuron. each neuron population further comprises a time scale generation circuit that controls timing for operation of said neuron population, wherein each neuron of said neuron population integrates neuronal firing events at a time scale corresponding to said neuron population. the neural network further comprises a plurality of synapses interconnecting the neurons, wherein each synapse interconnects a neuron with another neuron. at least one neuron receives neuronal firing events from an interconnected neuron that operates at a different time scale.",2017-07-25,"The title of the patent is multi-scale spatio-temporal neural network system and its abstract is embodiments of the invention relate to a multi-scale spatio-temporal neural network system. one embodiment comprises a neural network including multiple heterogeneous neuron populations that operate at different time scales. each neuron population comprises at least one digital neuron. each neuron population further comprises a time scale generation circuit that controls timing for operation of said neuron population, wherein each neuron of said neuron population integrates neuronal firing events at a time scale corresponding to said neuron population. the neural network further comprises a plurality of synapses interconnecting the neurons, wherein each synapse interconnects a neuron with another neuron. at least one neuron receives neuronal firing events from an interconnected neuron that operates at a different time scale. dated 2017-07-25"
9715654,multi-scale spatio-temporal neural network system,"embodiments of the invention relate to a multi-scale spatio-temporal neural network system. one embodiment comprises a neural network including multiple heterogeneous neuron populations that operate at different time scales. each neuron population comprises at least one digital neuron. each neuron population further comprises a time scale generation circuit that controls timing for operation of said neuron population, wherein each neuron of said neuron population integrates neuronal firing events at a time scale corresponding to said neuron population. the neural network further comprises a plurality of synapses interconnecting the neurons, wherein each synapse interconnects a neuron with another neuron. at least one neuron receives neuronal firing events from an interconnected neuron that operates at a different time scale.",2017-07-25,"The title of the patent is multi-scale spatio-temporal neural network system and its abstract is embodiments of the invention relate to a multi-scale spatio-temporal neural network system. one embodiment comprises a neural network including multiple heterogeneous neuron populations that operate at different time scales. each neuron population comprises at least one digital neuron. each neuron population further comprises a time scale generation circuit that controls timing for operation of said neuron population, wherein each neuron of said neuron population integrates neuronal firing events at a time scale corresponding to said neuron population. the neural network further comprises a plurality of synapses interconnecting the neurons, wherein each synapse interconnects a neuron with another neuron. at least one neuron receives neuronal firing events from an interconnected neuron that operates at a different time scale. dated 2017-07-25"
9715656,killing asymmetric resistive processing units for neural network training,"technical solutions are described for improving efficiency of training a resistive processing unit (rpu) array using a neural network training methodology. an example method includes reducing asymmetric rpus from the rpu array by determining an asymmetric value of an rpu from the rpu array, and burning the rpu in response to the asymmetry value being above a predetermined threshold. the rpu can be burned by causing an electric voltage across the rpu to be above a predetermined limit. the method further includes initiating the training methodology for the rpu array after the asymmetric rpus from the rpu array are reduced.",2017-07-25,"The title of the patent is killing asymmetric resistive processing units for neural network training and its abstract is technical solutions are described for improving efficiency of training a resistive processing unit (rpu) array using a neural network training methodology. an example method includes reducing asymmetric rpus from the rpu array by determining an asymmetric value of an rpu from the rpu array, and burning the rpu in response to the asymmetry value being above a predetermined threshold. the rpu can be burned by causing an electric voltage across the rpu to be above a predetermined limit. the method further includes initiating the training methodology for the rpu array after the asymmetric rpus from the rpu array are reduced. dated 2017-07-25"
9715660,transfer learning for deep neural network based hotword detection,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a deep neural network. one of the methods includes training a deep neural network with a first training set by adjusting values for each of a plurality of weights included in the neural network, and training the deep neural network to determine a probability that data received by the deep neural network has features similar to key features of one or more keywords or key phrases, the training comprising providing the deep neural network with a second training set and adjusting the values for a first subset of the plurality of weights, wherein the second training set includes data representing the key features of the one or more keywords or key phrases.",2017-07-25,"The title of the patent is transfer learning for deep neural network based hotword detection and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a deep neural network. one of the methods includes training a deep neural network with a first training set by adjusting values for each of a plurality of weights included in the neural network, and training the deep neural network to determine a probability that data received by the deep neural network has features similar to key features of one or more keywords or key phrases, the training comprising providing the deep neural network with a second training set and adjusting the values for a first subset of the plurality of weights, wherein the second training set includes data representing the key features of the one or more keywords or key phrases. dated 2017-07-25"
9721097,neural attention mechanisms for malware analysis,"as part of an analysis of the likelihood that a given input (e.g. a file, etc.) includes malicious code, a convolutional neural network can be used to review a sequence of chunks into which an input is divided to assess how best to navigate through the input and to classify parts of the input in a most optimal manner. at least some of the sequence of chunks can be further examined using a recurrent neural network in series with the convolutional neural network to determine how to progress through the sequence of chunks. a state of the at least some of the chunks examined using the recurrent neural network summarized to form an output indicative of the likelihood that the input includes malicious code. methods, systems, and articles of manufacture are also described.",2017-08-01,"The title of the patent is neural attention mechanisms for malware analysis and its abstract is as part of an analysis of the likelihood that a given input (e.g. a file, etc.) includes malicious code, a convolutional neural network can be used to review a sequence of chunks into which an input is divided to assess how best to navigate through the input and to classify parts of the input in a most optimal manner. at least some of the sequence of chunks can be further examined using a recurrent neural network in series with the convolutional neural network to determine how to progress through the sequence of chunks. a state of the at least some of the chunks examined using the recurrent neural network summarized to form an output indicative of the likelihood that the input includes malicious code. methods, systems, and articles of manufacture are also described. dated 2017-08-01"
9721190,large-scale classification in neural networks using hashing,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for classification using a neural network. one of the methods for processing an input through each of multiple layers of a neural network to generate an output, wherein each of the multiple layers of the neural network includes a respective multiple nodes includes for a particular layer of the multiple layers: receiving, by a classification system, an activation vector as input for the particular layer, selecting one or more nodes in the particular layer using the activation vector and a hash table that maps numeric values to nodes in the particular layer, and processing the activation vector using the selected nodes to generate an output for the particular layer.",2017-08-01,"The title of the patent is large-scale classification in neural networks using hashing and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for classification using a neural network. one of the methods for processing an input through each of multiple layers of a neural network to generate an output, wherein each of the multiple layers of the neural network includes a respective multiple nodes includes for a particular layer of the multiple layers: receiving, by a classification system, an activation vector as input for the particular layer, selecting one or more nodes in the particular layer using the activation vector and a hash table that maps numeric values to nodes in the particular layer, and processing the activation vector using the selected nodes to generate an output for the particular layer. dated 2017-08-01"
9721202,non-negative matrix factorization regularized by recurrent neural networks for audio processing,"sound processing techniques using recurrent neural networks are described. in one or more implementations, temporal dependencies are captured in sound data that are modeled through use of a recurrent neural network (rnn). the captured temporal dependencies are employed as part of feature extraction performed using nonnegative matrix factorization (nmf). one or more sound processing techniques are performed on the sound data based at least in part on the feature extraction.",2017-08-01,"The title of the patent is non-negative matrix factorization regularized by recurrent neural networks for audio processing and its abstract is sound processing techniques using recurrent neural networks are described. in one or more implementations, temporal dependencies are captured in sound data that are modeled through use of a recurrent neural network (rnn). the captured temporal dependencies are employed as part of feature extraction performed using nonnegative matrix factorization (nmf). one or more sound processing techniques are performed on the sound data based at least in part on the feature extraction. dated 2017-08-01"
9721203,performing kernel striding in hardware,"methods for receiving a request to process, on a hardware circuit, a neural network comprising a first convolutional neural network layer having a stride greater than one, and in response, generating instructions that cause the hardware circuit to, during processing of an input tensor, generate a layer output tensor equivalent to an output of the first convolutional neural network layer by processing the input tensor using a second convolutional neural network layer having a stride equal to one but that is otherwise equivalent to the first convolutional neural network layer to generate a first tensor, zeroing out elements of the first tensor that would not have been generated if the second convolutional neural network layer had the stride of the first convolutional neural network layer to generate a second tensor, and performing max pooling on the second tensor to generate the layer output tensor.",2017-08-01,"The title of the patent is performing kernel striding in hardware and its abstract is methods for receiving a request to process, on a hardware circuit, a neural network comprising a first convolutional neural network layer having a stride greater than one, and in response, generating instructions that cause the hardware circuit to, during processing of an input tensor, generate a layer output tensor equivalent to an output of the first convolutional neural network layer by processing the input tensor using a second convolutional neural network layer having a stride equal to one but that is otherwise equivalent to the first convolutional neural network layer to generate a first tensor, zeroing out elements of the first tensor that would not have been generated if the second convolutional neural network layer had the stride of the first convolutional neural network layer to generate a second tensor, and performing max pooling on the second tensor to generate the layer output tensor. dated 2017-08-01"
9721204,evaluation of a system including separable sub-systems over a multidimensional range,"an artificial neural network may be configured to test the impact of certain input parameters. to improve testing efficiency and to avoid test runs that may not alter system performance, the effect of input parameters on neurons or groups of neurons may be determined to classify the neurons into groups based on the impact of certain parameters on those groups. groups may be ordered serially and/or in parallel based on the interconnected nature of the groups and whether the output of neurons in one group may affect the operation of another. parameters not affecting group performance may be pruned as inputs to that particular group prior to running system tests, thereby conserving processing resources during testing.",2017-08-01,"The title of the patent is evaluation of a system including separable sub-systems over a multidimensional range and its abstract is an artificial neural network may be configured to test the impact of certain input parameters. to improve testing efficiency and to avoid test runs that may not alter system performance, the effect of input parameters on neurons or groups of neurons may be determined to classify the neurons into groups based on the impact of certain parameters on those groups. groups may be ordered serially and/or in parallel based on the interconnected nature of the groups and whether the output of neurons in one group may affect the operation of another. parameters not affecting group performance may be pruned as inputs to that particular group prior to running system tests, thereby conserving processing resources during testing. dated 2017-08-01"
9721562,generating representations of acoustic sequences,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating representation of acoustic sequences. one of the methods includes: receiving an acoustic sequence, the acoustic sequence comprising a respective acoustic feature representation at each of a plurality of time steps; processing the acoustic feature representation at an initial time step using an acoustic modeling neural network; for each subsequent time step of the plurality of time steps: receiving an output generated by the acoustic modeling neural network for a preceding time step, generating a modified input from the output generated by the acoustic modeling neural network for the preceding time step and the acoustic representation for the time step, and processing the modified input using the acoustic modeling neural network to generate an output for the time step; and generating a phoneme representation for the utterance from the outputs for each of the time steps.",2017-08-01,"The title of the patent is generating representations of acoustic sequences and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating representation of acoustic sequences. one of the methods includes: receiving an acoustic sequence, the acoustic sequence comprising a respective acoustic feature representation at each of a plurality of time steps; processing the acoustic feature representation at an initial time step using an acoustic modeling neural network; for each subsequent time step of the plurality of time steps: receiving an output generated by the acoustic modeling neural network for a preceding time step, generating a modified input from the output generated by the acoustic modeling neural network for the preceding time step and the acoustic representation for the time step, and processing the modified input using the acoustic modeling neural network to generate an output for the time step; and generating a phoneme representation for the utterance from the outputs for each of the time steps. dated 2017-08-01"
9724068,method for measuring intracranial elasticity,a novel method to noninvasively measure intracranial pressure (icp) and more generally brain elasticity is disclosed. icp is determined using an algorithm coupled on a simulated artificial neural network (sann) that calculates icp based on a determination of a set of interacted ultrasound signals (iuss) generated from multiple ultrasound pulses. the methods and systems of the present invention are capable of rapidly determining icp without manual review of epg waves by a technician.,2017-08-08,The title of the patent is method for measuring intracranial elasticity and its abstract is a novel method to noninvasively measure intracranial pressure (icp) and more generally brain elasticity is disclosed. icp is determined using an algorithm coupled on a simulated artificial neural network (sann) that calculates icp based on a determination of a set of interacted ultrasound signals (iuss) generated from multiple ultrasound pulses. the methods and systems of the present invention are capable of rapidly determining icp without manual review of epg waves by a technician. dated 2017-08-08
9727803,systems and methods for image object recognition based on location information and object categories,"systems, methods, and non-transitory computer-readable media can identify a set of regions corresponding to a geographical area. a collection of training images can be acquired. each training image in the collection can be associated with one or more respective recognized objects and with a respective region in the set of regions. histogram metrics for a plurality of object categories within each region in the set of regions can be determined based at least in part on the collection of training images. a neural network can be developed based at least in part on the histogram metrics for the plurality of object categories within each region in the set of regions and on the collection of training images.",2017-08-08,"The title of the patent is systems and methods for image object recognition based on location information and object categories and its abstract is systems, methods, and non-transitory computer-readable media can identify a set of regions corresponding to a geographical area. a collection of training images can be acquired. each training image in the collection can be associated with one or more respective recognized objects and with a respective region in the set of regions. histogram metrics for a plurality of object categories within each region in the set of regions can be determined based at least in part on the collection of training images. a neural network can be developed based at least in part on the histogram metrics for the plurality of object categories within each region in the set of regions and on the collection of training images. dated 2017-08-08"
9728184,restructuring deep neural network acoustic models,"a deep neural network (dnn) model used in an automatic speech recognition (asr) system is restructured. a restructured dnn model may include fewer parameters compared to the original dnn model. the restructured dnn model may include a monophone state output layer in addition to the senone output layer of the original dnn model. singular value decomposition (svd) can be applied to one or more weight matrices of the dnn model to reduce the size of the dnn model. the output layer of the dnn model may be restructured to include monophone states in addition to the senones (tied triphone states) which are included in the original dnn model. when the monophone states are included in the restructured dnn model, the posteriors of monophone states are used to select a small part of senones to be evaluated.",2017-08-08,"The title of the patent is restructuring deep neural network acoustic models and its abstract is a deep neural network (dnn) model used in an automatic speech recognition (asr) system is restructured. a restructured dnn model may include fewer parameters compared to the original dnn model. the restructured dnn model may include a monophone state output layer in addition to the senone output layer of the original dnn model. singular value decomposition (svd) can be applied to one or more weight matrices of the dnn model to reduce the size of the dnn model. the output layer of the dnn model may be restructured to include monophone states in addition to the senones (tied triphone states) which are included in the original dnn model. when the monophone states are included in the restructured dnn model, the posteriors of monophone states are used to select a small part of senones to be evaluated. dated 2017-08-08"
9730098,knowledge discovery and data mining-assisted multi-radio access technology control,"joint provisioning of cell sector capacity (csc) and a defined customer service level (csl) is provided utilizing a knowledge discovery and data mining-assisted multi-radio access technology controller. for example, strategic performance indexes, csc and csl, are identified. the relationships between csc, csl, ergodic channel capacity (ecc) and an interface load (il) for a radio network (rn) (or cell sector of an rn) are determined. extensive information associated with the rn is collected and neural networks analysis is employed to reduce the information to a manageable set including the specific information associated with ecc and il. the reduced set of information is mapped to ecc and il using eigenvalue analysis, and the relationships between the ecc, il, csc and csl are employed to determine the csc and csl for the rn (or cell sector of the rn). network assignments and/or parameters can be updated based on the results.",2017-08-08,"The title of the patent is knowledge discovery and data mining-assisted multi-radio access technology control and its abstract is joint provisioning of cell sector capacity (csc) and a defined customer service level (csl) is provided utilizing a knowledge discovery and data mining-assisted multi-radio access technology controller. for example, strategic performance indexes, csc and csl, are identified. the relationships between csc, csl, ergodic channel capacity (ecc) and an interface load (il) for a radio network (rn) (or cell sector of an rn) are determined. extensive information associated with the rn is collected and neural networks analysis is employed to reduce the information to a manageable set including the specific information associated with ecc and il. the reduced set of information is mapped to ecc and il using eigenvalue analysis, and the relationships between the ecc, il, csc and csl are employed to determine the csc and csl for the rn (or cell sector of the rn). network assignments and/or parameters can be updated based on the results. dated 2017-08-08"
9730643,method and system for anatomical object detection using marginal space deep neural networks,"a method and system for anatomical object detection using marginal space deep neural networks is disclosed. the pose parameter space for an anatomical object is divided into a series of marginal search spaces with increasing dimensionality. a respective sparse deep neural network is trained for each of the marginal search spaces, resulting in a series of trained sparse deep neural networks. each of the trained sparse deep neural networks is trained by injecting sparsity into a deep neural network by removing filter weights of the deep neural network.",2017-08-15,"The title of the patent is method and system for anatomical object detection using marginal space deep neural networks and its abstract is a method and system for anatomical object detection using marginal space deep neural networks is disclosed. the pose parameter space for an anatomical object is divided into a series of marginal search spaces with increasing dimensionality. a respective sparse deep neural network is trained for each of the marginal search spaces, resulting in a series of trained sparse deep neural networks. each of the trained sparse deep neural networks is trained by injecting sparsity into a deep neural network by removing filter weights of the deep neural network. dated 2017-08-15"
9734436,hash codes for images,"a method includes receiving, at a neural network, a subset of images of a plurality of images of a training image set. the method includes training the neural network by iteratively adjusting parameters of the neural network based on concurrent application of multiple loss functions to the subset of images. the multiple loss functions include a classification loss function and a hashing loss function. the classification loss function is associated with an image classification function that extracts image features from an image. the hashing loss function is associated with a hashing function that generates a hash code for the image.",2017-08-15,"The title of the patent is hash codes for images and its abstract is a method includes receiving, at a neural network, a subset of images of a plurality of images of a training image set. the method includes training the neural network by iteratively adjusting parameters of the neural network based on concurrent application of multiple loss functions to the subset of images. the multiple loss functions include a classification loss function and a hashing loss function. the classification loss function is associated with an image classification function that extracts image features from an image. the hashing loss function is associated with a hashing function that generates a hash code for the image. dated 2017-08-15"
9734447,generating accurate reason codes with complex non-linear modeling and neural networks,"a computer system computes a score for a received data exchange and, in accordance with a neural network and input variables determined by received current exchange and history data, the computed score indicates a condition suitable for a denial. a set of attribution scores are computed using an alternating decision tree model in response to a computed score that is greater than a predetermined score threshold value for the denial. the computed score is provided to an assessment unit and, if the computed score indicates a condition suitable for the denial and if attribution scores are computed, then a predetermined number of input variable categories from a rank-ordered list of input variable categories is also provided to the assessment unit of the computer system.",2017-08-15,"The title of the patent is generating accurate reason codes with complex non-linear modeling and neural networks and its abstract is a computer system computes a score for a received data exchange and, in accordance with a neural network and input variables determined by received current exchange and history data, the computed score indicates a condition suitable for a denial. a set of attribution scores are computed using an alternating decision tree model in response to a computed score that is greater than a predetermined score threshold value for the denial. the computed score is provided to an assessment unit and, if the computed score indicates a condition suitable for the denial and if attribution scores are computed, then a predetermined number of input variable categories from a rank-ordered list of input variable categories is also provided to the assessment unit of the computer system. dated 2017-08-15"
9734567,label-free non-reference image quality assessment via deep neural network,"a method for training a neural network to perform assessments of image quality is provided. the method includes: inputting into the neural network at least one set of images, each set including an image and at least one degraded version of the image; performing comparative ranking of each image in the at least one set of images; and training the neural network with the ranking information. a neural network and image signal processing tuning system are disclosed.",2017-08-15,"The title of the patent is label-free non-reference image quality assessment via deep neural network and its abstract is a method for training a neural network to perform assessments of image quality is provided. the method includes: inputting into the neural network at least one set of images, each set including an image and at least one degraded version of the image; performing comparative ranking of each image in the at least one set of images; and training the neural network with the ranking information. a neural network and image signal processing tuning system are disclosed. dated 2017-08-15"
9734824,system and method for applying a convolutional neural network to speech recognition,a system and method for applying a convolutional neural network (cnn) to speech recognition. the cnn may provide input to a hidden markov model and has at least one pair of a convolution layer and a pooling layer. the cnn operates along the frequency axis. the cnn has units that operate upon one or more local frequency bands of an acoustic signal. the cnn mitigates acoustic variation.,2017-08-15,The title of the patent is system and method for applying a convolutional neural network to speech recognition and its abstract is a system and method for applying a convolutional neural network (cnn) to speech recognition. the cnn may provide input to a hidden markov model and has at least one pair of a convolution layer and a pooling layer. the cnn operates along the frequency axis. the cnn has units that operate upon one or more local frequency bands of an acoustic signal. the cnn mitigates acoustic variation. dated 2017-08-15
9740214,nonlinear model predictive control for chemical looping process,"a control system for optimizing a chemical looping (“cl”) plant includes a reduced order mathematical model (“rom”) that is designed by eliminating mathematical terms that have minimal effect on the outcome. a non-linear optimizer provides various inputs to the rom and monitors the outputs to determine the optimum inputs that are then provided to the cl plant. an estimator estimates the values of various internal state variables of the cl plant. the system has one structure adapted to control a cl plant that only provides pressure measurements in the cl loops a and b, a second structure adapted to a cl plant that provides pressure measurements and solid levels in both loops a, and b, and a third structure adapted to control a cl plant that provides full information on internal state variables. a final structure provides a neural network nmpc controller to control operation of loops a and b.",2017-08-22,"The title of the patent is nonlinear model predictive control for chemical looping process and its abstract is a control system for optimizing a chemical looping (“cl”) plant includes a reduced order mathematical model (“rom”) that is designed by eliminating mathematical terms that have minimal effect on the outcome. a non-linear optimizer provides various inputs to the rom and monitors the outputs to determine the optimum inputs that are then provided to the cl plant. an estimator estimates the values of various internal state variables of the cl plant. the system has one structure adapted to control a cl plant that only provides pressure measurements in the cl loops a and b, a second structure adapted to a cl plant that provides pressure measurements and solid levels in both loops a, and b, and a third structure adapted to control a cl plant that provides full information on internal state variables. a final structure provides a neural network nmpc controller to control operation of loops a and b. dated 2017-08-22"
9740966,tagging similar images using neural network,"an approach is provided in which a knowledge manager selects an extraction layer from a convolutional neural network that was trained on an initial set of images. the knowledge manager processes subsequent images obtained from crawling a computer network that includes extracting image feature sets of the subsequent images from the selected extraction layer and generating tags from metadata associated with the subsequent images. in turn, the knowledge manager receives a new image, extracts a new image feature set from the selected extraction layer, and assigns one or more of the tags to the new image based upon evaluating the new image feature set to the image features sets of the subsequent images.",2017-08-22,"The title of the patent is tagging similar images using neural network and its abstract is an approach is provided in which a knowledge manager selects an extraction layer from a convolutional neural network that was trained on an initial set of images. the knowledge manager processes subsequent images obtained from crawling a computer network that includes extracting image feature sets of the subsequent images from the selected extraction layer and generating tags from metadata associated with the subsequent images. in turn, the knowledge manager receives a new image, extracts a new image feature set from the selected extraction layer, and assigns one or more of the tags to the new image based upon evaluating the new image feature set to the image features sets of the subsequent images. dated 2017-08-22"
9741107,full reference image quality assessment based on convolutional neural network,"embodiments generally relate to providing systems and methods for assessing image quality of a distorted image relative to a reference image. in one embodiment, the system comprises a convolutional neural network that accepts as an input the distorted image and the reference image, and provides as an output a metric of image quality. in another embodiment, the method comprises inputting the distorted image and the reference image to a convolutional neural network configured to process the distorted image and the reference image and provide as an output a metric of image quality.",2017-08-22,"The title of the patent is full reference image quality assessment based on convolutional neural network and its abstract is embodiments generally relate to providing systems and methods for assessing image quality of a distorted image relative to a reference image. in one embodiment, the system comprises a convolutional neural network that accepts as an input the distorted image and the reference image, and provides as an output a metric of image quality. in another embodiment, the method comprises inputting the distorted image and the reference image to a convolutional neural network configured to process the distorted image and the reference image and provide as an output a metric of image quality. dated 2017-08-22"
9747544,method and system for wastewater treatment based on dissolved oxygen control by fuzzy neural network,"a method and system for wastewater treatment based on dissolved oxygen control by a fuzzy neural network, the method for wastewater treatment comprising the following steps: (1) measuring art inlet water flow rate, an orp value in an anaerobic tank, a do value in an aerobic tank, an inlet water cod value, and an actual outlet water cod value; (2) collecting the measured sample data and sending them via a computer to a cod fuzzy neural network predictive model, so as to establish an outlet water cod predicted value, (3) comparing the outlet cod predicted value with the outlet water cod set value, so as to obtain an error and an error change rate, and using them as two input variables to adjust a suitable dissolved oxygen concentration. accordingly, the on-line prediction and real-time control of dissolved oxygen wastewater treatment are achieved. the accurate control of dissolved oxygen concentration by the present method for wastewater treatment can achieve a saving in energy consumption while ensuring stable running of the sewage treatment system, and the outlet water quality meets the national emission standards.",2017-08-29,"The title of the patent is method and system for wastewater treatment based on dissolved oxygen control by fuzzy neural network and its abstract is a method and system for wastewater treatment based on dissolved oxygen control by a fuzzy neural network, the method for wastewater treatment comprising the following steps: (1) measuring art inlet water flow rate, an orp value in an anaerobic tank, a do value in an aerobic tank, an inlet water cod value, and an actual outlet water cod value; (2) collecting the measured sample data and sending them via a computer to a cod fuzzy neural network predictive model, so as to establish an outlet water cod predicted value, (3) comparing the outlet cod predicted value with the outlet water cod set value, so as to obtain an error and an error change rate, and using them as two input variables to adjust a suitable dissolved oxygen concentration. accordingly, the on-line prediction and real-time control of dissolved oxygen wastewater treatment are achieved. the accurate control of dissolved oxygen concentration by the present method for wastewater treatment can achieve a saving in energy consumption while ensuring stable running of the sewage treatment system, and the outlet water quality meets the national emission standards. dated 2017-08-29"
9747546,neural network processor,"a circuit for performing neural network computations for a neural network comprising a plurality of neural network layers, the circuit comprising: a matrix computation unit configured to, for each of the plurality of neural network layers: receive a plurality of weight inputs and a plurality of activation inputs for the neural network layer, and generate a plurality of accumulated values based on the plurality of weight inputs and the plurality of activation inputs; and a vector computation unit communicatively coupled to the matrix computation unit and configured to, for each of the plurality of neural network layers: apply an activation function to each accumulated value generated by the matrix computation unit to generate a plurality of activated values for the neural network layer.",2017-08-29,"The title of the patent is neural network processor and its abstract is a circuit for performing neural network computations for a neural network comprising a plurality of neural network layers, the circuit comprising: a matrix computation unit configured to, for each of the plurality of neural network layers: receive a plurality of weight inputs and a plurality of activation inputs for the neural network layer, and generate a plurality of accumulated values based on the plurality of weight inputs and the plurality of activation inputs; and a vector computation unit communicatively coupled to the matrix computation unit and configured to, for each of the plurality of neural network layers: apply an activation function to each accumulated value generated by the matrix computation unit to generate a plurality of activated values for the neural network layer. dated 2017-08-29"
9747547,hardware enhancements to radial basis function with restricted coulomb energy learning and/or k-nearest neighbor based neural network classifiers,"a nonlinear neuron classifier comprising a neuron array including a plurality of neuron chips each including a plurality of neurons of variable length and variable depth, the chips processing input vectors of variable length and variable depth that are input into the classifier for comparison against vectors stored in the classifier, wherein an nsp flag is set for a plurality of the neurons to indicate that only that plurality of neurons is to participate in the vector calculations. a virtual content addressable memory flag is set for certain of the neuron chips to enable functions including fast readout of data from the chips. results of vector calculations are aggregated for fast readout for a host computer interfacing with the classifier.",2017-08-29,"The title of the patent is hardware enhancements to radial basis function with restricted coulomb energy learning and/or k-nearest neighbor based neural network classifiers and its abstract is a nonlinear neuron classifier comprising a neuron array including a plurality of neuron chips each including a plurality of neurons of variable length and variable depth, the chips processing input vectors of variable length and variable depth that are input into the classifier for comparison against vectors stored in the classifier, wherein an nsp flag is set for a plurality of the neurons to indicate that only that plurality of neurons is to participate in the vector calculations. a virtual content addressable memory flag is set for certain of the neuron chips to enable functions including fast readout of data from the chips. results of vector calculations are aggregated for fast readout for a host computer interfacing with the classifier. dated 2017-08-29"
9747548,rotating data for neural network computations,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for computing a layer output for a convolutional neural network layer, the method comprising: receiving a plurality of activation inputs; forming a plurality of vector inputs from the plurality of activation inputs, each vector input comprising values from a distinct region within the multi-dimensional matrix; sending the plurality of vector inputs to one or more cells along a first dimension of the systolic array; generating a plurality of rotated kernel structures from each of the plurality of kernel; sending each kernel structure and each rotated kernel structure to one or more cells along a second dimension of the systolic array; causing the systolic array to generate an accumulated output based on the plurality of value inputs and the plurality of kernels; and generating the layer output from the accumulated output.",2017-08-29,"The title of the patent is rotating data for neural network computations and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for computing a layer output for a convolutional neural network layer, the method comprising: receiving a plurality of activation inputs; forming a plurality of vector inputs from the plurality of activation inputs, each vector input comprising values from a distinct region within the multi-dimensional matrix; sending the plurality of vector inputs to one or more cells along a first dimension of the systolic array; generating a plurality of rotated kernel structures from each of the plurality of kernel; sending each kernel structure and each rotated kernel structure to one or more cells along a second dimension of the systolic array; causing the systolic array to generate an accumulated output based on the plurality of value inputs and the plurality of kernels; and generating the layer output from the accumulated output. dated 2017-08-29"
9750909,system and method for diagnosis and treatment of a breathing pattern of a patient,"described is a system including a sensor and a processing arrangement. the sensor measures data corresponding to a patient's breathing patterns. the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of a rem sleep state. in another embodiment, the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of one of the following states: (i) a wake state and (ii) a sleep state. in another embodiment, a neural network analyzes the data to determine whether the breathing patterns are indicative of one of the following states: (i) a rem sleep state, (ii) a wake state and (iii) a sleep state. in another embodiment, the processing arrangement analyzes the data to determine whether the breathing pattern is indicative of an arousal.",2017-09-05,"The title of the patent is system and method for diagnosis and treatment of a breathing pattern of a patient and its abstract is described is a system including a sensor and a processing arrangement. the sensor measures data corresponding to a patient's breathing patterns. the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of a rem sleep state. in another embodiment, the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of one of the following states: (i) a wake state and (ii) a sleep state. in another embodiment, a neural network analyzes the data to determine whether the breathing patterns are indicative of one of the following states: (i) a rem sleep state, (ii) a wake state and (iii) a sleep state. in another embodiment, the processing arrangement analyzes the data to determine whether the breathing pattern is indicative of an arousal. dated 2017-09-05"
9753949,region-specific image download probability modeling,"methods for prioritizing a set of images identified as responsive to an image search query from a user based on features of the images identified as relevant to a geographic region of the user are provided. in one aspect, the method includes submitting a plurality of images to a computer-operated convolutional neural network that is configured to analyze image pixel data for each of the plurality of images to identify features, in each of the plurality of images, influencing a download probability of the corresponding image in a plurality of geographic regions. the method also includes receiving, from the neural network and for each of the plurality of images, a download probability of each image for each of the plurality of geographic regions. systems and machine-readable media are also provided.",2017-09-05,"The title of the patent is region-specific image download probability modeling and its abstract is methods for prioritizing a set of images identified as responsive to an image search query from a user based on features of the images identified as relevant to a geographic region of the user are provided. in one aspect, the method includes submitting a plurality of images to a computer-operated convolutional neural network that is configured to analyze image pixel data for each of the plurality of images to identify features, in each of the plurality of images, influencing a download probability of the corresponding image in a plurality of geographic regions. the method also includes receiving, from the neural network and for each of the plurality of images, a download probability of each image for each of the plurality of geographic regions. systems and machine-readable media are also provided. dated 2017-09-05"
9753959,method and apparatus for constructing a neuroscience-inspired artificial neural network with visualization of neural pathways,"a method and apparatus for constructing one of a neuroscience-inspired artificial neural network and a neural network array comprises one of a neuroscience-inspired dynamic architecture, a dynamic artificial neural network array and a neural network array of electrodes associated with neural tissue such as a brain, the method and apparatus having a special purpose display processor. the special purpose display processor outputs a display over a period of selected reference time units to demonstrate a neural pathway from, for example, one or a plurality of input neurons through intermediate destination neurons to an output neuron in three dimensional space. the displayed neural network may comprise neurons and synapses in different colors and may be utilized, for example, to show the behavior of a neural network for classifying hand-written digits between values of 0 and 9 or recognizing vertical/horizontal lines in a grid image of lines.",2017-09-05,"The title of the patent is method and apparatus for constructing a neuroscience-inspired artificial neural network with visualization of neural pathways and its abstract is a method and apparatus for constructing one of a neuroscience-inspired artificial neural network and a neural network array comprises one of a neuroscience-inspired dynamic architecture, a dynamic artificial neural network array and a neural network array of electrodes associated with neural tissue such as a brain, the method and apparatus having a special purpose display processor. the special purpose display processor outputs a display over a period of selected reference time units to demonstrate a neural pathway from, for example, one or a plurality of input neurons through intermediate destination neurons to an output neuron in three dimensional space. the displayed neural network may comprise neurons and synapses in different colors and may be utilized, for example, to show the behavior of a neural network for classifying hand-written digits between values of 0 and 9 or recognizing vertical/horizontal lines in a grid image of lines. dated 2017-09-05"
9754204,"systems, methods and devices for vector control of permanent magnet synchronous machines using artificial neural networks","an example method for controlling an ac electrical machine can include providing a pwm converter operably connected between an electrical power source and the ac electrical machine and providing a neural network vector control system operably connected to the pwm converter. the control system can include a current-loop neural network configured to receive a plurality of inputs. the current-loop neural network can be configured to optimize the compensating dq-control voltage. the inputs can be d- and q-axis currents, d- and q-axis error signals, predicted d- and q-axis current signals, and a feedback compensating dq-control voltage. the d- and q-axis error signals can be a difference between the d- and q-axis currents and reference d- and q-axis currents, respectively. the method can further include outputting a compensating dq-control voltage from the current-loop neural network and controlling the pwm converter using the compensating dq-control voltage.",2017-09-05,"The title of the patent is systems, methods and devices for vector control of permanent magnet synchronous machines using artificial neural networks and its abstract is an example method for controlling an ac electrical machine can include providing a pwm converter operably connected between an electrical power source and the ac electrical machine and providing a neural network vector control system operably connected to the pwm converter. the control system can include a current-loop neural network configured to receive a plurality of inputs. the current-loop neural network can be configured to optimize the compensating dq-control voltage. the inputs can be d- and q-axis currents, d- and q-axis error signals, predicted d- and q-axis current signals, and a feedback compensating dq-control voltage. the d- and q-axis error signals can be a difference between the d- and q-axis currents and reference d- and q-axis currents, respectively. the method can further include outputting a compensating dq-control voltage from the current-loop neural network and controlling the pwm converter using the compensating dq-control voltage. dated 2017-09-05"
9754221,processor for implementing reinforcement learning operations,"a reinforcement learning processor specifically configured to execute reinforcement learning operations by the way of implementing an application-specific instruction set is envisaged. the application-specific instruction set incorporates ‘single instruction multiple agents (sima)’ instructions. sima type instructions are specifically designed to be implemented simultaneously on a plurality of reinforcement learning agents which interact with corresponding reinforcement learning environments. the sima type instructions are specifically configured to receive either a reinforcement learning agent id or a reinforcement learning environment id as the operand. the reinforcement learning processor uses neural network data paths to communicate with a neural network which in turn uses the actions, state-value functions, q-values and reward values generated by the reinforcement learning processor to approximate an optimal state-value function as well as an optimal reward function.",2017-09-05,"The title of the patent is processor for implementing reinforcement learning operations and its abstract is a reinforcement learning processor specifically configured to execute reinforcement learning operations by the way of implementing an application-specific instruction set is envisaged. the application-specific instruction set incorporates ‘single instruction multiple agents (sima)’ instructions. sima type instructions are specifically designed to be implemented simultaneously on a plurality of reinforcement learning agents which interact with corresponding reinforcement learning environments. the sima type instructions are specifically configured to receive either a reinforcement learning agent id or a reinforcement learning environment id as the operand. the reinforcement learning processor uses neural network data paths to communicate with a neural network which in turn uses the actions, state-value functions, q-values and reward values generated by the reinforcement learning processor to approximate an optimal state-value function as well as an optimal reward function. dated 2017-09-05"
9754351,systems and methods for processing content using convolutional neural networks,"systems, methods, and non-transitory computer-readable media can obtain a set of video frames at a first resolution. process the set of video frames using a convolutional neural network to output one or more signals, the convolutional neural network including (i) a set of two-dimensional convolutional layers and (ii) a set of three-dimensional convolutional layers, wherein the processing causes the set of video frames to be reduced to a second resolution. process the one or more signals using a set of three-dimensional de-convolutional layers of the convolutional neural network. obtain one or more outputs corresponding to the set of video frames from the convolutional neural network.",2017-09-05,"The title of the patent is systems and methods for processing content using convolutional neural networks and its abstract is systems, methods, and non-transitory computer-readable media can obtain a set of video frames at a first resolution. process the set of video frames using a convolutional neural network to output one or more signals, the convolutional neural network including (i) a set of two-dimensional convolutional layers and (ii) a set of three-dimensional convolutional layers, wherein the processing causes the set of video frames to be reduced to a second resolution. process the one or more signals using a set of three-dimensional de-convolutional layers of the convolutional neural network. obtain one or more outputs corresponding to the set of video frames from the convolutional neural network. dated 2017-09-05"
9754584,user specified keyword spotting using neural network feature extractor,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for recognizing keywords using a long short term memory neural network. one of the methods includes receiving, by a device for each of multiple variable length enrollment audio signals, a respective plurality of enrollment feature vectors that represent features of the respective variable length enrollment audio signal, processing each of the plurality of enrollment feature vectors using a long short term memory (lstm) neural network to generate a respective enrollment lstm output vector for each enrollment feature vector, and generating, for the respective variable length enrollment audio signal, a template fixed length representation for use in determining whether another audio signal encodes another spoken utterance of the enrollment phrase by combining at most a quantity k of the enrollment lstm output vectors for the enrollment audio signal.",2017-09-05,"The title of the patent is user specified keyword spotting using neural network feature extractor and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for recognizing keywords using a long short term memory neural network. one of the methods includes receiving, by a device for each of multiple variable length enrollment audio signals, a respective plurality of enrollment feature vectors that represent features of the respective variable length enrollment audio signal, processing each of the plurality of enrollment feature vectors using a long short term memory (lstm) neural network to generate a respective enrollment lstm output vector for each enrollment feature vector, and generating, for the respective variable length enrollment audio signal, a template fixed length representation for use in determining whether another audio signal encodes another spoken utterance of the enrollment phrase by combining at most a quantity k of the enrollment lstm output vectors for the enrollment audio signal. dated 2017-09-05"
9754668,digital perceptron,"in view of the neural network information parallel processing, a digital perceptron device analogous to the build-in neural network hardware systems for parallel processing digital signals directly by the processor's memory content and memory perception in one feed-forward step is disclosed. the digital perceptron device of the invention applies the configurable content and perceptive non-volatile memory arrays as the memory processor hardware. the input digital signals are then broadcasted into the non-volatile content memory array for a match to output the digital signals from the perceptive non-volatile memory array as the content-perceptive digital perceptron device.",2017-09-05,"The title of the patent is digital perceptron and its abstract is in view of the neural network information parallel processing, a digital perceptron device analogous to the build-in neural network hardware systems for parallel processing digital signals directly by the processor's memory content and memory perception in one feed-forward step is disclosed. the digital perceptron device of the invention applies the configurable content and perceptive non-volatile memory arrays as the memory processor hardware. the input digital signals are then broadcasted into the non-volatile content memory array for a match to output the digital signals from the perceptive non-volatile memory array as the content-perceptive digital perceptron device. dated 2017-09-05"
9755948,controlling an optical bypass switch in a data center based on a neural network output result,"a flow of packets is communicated through a data center including an electrical switch, an optical switch, and multiple racks each including multiple network devices. the optical switch can be controlled to receive packet traffic from a network device via a first optical link and to output that packet traffic to another network device via a second optical link. one network device includes a neural network that analyzes received packets of the flow. the optical switch is controlled to switch based on a result of the analysis performed. in one instance, the optical switch is controlled such that immediately prior to the switching no packet traffic passes from the first optical link and through the optical switch and to the second optical link but such that after the switching packet traffic does pass from the first optical link and through the optical switch and to the second optical link.",2017-09-05,"The title of the patent is controlling an optical bypass switch in a data center based on a neural network output result and its abstract is a flow of packets is communicated through a data center including an electrical switch, an optical switch, and multiple racks each including multiple network devices. the optical switch can be controlled to receive packet traffic from a network device via a first optical link and to output that packet traffic to another network device via a second optical link. one network device includes a neural network that analyzes received packets of the flow. the optical switch is controlled to switch based on a result of the analysis performed. in one instance, the optical switch is controlled such that immediately prior to the switching no packet traffic passes from the first optical link and through the optical switch and to the second optical link but such that after the switching packet traffic does pass from the first optical link and through the optical switch and to the second optical link. dated 2017-09-05"
9760827,neural network applications in resource constrained environments,systems and methods are disclosed for applying neural networks in resource-constrained environments. a system may include a sensor located in a resource-constrained environment configured to generate sensor data of the resource-constrained environment. the system may also include a first computing device not located in the resource-constrained environment configured to produce a neural network structure based on the sensor data. the system may further include a second computing device located in the resource-constrained environment configured to provide the sensor data as input to the neural network structure. the second computing device may be further configured to determine a state of the resource-constrained environment based on the input of the sensor data to the neural network structure.,2017-09-12,The title of the patent is neural network applications in resource constrained environments and its abstract is systems and methods are disclosed for applying neural networks in resource-constrained environments. a system may include a sensor located in a resource-constrained environment configured to generate sensor data of the resource-constrained environment. the system may also include a first computing device not located in the resource-constrained environment configured to produce a neural network structure based on the sensor data. the system may further include a second computing device located in the resource-constrained environment configured to provide the sensor data as input to the neural network structure. the second computing device may be further configured to determine a state of the resource-constrained environment based on the input of the sensor data to the neural network structure. dated 2017-09-12
9761131,systems and methods involving features of adaptive and/or autonomous traffic control,"systems and method are disclosed for adaptive and/or autonomous traffic control. in one illustrative implementation, there is provided a method for processing traffic information. moreover, the method may include receiving data regarding travel of vehicles associated with an intersection, using neural network technology to recognize types and/or states of traffic, and using the neural network technology to process/determine/memorize optimal traffic flow decisions as a function of experience information. exemplary implementations may also include using the neural network technology to achieve efficient traffic flow via recognition of the optimal traffic flow decisions.",2017-09-12,"The title of the patent is systems and methods involving features of adaptive and/or autonomous traffic control and its abstract is systems and method are disclosed for adaptive and/or autonomous traffic control. in one illustrative implementation, there is provided a method for processing traffic information. moreover, the method may include receiving data regarding travel of vehicles associated with an intersection, using neural network technology to recognize types and/or states of traffic, and using the neural network technology to process/determine/memorize optimal traffic flow decisions as a function of experience information. exemplary implementations may also include using the neural network technology to achieve efficient traffic flow via recognition of the optimal traffic flow decisions. dated 2017-09-12"
9761221,order statistic techniques for neural networks,"according to some aspects, a method of classifying speech recognition results is provided, using a neural network comprising a plurality of interconnected network units, each network unit having one or more weight values, the method comprising using at least one computer, performing acts of providing a first vector as input to a first network layer comprising one or more network units of the neural network, transforming, by a first network unit of the one or more network units, the input vector to produce a plurality of values, the transformation being based at least in part on a plurality of weight values of the first network unit, sorting the plurality of values to produce a sorted plurality of values, and providing the sorted plurality of values as input to a second network layer of the neural network.",2017-09-12,"The title of the patent is order statistic techniques for neural networks and its abstract is according to some aspects, a method of classifying speech recognition results is provided, using a neural network comprising a plurality of interconnected network units, each network unit having one or more weight values, the method comprising using at least one computer, performing acts of providing a first vector as input to a first network layer comprising one or more network units of the neural network, transforming, by a first network unit of the one or more network units, the input vector to produce a plurality of values, the transformation being based at least in part on a plurality of weight values of the first network unit, sorting the plurality of values to produce a sorted plurality of values, and providing the sorted plurality of values as input to a second network layer of the neural network. dated 2017-09-12"
9764151,neural network system for the evaluation and the adaptation of antitachycardia therapy by an implantable defibrillator,"the system includes an active medical device with means for delivering defibrillation shocks; means for continuous collection of the patient current cardiac activity parameters; and evaluator means with neuronal analysis comprising a neural network with at least two layers. this neural network comprises upstream three neural sub-networks receiving the respective parameters divided into separate sub-groups corresponding to classes of arrhythmogenic factors; and downstream an output neuron coupled to the three sub-networks and capable of outputting an index of risk of ventricular arrhythmia. the risk index is compared with a given threshold, to enable or disable at least one function of the device in case of crossing of the threshold.",2017-09-19,"The title of the patent is neural network system for the evaluation and the adaptation of antitachycardia therapy by an implantable defibrillator and its abstract is the system includes an active medical device with means for delivering defibrillation shocks; means for continuous collection of the patient current cardiac activity parameters; and evaluator means with neuronal analysis comprising a neural network with at least two layers. this neural network comprises upstream three neural sub-networks receiving the respective parameters divided into separate sub-groups corresponding to classes of arrhythmogenic factors; and downstream an output neuron coupled to the three sub-networks and capable of outputting an index of risk of ventricular arrhythmia. the risk index is compared with a given threshold, to enable or disable at least one function of the device in case of crossing of the threshold. dated 2017-09-19"
9767381,similarity-based detection of prominent objects using deep cnn pooling layers as features,"a system and method provide object localization in a query image based on a global representation of the image generated with a model derived from a convolutional neural network. representations of annotated images and a query image are each generated based on activations output by a layer of the model which precedes the fully-connected layers of the neural network. a similarity is computed between the query image representation and each of the annotated image representations to identify a subset of the annotated images having the highest computed similarity. object location information from at least one of the subset of annotated images is transferred to the query image and information is output, based on the transferred object location information.",2017-09-19,"The title of the patent is similarity-based detection of prominent objects using deep cnn pooling layers as features and its abstract is a system and method provide object localization in a query image based on a global representation of the image generated with a model derived from a convolutional neural network. representations of annotated images and a query image are each generated based on activations output by a layer of the model which precedes the fully-connected layers of the neural network. a similarity is computed between the query image representation and each of the annotated image representations to identify a subset of the annotated images having the highest computed similarity. object location information from at least one of the subset of annotated images is transferred to the query image and information is output, based on the transferred object location information. dated 2017-09-19"
9767407,"weighting device, neural network, and operating method of the weighting device","provided are a weighting device that may be driven at a low voltage and is capable of embodying multi-level weights, a neural network, and a method of operating the weighting device. the weighting device includes a switching layer that may switch between a high resistance state and a low resistance state based on a voltage applied thereto and a charge trap material layer that traps or discharges charges according to a resistance state of the switching layer. the weighting device may be used for controlling a weight in a neural network.",2017-09-19,"The title of the patent is weighting device, neural network, and operating method of the weighting device and its abstract is provided are a weighting device that may be driven at a low voltage and is capable of embodying multi-level weights, a neural network, and a method of operating the weighting device. the weighting device includes a switching layer that may switch between a high resistance state and a low resistance state based on a voltage applied thereto and a charge trap material layer that traps or discharges charges according to a resistance state of the switching layer. the weighting device may be used for controlling a weight in a neural network. dated 2017-09-19"
9767409,latent feature based tag routing,"features are disclosed for identifying and routing items for tagging using a latent feature model, such as a recurrent neural network language model (rnnlm). the model may be trained to identify latent features for catalog items such as movies, books, food items, beverages, and the like. based on similarities in latent features, tags previous assigned to items may be applied to untagged items. application may be manual or automatic. in either case, resources need to be balances to ensure efficient tagging of items. the included features help to identify and direct these limited tagging resources.",2017-09-19,"The title of the patent is latent feature based tag routing and its abstract is features are disclosed for identifying and routing items for tagging using a latent feature model, such as a recurrent neural network language model (rnnlm). the model may be trained to identify latent features for catalog items such as movies, books, food items, beverages, and the like. based on similarities in latent features, tags previous assigned to items may be applied to untagged items. application may be manual or automatic. in either case, resources need to be balances to ensure efficient tagging of items. the included features help to identify and direct these limited tagging resources. dated 2017-09-19"
9767410,rank-constrained neural networks,"this specification describes, among other things, a computer-implemented method. the method can include training a baseline neural network using a first set of training data. for each node in a subset of interconnected nodes in the baseline neural network, a rank-k approximation of a filter for the node can be computed. a subset of nodes in a rank-constrained neural network can then be initialized with the rank-k approximations of the filters from the baseline neural network. the subset of nodes in the rank-constrained neural network can correspond to the subset of nodes in the baseline neural network. after initializing, the rank-constrained neural network can be trained using a second set of training data while maintaining a rank-k filter topology for the subset of nodes in the rank-constrained neural network.",2017-09-19,"The title of the patent is rank-constrained neural networks and its abstract is this specification describes, among other things, a computer-implemented method. the method can include training a baseline neural network using a first set of training data. for each node in a subset of interconnected nodes in the baseline neural network, a rank-k approximation of a filter for the node can be computed. a subset of nodes in a rank-constrained neural network can then be initialized with the rank-k approximations of the filters from the baseline neural network. the subset of nodes in the rank-constrained neural network can correspond to the subset of nodes in the baseline neural network. after initializing, the rank-constrained neural network can be trained using a second set of training data while maintaining a rank-k filter topology for the subset of nodes in the rank-constrained neural network. dated 2017-09-19"
9767557,method and system for vascular disease detection using recurrent neural networks,"a method and apparatus for vascular disease detection and characterization using a recurrent neural network (rnn) is disclosed. a plurality of 2d cross-section image patches are extracted from a 3d computed tomography angiography (cta) image, each extracted at a respective sampling point along a vessel centerline of a vessel of interest in the 3d cta image. vascular abnormalities in the vessel of interest are detected and characterized by classifying each of the sampling points along the vessel centerline based on the plurality of 2d cross-section image patches using a trained rnn.",2017-09-19,"The title of the patent is method and system for vascular disease detection using recurrent neural networks and its abstract is a method and apparatus for vascular disease detection and characterization using a recurrent neural network (rnn) is disclosed. a plurality of 2d cross-section image patches are extracted from a 3d computed tomography angiography (cta) image, each extracted at a respective sampling point along a vessel centerline of a vessel of interest in the 3d cta image. vascular abnormalities in the vessel of interest are detected and characterized by classifying each of the sampling points along the vessel centerline based on the plurality of 2d cross-section image patches using a trained rnn. dated 2017-09-19"
9767565,synthesizing training data for broad area geospatial object detection,"a system for broad area geospatial object recognition, identification, classification, location and quantification, comprising an image manipulation module to create synthetically-generated images to imitate and augment an existing quantity of orthorectified geospatial images; together with a deep learning module and a convolutional neural network serving as an image analysis module, to analyze a large corpus of orthorectified geospatial images, identify and demarcate a searched object of interest from within the corpus, locate and quantify the identified or classified objects from the corpus of geospatial imagery available to the system. the system reports results in a requestor's preferred format.",2017-09-19,"The title of the patent is synthesizing training data for broad area geospatial object detection and its abstract is a system for broad area geospatial object recognition, identification, classification, location and quantification, comprising an image manipulation module to create synthetically-generated images to imitate and augment an existing quantity of orthorectified geospatial images; together with a deep learning module and a convolutional neural network serving as an image analysis module, to analyze a large corpus of orthorectified geospatial images, identify and demarcate a searched object of interest from within the corpus, locate and quantify the identified or classified objects from the corpus of geospatial imagery available to the system. the system reports results in a requestor's preferred format. dated 2017-09-19"
9773195,user-configurable settings for content obfuscation,"an aspect of providing user-configurable settings for content obfuscation includes, for each media segment in a media file, inputting the media segment to a neural network, applying a classifier to features output by the neural network, and determining from results of the classifier images in the media segment that contain the sensitive characteristics. the classifier specifies images that are predetermined to include sensitive characteristics. an aspect further includes assigning a tag to each of the images in the media segment that contain the sensitive characteristics. the tag indicates a type of sensitivity. an aspect also includes receiving at least one user-defined sensitivity, the user-defined sensitivity indicating an action or condition that is considered objectionable to a user, identifying a subset of the tagged images that correlate to the user-defined sensitivity, and visually modifying, during playback of the media file, an appearance of the subset of the tagged images.",2017-09-26,"The title of the patent is user-configurable settings for content obfuscation and its abstract is an aspect of providing user-configurable settings for content obfuscation includes, for each media segment in a media file, inputting the media segment to a neural network, applying a classifier to features output by the neural network, and determining from results of the classifier images in the media segment that contain the sensitive characteristics. the classifier specifies images that are predetermined to include sensitive characteristics. an aspect further includes assigning a tag to each of the images in the media segment that contain the sensitive characteristics. the tag indicates a type of sensitivity. an aspect also includes receiving at least one user-defined sensitivity, the user-defined sensitivity indicating an action or condition that is considered objectionable to a user, identifying a subset of the tagged images that correlate to the user-defined sensitivity, and visually modifying, during playback of the media file, an appearance of the subset of the tagged images. dated 2017-09-26"
9773196,utilizing deep learning for automatic digital image segmentation and stylization,"systems and methods are disclosed for segregating target individuals represented in a probe digital image from background pixels in the probe digital image. in particular, in one or more embodiments, the disclosed systems and methods train a neural network based on two or more of training position channels, training shape input channels, training color channels, or training object data. moreover, in one or more embodiments, the disclosed systems and methods utilize the trained neural network to select a target individual in a probe digital image. specifically, in one or more embodiments, the disclosed systems and methods generate position channels, training shape input channels, and color channels corresponding the probe digital image, and utilize the generated channels in conjunction with the trained neural network to select the target individual.",2017-09-26,"The title of the patent is utilizing deep learning for automatic digital image segmentation and stylization and its abstract is systems and methods are disclosed for segregating target individuals represented in a probe digital image from background pixels in the probe digital image. in particular, in one or more embodiments, the disclosed systems and methods train a neural network based on two or more of training position channels, training shape input channels, training color channels, or training object data. moreover, in one or more embodiments, the disclosed systems and methods utilize the trained neural network to select a target individual in a probe digital image. specifically, in one or more embodiments, the disclosed systems and methods generate position channels, training shape input channels, and color channels corresponding the probe digital image, and utilize the generated channels in conjunction with the trained neural network to select the target individual. dated 2017-09-26"
9779354,learning method and recording medium,"learning method includes performing a first process in which a coarse class classifier configured with a first neural network is made to classify a plurality of images given as a set of images each attached with a label indicating a detailed class into a plurality of coarse classes including a plurality of detailed classes and is then made to learn a first feature that is a feature common in each of the coarse classes, and performing a second process in which a detailed class classifier, configured with a second neural network that is the same in terms of layers other than the final layer as but different in terms of the final layer from the first neural network made to perform the learning in the first process, is made to classify the set of images into detailed classes and learn a second feature of each detailed class.",2017-10-03,"The title of the patent is learning method and recording medium and its abstract is learning method includes performing a first process in which a coarse class classifier configured with a first neural network is made to classify a plurality of images given as a set of images each attached with a label indicating a detailed class into a plurality of coarse classes including a plurality of detailed classes and is then made to learn a first feature that is a feature common in each of the coarse classes, and performing a second process in which a detailed class classifier, configured with a second neural network that is the same in terms of layers other than the final layer as but different in terms of the final layer from the first neural network made to perform the learning in the first process, is made to classify the set of images into detailed classes and learn a second feature of each detailed class. dated 2017-10-03"
9779355,back propagation gates and storage capacitor for neural networks,"technical solutions are described for implementing a neural network. an example system includes a crosspoint array including a plurality of nodes, each node representing a weight assigned to a neuron of the neural network. the system also includes a capacitor associated with a set of nodes from the plurality of nodes, where the capacitor is configured to store a current value corresponding to a sum of outputs from each respective node from the set of nodes. the system also includes a clocking circuit that initiates a forward pass to propagate the current value stored in the capacitor to a subsequent layer of the neural network. the clocking circuit further initiates a backward pass to propagate the current value stored in the capacitor to a preceding layer of the neural network. the clocking circuit further initiates a weight-update pass to update the weights in the neural network.",2017-10-03,"The title of the patent is back propagation gates and storage capacitor for neural networks and its abstract is technical solutions are described for implementing a neural network. an example system includes a crosspoint array including a plurality of nodes, each node representing a weight assigned to a neuron of the neural network. the system also includes a capacitor associated with a set of nodes from the plurality of nodes, where the capacitor is configured to store a current value corresponding to a sum of outputs from each respective node from the set of nodes. the system also includes a clocking circuit that initiates a forward pass to propagate the current value stored in the capacitor to a subsequent layer of the neural network. the clocking circuit further initiates a backward pass to propagate the current value stored in the capacitor to a preceding layer of the neural network. the clocking circuit further initiates a weight-update pass to update the weights in the neural network. dated 2017-10-03"
9779492,"retinal image quality assessment, error identification and automatic quality correction","automatically determining image quality of a machine generated image may generate a local saliency map of the image to obtain a set of unsupervised features. the image is run through a trained convolutional neural network (cnn) to extract a set of supervised features from a fully connected layer of the cnn, the image convolved with a set of learned kernels from the cnn to obtain a complementary set of supervised features. the set of unsupervised features and the complementary set of supervised features are combined, and a first decision on gradability of the image is predicted. a second decision on gradability of the image is predicted based on the set of supervised features. whether the image is gradable is determined based on a weighted combination of the first decision and the second decision.",2017-10-03,"The title of the patent is retinal image quality assessment, error identification and automatic quality correction and its abstract is automatically determining image quality of a machine generated image may generate a local saliency map of the image to obtain a set of unsupervised features. the image is run through a trained convolutional neural network (cnn) to extract a set of supervised features from a fully connected layer of the cnn, the image convolved with a set of learned kernels from the cnn to obtain a complementary set of supervised features. the set of unsupervised features and the complementary set of supervised features are combined, and a first decision on gradability of the image is predicted. a second decision on gradability of the image is predicted based on the set of supervised features. whether the image is gradable is determined based on a weighted combination of the first decision and the second decision. dated 2017-10-03"
9779727,mixed speech recognition,"the claimed subject matter includes a system and method for recognizing mixed speech from a source. the method includes training a first neural network to recognize the speech signal spoken by the speaker with a higher level of a speech characteristic from a mixed speech sample. the method also includes training a second neural network to recognize the speech signal spoken by the speaker with a lower level of the speech characteristic from the mixed speech sample. additionally, the method includes decoding the mixed speech sample with the first neural network and the second neural network by optimizing the joint likelihood of observing the two speech signals considering the probability that a specific frame is a switching point of the speech characteristic.",2017-10-03,"The title of the patent is mixed speech recognition and its abstract is the claimed subject matter includes a system and method for recognizing mixed speech from a source. the method includes training a first neural network to recognize the speech signal spoken by the speaker with a higher level of a speech characteristic from a mixed speech sample. the method also includes training a second neural network to recognize the speech signal spoken by the speaker with a lower level of the speech characteristic from the mixed speech sample. additionally, the method includes decoding the mixed speech sample with the first neural network and the second neural network by optimizing the joint likelihood of observing the two speech signals considering the probability that a specific frame is a switching point of the speech characteristic. dated 2017-10-03"
9779759,device impairment detection,examples described herein involve detecting known impairments or other known conditions using a neural network. an example implementation involves receiving data indicating a response of a playback device as captured by a microphone. the implementation also involves determining an input vector by projecting a response vector that represents the response of the playback device onto a principle component matrix representing variance caused by one or more known impairments. the implementation further involves providing the determined input vector to a neural network that includes an output layer comprising neurons that correspond to respective known impairments. the implementation involves detecting that the input vector caused one or more neurons of the neural network to fire such that the neural network indicates that a particular known impairment is affecting the microphone and/or the playback device and adjusting operation of the playback device and/or the microphone to offset the particular known impairment.,2017-10-03,The title of the patent is device impairment detection and its abstract is examples described herein involve detecting known impairments or other known conditions using a neural network. an example implementation involves receiving data indicating a response of a playback device as captured by a microphone. the implementation also involves determining an input vector by projecting a response vector that represents the response of the playback device onto a principle component matrix representing variance caused by one or more known impairments. the implementation further involves providing the determined input vector to a neural network that includes an output layer comprising neurons that correspond to respective known impairments. the implementation involves detecting that the input vector caused one or more neurons of the neural network to fire such that the neural network indicates that a particular known impairment is affecting the microphone and/or the playback device and adjusting operation of the playback device and/or the microphone to offset the particular known impairment. dated 2017-10-03
9785855,coarse-to-fine cascade adaptations for license plate recognition with convolutional neural networks,"methods and systems for license plate recognition utilizing a trained neural network. in an example embodiment, a neural network can be subject to operations involving iteratively training and adapting the neural network for a particular task such as, for example, text recognition in the context of a license plate recognition application. the neural network can be trained to perform generic text recognition utilizing a plurality of training samples. the neural network can be applied to a cropped image of a license plate in order to recognize text and produce a license plate transcription with respect to the license plate. an example of such a neural network is a cnn (convolutional neural. network).",2017-10-10,"The title of the patent is coarse-to-fine cascade adaptations for license plate recognition with convolutional neural networks and its abstract is methods and systems for license plate recognition utilizing a trained neural network. in an example embodiment, a neural network can be subject to operations involving iteratively training and adapting the neural network for a particular task such as, for example, text recognition in the context of a license plate recognition application. the neural network can be trained to perform generic text recognition utilizing a plurality of training samples. the neural network can be applied to a cropped image of a license plate in order to recognize text and produce a license plate transcription with respect to the license plate. an example of such a neural network is a cnn (convolutional neural. network). dated 2017-10-10"
9785886,cooperative execution of a genetic algorithm with an efficient training algorithm for data-driven model creation,"a method includes, based on a fitness function, selecting a subset of models from a plurality of models. the plurality of models is generated based on a genetic algorithm and corresponds to a first epoch of the genetic algorithm. each of the plurality of models includes data representative of a neural network. the method also includes performing at least one genetic operation of the genetic algorithm with respect to at least one model of the subset to generate a trainable model and sending the trainable model to an optimization trainer. the method includes adding a trained model received from the optimization trainer as input to a second epoch of the genetic algorithm that is subsequent to the first epoch.",2017-10-10,"The title of the patent is cooperative execution of a genetic algorithm with an efficient training algorithm for data-driven model creation and its abstract is a method includes, based on a fitness function, selecting a subset of models from a plurality of models. the plurality of models is generated based on a genetic algorithm and corresponds to a first epoch of the genetic algorithm. each of the plurality of models includes data representative of a neural network. the method also includes performing at least one genetic operation of the genetic algorithm with respect to at least one model of the subset to generate a trainable model and sending the trainable model to an optimization trainer. the method includes adding a trained model received from the optimization trainer as input to a second epoch of the genetic algorithm that is subsequent to the first epoch. dated 2017-10-10"
9786270,generating acoustic models,"methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating acoustic models. in some implementations, a first neural network trained as an acoustic model using the connectionist temporal classification algorithm is obtained. output distributions from the first neural network are obtained for an utterance. a second neural network is trained as an acoustic model using the output distributions produced by the first neural network as output targets for the second neural network. an automated speech recognizer configured to use the trained second neural network is provided.",2017-10-10,"The title of the patent is generating acoustic models and its abstract is methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating acoustic models. in some implementations, a first neural network trained as an acoustic model using the connectionist temporal classification algorithm is obtained. output distributions from the first neural network are obtained for an utterance. a second neural network is trained as an acoustic model using the output distributions produced by the first neural network as output targets for the second neural network. an automated speech recognizer configured to use the trained second neural network is provided. dated 2017-10-10"
9788030,television system with aided user program searching,a system having an adaptive browse feature and an adaptive flip feature is provided. the adaptive browse and flip features may be selected to receive program viewing suggestions. the system may provide a suggestion by displaying an adaptive browse region or adaptive flip region including a program suggestion. the system identifies programs to suggest based on a user=s viewing activity. the system uses different algorithms that are user-selectable and user-adjustable to identify program suggestions. the system may query a program guide database to build a list of programs having attributes similar to the attributes of the current program or the last viewed program. the system may use an adaptive learning algorithm such as a neural network. the neural network may be trained by the program guide by monitoring user-viewing activity. each algorithm may be personalized for multiple users.,2017-10-10,The title of the patent is television system with aided user program searching and its abstract is a system having an adaptive browse feature and an adaptive flip feature is provided. the adaptive browse and flip features may be selected to receive program viewing suggestions. the system may provide a suggestion by displaying an adaptive browse region or adaptive flip region including a program suggestion. the system identifies programs to suggest based on a user=s viewing activity. the system uses different algorithms that are user-selectable and user-adjustable to identify program suggestions. the system may query a program guide database to build a list of programs having attributes similar to the attributes of the current program or the last viewed program. the system may use an adaptive learning algorithm such as a neural network. the neural network may be trained by the program guide by monitoring user-viewing activity. each algorithm may be personalized for multiple users. dated 2017-10-10
9791638,low power integrated analog mathematical engine,"a method for creating on chip analog mathematical engines is provided utilizing a neural network with a switched capacitor structure to implement coefficients for weighted connections and error functions for the neural network. the neural networks are capable of any transfer function, learning, doing pattern recognition, clustering, control or many other functions. the switched capacitor charge controls allow for nodal control of charge transfer based switched capacitor circuits. the method reduces reliance on passive component programmable arrays to produce programmable switched capacitor circuit coefficients. the switched capacitor circuits are dynamically scaled without having to rely on switched in unit passives, such as unit capacitors, and the complexities of switching these capacitors into and out of circuit. the current, and thus the charge transferred is controlled at a nodal level, and the current rather than the capacitors are scaled providing a more accurate result in addition to saving silicon area.",2017-10-17,"The title of the patent is low power integrated analog mathematical engine and its abstract is a method for creating on chip analog mathematical engines is provided utilizing a neural network with a switched capacitor structure to implement coefficients for weighted connections and error functions for the neural network. the neural networks are capable of any transfer function, learning, doing pattern recognition, clustering, control or many other functions. the switched capacitor charge controls allow for nodal control of charge transfer based switched capacitor circuits. the method reduces reliance on passive component programmable arrays to produce programmable switched capacitor circuit coefficients. the switched capacitor circuits are dynamically scaled without having to rely on switched in unit passives, such as unit capacitors, and the complexities of switching these capacitors into and out of circuit. the current, and thus the charge transferred is controlled at a nodal level, and the current rather than the capacitors are scaled providing a more accurate result in addition to saving silicon area. dated 2017-10-17"
9792492,extracting gradient features from neural networks,"a method for extracting a representation from an image includes inputting an image to a pre-trained neural network. the gradient of a loss function is computed with respect to parameters of the neural network, for the image. a gradient representation is extracted for the image based on the computed gradients, which can be used, for example, for classification or retrieval.",2017-10-17,"The title of the patent is extracting gradient features from neural networks and its abstract is a method for extracting a representation from an image includes inputting an image to a pre-trained neural network. the gradient of a loss function is computed with respect to parameters of the neural network, for the image. a gradient representation is extracted for the image based on the computed gradients, which can be used, for example, for classification or retrieval. dated 2017-10-17"
9792530,generating and using a knowledge base for image classification,"a knowledge base (kb) is generated and used to classify images. the knowledge base includes a number subcategories of a specified category. instead of obtaining images just based on a category name, structured and unstructured data sources are used to identify subcategories of the category. subcategories that are determined to not be relevant to the category may be removed. the remaining data may be used to generate the kb. after identifying the relevant subcategories, representative images are obtained from one or more image sources based on the subcategories identified by the kb. the obtained images and the kb are then used to train an image classifier, such as a neural network or some other machine learning mechanism. after training, the neural network might, for example, classify an object within an image of a car, as a car, but also as a particular brand and model type.",2017-10-17,"The title of the patent is generating and using a knowledge base for image classification and its abstract is a knowledge base (kb) is generated and used to classify images. the knowledge base includes a number subcategories of a specified category. instead of obtaining images just based on a category name, structured and unstructured data sources are used to identify subcategories of the category. subcategories that are determined to not be relevant to the category may be removed. the remaining data may be used to generate the kb. after identifying the relevant subcategories, representative images are obtained from one or more image sources based on the subcategories identified by the kb. the obtained images and the kb are then used to train an image classifier, such as a neural network or some other machine learning mechanism. after training, the neural network might, for example, classify an object within an image of a car, as a car, but also as a particular brand and model type. dated 2017-10-17"
9792547,neural network circuit and learning method for neural network circuit,"a neural network circuit includes an error calculating circuit that generates an error voltage signal having a magnitude in accordance with a time difference between an output signal and a teaching signal corresponding to the output signal. a weight change pulse voltage signal is input to a synapse circuit of a neural network circuit element including a neuron circuit that output the weight change pulse voltage signal, and a switching pulse voltage signal is input to a synapse circuit of a neural network circuit element other than the neural network circuit element including the neuron circuit that output the switching pulse voltage signal. the neural network circuit element changes the amplitude of the weight change pulse voltage signal on the basis of the error voltage signal generated by the error calculating circuit.",2017-10-17,"The title of the patent is neural network circuit and learning method for neural network circuit and its abstract is a neural network circuit includes an error calculating circuit that generates an error voltage signal having a magnitude in accordance with a time difference between an output signal and a teaching signal corresponding to the output signal. a weight change pulse voltage signal is input to a synapse circuit of a neural network circuit element including a neuron circuit that output the weight change pulse voltage signal, and a switching pulse voltage signal is input to a synapse circuit of a neural network circuit element other than the neural network circuit element including the neuron circuit that output the switching pulse voltage signal. the neural network circuit element changes the amplitude of the weight change pulse voltage signal on the basis of the error voltage signal generated by the error calculating circuit. dated 2017-10-17"
9792897,phoneme-expert assisted speech recognition and re-synthesis,"various implementations disclosed herein include an expert-assisted phoneme recognition neural network system configured to recognize phonemes within continuous large vocabulary speech sequences without using language specific models (“left-context”), look-ahead (“right-context”) information, or multi-pass sequence processing, and while operating within the resource constraints of low-power and real-time devices. to these ends, in various implementations, an expert-assisted phoneme recognition neural network system as described herein utilizes a-priori phonetic knowledge. phonetics is concerned with the configuration of the human vocal tract while speaking and acoustic consequences on vocalizations. while similar sounding phonemes are difficult to detect and are frequently misidentified by previously known neural networks, phonetic knowledge gives insight into what aspects of sound acoustics contain the strongest contrast between similar sounding phonemes. utilizing features that emphasize the respective second formants allows for more robust sound discrimination between these problematic phonemes.",2017-10-17,"The title of the patent is phoneme-expert assisted speech recognition and re-synthesis and its abstract is various implementations disclosed herein include an expert-assisted phoneme recognition neural network system configured to recognize phonemes within continuous large vocabulary speech sequences without using language specific models (“left-context”), look-ahead (“right-context”) information, or multi-pass sequence processing, and while operating within the resource constraints of low-power and real-time devices. to these ends, in various implementations, an expert-assisted phoneme recognition neural network system as described herein utilizes a-priori phonetic knowledge. phonetics is concerned with the configuration of the human vocal tract while speaking and acoustic consequences on vocalizations. while similar sounding phonemes are difficult to detect and are frequently misidentified by previously known neural networks, phonetic knowledge gives insight into what aspects of sound acoustics contain the strongest contrast between similar sounding phonemes. utilizing features that emphasize the respective second formants allows for more robust sound discrimination between these problematic phonemes. dated 2017-10-17"
9792900,generation of phoneme-experts for speech recognition,"various implementations disclosed herein include an expert-assisted phoneme recognition neural network system configured to recognize phonemes within continuous large vocabulary speech sequences without using language specific models (“left-context”), look-ahead (“right-context”) information, or multi-pass sequence processing, and while operating within the resource constraints of low-power and real-time devices. to these ends, in various implementations, an expert-assisted phoneme recognition neural network system as described herein utilizes a-priori phonetic knowledge. phonetics is concerned with the configuration of the human vocal tract while speaking and acoustic consequences on vocalizations. while similar sounding phonemes are difficult to detect and are frequently misidentified by previously known neural networks, phonetic knowledge gives insight into what aspects of sound acoustics contain the strongest contrast between similar sounding phonemes. utilizing features that emphasize the respective second formants allows for more robust sound discrimination between these problematic phonemes.",2017-10-17,"The title of the patent is generation of phoneme-experts for speech recognition and its abstract is various implementations disclosed herein include an expert-assisted phoneme recognition neural network system configured to recognize phonemes within continuous large vocabulary speech sequences without using language specific models (“left-context”), look-ahead (“right-context”) information, or multi-pass sequence processing, and while operating within the resource constraints of low-power and real-time devices. to these ends, in various implementations, an expert-assisted phoneme recognition neural network system as described herein utilizes a-priori phonetic knowledge. phonetics is concerned with the configuration of the human vocal tract while speaking and acoustic consequences on vocalizations. while similar sounding phonemes are difficult to detect and are frequently misidentified by previously known neural networks, phonetic knowledge gives insight into what aspects of sound acoustics contain the strongest contrast between similar sounding phonemes. utilizing features that emphasize the respective second formants allows for more robust sound discrimination between these problematic phonemes. dated 2017-10-17"
9795342,system and method for predicting acute cardiopulmonary events and survivability of a patient,"a method of producing an artificial neural network capable of predicting the survivability of a patient, including: storing in an electronic database patient health data comprising a plurality of sets of data, each set having at least one of a first parameter relating to heart rate variability data and a second parameter relating to vital sign data, each set further having a third parameter relating to patient survivability; providing a network of nodes interconnected to form an artificial neural network, the nodes comprising a plurality of artificial neurons, each artificial neuron having at least one input with an associated weight; and training the artificial neural network using the patient health data such that the associated weight of the at least one input of each artificial neuron is adjusted in response to respective first, second and third parameters of different sets of data from the patient health data.",2017-10-24,"The title of the patent is system and method for predicting acute cardiopulmonary events and survivability of a patient and its abstract is a method of producing an artificial neural network capable of predicting the survivability of a patient, including: storing in an electronic database patient health data comprising a plurality of sets of data, each set having at least one of a first parameter relating to heart rate variability data and a second parameter relating to vital sign data, each set further having a third parameter relating to patient survivability; providing a network of nodes interconnected to form an artificial neural network, the nodes comprising a plurality of artificial neurons, each artificial neuron having at least one input with an associated weight; and training the artificial neural network using the patient health data such that the associated weight of the at least one input of each artificial neuron is adjusted in response to respective first, second and third parameters of different sets of data from the patient health data. dated 2017-10-24"
9797799,intelligent adaptive system and method for monitoring leakage of oil pipeline networks based on big data,"the present invention relates to an intelligent adaptive system and method for monitoring leakage of oil pipeline networks based on big data. the present invention effectively analyzes a large amount of data collected on site within a reasonable time period and obtains a state of a pipeline network by an intelligent adaptive method, thereby obtaining a topological structure of a pipeline network. the present invention specifically adopts a flow balance method in combination with information conformance theory to analyze whether the pipeline network has leakage; small amount of leakage and slow leakage can be perfectly and accurately alarmed upon detection; as a generalized regression neural network is adopted to locate a leakage of the pipeline network, an accuracy of a result is increased. therefore, the present invention adopts a policy and intelligent adaptive method based on big data to solve problems of detecting and locating leakage of the pipeline network.",2017-10-24,"The title of the patent is intelligent adaptive system and method for monitoring leakage of oil pipeline networks based on big data and its abstract is the present invention relates to an intelligent adaptive system and method for monitoring leakage of oil pipeline networks based on big data. the present invention effectively analyzes a large amount of data collected on site within a reasonable time period and obtains a state of a pipeline network by an intelligent adaptive method, thereby obtaining a topological structure of a pipeline network. the present invention specifically adopts a flow balance method in combination with information conformance theory to analyze whether the pipeline network has leakage; small amount of leakage and slow leakage can be perfectly and accurately alarmed upon detection; as a generalized regression neural network is adopted to locate a leakage of the pipeline network, an accuracy of a result is increased. therefore, the present invention adopts a policy and intelligent adaptive method based on big data to solve problems of detecting and locating leakage of the pipeline network. dated 2017-10-24"
9798612,artifact correction using neural networks,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for correcting a corrupted data sample using a trained deep neural network, the method including obtaining a feature representation of a corrupted data sample; and modifying the feature representation of the corrupted data sample to generate a feature representation of a corrected data sample by iteratively processing a current version of the feature representation of the corrupted data sample using the trained deep neural network to generate a current corruption score for the current version of the feature representation of the corrupted data sample and generating a less-corrupted version of the feature representation by performing an iteration of gradient descent against the current version of the feature representation of the corrupted data sample to reduce the current corruption score.",2017-10-24,"The title of the patent is artifact correction using neural networks and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for correcting a corrupted data sample using a trained deep neural network, the method including obtaining a feature representation of a corrupted data sample; and modifying the feature representation of the corrupted data sample to generate a feature representation of a corrected data sample by iteratively processing a current version of the feature representation of the corrupted data sample using the trained deep neural network to generate a current corruption score for the current version of the feature representation of the corrupted data sample and generating a less-corrupted version of the feature representation by performing an iteration of gradient descent against the current version of the feature representation of the corrupted data sample to reduce the current corruption score. dated 2017-10-24"
9798751,method and apparatus for constructing a neuroscience-inspired artificial neural network,"a method and apparatus for constructing a neuroscience-inspired dynamic architecture (nida) for an artificial neural network is disclosed. the method comprises constructing, in one embodiment, an artificial neural network embodiment in a multi-dimensional space in memory such that a neuron is connected by a synapse to another neuron. the neuron and the synapse each have parameters and have features of long-term potentiation and long-term depression. furthermore, crossover and mutation are employed to select children of parents. through learning, an initial network may evolve into a different network when nida is applied to solve different problems of control, anomaly detection and classification over selected time units. the apparatus comprises in one embodiment a computational neuroscience-inspired artificial neural network with at least one affective network coupled to receive input data from an environment and to output data to the environment.",2017-10-24,"The title of the patent is method and apparatus for constructing a neuroscience-inspired artificial neural network and its abstract is a method and apparatus for constructing a neuroscience-inspired dynamic architecture (nida) for an artificial neural network is disclosed. the method comprises constructing, in one embodiment, an artificial neural network embodiment in a multi-dimensional space in memory such that a neuron is connected by a synapse to another neuron. the neuron and the synapse each have parameters and have features of long-term potentiation and long-term depression. furthermore, crossover and mutation are employed to select children of parents. through learning, an initial network may evolve into a different network when nida is applied to solve different problems of control, anomaly detection and classification over selected time units. the apparatus comprises in one embodiment a computational neuroscience-inspired artificial neural network with at least one affective network coupled to receive input data from an environment and to output data to the environment. dated 2017-10-24"
9799098,method and apparatus for image processing,"identifying objects in images is a difficult problem, particularly in cases an original image is noisy or has areas narrow in color or grayscale gradient. a technique employing a convolutional network has been identified to identify objects in such images in an automated and rapid manner. one example embodiment trains a convolutional network including multiple layers of filters. the filters are trained by learning and are arranged in successive layers and produce images having at least a same resolution as an original image. the filters are trained as a function of the original image or a desired image labeling; the image labels of objects identified in the original image are reported and may be used for segmentation. the technique can be applied to images of neural circuitry or electron microscopy, for example. the same technique can also be applied to correction of photographs or videos.",2017-10-24,"The title of the patent is method and apparatus for image processing and its abstract is identifying objects in images is a difficult problem, particularly in cases an original image is noisy or has areas narrow in color or grayscale gradient. a technique employing a convolutional network has been identified to identify objects in such images in an automated and rapid manner. one example embodiment trains a convolutional network including multiple layers of filters. the filters are trained by learning and are arranged in successive layers and produce images having at least a same resolution as an original image. the filters are trained as a function of the original image or a desired image labeling; the image labels of objects identified in the original image are reported and may be used for segmentation. the technique can be applied to images of neural circuitry or electron microscopy, for example. the same technique can also be applied to correction of photographs or videos. dated 2017-10-24"
9799327,speech recognition with attention-based recurrent neural networks,"methods, systems, and apparatus, including computer programs encoded on computer storage media for speech recognition. one method includes obtaining an input acoustic sequence, the input acoustic sequence representing an utterance, and the input acoustic sequence comprising a respective acoustic feature representation at each of a first number of time steps; processing the input acoustic sequence using a first neural network to convert the input acoustic sequence into an alternative representation for the input acoustic sequence; processing the alternative representation for the input acoustic sequence using an attention-based recurrent neural network (rnn) to generate, for each position in an output sequence order, a set of substring scores that includes a respective substring score for each substring in a set of substrings; and generating a sequence of substrings that represent a transcription of the utterance.",2017-10-24,"The title of the patent is speech recognition with attention-based recurrent neural networks and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media for speech recognition. one method includes obtaining an input acoustic sequence, the input acoustic sequence representing an utterance, and the input acoustic sequence comprising a respective acoustic feature representation at each of a first number of time steps; processing the input acoustic sequence using a first neural network to convert the input acoustic sequence into an alternative representation for the input acoustic sequence; processing the alternative representation for the input acoustic sequence using an attention-based recurrent neural network (rnn) to generate, for each position in an output sequence order, a set of substring scores that includes a respective substring score for each substring in a set of substrings; and generating a sequence of substrings that represent a transcription of the utterance. dated 2017-10-24"
9802599,vehicle lane placement,"a system of one or more computers can be configured to perform particular operations or actions by virtue of having software, firmware, hardware, or a combination of them installed on the system that in operation causes or cause the system to perform the actions. one or more computer programs can be configured to perform particular operations or actions by virtue of including instructions that, when executed by data processing apparatus, cause the apparatus to perform the actions. one general aspect includes sends video data and light detecting and ranging (lidar) data to a recurrent neural network (rnn) that includes feedback elements to identify a roadway feature. the system also sends the data to a dynamic convolutional neural network (dcnn) to identify the feature. output values are sent to a softmax decision network to aggregate the rnn and the dcnn output values and determine a vehicle positional location on the roadway.",2017-10-31,"The title of the patent is vehicle lane placement and its abstract is a system of one or more computers can be configured to perform particular operations or actions by virtue of having software, firmware, hardware, or a combination of them installed on the system that in operation causes or cause the system to perform the actions. one or more computer programs can be configured to perform particular operations or actions by virtue of including instructions that, when executed by data processing apparatus, cause the apparatus to perform the actions. one general aspect includes sends video data and light detecting and ranging (lidar) data to a recurrent neural network (rnn) that includes feedback elements to identify a roadway feature. the system also sends the data to a dynamic convolutional neural network (dcnn) to identify the feature. output values are sent to a softmax decision network to aggregate the rnn and the dcnn output values and determine a vehicle positional location on the roadway. dated 2017-10-31"
9805303,rotating data for neural network computations,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for computing a layer output for a convolutional neural network layer, the method comprising: receiving a plurality of activation inputs; forming a plurality of vector inputs from the plurality of activation inputs, each vector input comprising values from a distinct region within the multi-dimensional matrix; sending the plurality of vector inputs to one or more cells along a first dimension of the systolic array; generating a plurality of rotated kernel structures from each of the plurality of kernel; sending each kernel structure and each rotated kernel structure to one or more cells along a second dimension of the systolic array; causing the systolic array to generate an accumulated output based on the plurality of value inputs and the plurality of kernels; and generating the layer output from the accumulated output.",2017-10-31,"The title of the patent is rotating data for neural network computations and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for computing a layer output for a convolutional neural network layer, the method comprising: receiving a plurality of activation inputs; forming a plurality of vector inputs from the plurality of activation inputs, each vector input comprising values from a distinct region within the multi-dimensional matrix; sending the plurality of vector inputs to one or more cells along a first dimension of the systolic array; generating a plurality of rotated kernel structures from each of the plurality of kernel; sending each kernel structure and each rotated kernel structure to one or more cells along a second dimension of the systolic array; causing the systolic array to generate an accumulated output based on the plurality of value inputs and the plurality of kernels; and generating the layer output from the accumulated output. dated 2017-10-31"
9805304,prefetching weights for use in a neural network processor,"a circuit for performing neural network computations for a neural network, the circuit comprising: a systolic array comprising a plurality of cells; a weight fetcher unit configured to, for each of the plurality of neural network layers: send, for the neural network layer, a plurality of weight inputs to cells along a first dimension of the systolic array; and a plurality of weight sequencer units, each weight sequencer unit coupled to a distinct cell along the first dimension of the systolic array, the plurality of weight sequencer units configured to, for each of the plurality of neural network layers: shift, for the neural network layer, the plurality of weight inputs to cells along the second dimension of the systolic array over a plurality of clock cycles and where each cell is configured to compute a product of an activation input and a respective weight input using multiplication circuitry.",2017-10-31,"The title of the patent is prefetching weights for use in a neural network processor and its abstract is a circuit for performing neural network computations for a neural network, the circuit comprising: a systolic array comprising a plurality of cells; a weight fetcher unit configured to, for each of the plurality of neural network layers: send, for the neural network layer, a plurality of weight inputs to cells along a first dimension of the systolic array; and a plurality of weight sequencer units, each weight sequencer unit coupled to a distinct cell along the first dimension of the systolic array, the plurality of weight sequencer units configured to, for each of the plurality of neural network layers: shift, for the neural network layer, the plurality of weight inputs to cells along the second dimension of the systolic array over a plurality of clock cycles and where each cell is configured to compute a product of an activation input and a respective weight input using multiplication circuitry. dated 2017-10-31"
9805716,apparatus and method for large vocabulary continuous speech recognition,"provided is an apparatus for large vocabulary continuous speech recognition (lvcsr) based on a context-dependent deep neural network hidden markov model (cd-dnn-hmm) algorithm. the apparatus may include an extractor configured to extract acoustic model-state level information corresponding to an input speech signal from a training data model set using at least one of a first feature vector based on a gammatone filterbank signal analysis algorithm and a second feature vector based on a bottleneck algorithm, and a speech recognizer configured to provide a result of recognizing the input speech signal based on the extracted acoustic model-state level information.",2017-10-31,"The title of the patent is apparatus and method for large vocabulary continuous speech recognition and its abstract is provided is an apparatus for large vocabulary continuous speech recognition (lvcsr) based on a context-dependent deep neural network hidden markov model (cd-dnn-hmm) algorithm. the apparatus may include an extractor configured to extract acoustic model-state level information corresponding to an input speech signal from a training data model set using at least one of a first feature vector based on a gammatone filterbank signal analysis algorithm and a second feature vector based on a bottleneck algorithm, and a speech recognizer configured to provide a result of recognizing the input speech signal based on the extracted acoustic model-state level information. dated 2017-10-31"
9806072,super cmos devices on a microelectronics system,"this application is directed to a low cost ic solution that provides super cmos microelectronics macros. hereinafter, scmos refers to super cmos and schottky cmos. scmos device solutions includes a niche circuit element, such as complementary low threshold schottky barrier diode pairs (sbd) made by selected metal barrier contacts (co, ti, ni or other metal atoms or compounds) to p- and n- si beds of the cmos transistors. a dtl like new circuit topology and designed wide contents of broad product libraries, which used the integrated sbd and transistors (bjt, cmos, and flash versions) as basic components. the macros are composed of diodes that are selectively attached to the diffusion bed of the transistors, configuring them to form (i) generic logic gates, (ii) functional blocks of microprocessors and microcontrollers such as but not limited to data paths, multipliers, muliplier-accumaltors, (ii) memory cells and control circuits of various types (sram's with single or multiple read/write port(s), binary and ternary cam's), (iii) multiplexers, crossbar switches, switch matrices in network processors, graphics processors and other processors to implement a variety of communication protocols and algorithms of data processing engines for (iv) analytics, (v) block-chain and encryption-based security engines (vi) artificial neural networks with specific circuits to emulate or to implement a self-learning data processor similar to or derived from the neurons and synapses of human or animal brains, (vii) analog circuits and functional blocks from simple to the complicated including but not limited to power conversion, control and management either based on charge pumps or inductors, sensor signal amplifiers and conditioners, interface drivers, wireline data transceivers, oscillators and clock synthesizers with phase and/or delay locked loops, temperature monitors and controllers; all the above are built from discrete components to all grades of vlsi chips. solar photovoltaic electricity conversion, bio-lab-on-a-chip, hyperspectral imaging (capture/sensing and processing), wireless communication with various transceiver and/or transponder circuits for ranges of frequency that extend beyond a few 100 mhz, up to multi-thz, ambient energy harvesting either mechanical vibrations or antenna-based electromagnetic are newly extended or nacent fields of the scmos ic applications.",2017-10-31,"The title of the patent is super cmos devices on a microelectronics system and its abstract is this application is directed to a low cost ic solution that provides super cmos microelectronics macros. hereinafter, scmos refers to super cmos and schottky cmos. scmos device solutions includes a niche circuit element, such as complementary low threshold schottky barrier diode pairs (sbd) made by selected metal barrier contacts (co, ti, ni or other metal atoms or compounds) to p- and n- si beds of the cmos transistors. a dtl like new circuit topology and designed wide contents of broad product libraries, which used the integrated sbd and transistors (bjt, cmos, and flash versions) as basic components. the macros are composed of diodes that are selectively attached to the diffusion bed of the transistors, configuring them to form (i) generic logic gates, (ii) functional blocks of microprocessors and microcontrollers such as but not limited to data paths, multipliers, muliplier-accumaltors, (ii) memory cells and control circuits of various types (sram's with single or multiple read/write port(s), binary and ternary cam's), (iii) multiplexers, crossbar switches, switch matrices in network processors, graphics processors and other processors to implement a variety of communication protocols and algorithms of data processing engines for (iv) analytics, (v) block-chain and encryption-based security engines (vi) artificial neural networks with specific circuits to emulate or to implement a self-learning data processor similar to or derived from the neurons and synapses of human or animal brains, (vii) analog circuits and functional blocks from simple to the complicated including but not limited to power conversion, control and management either based on charge pumps or inductors, sensor signal amplifiers and conditioners, interface drivers, wireline data transceivers, oscillators and clock synthesizers with phase and/or delay locked loops, temperature monitors and controllers; all the above are built from discrete components to all grades of vlsi chips. solar photovoltaic electricity conversion, bio-lab-on-a-chip, hyperspectral imaging (capture/sensing and processing), wireless communication with various transceiver and/or transponder circuits for ranges of frequency that extend beyond a few 100 mhz, up to multi-thz, ambient energy harvesting either mechanical vibrations or antenna-based electromagnetic are newly extended or nacent fields of the scmos ic applications. dated 2017-10-31"
9807473,jointly modeling embedding and translation to bridge video and language,"video description generation using neural network training based on relevance and coherence is described. in some examples, long short-term memory with visual-semantic embedding (lstm-e) can maximize the probability of generating the next word given previous words and visual content and can create a visual-semantic embedding space for enforcing the relationship between the semantics of an entire sentence and visual content. lstm-e can include a 2-d and/or 3-d deep convolutional neural networks for learning powerful video representation, a deep recurrent neural network for generating sentences, and a joint embedding model for exploring the relationships between visual content and sentence semantics.",2017-10-31,"The title of the patent is jointly modeling embedding and translation to bridge video and language and its abstract is video description generation using neural network training based on relevance and coherence is described. in some examples, long short-term memory with visual-semantic embedding (lstm-e) can maximize the probability of generating the next word given previous words and visual content and can create a visual-semantic embedding space for enforcing the relationship between the semantics of an entire sentence and visual content. lstm-e can include a 2-d and/or 3-d deep convolutional neural networks for learning powerful video representation, a deep recurrent neural network for generating sentences, and a joint embedding model for exploring the relationships between visual content and sentence semantics. dated 2017-10-31"
9808216,material decomposition of multi-spectral x-ray projections using neural networks,"a method of processing x-ray images comprises training an artificial neural network to process multi-spectral x-ray projections to determine composition information about an object in terms of equivalent thickness of at least one basis material. the method further comprises providing a multi-spectral x-ray projection of an object, wherein the multi-spectral x-ray projection of the object contains energy content information describing the energy content of the multi-spectral x-ray projection. the multi-spectral x-ray projection is then processed with the artificial neural network to determine composition information about the object, and then the composition information about the object is provided.",2017-11-07,"The title of the patent is material decomposition of multi-spectral x-ray projections using neural networks and its abstract is a method of processing x-ray images comprises training an artificial neural network to process multi-spectral x-ray projections to determine composition information about an object in terms of equivalent thickness of at least one basis material. the method further comprises providing a multi-spectral x-ray projection of an object, wherein the multi-spectral x-ray projection of the object contains energy content information describing the energy content of the multi-spectral x-ray projection. the multi-spectral x-ray projection is then processed with the artificial neural network to determine composition information about the object, and then the composition information about the object is provided. dated 2017-11-07"
9811775,parallelizing neural networks during training,a parallel convolutional neural network is provided. the cnn is implemented by a plurality of convolutional neural networks each on a respective processing node. each cnn has a plurality of layers. a subset of the layers are interconnected between processing nodes such that activations are fed forward across nodes. the remaining subset is not so interconnected.,2017-11-07,The title of the patent is parallelizing neural networks during training and its abstract is a parallel convolutional neural network is provided. the cnn is implemented by a plurality of convolutional neural networks each on a respective processing node. each cnn has a plurality of layers. a subset of the layers are interconnected between processing nodes such that activations are fed forward across nodes. the remaining subset is not so interconnected. dated 2017-11-07
9813048,electronic comparison systems,"an electronic comparison system includes input stages that successively provide bits of code words. one-shots connected to respective stages successively provide a first bit value until receiving a bit having a non-preferred value concurrently with an enable signal, and then provide a second, different bit value. an enable circuit provides the enable signal if at least one of the one-shots is providing the first bit value. a neural network system includes a crossbar with row and column electrodes and resistive memory elements at their intersections. a writing circuit stores weights in the elements. a signal source applies signals to the row electrodes. comparators compare signals on the column electrodes to corresponding references using domain-wall neurons and store bit values in cmos latches by comparison with a threshold.",2017-11-07,"The title of the patent is electronic comparison systems and its abstract is an electronic comparison system includes input stages that successively provide bits of code words. one-shots connected to respective stages successively provide a first bit value until receiving a bit having a non-preferred value concurrently with an enable signal, and then provide a second, different bit value. an enable circuit provides the enable signal if at least one of the one-shots is providing the first bit value. a neural network system includes a crossbar with row and column electrodes and resistive memory elements at their intersections. a writing circuit stores weights in the elements. a signal source applies signals to the row electrodes. comparators compare signals on the column electrodes to corresponding references using domain-wall neurons and store bit values in cmos latches by comparison with a threshold. dated 2017-11-07"
9813246,encryption using biometric image-based key,"methods and systems according to the present disclosure improve upon known biometric security systems by not permanently storing (e.g., for later comparison as in known systems) the actual image of the biometric characteristic. instead, an image of a biometric identifier (e.g., retina, fingerprint, etc.) may be used to form a key which may be used to secure and provide access to data. the key may be formed, in embodiments, using a neural network and/or a random input (e.g., a vector of random characters), for example. the image of the biometric identifier may be discarded, and thus may not be vulnerable to theft. in an embodiment, the key may be used in a key-based encryption system.",2017-11-07,"The title of the patent is encryption using biometric image-based key and its abstract is methods and systems according to the present disclosure improve upon known biometric security systems by not permanently storing (e.g., for later comparison as in known systems) the actual image of the biometric characteristic. instead, an image of a biometric identifier (e.g., retina, fingerprint, etc.) may be used to form a key which may be used to secure and provide access to data. the key may be formed, in embodiments, using a neural network and/or a random input (e.g., a vector of random characters), for example. the image of the biometric identifier may be discarded, and thus may not be vulnerable to theft. in an embodiment, the key may be used in a key-based encryption system. dated 2017-11-07"
9813810,multi-microphone neural network for sound recognition,"a neural network is provided for recognition and enhancement of multi-channel sound signals received by multiple microphones, which need not be aligned in a linear array in a given environment. directions and distances of sound sources may also be detected by the neural network without the need for a beamformer connected to the microphones. the neural network may be trained by knowledge gained from free-field array impulse responses obtained in an anechoic chamber, array impulse responses that model simulated environments of different reverberation times, and array impulse responses obtained in actual environments.",2017-11-07,"The title of the patent is multi-microphone neural network for sound recognition and its abstract is a neural network is provided for recognition and enhancement of multi-channel sound signals received by multiple microphones, which need not be aligned in a linear array in a given environment. directions and distances of sound sources may also be detected by the neural network without the need for a beamformer connected to the microphones. the neural network may be trained by knowledge gained from free-field array impulse responses obtained in an anechoic chamber, array impulse responses that model simulated environments of different reverberation times, and array impulse responses obtained in actual environments. dated 2017-11-07"
9817847,neural network image curation control,"neural network image curation techniques are described. in one or more implementations, curation is controlled of images that represent a repository of images. a plurality of images of the repository are curated by one or more computing devices to select representative images of the repository. the curation includes calculating a score based on image and face aesthetics, jointly, for each of the plurality of images through processing by a neural network, ranking the plurality of images based on respective said scores, and selecting one or more of the plurality of images as one of the representative images of the repository based on the ranking and a determination that the one or more said images are not visually similar to images that have already been selected as one of the representative images of the repository.",2017-11-14,"The title of the patent is neural network image curation control and its abstract is neural network image curation techniques are described. in one or more implementations, curation is controlled of images that represent a repository of images. a plurality of images of the repository are curated by one or more computing devices to select representative images of the repository. the curation includes calculating a score based on image and face aesthetics, jointly, for each of the plurality of images through processing by a neural network, ranking the plurality of images based on respective said scores, and selecting one or more of the plurality of images as one of the representative images of the repository based on the ranking and a determination that the one or more said images are not visually similar to images that have already been selected as one of the representative images of the repository. dated 2017-11-14"
9818057,neural network circuit and learning method thereof,"in a neural network circuit element, a neuron circuit includes a waveform generating circuit for generating an analog pulse voltage, and a switching pulse voltage which is input as a first input signal to another neural network circuit element; a synapse circuit is configured such that the analog pulse voltage generated in the neuron circuit of the neural network circuit element including the synapse circuit is input to a third terminal of a variable resistance element of the synapse circuit, for a permissible input period, in the first input signal from another neural network circuit element; and the synapse circuit is configured such that the resistance value of the variable resistance element is changed in response to an electric potential difference between a first terminal and the third terminal, which occurs depending on a magnitude of the analog pulse voltage for the permissible input period.",2017-11-14,"The title of the patent is neural network circuit and learning method thereof and its abstract is in a neural network circuit element, a neuron circuit includes a waveform generating circuit for generating an analog pulse voltage, and a switching pulse voltage which is input as a first input signal to another neural network circuit element; a synapse circuit is configured such that the analog pulse voltage generated in the neuron circuit of the neural network circuit element including the synapse circuit is input to a third terminal of a variable resistance element of the synapse circuit, for a permissible input period, in the first input signal from another neural network circuit element; and the synapse circuit is configured such that the resistance value of the variable resistance element is changed in response to an electric potential difference between a first terminal and the third terminal, which occurs depending on a magnitude of the analog pulse voltage for the permissible input period. dated 2017-11-14"
9818059,exploiting input data sparsity in neural network compute units,"a computer-implemented method includes receiving, by a computing device, input activations and determining, by a controller of the computing device, whether each of the input activations has either a zero value or a non-zero value. the method further includes storing, in a memory bank of the computing device, at least one of the input activations. storing the at least one input activation includes generating an index comprising one or more memory address locations that have input activation values that are non-zero values. the method still further includes providing, by the controller and from the memory bank, at least one input activation onto a data bus that is accessible by one or more units of a computational array. the activations are provided, at least in part, from a memory address location associated with the index.",2017-11-14,"The title of the patent is exploiting input data sparsity in neural network compute units and its abstract is a computer-implemented method includes receiving, by a computing device, input activations and determining, by a controller of the computing device, whether each of the input activations has either a zero value or a non-zero value. the method further includes storing, in a memory bank of the computing device, at least one of the input activations. storing the at least one input activation includes generating an index comprising one or more memory address locations that have input activation values that are non-zero values. the method still further includes providing, by the controller and from the memory bank, at least one input activation onto a data bus that is accessible by one or more units of a computational array. the activations are provided, at least in part, from a memory address location associated with the index. dated 2017-11-14"
9818409,context-dependent modeling of phonemes,"methods, systems, and apparatus, including computer programs encoded on computer storage media for modeling phonemes. one method includes receiving an acoustic sequence, the acoustic sequence representing an utterance, and the acoustic sequence comprising a respective acoustic feature representation at each of a plurality of time steps; for each of the plurality of time steps: processing the acoustic feature representation through each of one or more recurrent neural network layers to generate a recurrent output; processing the recurrent output using a softmax output layer to generate a set of scores, the set of scores comprising a respective score for each of a plurality of context dependent vocabulary phonemes, the score for each context dependent vocabulary phoneme representing a likelihood that the context dependent vocabulary phoneme represents the utterance at the time step; and determining, from the scores for the plurality of time steps, a context dependent phoneme representation of the sequence.",2017-11-14,"The title of the patent is context-dependent modeling of phonemes and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media for modeling phonemes. one method includes receiving an acoustic sequence, the acoustic sequence representing an utterance, and the acoustic sequence comprising a respective acoustic feature representation at each of a plurality of time steps; for each of the plurality of time steps: processing the acoustic feature representation through each of one or more recurrent neural network layers to generate a recurrent output; processing the recurrent output using a softmax output layer to generate a set of scores, the set of scores comprising a respective score for each of a plurality of context dependent vocabulary phonemes, the score for each context dependent vocabulary phoneme representing a likelihood that the context dependent vocabulary phoneme represents the utterance at the time step; and determining, from the scores for the plurality of time steps, a context dependent phoneme representation of the sequence. dated 2017-11-14"
9818410,speech recognition with acoustic models,"methods, systems, and apparatus, including computer programs encoded on computer storage media for learning pronunciations from acoustic sequences. one method includes receiving an acoustic sequence, the acoustic sequence representing an utterance, and the acoustic sequence comprising a sequence of multiple frames of acoustic data at each of a plurality of time steps; stacking one or more frames of acoustic data to generate a sequence of modified frames of acoustic data; processing the sequence of modified frames of acoustic data through an acoustic modeling neural network comprising one or more recurrent neural network (rnn) layers and a final ctc output layer to generate a neural network output, wherein processing the sequence of modified frames of acoustic data comprises: subsampling the modified frames of acoustic data; and processing each subsampled modified frame of acoustic data through the acoustic modeling neural network.",2017-11-14,"The title of the patent is speech recognition with acoustic models and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media for learning pronunciations from acoustic sequences. one method includes receiving an acoustic sequence, the acoustic sequence representing an utterance, and the acoustic sequence comprising a sequence of multiple frames of acoustic data at each of a plurality of time steps; stacking one or more frames of acoustic data to generate a sequence of modified frames of acoustic data; processing the sequence of modified frames of acoustic data through an acoustic modeling neural network comprising one or more recurrent neural network (rnn) layers and a final ctc output layer to generate a neural network output, wherein processing the sequence of modified frames of acoustic data comprises: subsampling the modified frames of acoustic data; and processing each subsampled modified frame of acoustic data through the acoustic modeling neural network. dated 2017-11-14"
9824304,determination of font similarity,"font recognition and similarity determination techniques and systems are described. in a first example, localization techniques are described to train a model using machine learning (e.g., a convolutional neural network) using training images. the model is then used to localize text in a subsequently received image, and may do so automatically and without user intervention, e.g., without specifying any of the edges of a bounding box. in a second example, a deep neural network is directly learned as an embedding function of a model that is usable to determine font similarity. in a third example, techniques are described that leverage attributes described in metadata associated with fonts as part of font recognition and similarity determinations.",2017-11-21,"The title of the patent is determination of font similarity and its abstract is font recognition and similarity determination techniques and systems are described. in a first example, localization techniques are described to train a model using machine learning (e.g., a convolutional neural network) using training images. the model is then used to localize text in a subsequently received image, and may do so automatically and without user intervention, e.g., without specifying any of the edges of a bounding box. in a second example, a deep neural network is directly learned as an embedding function of a model that is usable to determine font similarity. in a third example, techniques are described that leverage attributes described in metadata associated with fonts as part of font recognition and similarity determinations. dated 2017-11-21"
9824692,end-to-end speaker recognition using deep neural network,"the present invention is directed to a deep neural network (dnn) having a triplet network architecture, which is suitable to perform speaker recognition. in particular, the dnn includes three feed-forward neural networks, which are trained according to a batch process utilizing a cohort set of negative training samples. after each batch of training samples is processed, the dnn may be trained according to a loss function, e.g., utilizing a cosine measure of similarity between respective samples, along with positive and negative margins, to provide a robust representation of voiceprints.",2017-11-21,"The title of the patent is end-to-end speaker recognition using deep neural network and its abstract is the present invention is directed to a deep neural network (dnn) having a triplet network architecture, which is suitable to perform speaker recognition. in particular, the dnn includes three feed-forward neural networks, which are trained according to a batch process utilizing a cohort set of negative training samples. after each batch of training samples is processed, the dnn may be trained according to a loss function, e.g., utilizing a cosine measure of similarity between respective samples, along with positive and negative margins, to provide a robust representation of voiceprints. dated 2017-11-21"
9830315,sequence-based structured prediction for semantic parsing,"a system and method are provided which employ a neural network model which has been trained to predict a sequentialized form for an input text sequence. the sequentialized form includes a sequence of symbols. the neural network model includes an encoder which generates a representation of the input text sequence based on a representation of n-grams in the text sequence and a decoder which sequentially predicts a next symbol of the sequentialized form based on the representation and a predicted prefix of the sequentialized form. given an input text sequence, a sequentialized form is predicted with the trained neural network model. the sequentialized form is converted to a structured form and information based on the structured form is output.",2017-11-28,"The title of the patent is sequence-based structured prediction for semantic parsing and its abstract is a system and method are provided which employ a neural network model which has been trained to predict a sequentialized form for an input text sequence. the sequentialized form includes a sequence of symbols. the neural network model includes an encoder which generates a representation of the input text sequence based on a representation of n-grams in the text sequence and a decoder which sequentially predicts a next symbol of the sequentialized form based on the representation and a predicted prefix of the sequentialized form. given an input text sequence, a sequentialized form is predicted with the trained neural network model. the sequentialized form is converted to a structured form and information based on the structured form is output. dated 2017-11-28"
9830526,generating image features based on robust feature-learning,"techniques for increasing robustness of a convolutional neural network based on training that uses multiple datasets and multiple tasks are described. for example, a computer system trains the convolutional neural network across multiple datasets and multiple tasks. the convolutional neural network is configured for learning features from images and accordingly generating feature vectors. by using multiple datasets and multiple tasks, the robustness of the convolutional neural network is increased. a feature vector of an image is used to apply an image-related operation to the image. for example, the image is classified, indexed, or objects in the image are tagged based on the feature vector. because the robustness is increased, the accuracy of the generating feature vectors is also increased. hence, the overall quality of an image service is enhanced, where the image service relies on the image-related operation.",2017-11-28,"The title of the patent is generating image features based on robust feature-learning and its abstract is techniques for increasing robustness of a convolutional neural network based on training that uses multiple datasets and multiple tasks are described. for example, a computer system trains the convolutional neural network across multiple datasets and multiple tasks. the convolutional neural network is configured for learning features from images and accordingly generating feature vectors. by using multiple datasets and multiple tasks, the robustness of the convolutional neural network is increased. a feature vector of an image is used to apply an image-related operation to the image. for example, the image is classified, indexed, or objects in the image are tagged based on the feature vector. because the robustness is increased, the accuracy of the generating feature vectors is also increased. hence, the overall quality of an image service is enhanced, where the image service relies on the image-related operation. dated 2017-11-28"
9830529,end-to-end saliency mapping via probability distribution prediction,a method for generating a system for predicting saliency in an image and method of use of the prediction system are described. attention maps for each of a set of training images are used to train the system. the training includes passing the training images though a neural network and optimizing an objective function over the training set which is based on a distance measure computed between a first probability distribution computed for a saliency map output by the neural network and a second probability distribution computed for the attention map for the respective training image. the trained neural network is suited to generation of saliency maps for new images.,2017-11-28,The title of the patent is end-to-end saliency mapping via probability distribution prediction and its abstract is a method for generating a system for predicting saliency in an image and method of use of the prediction system are described. attention maps for each of a set of training images are used to train the system. the training includes passing the training images though a neural network and optimizing an objective function over the training set which is based on a distance measure computed between a first probability distribution computed for a saliency map output by the neural network and a second probability distribution computed for the attention map for the respective training image. the trained neural network is suited to generation of saliency maps for new images. dated 2017-11-28
9830534,object recognition,"approaches introduce a pre-processing and post-processing framework to a neural network-based approach to identify items represented in an image. for example, a classifier that is trained on several categories can be provided. an image that includes a representation of an item of interest is obtained. rotated versions of the image are generated and each of a subset of the rotated images is analyzed to determine a probability that a respective image includes an instance of a particular category. the probabilities can be used to determine a probability distribution of output category data, and the data can be analyzed to select an image of the rotated versions of the image. thereafter, a categorization tree can then be utilized, whereby for the item of interest represented the image, the category of the item can be determined. the determined category can be provided to an item retrieval algorithm to determine primary content for the item of interest. this information also can be used to determine recommendations, advertising, or other supplemental content, within a specific category, to be displayed with the primary content.",2017-11-28,"The title of the patent is object recognition and its abstract is approaches introduce a pre-processing and post-processing framework to a neural network-based approach to identify items represented in an image. for example, a classifier that is trained on several categories can be provided. an image that includes a representation of an item of interest is obtained. rotated versions of the image are generated and each of a subset of the rotated images is analyzed to determine a probability that a respective image includes an instance of a particular category. the probabilities can be used to determine a probability distribution of output category data, and the data can be analyzed to select an image of the rotated versions of the image. thereafter, a categorization tree can then be utilized, whereby for the item of interest represented the image, the category of the item can be determined. the determined category can be provided to an item retrieval algorithm to determine primary content for the item of interest. this information also can be used to determine recommendations, advertising, or other supplemental content, within a specific category, to be displayed with the primary content. dated 2017-11-28"
9830709,video analysis with convolutional attention recurrent neural networks,a method of processing data within a convolutional attention recurrent neural network (rnn) includes generating a current multi-dimensional attention map. the current multi-dimensional attention map indicates areas of interest in a first frame from a sequence of spatio-temporal data. the method further includes receiving a multi-dimensional feature map. the method also includes convolving the current multi-dimensional attention map and the multi-dimensional feature map to obtain a multi-dimensional hidden state and a next multi-dimensional attention map. the method identifies a class of interest in the first frame based on the multi-dimensional hidden state and training data.,2017-11-28,The title of the patent is video analysis with convolutional attention recurrent neural networks and its abstract is a method of processing data within a convolutional attention recurrent neural network (rnn) includes generating a current multi-dimensional attention map. the current multi-dimensional attention map indicates areas of interest in a first frame from a sequence of spatio-temporal data. the method further includes receiving a multi-dimensional feature map. the method also includes convolving the current multi-dimensional attention map and the multi-dimensional feature map to obtain a multi-dimensional hidden state and a next multi-dimensional attention map. the method identifies a class of interest in the first frame based on the multi-dimensional hidden state and training data. dated 2017-11-28
9836484,systems and methods that leverage deep learning to selectively store images at a mobile image capture device,"the present disclosure provides an image capture, curation, and editing system that includes a resource-efficient mobile image capture device that continuously captures images. the mobile image capture device is operable to input an image into at least one neural network and to receive at least one descriptor of the desirability of a scene depicted by the image as an output of the at least one neural network. the mobile image capture device is operable to determine, based at least in part on the at least one descriptor of the desirability of the scene of the image, whether to store a second copy of such image in a non-volatile memory of the mobile image capture device or to discard a first copy of such image from a temporary image buffer without storing the second copy of such image in the non-volatile memory.",2017-12-05,"The title of the patent is systems and methods that leverage deep learning to selectively store images at a mobile image capture device and its abstract is the present disclosure provides an image capture, curation, and editing system that includes a resource-efficient mobile image capture device that continuously captures images. the mobile image capture device is operable to input an image into at least one neural network and to receive at least one descriptor of the desirability of a scene depicted by the image as an output of the at least one neural network. the mobile image capture device is operable to determine, based at least in part on the at least one descriptor of the desirability of the scene of the image, whether to store a second copy of such image in a non-volatile memory of the mobile image capture device or to discard a first copy of such image from a temporary image buffer without storing the second copy of such image in the non-volatile memory. dated 2017-12-05"
9836641,generating numeric embeddings of images,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating numeric embeddings of images. one of the methods includes obtaining training images; generating a plurality of triplets of training images; and training a neural network on each of the triplets to determine trained values of a plurality of parameters of the neural network, wherein training the neural network comprises, for each of the triplets: processing the anchor image in the triplet using the neural network to generate a numeric embedding of the anchor image; processing the positive image in the triplet using the neural network to generate a numeric embedding of the positive image; processing the negative image in the triplet using the neural network to generate a numeric embedding of the negative image; computing a triplet loss; and adjusting the current values of the parameters of the neural network using the triplet loss.",2017-12-05,"The title of the patent is generating numeric embeddings of images and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating numeric embeddings of images. one of the methods includes obtaining training images; generating a plurality of triplets of training images; and training a neural network on each of the triplets to determine trained values of a plurality of parameters of the neural network, wherein training the neural network comprises, for each of the triplets: processing the anchor image in the triplet using the neural network to generate a numeric embedding of the anchor image; processing the positive image in the triplet using the neural network to generate a numeric embedding of the positive image; processing the negative image in the triplet using the neural network to generate a numeric embedding of the negative image; computing a triplet loss; and adjusting the current values of the parameters of the neural network using the triplet loss. dated 2017-12-05"
9836691,neural network instruction set architecture,"a computer-implemented method that includes receiving, by a processing unit, an instruction that specifies data values for performing a tensor computation. in response to receiving the instruction, the method may include, performing, by the processing unit, the tensor computation by executing a loop nest comprising a plurality of loops, wherein a structure of the loop nest is defined based on one or more of the data values of the instruction. the tensor computation can be at least a portion of a computation of a neural network layer. the data values specified by the instruction may comprise a value that specifies a type of the neural network layer, and the structure of the loop nest can be defined at least in part by the type of the neural network layer.",2017-12-05,"The title of the patent is neural network instruction set architecture and its abstract is a computer-implemented method that includes receiving, by a processing unit, an instruction that specifies data values for performing a tensor computation. in response to receiving the instruction, the method may include, performing, by the processing unit, the tensor computation by executing a loop nest comprising a plurality of loops, wherein a structure of the loop nest is defined based on one or more of the data values of the instruction. the tensor computation can be at least a portion of a computation of a neural network layer. the data values specified by the instruction may comprise a value that specifies a type of the neural network layer, and the structure of the loop nest can be defined at least in part by the type of the neural network layer. dated 2017-12-05"
9836692,implementing neural networks in fixed point arithmetic computing systems,"methods, systems, and computer storage media for implementing neural networks in fixed point arithmetic computing systems. in one aspect, a method includes the actions of receiving a request to process a neural network using a processing system that performs neural network computations using fixed point arithmetic; for each node of each layer of the neural network, determining a respective scaling value for the node from the respective set of floating point weight values for the node; and converting each floating point weight value of the node into a corresponding fixed point weight value using the respective scaling value for the node to generate a set of fixed point weight values for the node; and providing the sets of fixed point floating point weight values for the nodes to the processing system for use in processing inputs using the neural network.",2017-12-05,"The title of the patent is implementing neural networks in fixed point arithmetic computing systems and its abstract is methods, systems, and computer storage media for implementing neural networks in fixed point arithmetic computing systems. in one aspect, a method includes the actions of receiving a request to process a neural network using a processing system that performs neural network computations using fixed point arithmetic; for each node of each layer of the neural network, determining a respective scaling value for the node from the respective set of floating point weight values for the node; and converting each floating point weight value of the node into a corresponding fixed point weight value using the respective scaling value for the node to generate a set of fixed point weight values for the node; and providing the sets of fixed point floating point weight values for the nodes to the processing system for use in processing inputs using the neural network. dated 2017-12-05"
9836819,systems and methods for selective retention and editing of images captured by mobile image capture device,"the present disclosure provides an image capture, curation, and editing system that includes a resource-efficient mobile image capture device that continuously captures images. the mobile image capture device is operable to input an image into at least one neural network and to receive at least one descriptor of the desirability of a scene depicted by the image as an output of the at least one neural network. the mobile image capture device is operable to determine, based at least in part on the at least one descriptor of the desirability of the scene of the image, whether to store a second copy of such image and/or one or more contemporaneously captured images in a non-volatile memory of the mobile image capture device or to discard a first copy of such image from a temporary image buffer without storing the second copy of such image in the non-volatile memory.",2017-12-05,"The title of the patent is systems and methods for selective retention and editing of images captured by mobile image capture device and its abstract is the present disclosure provides an image capture, curation, and editing system that includes a resource-efficient mobile image capture device that continuously captures images. the mobile image capture device is operable to input an image into at least one neural network and to receive at least one descriptor of the desirability of a scene depicted by the image as an output of the at least one neural network. the mobile image capture device is operable to determine, based at least in part on the at least one descriptor of the desirability of the scene of the image, whether to store a second copy of such image and/or one or more contemporaneously captured images in a non-volatile memory of the mobile image capture device or to discard a first copy of such image from a temporary image buffer without storing the second copy of such image in the non-volatile memory. dated 2017-12-05"
9836820,image upsampling using global and local constraints,a method upsamples an image using a non-linear fully connected neural network to produce only global details of an upsampled image and interpolates the image to produce a smooth upsampled image. the method concatenates the global details and the smooth upsampled image into a tensor and applies a sequence of nonlinear convolutions to the tensor using a convolutional neural network to produce the upsampled image.,2017-12-05,The title of the patent is image upsampling using global and local constraints and its abstract is a method upsamples an image using a non-linear fully connected neural network to produce only global details of an upsampled image and interpolates the image to produce a smooth upsampled image. the method concatenates the global details and the smooth upsampled image into a tensor and applies a sequence of nonlinear convolutions to the tensor using a convolutional neural network to produce the upsampled image. dated 2017-12-05
9836853,three-dimensional convolutional neural networks for video highlight detection,"a three-dimensional convolutional neural network may include a preliminary layer group, one or more intermediate layer groups, a final layer group, and/or other layers/layer groups. the preliminary layer group may include an input layer, a preliminary three-dimensional padding layer, a preliminary three-dimensional convolution layer, a preliminary activation layer, a preliminary normalization layer, and a preliminary downsampling layer. one or more intermediate layer groups may include an intermediate three-dimensional squeeze layer, a first intermediate normalization layer, an intermediate three-dimensional padding layer, a first intermediate three-dimensional expand layer, a second intermediate three-dimensional expand layer, an intermediate concatenation layer, a second intermediate normalization layer, an intermediate activation layer, and an intermediate combination layer. the final layer group may include a final dropout layer, a final three-dimensional convolution layer, a final activation layer, a final normalization layer, a final three-dimensional downsampling layer, and a final flatten layer.",2017-12-05,"The title of the patent is three-dimensional convolutional neural networks for video highlight detection and its abstract is a three-dimensional convolutional neural network may include a preliminary layer group, one or more intermediate layer groups, a final layer group, and/or other layers/layer groups. the preliminary layer group may include an input layer, a preliminary three-dimensional padding layer, a preliminary three-dimensional convolution layer, a preliminary activation layer, a preliminary normalization layer, and a preliminary downsampling layer. one or more intermediate layer groups may include an intermediate three-dimensional squeeze layer, a first intermediate normalization layer, an intermediate three-dimensional padding layer, a first intermediate three-dimensional expand layer, a second intermediate three-dimensional expand layer, an intermediate concatenation layer, a second intermediate normalization layer, an intermediate activation layer, and an intermediate combination layer. the final layer group may include a final dropout layer, a final three-dimensional convolution layer, a final activation layer, a final normalization layer, a final three-dimensional downsampling layer, and a final flatten layer. dated 2017-12-05"
9842106,method and system for role dependent context sensitive spoken and textual language understanding with neural networks,"a method and system processes utterances that are acquired either from an automatic speech recognition (asr) system or text. the utterances have associated identities of each party, such as role a utterances and role b utterances. the information corresponding to utterances, such as word sequence and identity, are converted to features. each feature is received in an input layer of a neural network (nn). a dimensionality of each feature is reduced, in a projection layer of the nn, to produce a reduced dimensional feature. the reduced dimensional feature is processed to provide probabilities of labels for the utterances.",2017-12-12,"The title of the patent is method and system for role dependent context sensitive spoken and textual language understanding with neural networks and its abstract is a method and system processes utterances that are acquired either from an automatic speech recognition (asr) system or text. the utterances have associated identities of each party, such as role a utterances and role b utterances. the information corresponding to utterances, such as word sequence and identity, are converted to features. each feature is received in an input layer of a neural network (nn). a dimensionality of each feature is reduced, in a projection layer of the nn, to produce a reduced dimensional feature. the reduced dimensional feature is processed to provide probabilities of labels for the utterances. dated 2017-12-12"
9842293,batch processing in a neural network processor,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating a respective neural network output for each of a plurality of inputs, the method comprising, for each of the neural network layers: receiving a plurality of inputs to be processed at the neural network layer; forming one or more batches of inputs from the plurality of inputs, each batch having a number of inputs up to the respective batch size for the neural network layer; selecting a number of the one or more batches of inputs to process, where a count of the inputs in the number of the one or more batches is greater than or equal to the respective associated batch size of a subsequent layer in the sequence; and processing the number of the one or more batches of inputs to generate the respective neural network layer output.",2017-12-12,"The title of the patent is batch processing in a neural network processor and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating a respective neural network output for each of a plurality of inputs, the method comprising, for each of the neural network layers: receiving a plurality of inputs to be processed at the neural network layer; forming one or more batches of inputs from the plurality of inputs, each batch having a number of inputs up to the respective batch size for the neural network layer; selecting a number of the one or more batches of inputs to process, where a count of the inputs in the number of the one or more batches is greater than or equal to the respective associated batch size of a subsequent layer in the sequence; and processing the number of the one or more batches of inputs to generate the respective neural network layer output. dated 2017-12-12"
9842585,multilingual deep neural network,"described herein are various technologies pertaining to a multilingual deep neural network (mdnn). the mdnn includes a plurality of hidden layers, wherein values for weight parameters of the plurality of hidden layers are learned during a training phase based upon training data in terms of acoustic raw features for multiple languages. the mdnn further includes softmax layers that are trained for each target language separately, making use of the hidden layer values trained jointly with multiple source languages. the mdnn is adaptable, such that a new softmax layer may be added on top of the existing hidden layers, where the new softmax layer corresponds to a new target language.",2017-12-12,"The title of the patent is multilingual deep neural network and its abstract is described herein are various technologies pertaining to a multilingual deep neural network (mdnn). the mdnn includes a plurality of hidden layers, wherein values for weight parameters of the plurality of hidden layers are learned during a training phase based upon training data in terms of acoustic raw features for multiple languages. the mdnn further includes softmax layers that are trained for each target language separately, making use of the hidden layer values trained jointly with multiple source languages. the mdnn is adaptable, such that a new softmax layer may be added on top of the existing hidden layers, where the new softmax layer corresponds to a new target language. dated 2017-12-12"
9842610,training deep neural network for acoustic modeling in speech recognition,a method is provided for training a deep neural network (dnn) for acoustic modeling in speech recognition. the method includes reading central frames and side frames as input frames from a memory. the side frames are preceding side frames preceding the central frames and/or succeeding side frames succeeding the central frames. the method further includes executing pre-training for only the central frames or both the central frames and the side frames and fine-tuning for the central frames and the side frames so as to emphasize connections between acoustic features in the central frames and units of the bottom layer in hidden layer of the dnn.,2017-12-12,The title of the patent is training deep neural network for acoustic modeling in speech recognition and its abstract is a method is provided for training a deep neural network (dnn) for acoustic modeling in speech recognition. the method includes reading central frames and side frames as input frames from a memory. the side frames are preceding side frames preceding the central frames and/or succeeding side frames succeeding the central frames. the method further includes executing pre-training for only the central frames or both the central frames and the side frames and fine-tuning for the central frames and the side frames so as to emphasize connections between acoustic features in the central frames and units of the bottom layer in hidden layer of the dnn. dated 2017-12-12
9843837,cross-platform analysis,"a method includes receiving, at a processor, a first data stream from a first platform and a second data stream from a second platform. the first data stream includes content and the second data stream includes the content. the method also includes performing an analysis operation on the first data stream and the second data stream to interpret the content. performing the analysis operation includes performing a statistical analysis on the first data stream and the second data stream using one or more artificial neural network (ann) nodes of an analytical network. performing the analysis operation also includes performing a syntactic analysis on the first data stream and the second data stream using one or more markov logic network (mln) nodes of the analytical network.",2017-12-12,"The title of the patent is cross-platform analysis and its abstract is a method includes receiving, at a processor, a first data stream from a first platform and a second data stream from a second platform. the first data stream includes content and the second data stream includes the content. the method also includes performing an analysis operation on the first data stream and the second data stream to interpret the content. performing the analysis operation includes performing a statistical analysis on the first data stream and the second data stream using one or more artificial neural network (ann) nodes of an analytical network. performing the analysis operation also includes performing a syntactic analysis on the first data stream and the second data stream using one or more markov logic network (mln) nodes of the analytical network. dated 2017-12-12"
9846444,method for controlling and adjusting fans of electronic apparatus,"a method for controlling and adjusting fans of the electronic apparatus is disclosed and comprises following steps: controlling a fan to operate according to a default value after an electronic apparatus is booted; obtaining operating watt value of a cpu after detecting that the temperature of the cpu reaches a set-point; inquiring one of a learning table and a pre-established neural-network data array for obtaining a pwm value and p,i,d parameters corresponding to the operating watt value; performing an error adjustment process to the obtained pwm value through a pid controller; controlling the operation of the fan according to the adjusted pwm value; storing the adjusted pwm value to the learning table if the temperature of the cpu equals the set-point; and, continuing to obtain, adjust and store the pwm value to continuously control the operation of the fan before the electronic apparatus is powered off.",2017-12-19,"The title of the patent is method for controlling and adjusting fans of electronic apparatus and its abstract is a method for controlling and adjusting fans of the electronic apparatus is disclosed and comprises following steps: controlling a fan to operate according to a default value after an electronic apparatus is booted; obtaining operating watt value of a cpu after detecting that the temperature of the cpu reaches a set-point; inquiring one of a learning table and a pre-established neural-network data array for obtaining a pwm value and p,i,d parameters corresponding to the operating watt value; performing an error adjustment process to the obtained pwm value through a pid controller; controlling the operation of the fan according to the adjusted pwm value; storing the adjusted pwm value to the learning table if the temperature of the cpu equals the set-point; and, continuing to obtain, adjust and store the pwm value to continuously control the operation of the fan before the electronic apparatus is powered off. dated 2017-12-19"
9846807,detecting eye corners,"a method and system for detecting eye corners using neural network classifiers is described. after an eye image is received, the eye image may be processed by at least two neural network classifiers including an inner eye corner neural network classifier and an outer eye corner neural network classifier. the neural network classifiers provide periocular information including a distance or coordinates of an eye corner location from a center of an iris of the eye, and an outcome of whether the eye corner is an inner eye corner or an outer eye corner. output from the various neural network classifiers are combined to generate a decision on the location of eye corners in an eye image.",2017-12-19,"The title of the patent is detecting eye corners and its abstract is a method and system for detecting eye corners using neural network classifiers is described. after an eye image is received, the eye image may be processed by at least two neural network classifiers including an inner eye corner neural network classifier and an outer eye corner neural network classifier. the neural network classifiers provide periocular information including a distance or coordinates of an eye corner location from a center of an iris of the eye, and an outcome of whether the eye corner is an inner eye corner or an outer eye corner. output from the various neural network classifiers are combined to generate a decision on the location of eye corners in an eye image. dated 2017-12-19"
9846836,modeling interestingness with deep neural networks,"an “interestingness modeler” uses deep neural networks to learn deep semantic models (dsm) of “interestingness.” the dsm, consisting of two branches of deep neural networks or their convolutional versions, identifies and predicts target documents that would interest users reading source documents. the learned model observes, identifies, and detects naturally occurring signals of interestingness in click transitions between source and target documents derived from web browser logs. interestingness is modeled with deep neural networks that map source-target document pairs to feature vectors in a latent space, trained on document transitions in view of a “context” and optional “focus” of source and target documents. network parameters are learned to minimize distances between source documents and their corresponding “interesting” targets in that space. the resulting interestingness model has applicable uses, including, but not limited to, contextual entity searches, automatic text highlighting, prefetching documents of likely interest, automated content recommendation, automated advertisement placement, etc.",2017-12-19,"The title of the patent is modeling interestingness with deep neural networks and its abstract is an “interestingness modeler” uses deep neural networks to learn deep semantic models (dsm) of “interestingness.” the dsm, consisting of two branches of deep neural networks or their convolutional versions, identifies and predicts target documents that would interest users reading source documents. the learned model observes, identifies, and detects naturally occurring signals of interestingness in click transitions between source and target documents derived from web browser logs. interestingness is modeled with deep neural networks that map source-target document pairs to feature vectors in a latent space, trained on document transitions in view of a “context” and optional “focus” of source and target documents. network parameters are learned to minimize distances between source documents and their corresponding “interesting” targets in that space. the resulting interestingness model has applicable uses, including, but not limited to, contextual entity searches, automatic text highlighting, prefetching documents of likely interest, automated content recommendation, automated advertisement placement, etc. dated 2017-12-19"
9846837,neural network instruction set architecture,"a computer-implemented method that includes receiving, by a processing unit, an instruction that specifies data values for performing a tensor computation. in response to receiving the instruction, the method may include, performing, by the processing unit, the tensor computation by executing a loop nest comprising a plurality of loops, wherein a structure of the loop nest is defined based on one or more of the data values of the instruction. the tensor computation can be at least a portion of a computation of a neural network layer. the data values specified by the instruction may comprise a value that specifies a type of the neural network layer, and the structure of the loop nest can be defined at least in part by the type of the neural network layer.",2017-12-19,"The title of the patent is neural network instruction set architecture and its abstract is a computer-implemented method that includes receiving, by a processing unit, an instruction that specifies data values for performing a tensor computation. in response to receiving the instruction, the method may include, performing, by the processing unit, the tensor computation by executing a loop nest comprising a plurality of loops, wherein a structure of the loop nest is defined based on one or more of the data values of the instruction. the tensor computation can be at least a portion of a computation of a neural network layer. the data values specified by the instruction may comprise a value that specifies a type of the neural network layer, and the structure of the loop nest can be defined at least in part by the type of the neural network layer. dated 2017-12-19"
9846839,"systems and methods for real-time forecasting and predicting of electrical peaks and managing the energy, health, reliability, and performance of electrical power systems based on an artificial adaptive neural network","a system for utilizing a neural network to make real-time predictions about the health, reliability, and performance of a monitored system are disclosed. the system includes a data acquisition component, a power analytics server and a client terminal. the data acquisition component acquires real-time data output from the electrical system. the power analytics server is comprised of a virtual system modeling engine, an analytics engine, an adaptive prediction engine. the virtual system modeling engine generates predicted data output for the electrical system. the analytics engine monitors real-time data output and predicted data output of the electrical system. the adaptive prediction engine can be configured to forecast an aspect of the monitored system using a neural network algorithm. the adaptive prediction engine is further configured to process the real-time data output and automatically optimize the neural network algorithm by minimizing a measure of error between the real-time data output and an estimated data output predicted by the neural network algorithm.",2017-12-19,"The title of the patent is systems and methods for real-time forecasting and predicting of electrical peaks and managing the energy, health, reliability, and performance of electrical power systems based on an artificial adaptive neural network and its abstract is a system for utilizing a neural network to make real-time predictions about the health, reliability, and performance of a monitored system are disclosed. the system includes a data acquisition component, a power analytics server and a client terminal. the data acquisition component acquires real-time data output from the electrical system. the power analytics server is comprised of a virtual system modeling engine, an analytics engine, an adaptive prediction engine. the virtual system modeling engine generates predicted data output for the electrical system. the analytics engine monitors real-time data output and predicted data output of the electrical system. the adaptive prediction engine can be configured to forecast an aspect of the monitored system using a neural network algorithm. the adaptive prediction engine is further configured to process the real-time data output and automatically optimize the neural network algorithm by minimizing a measure of error between the real-time data output and an estimated data output predicted by the neural network algorithm. dated 2017-12-19"
9846840,semantic class localization in images,"semantic class localization techniques and systems are described. in one or more implementation, a technique is employed to back communicate relevancies of aggregations back through layers of a neural network. through use of these relevancies, activation relevancy maps are created that describe relevancy of portions of the image to the classification of the image as corresponding to a semantic class. in this way, the semantic class is localized to portions of the image. this may be performed through communication of positive and not negative relevancies, use of contrastive attention maps to different between semantic classes and even within a same semantic class through use of a self-contrastive technique.",2017-12-19,"The title of the patent is semantic class localization in images and its abstract is semantic class localization techniques and systems are described. in one or more implementation, a technique is employed to back communicate relevancies of aggregations back through layers of a neural network. through use of these relevancies, activation relevancy maps are created that describe relevancy of portions of the image to the classification of the image as corresponding to a semantic class. in this way, the semantic class is localized to portions of the image. this may be performed through communication of positive and not negative relevancies, use of contrastive attention maps to different between semantic classes and even within a same semantic class through use of a self-contrastive technique. dated 2017-12-19"
9847974,image document processing in a client-server system including privacy-preserving text recognition,"disclosed are devices and methods for processing an image document in a client-server environment such that privacy of text information contained in the image document is preserved. specifically, in a client-server environment, an image document can be processed using a local computerized device of a client to create an obfuscated document by identifying word images in the image document and scrambling those word images. the obfuscated document can be received by a server of a service provider over a network (e.g., the internet) and processed by previously trained software (e.g., a previously trained convolutional neural network (cnn)) to recognize specific words represented by the scrambled images in the obfuscated document without having to reconstruct the image document. since the image document is neither communicated over the network, nor reconstructed and stored on the server, privacy concerns are minimized.",2017-12-19,"The title of the patent is image document processing in a client-server system including privacy-preserving text recognition and its abstract is disclosed are devices and methods for processing an image document in a client-server environment such that privacy of text information contained in the image document is preserved. specifically, in a client-server environment, an image document can be processed using a local computerized device of a client to create an obfuscated document by identifying word images in the image document and scrambling those word images. the obfuscated document can be received by a server of a service provider over a network (e.g., the internet) and processed by previously trained software (e.g., a previously trained convolutional neural network (cnn)) to recognize specific words represented by the scrambled images in the obfuscated document without having to reconstruct the image document. since the image document is neither communicated over the network, nor reconstructed and stored on the server, privacy concerns are minimized. dated 2017-12-19"
9852006,consolidating multiple neurosynaptic core circuits into one reconfigurable memory block maintaining neuronal information for the core circuits,"embodiments of the invention relate to a neural network circuit comprising a memory block for maintaining neuronal data for multiple neurons, a scheduler for maintaining incoming firing events targeting the neurons, and a computational logic unit for updating the neuronal data for the neurons by processing the firing events. the network circuit further comprises at least one permutation logic unit enabling data exchange between the computational logic unit and at least one of the memory block and the scheduler. the network circuit further comprises a controller for controlling the computational logic unit, the memory block, the scheduler, and each permutation logic unit.",2017-12-26,"The title of the patent is consolidating multiple neurosynaptic core circuits into one reconfigurable memory block maintaining neuronal information for the core circuits and its abstract is embodiments of the invention relate to a neural network circuit comprising a memory block for maintaining neuronal data for multiple neurons, a scheduler for maintaining incoming firing events targeting the neurons, and a computational logic unit for updating the neuronal data for the neurons by processing the firing events. the network circuit further comprises at least one permutation logic unit enabling data exchange between the computational logic unit and at least one of the memory block and the scheduler. the network circuit further comprises a controller for controlling the computational logic unit, the memory block, the scheduler, and each permutation logic unit. dated 2017-12-26"
9852019,system and method for abnormality detection,"a system and method for use in data analysis are provided. the system comprises a data processing utility configured to receive and process input data, comprising: plurality of neural network modules capable for operating in a training mode and in a data processing mode in accordance with the training; a network training utility configured for operating the neural network modules in the training mode utilizing selected set of training data pieces for sequentially training of the neural network modules in a cascade order to reduce an error value with respect to the selected set of the training data pieces for each successive neural network module in the cascade; and an abnormality detection utility configured for sequentially operating said neural network modules for processing input data, and classifying said input data as abnormal upon identifying that all the neural network modules provide error values being above corresponding abnormality detection thresholds.",2017-12-26,"The title of the patent is system and method for abnormality detection and its abstract is a system and method for use in data analysis are provided. the system comprises a data processing utility configured to receive and process input data, comprising: plurality of neural network modules capable for operating in a training mode and in a data processing mode in accordance with the training; a network training utility configured for operating the neural network modules in the training mode utilizing selected set of training data pieces for sequentially training of the neural network modules in a cascade order to reduce an error value with respect to the selected set of the training data pieces for each successive neural network module in the cascade; and an abnormality detection utility configured for sequentially operating said neural network modules for processing input data, and classifying said input data as abnormal upon identifying that all the neural network modules provide error values being above corresponding abnormality detection thresholds. dated 2017-12-26"
9852492,face detection,"briefly, embodiments of methods and/or systems of detecting and image of a human face in a digital image are disclosed. for one embodiment, as an example, parameters of a neural network may be developed to generate object labels for digital images. the developed parameters may be refined by a neural network to generate signal sample value levels corresponding to probability that a human face may be depicted at a localized region of a digital image.",2017-12-26,"The title of the patent is face detection and its abstract is briefly, embodiments of methods and/or systems of detecting and image of a human face in a digital image are disclosed. for one embodiment, as an example, parameters of a neural network may be developed to generate object labels for digital images. the developed parameters may be refined by a neural network to generate signal sample value levels corresponding to probability that a human face may be depicted at a localized region of a digital image. dated 2017-12-26"
9857271,life-time management of downhole tools and components,"systems, methods and devices for evaluating a condition of a downhole component of a drillstring. methods include estimating a value of a tool parameter of the component at at least one selected position on the drillstring; and using the estimated value to evaluate the condition of the downhole component. the estimating is done using a trained artificial neural network that receives information from at least one sensor that is positionally offset from the selected position. the method may further include creating a record representing information from estimated values of the tool parameter at the at least one selected position over time. the at least one selected position may include a plurality of positions, such as positions at intervals along the component, including substantially continuously along the component.",2018-01-02,"The title of the patent is life-time management of downhole tools and components and its abstract is systems, methods and devices for evaluating a condition of a downhole component of a drillstring. methods include estimating a value of a tool parameter of the component at at least one selected position on the drillstring; and using the estimated value to evaluate the condition of the downhole component. the estimating is done using a trained artificial neural network that receives information from at least one sensor that is positionally offset from the selected position. the method may further include creating a record representing information from estimated values of the tool parameter at the at least one selected position over time. the at least one selected position may include a plurality of positions, such as positions at intervals along the component, including substantially continuously along the component. dated 2018-01-02"
9858263,semantic parsing using deep neural networks for predicting canonical forms,"a method for predicting a canonical form for an input text sequence includes predicting the canonical form with a neural network model. the model includes an encoder, which generates a first representation of the input text sequence based on a representation of n-grams in the text sequence and a second representation of the input text sequence generated by a first neural network. the model also includes a decoder which sequentially predicts terms of the canonical form based on the first and second representations and a predicted prefix of the canonical form. the canonical form can be used, for example, to query a knowledge base or to generate a next utterance in a discourse.",2018-01-02,"The title of the patent is semantic parsing using deep neural networks for predicting canonical forms and its abstract is a method for predicting a canonical form for an input text sequence includes predicting the canonical form with a neural network model. the model includes an encoder, which generates a first representation of the input text sequence based on a representation of n-grams in the text sequence and a second representation of the input text sequence generated by a first neural network. the model also includes a decoder which sequentially predicts terms of the canonical form based on the first and second representations and a predicted prefix of the canonical form. the canonical form can be used, for example, to query a knowledge base or to generate a next utterance in a discourse. dated 2018-01-02"
9858265,systems and methods for determining context switching in conversation,"systems and methods are described to address shortcomings in a conventional conversation system via a novel technique utilizing artificial neural networks to train the conversation system whether or not to continue context. in some aspects, an interactive media guidance application determines a type of conversation continuity in a natural language conversation comprising first and second queries. the interactive media guidance application determines a first token in the first query and a second token in the second query. the interactive media guidance application identifies entity data for the first and second tokens. the interactive media guidance application retrieves, from a knowledge graph, graph connections between the entity data for the first and second tokens. the interactive media guidance application applies this data as inputs to an artificial neural network. the interactive media guidance application determines an output that indicates the type of conversation continuity between the first and second queries.",2018-01-02,"The title of the patent is systems and methods for determining context switching in conversation and its abstract is systems and methods are described to address shortcomings in a conventional conversation system via a novel technique utilizing artificial neural networks to train the conversation system whether or not to continue context. in some aspects, an interactive media guidance application determines a type of conversation continuity in a natural language conversation comprising first and second queries. the interactive media guidance application determines a first token in the first query and a second token in the second query. the interactive media guidance application identifies entity data for the first and second tokens. the interactive media guidance application retrieves, from a knowledge graph, graph connections between the entity data for the first and second tokens. the interactive media guidance application applies this data as inputs to an artificial neural network. the interactive media guidance application determines an output that indicates the type of conversation continuity between the first and second queries. dated 2018-01-02"
9858484,systems and methods for determining video feature descriptors based on convolutional neural networks,"systems, methods, and non-transitory computer-readable media can acquire video content for which video feature descriptors are to be determined. the video content can be processed based at least in part on a convolutional neural network including a set of two-dimensional convolutional layers and a set of three-dimensional convolutional layers. one or more outputs can be generated from the convolutional neural network. a plurality of video feature descriptors for the video content can be determined based at least in part on the one or more outputs from the convolutional neural network.",2018-01-02,"The title of the patent is systems and methods for determining video feature descriptors based on convolutional neural networks and its abstract is systems, methods, and non-transitory computer-readable media can acquire video content for which video feature descriptors are to be determined. the video content can be processed based at least in part on a convolutional neural network including a set of two-dimensional convolutional layers and a set of three-dimensional convolutional layers. one or more outputs can be generated from the convolutional neural network. a plurality of video feature descriptors for the video content can be determined based at least in part on the one or more outputs from the convolutional neural network. dated 2018-01-02"
9858496,object detection and classification in images,"systems, methods, and computer-readable media for providing fast and accurate object detection and classification in images are described herein. in some examples, a computing device can receive an input image. the computing device can process the image, and generate a convolutional feature map. in some configurations, the convolutional feature map can be processed through a region proposal network (rpn) to generate proposals for candidate objects in the image. in various examples, the computing device can process the convolutional feature map with the proposals through a fast region-based convolutional neural network (frcn) proposal classifier to determine a class of each object in the image and a confidence score associated therewith. the computing device can then provide a requestor with an output including the object classification and/or confidence score.",2018-01-02,"The title of the patent is object detection and classification in images and its abstract is systems, methods, and computer-readable media for providing fast and accurate object detection and classification in images are described herein. in some examples, a computing device can receive an input image. the computing device can process the image, and generate a convolutional feature map. in some configurations, the convolutional feature map can be processed through a region proposal network (rpn) to generate proposals for candidate objects in the image. in various examples, the computing device can process the convolutional feature map with the proposals through a fast region-based convolutional neural network (frcn) proposal classifier to determine a class of each object in the image and a confidence score associated therewith. the computing device can then provide a requestor with an output including the object classification and/or confidence score. dated 2018-01-02"
9858524,generating natural language descriptions of images,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating descriptions of input images. one of the methods includes obtaining an input image; processing the input image using a first neural network to generate an alternative representation for the input image; and processing the alternative representation for the input image using a second neural network to generate a sequence of a plurality of words in a target natural language that describes the input image.",2018-01-02,"The title of the patent is generating natural language descriptions of images and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating descriptions of input images. one of the methods includes obtaining an input image; processing the input image using a first neural network to generate an alternative representation for the input image; and processing the alternative representation for the input image using a second neural network to generate a sequence of a plurality of words in a target natural language that describes the input image. dated 2018-01-02"
9858525,system for training networks for semantic segmentation,"disclosed herein are technologies directed to training a neural network to perform semantic segmentation. a system receives a training image, and using the training image, candidate masks are generated. the candidate masks are ranked and a set of the ranked candidate masks are selected for further processing. one of the set of the ranked candidate masks is selected to train the neural network. the one of the set of the set of the ranked candidate masks is also used as an input to train the neural network in a further training evolution. in some examples, the one of the set of the ranked candidate masks is selected randomly to reduce the likelihood of ending up in poor local optima that result in poor training inputs.",2018-01-02,"The title of the patent is system for training networks for semantic segmentation and its abstract is disclosed herein are technologies directed to training a neural network to perform semantic segmentation. a system receives a training image, and using the training image, candidate masks are generated. the candidate masks are ranked and a set of the ranked candidate masks are selected for further processing. one of the set of the ranked candidate masks is selected to train the neural network. the one of the set of the set of the ranked candidate masks is also used as an input to train the neural network in a further training evolution. in some examples, the one of the set of the ranked candidate masks is selected randomly to reduce the likelihood of ending up in poor local optima that result in poor training inputs. dated 2018-01-02"
9858919,speaker adaptation of neural network acoustic models using i-vectors,"a method includes providing a deep neural network acoustic model, receiving audio data including one or more utterances of a speaker, extracting a plurality of speech recognition features from the one or more utterances of the speaker, creating a speaker identity vector for the speaker based on the extracted speech recognition features, and adapting the deep neural network acoustic model for automatic speech recognition using the extracted speech recognition features and the speaker identity vector.",2018-01-02,"The title of the patent is speaker adaptation of neural network acoustic models using i-vectors and its abstract is a method includes providing a deep neural network acoustic model, receiving audio data including one or more utterances of a speaker, extracting a plurality of speech recognition features from the one or more utterances of the speaker, creating a speaker identity vector for the speaker based on the extracted speech recognition features, and adapting the deep neural network acoustic model for automatic speech recognition using the extracted speech recognition features and the speaker identity vector. dated 2018-01-02"
9860636,directional microphone device and signal processing techniques,"methods and apparatus relating to microphone devices and signal processing techniques are provided. in an example, a microphone device can detect sound, as well as enhance an ability to perceive at least a general direction from which the sound arrives at the microphone device. in an example, a case of the microphone device has an external surface which at least partially defines funnel-shaped surfaces. each funnel-shaped surface is configured to direct the sound to a respective microphone diaphragm to produce an auralized multi-microphone output. the funnel-shaped surfaces are configured to cause direction-dependent variations in spectral notches and frequency response of the sound as received by the microphone diaphragms. a neural network can device-shape the auralized multi-microphone output to create a binaural output. the binaural output can be auralized with respect to a human listener.",2018-01-02,"The title of the patent is directional microphone device and signal processing techniques and its abstract is methods and apparatus relating to microphone devices and signal processing techniques are provided. in an example, a microphone device can detect sound, as well as enhance an ability to perceive at least a general direction from which the sound arrives at the microphone device. in an example, a case of the microphone device has an external surface which at least partially defines funnel-shaped surfaces. each funnel-shaped surface is configured to direct the sound to a respective microphone diaphragm to produce an auralized multi-microphone output. the funnel-shaped surfaces are configured to cause direction-dependent variations in spectral notches and frequency response of the sound as received by the microphone diaphragms. a neural network can device-shape the auralized multi-microphone output to create a binaural output. the binaural output can be auralized with respect to a human listener. dated 2018-01-02"
9862924,neural progenitor cell differentiation,"differentiation and stability of neural stem cells can be enhanced by in vitro or in vivo culturing with one or more extracellular matrix (ecm) compositions, such as collagen i, iv, laminin and/or a heparan sulfate proteoglycan. in one aspect of the invention, adult mammalian enteric neuronal progenitor cells can be induced to differentiate on various substrates derived from components or combinations of neural ecm compositions. collagen i and iv supported neuronal differentiation and extensive glial differentiation individually and in combination. addition of laminin or heparan sulfate to collagen substrates unexpectedly improved neuronal differentiation, increasing neuron number, branching of neuronal processes, and initiation of neuronal network formation. in another aspect, neuronal subtype differentiation was affected by varying ecm compositions in hydrogels overlaid on intestinal smooth muscle sheets. the matrix compositions of the present invention can be used to tissue engineer transplantable innervated gi smooth muscle constructs to remedy aganglionic disorders.",2018-01-09,"The title of the patent is neural progenitor cell differentiation and its abstract is differentiation and stability of neural stem cells can be enhanced by in vitro or in vivo culturing with one or more extracellular matrix (ecm) compositions, such as collagen i, iv, laminin and/or a heparan sulfate proteoglycan. in one aspect of the invention, adult mammalian enteric neuronal progenitor cells can be induced to differentiate on various substrates derived from components or combinations of neural ecm compositions. collagen i and iv supported neuronal differentiation and extensive glial differentiation individually and in combination. addition of laminin or heparan sulfate to collagen substrates unexpectedly improved neuronal differentiation, increasing neuron number, branching of neuronal processes, and initiation of neuronal network formation. in another aspect, neuronal subtype differentiation was affected by varying ecm compositions in hydrogels overlaid on intestinal smooth muscle sheets. the matrix compositions of the present invention can be used to tissue engineer transplantable innervated gi smooth muscle constructs to remedy aganglionic disorders. dated 2018-01-09"
9864912,large margin high-order deep learning with auxiliary tasks for video-based anomaly detection,"a video camera is provided for video-based anomaly detection that includes at least one imaging sensor configured to capture video sequences in a workplace environment having a plurality of machines therein. the video camera further includes a processor. the processor is configured to generate one or more predictions of an impending anomaly affecting at least one item selected from the group consisting of (i) at least one of the plurality of machines and (ii) at least one operator of the at least one of the plurality of machines, using a deep high-order convolutional neural network (dhocnn)-based model applied to the video sequences. the dhocnn-based model has a one-class svm as a loss layer of the model. the processor is further configured to generate a signal for initiating an action to the at least one of the plurality of machines to mitigate expected harm to the at least one item.",2018-01-09,"The title of the patent is large margin high-order deep learning with auxiliary tasks for video-based anomaly detection and its abstract is a video camera is provided for video-based anomaly detection that includes at least one imaging sensor configured to capture video sequences in a workplace environment having a plurality of machines therein. the video camera further includes a processor. the processor is configured to generate one or more predictions of an impending anomaly affecting at least one item selected from the group consisting of (i) at least one of the plurality of machines and (ii) at least one operator of the at least one of the plurality of machines, using a deep high-order convolutional neural network (dhocnn)-based model applied to the video sequences. the dhocnn-based model has a one-class svm as a loss layer of the model. the processor is further configured to generate a signal for initiating an action to the at least one of the plurality of machines to mitigate expected harm to the at least one item. dated 2018-01-09"
9864933,"artificially intelligent systems, devices, and methods for learning and/or using visual surrounding for autonomous object operation","aspects of the disclosure generally relate to computing devices and/or systems, and may be generally directed to devices, systems, methods, and/or applications for learning operation of an application or an object of an application in various visual surroundings, storing this knowledge in a knowledgebase (i.e. neural network, graph, sequences, etc.), and enabling autonomous operation of the application or the object of the application.",2018-01-09,"The title of the patent is artificially intelligent systems, devices, and methods for learning and/or using visual surrounding for autonomous object operation and its abstract is aspects of the disclosure generally relate to computing devices and/or systems, and may be generally directed to devices, systems, methods, and/or applications for learning operation of an application or an object of an application in various visual surroundings, storing this knowledge in a knowledgebase (i.e. neural network, graph, sequences, etc.), and enabling autonomous operation of the application or the object of the application. dated 2018-01-09"
9864951,randomized latent feature learning,"features are disclosed for identifying randomized latent feature language modeling, such as a recurrent neural network language modeling (rnnlm). sequences of item identifiers may be provided as the language for training the language model where the item identifiers are the words of the language. to avoid localization bias, the sequences may be randomized prior to or during the training process to provide more accurate prediction models.",2018-01-09,"The title of the patent is randomized latent feature learning and its abstract is features are disclosed for identifying randomized latent feature language modeling, such as a recurrent neural network language modeling (rnnlm). sequences of item identifiers may be provided as the language for training the language model where the item identifiers are the words of the language. to avoid localization bias, the sequences may be randomized prior to or during the training process to provide more accurate prediction models. dated 2018-01-09"
9866580,forecasting and classifying cyber-attacks using neural embeddings,"a first collection including a first feature vector and a q&a feature vector is constructed. a second collection is constructed from the first collection by inserting noise in at least one of the vectors. a third collection is constructed by crossing over at least one the vectors of the second collection with a corresponding vector of a fourth collection, migrating at least one of the vectors of the second collection with a corresponding vector of a fifth collection, or both. using a forecasting configuration, a vector of the third collection is aged to generate a changed feature vector, the changed feature vector containing feature values expected at a future time. the changed feature vector is input into a trained neural network to predict a probability of the cyber-attack occurring at the future time.",2018-01-09,"The title of the patent is forecasting and classifying cyber-attacks using neural embeddings and its abstract is a first collection including a first feature vector and a q&a feature vector is constructed. a second collection is constructed from the first collection by inserting noise in at least one of the vectors. a third collection is constructed by crossing over at least one the vectors of the second collection with a corresponding vector of a fourth collection, migrating at least one of the vectors of the second collection with a corresponding vector of a fifth collection, or both. using a forecasting configuration, a vector of the third collection is aged to generate a changed feature vector, the changed feature vector containing feature values expected at a future time. the changed feature vector is input into a trained neural network to predict a probability of the cyber-attack occurring at the future time. dated 2018-01-09"
9866984,method for generating surround channel audio,"a method includes extracting a difference value through extraction of features of a front audio channel signal and a surround channel of multichannel sound content by setting the front audio channel signal and the surround channel as input and output channel signals, respectively, training a deep neural network (dnn) model by setting the input channel signal and the difference value as an input and an output of the dnn model, respectively, normalizing a frequency-domain signal of the input channel signal by converting the input channel signal into the frequency-domain signal, and extracting estimated difference values by decoding the normalized frequency-domain signal through the dnn model, deriving an estimated spectral amplitude of the surround channel based on the front audio channel signal and the difference value, and deriving an audio signal of a final surround channel by converting the estimated spectral amplitude of the surround channel into the time domain.",2018-01-09,"The title of the patent is method for generating surround channel audio and its abstract is a method includes extracting a difference value through extraction of features of a front audio channel signal and a surround channel of multichannel sound content by setting the front audio channel signal and the surround channel as input and output channel signals, respectively, training a deep neural network (dnn) model by setting the input channel signal and the difference value as an input and an output of the dnn model, respectively, normalizing a frequency-domain signal of the input channel signal by converting the input channel signal into the frequency-domain signal, and extracting estimated difference values by decoding the normalized frequency-domain signal through the dnn model, deriving an estimated spectral amplitude of the surround channel based on the front audio channel signal and the difference value, and deriving an audio signal of a final surround channel by converting the estimated spectral amplitude of the surround channel into the time domain. dated 2018-01-09"
9867561,systems and methods for determining whether regional oximetry sensors are properly positioned,"methods and systems are presented for determining whether a regional oximetry sensor is properly positioned on a subject. first and second metric values may be determined based on respective first and second light signals. the first and second metric values and a relationship between the first and second metrics are used to determine whether the sensor is properly positioned on the subject. the first and second metrics may form a pair of metrics, and whether the sensor is properly positioned on the subject may be determined based on whether the pair of metrics falls within a sensor-on region. in some embodiments, a plurality of metrics may be determined based on a plurality of received physiological signals. the plurality of metrics may be combined, using, for example, a neural network, to determine whether the regional oximetry sensor is properly positioned on a subject.",2018-01-16,"The title of the patent is systems and methods for determining whether regional oximetry sensors are properly positioned and its abstract is methods and systems are presented for determining whether a regional oximetry sensor is properly positioned on a subject. first and second metric values may be determined based on respective first and second light signals. the first and second metric values and a relationship between the first and second metrics are used to determine whether the sensor is properly positioned on the subject. the first and second metrics may form a pair of metrics, and whether the sensor is properly positioned on the subject may be determined based on whether the pair of metrics falls within a sensor-on region. in some embodiments, a plurality of metrics may be determined based on a plurality of received physiological signals. the plurality of metrics may be combined, using, for example, a neural network, to determine whether the regional oximetry sensor is properly positioned on a subject. dated 2018-01-16"
9867955,system and method for diagnosis and treatment of a breathing pattern of a patient,"described is a system including a sensor and a processing arrangement. the sensor measures data corresponding to a patient's breathing patterns. the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of a rem sleep state. in another embodiment, the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of one of the following states: (i) a wake state and (ii) a sleep state. in another embodiment, a neural network analyzes the data to determine whether the breathing patterns are indicative of one of the following states: (i) a rem sleep state, (ii) a wake state and (iii) a sleep state. in another embodiment, the processing arrangement analyzes the data to determine whether the breathing pattern is indicative of an arousal.",2018-01-16,"The title of the patent is system and method for diagnosis and treatment of a breathing pattern of a patient and its abstract is described is a system including a sensor and a processing arrangement. the sensor measures data corresponding to a patient's breathing patterns. the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of a rem sleep state. in another embodiment, the processing arrangement analyzes the breathing patterns to determine whether the breathing patterns are indicative of one of the following states: (i) a wake state and (ii) a sleep state. in another embodiment, a neural network analyzes the data to determine whether the breathing patterns are indicative of one of the following states: (i) a rem sleep state, (ii) a wake state and (iii) a sleep state. in another embodiment, the processing arrangement analyzes the data to determine whether the breathing pattern is indicative of an arousal. dated 2018-01-16"
9870768,subject estimation system for estimating subject of dialog,"a subject estimation system includes a convolutional neural network to estimate a subject label of a dialog. the convolution neural network includes: one or more topic-dependent convolutional layers and one topic-independent convolutional layer, each of the one or more topic-dependent convolutional layers performing, on an input of a word-string vector sequence corresponding to dialog text transcribed from a dialog, a convolution operation dependent on a topic, and the topic-independent convolutional layer performing, on the input of the word-string vector sequence, a convolution operation not dependent on the topic; a pooling layer performing pooling process on outputs of the convolutional layer; and a fully connected layer performing full connection process on outputs of the pooling layer.",2018-01-16,"The title of the patent is subject estimation system for estimating subject of dialog and its abstract is a subject estimation system includes a convolutional neural network to estimate a subject label of a dialog. the convolution neural network includes: one or more topic-dependent convolutional layers and one topic-independent convolutional layer, each of the one or more topic-dependent convolutional layers performing, on an input of a word-string vector sequence corresponding to dialog text transcribed from a dialog, a convolution operation dependent on a topic, and the topic-independent convolutional layer performing, on the input of the word-string vector sequence, a convolution operation not dependent on the topic; a pooling layer performing pooling process on outputs of the convolutional layer; and a fully connected layer performing full connection process on outputs of the pooling layer. dated 2018-01-16"
9872989,system and method for neuromorphic controlled adaptive pacing of respiratory muscles and nerves,"adaptive systems and methods for automatically determining and continuously updating stimulation parameters for adjusting ventilation to accommodate a patient's specific physiology, metabolic needs, and muscle state are disclosed herein. having a closed loop implementation, the system may comprise a controller including a neuromorphic controlled adaptive feed-forward pattern generator/pattern shaper (pg/ps) assembly, which controls respiratory muscle movement using electrical stimulation. this pg/ps assembly comprises a biomimetic design where the pattern generator includes a neural network mimicking the simplified connectivity pattern of respiratory related neurons in the brain stem to produce a rhythmic breathing pattern frequency and the pattern shaper includes a neural network mimicking the simplified connectivity pattern of neurons to produce a stimulus control signal. this biomimetic design for the controller automatically customizes and continually updates stimulation parameters to achieve a desired breathing pattern and, thereby, slow the development of muscle fatigue.",2018-01-23,"The title of the patent is system and method for neuromorphic controlled adaptive pacing of respiratory muscles and nerves and its abstract is adaptive systems and methods for automatically determining and continuously updating stimulation parameters for adjusting ventilation to accommodate a patient's specific physiology, metabolic needs, and muscle state are disclosed herein. having a closed loop implementation, the system may comprise a controller including a neuromorphic controlled adaptive feed-forward pattern generator/pattern shaper (pg/ps) assembly, which controls respiratory muscle movement using electrical stimulation. this pg/ps assembly comprises a biomimetic design where the pattern generator includes a neural network mimicking the simplified connectivity pattern of respiratory related neurons in the brain stem to produce a rhythmic breathing pattern frequency and the pattern shaper includes a neural network mimicking the simplified connectivity pattern of neurons to produce a stimulus control signal. this biomimetic design for the controller automatically customizes and continually updates stimulation parameters to achieve a desired breathing pattern and, thereby, slow the development of muscle fatigue. dated 2018-01-23"
9874081,detection of influxes and losses while drilling from a floating vessel,"a system for detecting fluid influxes and losses can include a sensor which detects floating vessel movement, and a neural network which receives a sensor output, and which outputs a predicted flow rate from a wellbore. a method can include isolating the wellbore from atmosphere with an annular sealing device which seals against a drill string, inputting to a neural network an output of a sensor which detects vessel movement, the neural network outputting a predicted flow rate from the wellbore, and determining whether the fluid influx or loss has occurred by comparing the predicted flow rate to an actual flow rate from the wellbore. another method can include inputting to a neural network actual flow rates into and out of the wellbore, and an output of a sensor which detects vessel movement, and training the neural network to output a predicted flow rate from the wellbore.",2018-01-23,"The title of the patent is detection of influxes and losses while drilling from a floating vessel and its abstract is a system for detecting fluid influxes and losses can include a sensor which detects floating vessel movement, and a neural network which receives a sensor output, and which outputs a predicted flow rate from a wellbore. a method can include isolating the wellbore from atmosphere with an annular sealing device which seals against a drill string, inputting to a neural network an output of a sensor which detects vessel movement, the neural network outputting a predicted flow rate from the wellbore, and determining whether the fluid influx or loss has occurred by comparing the predicted flow rate to an actual flow rate from the wellbore. another method can include inputting to a neural network actual flow rates into and out of the wellbore, and an output of a sensor which detects vessel movement, and training the neural network to output a predicted flow rate from the wellbore. dated 2018-01-23"
9875429,font attributes for font recognition and similarity,"font recognition and similarity determination techniques and systems are described. in a first example, localization techniques are described to train a model using machine learning (e.g., a convolutional neural network) using training images. the model is then used to localize text in a subsequently received image, and may do so automatically and without user intervention, e.g., without specifying any of the edges of a bounding box. in a second example, a deep neural network is directly learned as an embedding function of a model that is usable to determine font similarity. in a third example, techniques are described that leverage attributes described in metadata associated with fonts as part of font recognition and similarity determinations.",2018-01-23,"The title of the patent is font attributes for font recognition and similarity and its abstract is font recognition and similarity determination techniques and systems are described. in a first example, localization techniques are described to train a model using machine learning (e.g., a convolutional neural network) using training images. the model is then used to localize text in a subsequently received image, and may do so automatically and without user intervention, e.g., without specifying any of the edges of a bounding box. in a second example, a deep neural network is directly learned as an embedding function of a model that is usable to determine font similarity. in a third example, techniques are described that leverage attributes described in metadata associated with fonts as part of font recognition and similarity determinations. dated 2018-01-23"
9875440,intelligent control with hierarchical stacked neural networks,"a method of processing information is provided. the method involves receiving a message; processing the message with a trained artificial neural network based processor, having at least one set of outputs which represent information in a non-arbitrary organization of actions based on an architecture of the artificial neural network based processor and the training; representing as a noise vector at least one data pattern in the message which is incompletely represented in the non-arbitrary organization of actions; analyzing the noise vector distinctly from the trained artificial neural network; searching at least one database; and generating an output in dependence on said analyzing and said searching.",2018-01-23,"The title of the patent is intelligent control with hierarchical stacked neural networks and its abstract is a method of processing information is provided. the method involves receiving a message; processing the message with a trained artificial neural network based processor, having at least one set of outputs which represent information in a non-arbitrary organization of actions based on an architecture of the artificial neural network based processor and the training; representing as a noise vector at least one data pattern in the message which is incompletely represented in the non-arbitrary organization of actions; analyzing the noise vector distinctly from the trained artificial neural network; searching at least one database; and generating an output in dependence on said analyzing and said searching. dated 2018-01-23"
9875737,pre-training apparatus and method for speech recognition,"a pre-training apparatus and method for recognition speech, which initialize, by layers, a deep neural network to correct a node connection weight. the pre-training apparatus for speech recognition includes an input unit configured to receive speech data, a model generation unit configured to initialize a connection weight of a deep neural network, based on the speech data, and an output unit configured to output information about the connection weight. in order for a state of a phoneme unit corresponding to the speech data to be output, the model generation unit trains the connection weight by piling a plurality of hidden layers according to a determined structure of the deep neural network, applies an output layer to a certain layer between the plurality of hidden layers to correct the trained connection weight in each of the plurality of hidden layers, thereby initializing the connection weight.",2018-01-23,"The title of the patent is pre-training apparatus and method for speech recognition and its abstract is a pre-training apparatus and method for recognition speech, which initialize, by layers, a deep neural network to correct a node connection weight. the pre-training apparatus for speech recognition includes an input unit configured to receive speech data, a model generation unit configured to initialize a connection weight of a deep neural network, based on the speech data, and an output unit configured to output information about the connection weight. in order for a state of a phoneme unit corresponding to the speech data to be output, the model generation unit trains the connection weight by piling a plurality of hidden layers according to a determined structure of the deep neural network, applies an output layer to a certain layer between the plurality of hidden layers to correct the trained connection weight in each of the plurality of hidden layers, thereby initializing the connection weight. dated 2018-01-23"
9875747,device specific multi-channel data compression,a sensor device may include a computing device in communication with multiple microphones. a neural network executing on the computing device may receive audio signals from each microphone. one microphone signal may serve as a reference signal. the neural network may extract differences in signal characteristics of the other microphone signals as compared to the reference signal. the neural network may combine these signal differences into a lossy compressed signal. the sensor device may transmit the lossy compressed signal and the lossless reference signal to a remote neural network executing in a cloud computing environment for decompression and sound recognition analysis.,2018-01-23,The title of the patent is device specific multi-channel data compression and its abstract is a sensor device may include a computing device in communication with multiple microphones. a neural network executing on the computing device may receive audio signals from each microphone. one microphone signal may serve as a reference signal. the neural network may extract differences in signal characteristics of the other microphone signals as compared to the reference signal. the neural network may combine these signal differences into a lossy compressed signal. the sensor device may transmit the lossy compressed signal and the lossless reference signal to a remote neural network executing in a cloud computing environment for decompression and sound recognition analysis. dated 2018-01-23
9881208,neural network based recognition of mathematical expressions,"provided are methods and system for recognizing characters such as mathematical expressions or chemical formulas. an example method comprises the steps of receiving and processing an image by a pre-processing module to obtain one or more candidate regions, extracting features of each of the candidate regions by a feature extracting module such as a convolutional neural network (cnn), encoding the features into a distributive representation for each of the candidate regions separately using an encoding module such as a first long short-term memory (lstm) based neural network, decoding the distributive representation into output representations using a decoding module such as a second lstm-based recurrent neural network, and combining the output representations into an output expression, which is outputted in a computer-readable format or a markup language.",2018-01-30,"The title of the patent is neural network based recognition of mathematical expressions and its abstract is provided are methods and system for recognizing characters such as mathematical expressions or chemical formulas. an example method comprises the steps of receiving and processing an image by a pre-processing module to obtain one or more candidate regions, extracting features of each of the candidate regions by a feature extracting module such as a convolutional neural network (cnn), encoding the features into a distributive representation for each of the candidate regions separately using an encoding module such as a first long short-term memory (lstm) based neural network, decoding the distributive representation into output representations using a decoding module such as a second lstm-based recurrent neural network, and combining the output representations into an output expression, which is outputted in a computer-readable format or a markup language. dated 2018-01-30"
9881251,structural plasticity in spiking neural networks with symmetric dual of an electronic neuron,"a neural system comprises multiple neurons interconnected via synapse devices. each neuron integrates input signals arriving on its dendrite, generates a spike in response to the integrated input signals exceeding a threshold, and sends the spike to the interconnected neurons via its axon. the system further includes multiple noruens, each noruen is interconnected via the interconnect network with those neurons that the noruen's corresponding neuron sends its axon to. each noruen integrates input spikes from connected spiking neurons and generates a spike in response to the integrated input spikes exceeding a threshold. there can be one noruen for every corresponding neuron. for a first neuron connected via its axon via a synapse to dendrite of a second neuron, a noruen corresponding to the second neuron is connected via its axon through the same synapse to dendrite of the noruen corresponding to the first neuron.",2018-01-30,"The title of the patent is structural plasticity in spiking neural networks with symmetric dual of an electronic neuron and its abstract is a neural system comprises multiple neurons interconnected via synapse devices. each neuron integrates input signals arriving on its dendrite, generates a spike in response to the integrated input signals exceeding a threshold, and sends the spike to the interconnected neurons via its axon. the system further includes multiple noruens, each noruen is interconnected via the interconnect network with those neurons that the noruen's corresponding neuron sends its axon to. each noruen integrates input spikes from connected spiking neurons and generates a spike in response to the integrated input spikes exceeding a threshold. there can be one noruen for every corresponding neuron. for a first neuron connected via its axon via a synapse to dendrite of a second neuron, a noruen corresponding to the second neuron is connected via its axon through the same synapse to dendrite of the noruen corresponding to the first neuron. dated 2018-01-30"
9881253,synaptic neural network core based sensor system,"a sensor system comprises: an energy storage device; an intermittent energy release device electrically coupled to the energy storage device, wherein the intermittent energy release device causes the energy storage device to release stored energy intermittently; a sensor electrically coupled to the energy storage device; a register electrically coupled to the sensor, wherein the register stores readings from the sensor; a synaptic neural network core electrically coupled to the sensor, wherein the synaptic neural network core converts the readings from the sensor into a synthetic context-based object that is derived from the readings and a context object; a transponder electrically coupled to the synaptic neural network core; and a storage buffer within the transponder, wherein the storage buffer stores the synthetic context-based object for transmission by the transponder to a monitoring system.",2018-01-30,"The title of the patent is synaptic neural network core based sensor system and its abstract is a sensor system comprises: an energy storage device; an intermittent energy release device electrically coupled to the energy storage device, wherein the intermittent energy release device causes the energy storage device to release stored energy intermittently; a sensor electrically coupled to the energy storage device; a register electrically coupled to the sensor, wherein the register stores readings from the sensor; a synaptic neural network core electrically coupled to the sensor, wherein the synaptic neural network core converts the readings from the sensor into a synthetic context-based object that is derived from the readings and a context object; a transponder electrically coupled to the synaptic neural network core; and a storage buffer within the transponder, wherein the storage buffer stores the synthetic context-based object for transmission by the transponder to a monitoring system. dated 2018-01-30"
9881372,method and system for vascular disease detection using recurrent neural networks,"a method and apparatus for vascular disease detection and characterization using a recurrent neural network (rnn) is disclosed. a plurality of 2d cross-section image patches are extracted from a 3d computed tomography angiography (cta) image, each extracted at a respective sampling point along a vessel centerline of a vessel of interest in the 3d cta image. vascular abnormalities in the vessel of interest are detected and characterized by classifying each of the sampling points along the vessel centerline based on the plurality of 2d cross-section image patches using a trained rnn.",2018-01-30,"The title of the patent is method and system for vascular disease detection using recurrent neural networks and its abstract is a method and apparatus for vascular disease detection and characterization using a recurrent neural network (rnn) is disclosed. a plurality of 2d cross-section image patches are extracted from a 3d computed tomography angiography (cta) image, each extracted at a respective sampling point along a vessel centerline of a vessel of interest in the 3d cta image. vascular abnormalities in the vessel of interest are detected and characterized by classifying each of the sampling points along the vessel centerline based on the plurality of 2d cross-section image patches using a trained rnn. dated 2018-01-30"
9881615,speech recognition apparatus and method,"a speech recognition apparatus and method. the speech recognition apparatus includes a first recognizer configured to generate a first recognition result of an audio signal, in a first linguistic recognition unit, by using an acoustic model, a second recognizer configured to generate a second recognition result of the audio signal, in a second linguistic recognition unit, by using a language model, and a combiner configured to combine the first recognition result and the second recognition result to generate a final recognition result in the second linguistic recognition unit and to reflect the final recognition result in the language model. the first linguistic recognition unit may be a same linguistic unit type as the second linguistic recognition unit. the first recognizer and the second recognizer are configured in a same neural network and simultaneously/collectively trained in the neural network using audio training data provided to the first recognizer.",2018-01-30,"The title of the patent is speech recognition apparatus and method and its abstract is a speech recognition apparatus and method. the speech recognition apparatus includes a first recognizer configured to generate a first recognition result of an audio signal, in a first linguistic recognition unit, by using an acoustic model, a second recognizer configured to generate a second recognition result of the audio signal, in a second linguistic recognition unit, by using a language model, and a combiner configured to combine the first recognition result and the second recognition result to generate a final recognition result in the second linguistic recognition unit and to reflect the final recognition result in the language model. the first linguistic recognition unit may be a same linguistic unit type as the second linguistic recognition unit. the first recognizer and the second recognizer are configured in a same neural network and simultaneously/collectively trained in the neural network using audio training data provided to the first recognizer. dated 2018-01-30"
9886663,compiling network descriptions to multiple platforms,"a method of generating executable code for a target platform in a neural network includes receiving a spiking neural network description. the method also includes receiving platform-specific instructions for one or more target platforms. further, the method includes, generating executable code for the target platform(s) based on the platform-specific instructions and the network description.",2018-02-06,"The title of the patent is compiling network descriptions to multiple platforms and its abstract is a method of generating executable code for a target platform in a neural network includes receiving a spiking neural network description. the method also includes receiving platform-specific instructions for one or more target platforms. further, the method includes, generating executable code for the target platform(s) based on the platform-specific instructions and the network description. dated 2018-02-06"
9886758,annotation of skin image using learned feature representation,a method for annotation of skin images includes receiving a plurality of dermatoscopic images. each of the dermatoscopic includes a region of lesion skin and a region of normal skin. a first convolutional neural network is trained according to an interior of the region of lesion skin using each of the plurality of dermatoscopic images. a second convolutional neural network is trained according to a boundary between the region of lesion skin and the region of normal skin. an additional dermatoscopic image is acquired. the first and second convolutional neural networks are used to identify a region of lesion skin within the acquired additional dermatoscopic image.,2018-02-06,The title of the patent is annotation of skin image using learned feature representation and its abstract is a method for annotation of skin images includes receiving a plurality of dermatoscopic images. each of the dermatoscopic includes a region of lesion skin and a region of normal skin. a first convolutional neural network is trained according to an interior of the region of lesion skin using each of the plurality of dermatoscopic images. a second convolutional neural network is trained according to a boundary between the region of lesion skin and the region of normal skin. an additional dermatoscopic image is acquired. the first and second convolutional neural networks are used to identify a region of lesion skin within the acquired additional dermatoscopic image. dated 2018-02-06
9886771,heat map of vehicle damage,"an image processing system and/or method obtains source images in which a damaged vehicle is represented, and performs image processing techniques to determine, predict, estimate, and/or detect damage that has occurred at various locations on the vehicle. the image processing techniques may include generating a composite image of the damaged vehicle, aligning and/or isolating the image, applying convolutional neural network techniques to the image to generate damage parameter values, where each value corresponds to damage in a particular location of vehicle, and/or other techniques. based on the damage values, the image processing system/method generates and displays a heat map for the vehicle, where each color and/or color gradation corresponds to respective damage at a respective location on the vehicle. the heat map may be manipulatable by the user, and may include user controls for displaying additional information corresponding to the damage at a particular location on the vehicle.",2018-02-06,"The title of the patent is heat map of vehicle damage and its abstract is an image processing system and/or method obtains source images in which a damaged vehicle is represented, and performs image processing techniques to determine, predict, estimate, and/or detect damage that has occurred at various locations on the vehicle. the image processing techniques may include generating a composite image of the damaged vehicle, aligning and/or isolating the image, applying convolutional neural network techniques to the image to generate damage parameter values, where each value corresponds to damage in a particular location of vehicle, and/or other techniques. based on the damage values, the image processing system/method generates and displays a heat map for the vehicle, where each color and/or color gradation corresponds to respective damage at a respective location on the vehicle. the heat map may be manipulatable by the user, and may include user controls for displaying additional information corresponding to the damage at a particular location on the vehicle. dated 2018-02-06"
9886948,neural network processing of multiple feature streams using max pooling and restricted connectivity,"features are disclosed for improving the robustness of a neural network by using multiple (e.g., two or more) feature streams, combing data from the feature streams, and comparing the combined data to data from a subset of the feature streams (e.g., comparing values from the combined feature stream to values from one of the component feature streams of the combined feature stream). the neural network can include a component or layer that selects the data with the highest value, which can suppress or exclude some or all corrupted data from the combined feature stream. subsequent layers of the neural network can restrict connections from the combined feature stream to a component feature stream to reduce the possibility that a corrupted combined feature stream will corrupt the component feature stream.",2018-02-06,"The title of the patent is neural network processing of multiple feature streams using max pooling and restricted connectivity and its abstract is features are disclosed for improving the robustness of a neural network by using multiple (e.g., two or more) feature streams, combing data from the feature streams, and comparing the combined data to data from a subset of the feature streams (e.g., comparing values from the combined feature stream to values from one of the component feature streams of the combined feature stream). the neural network can include a component or layer that selects the data with the highest value, which can suppress or exclude some or all corrupted data from the combined feature stream. subsequent layers of the neural network can restrict connections from the combined feature stream to a component feature stream to reduce the possibility that a corrupted combined feature stream will corrupt the component feature stream. dated 2018-02-06"
9886949,adaptive audio enhancement for multichannel speech recognition,"methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for neural network adaptive beamforming for multichannel speech recognition are disclosed. in one aspect, a method includes the actions of receiving a first channel of audio data corresponding to an utterance and a second channel of audio data corresponding to the utterance. the actions further include generating a first set of filter parameters for a first filter based on the first channel of audio data and the second channel of audio data and a second set of filter parameters for a second filter based on the first channel of audio data and the second channel of audio data. the actions further include generating a single combined channel of audio data. the actions further include inputting the audio data to a neural network. the actions further include providing a transcription for the utterance.",2018-02-06,"The title of the patent is adaptive audio enhancement for multichannel speech recognition and its abstract is methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for neural network adaptive beamforming for multichannel speech recognition are disclosed. in one aspect, a method includes the actions of receiving a first channel of audio data corresponding to an utterance and a second channel of audio data corresponding to the utterance. the actions further include generating a first set of filter parameters for a first filter based on the first channel of audio data and the second channel of audio data and a second set of filter parameters for a second filter based on the first channel of audio data and the second channel of audio data. the actions further include generating a single combined channel of audio data. the actions further include inputting the audio data to a neural network. the actions further include providing a transcription for the utterance. dated 2018-02-06"
9886957,system and method for voice recognition,"a voice recognition system and method are provided. the voice recognition system includes a voice input unit configured to receive learning voice data and a target label including consonant and vowel (letter) information representing the learning voice data, and divide the learning voice data into windows having a predetermined size; a first voice recognition unit configured to learn features of the divided windows using a first neural network model and the target label; a second voice recognition unit configured to learn a time-series pattern of the extracted features using a second neural network model; and a text output unit configured to convert target voice data input to the voice input unit into a text based on learning results of the first voice recognition unit and the second voice recognition unit, and output the text.",2018-02-06,"The title of the patent is system and method for voice recognition and its abstract is a voice recognition system and method are provided. the voice recognition system includes a voice input unit configured to receive learning voice data and a target label including consonant and vowel (letter) information representing the learning voice data, and divide the learning voice data into windows having a predetermined size; a first voice recognition unit configured to learn features of the divided windows using a first neural network model and the target label; a second voice recognition unit configured to learn a time-series pattern of the extracted features using a second neural network model; and a text output unit configured to convert target voice data input to the voice input unit into a text based on learning results of the first voice recognition unit and the second voice recognition unit, and output the text. dated 2018-02-06"
9892344,activation layers for deep learning networks,"tasks such as object classification from image data can take advantage of a deep learning process using convolutional neural networks. these networks can include a convolutional layer followed by an activation layer, or activation unit, among other potential layers. improved accuracy can be obtained by using a generalized linear unit (glu) as an activation unit in such a network, where a glu is linear for both positive and negative inputs, and is defined by a positive slope, a negative slope, and a bias. these parameters can be learned for each channel or a block of channels, and stacking those types of activation units can further improve accuracy.",2018-02-13,"The title of the patent is activation layers for deep learning networks and its abstract is tasks such as object classification from image data can take advantage of a deep learning process using convolutional neural networks. these networks can include a convolutional layer followed by an activation layer, or activation unit, among other potential layers. improved accuracy can be obtained by using a generalized linear unit (glu) as an activation unit in such a network, where a glu is linear for both positive and negative inputs, and is defined by a positive slope, a negative slope, and a bias. these parameters can be learned for each channel or a block of channels, and stacking those types of activation units can further improve accuracy. dated 2018-02-13"
9892731,methods for speech enhancement and speech recognition using neural networks,"the present invention relates to implementing a system and method to improve speech recognition and speech enhancement of noisy speech. the present invention discloses a way to improve the noise robustness of a speech recognition system by providing additional input to a neural network speech classifier. the additional information characterizes the noise environment of the speech. the present invention further discloses a speech separation system that uses the output of the neural network. the speech separation system employs models for the speech and for the distractor or noise. the neural network is used to identify the most likely combinations of speech and noise. furthermore, a system for efficiently finding the most likely clean speech log-spectrum value is disclosed.",2018-02-13,"The title of the patent is methods for speech enhancement and speech recognition using neural networks and its abstract is the present invention relates to implementing a system and method to improve speech recognition and speech enhancement of noisy speech. the present invention discloses a way to improve the noise robustness of a speech recognition system by providing additional input to a neural network speech classifier. the additional information characterizes the noise environment of the speech. the present invention further discloses a speech separation system that uses the output of the neural network. the speech separation system employs models for the speech and for the distractor or noise. the neural network is used to identify the most likely combinations of speech and noise. furthermore, a system for efficiently finding the most likely clean speech log-spectrum value is disclosed. dated 2018-02-13"
9896656,neural progenitor cell differentiation,"differentiation and stability of neural stem cells can be enhanced by in vitro or in vivo culturing with one or more extracellular matrix (ecm) compositions, such as collagen i, iv, laminin and/or a heparan sulfate proteoglycan. in one aspect of the invention, adult mammalian enteric neuronal progenitor cells can be induced to differentiate on various substrates derived from components or combinations of neural ecm compositions. collagen i and iv supported neuronal differentiation and extensive glial differentiation individually and in combination. addition of laminin or heparan sulfate to collagen substrates unexpectedly improved neuronal differentiation, increasing neuron number, branching of neuronal processes, and initiation of neuronal network formation. in another aspect, neuronal subtype differentiation was affected by varying ecm compositions in hydrogels overlaid on intestinal smooth muscle sheets. the matrix compositions of the present invention can be used to tissue engineer transplantable innervated gi smooth muscle constructs to remedy aganglionic disorders.",2018-02-20,"The title of the patent is neural progenitor cell differentiation and its abstract is differentiation and stability of neural stem cells can be enhanced by in vitro or in vivo culturing with one or more extracellular matrix (ecm) compositions, such as collagen i, iv, laminin and/or a heparan sulfate proteoglycan. in one aspect of the invention, adult mammalian enteric neuronal progenitor cells can be induced to differentiate on various substrates derived from components or combinations of neural ecm compositions. collagen i and iv supported neuronal differentiation and extensive glial differentiation individually and in combination. addition of laminin or heparan sulfate to collagen substrates unexpectedly improved neuronal differentiation, increasing neuron number, branching of neuronal processes, and initiation of neuronal network formation. in another aspect, neuronal subtype differentiation was affected by varying ecm compositions in hydrogels overlaid on intestinal smooth muscle sheets. the matrix compositions of the present invention can be used to tissue engineer transplantable innervated gi smooth muscle constructs to remedy aganglionic disorders. dated 2018-02-20"
9898688,vision enhanced drones for precision farming,"methods, apparatuses and systems may provide for a neural network that analyzes and classifies agricultural conditions based on depth data and color data recorded by one or more drones, and generates an annotated three dimensional (3d) map with the agricultural conditions. additionally, an object recognition model may be trained for use by a drone controller to trigger drones to conduct a collection of depth data at an increased proximity to crop-related objects based on agricultural conditions.",2018-02-20,"The title of the patent is vision enhanced drones for precision farming and its abstract is methods, apparatuses and systems may provide for a neural network that analyzes and classifies agricultural conditions based on depth data and color data recorded by one or more drones, and generates an annotated three dimensional (3d) map with the agricultural conditions. additionally, an object recognition model may be trained for use by a drone controller to trigger drones to conduct a collection of depth data at an increased proximity to crop-related objects based on agricultural conditions. dated 2018-02-20"
9900338,forecasting and classifying cyber-attacks using neural embeddings based on pattern of life data,"a first collection including a pattern of life (pol) feature vector and a q&a feature vector is constructed. a second collection is constructed from the first collection by inserting noise in at least one of the vectors. a third collection is constructed by crossing over at least one of vectors of the second collection with a corresponding vector of a fourth collection, migrating at least one of the vectors of the second collection with a corresponding vector of a fifth collection. using a forecasting configuration, a pol feature vector of the third collection is aged to generate a changed pol feature vector containing pol feature values expected at a future time. the changed pol feature vector is input into a trained neural network to predict a probability of the cyber-attack occurring at the future time.",2018-02-20,"The title of the patent is forecasting and classifying cyber-attacks using neural embeddings based on pattern of life data and its abstract is a first collection including a pattern of life (pol) feature vector and a q&a feature vector is constructed. a second collection is constructed from the first collection by inserting noise in at least one of the vectors. a third collection is constructed by crossing over at least one of vectors of the second collection with a corresponding vector of a fourth collection, migrating at least one of the vectors of the second collection with a corresponding vector of a fifth collection. using a forecasting configuration, a pol feature vector of the third collection is aged to generate a changed pol feature vector containing pol feature values expected at a future time. the changed pol feature vector is input into a trained neural network to predict a probability of the cyber-attack occurring at the future time. dated 2018-02-20"
9903720,method and apparatus for creating cost data for use in generating a route across an electronic map,"a method is disclosed involving receiving gps data from persona portable training devices of users when traversing an off-road segment of an electronic map together with associated data indicative of a heart rate of a user during the movements. the position and heart rate data for each user traversing the segment are processed using data indicative of a fitness profile for the user. the resulting data is used to determine a normalized cost to be associated with the segment, indicative of the difficulty in traversing the segment. the cost data is generated using a neural network. the resulting cost data for different segments in a network of segments is used to generate route suggestions for users based upon desired workout intensity, fitness levels, etc.",2018-02-27,"The title of the patent is method and apparatus for creating cost data for use in generating a route across an electronic map and its abstract is a method is disclosed involving receiving gps data from persona portable training devices of users when traversing an off-road segment of an electronic map together with associated data indicative of a heart rate of a user during the movements. the position and heart rate data for each user traversing the segment are processed using data indicative of a fitness profile for the user. the resulting data is used to determine a normalized cost to be associated with the segment, indicative of the difficulty in traversing the segment. the cost data is generated using a neural network. the resulting cost data for different segments in a network of segments is used to generate route suggestions for users based upon desired workout intensity, fitness levels, etc. dated 2018-02-27"
9903963,"method, apparatus and system of the correction of energy crosstalk in dual-isotopes simultaneous acquisition","the present invention relates the system of the correction of energy crosstalk in dual-isotopes simultaneous acquisition (disa), the system includes a collimator, a metal thin film, a detecting unit, an analyzing unit and a display unit for analyzing energy distribution charts of the dual-isotopes, and using specific equations or artificial neural network methods or independent component analysis to compare the energy distribution charts which are with and without metal thin film the invention uses the metal thin film to remove the energy contamination from dual-isotopes simultaneous acquisition whose photopeak energies are close, the invention effectively separates the energy distribution charts without energy crosstalk, therefore, the system improves diagnostic imaging and relieves patient's discomfort.",2018-02-27,"The title of the patent is method, apparatus and system of the correction of energy crosstalk in dual-isotopes simultaneous acquisition and its abstract is the present invention relates the system of the correction of energy crosstalk in dual-isotopes simultaneous acquisition (disa), the system includes a collimator, a metal thin film, a detecting unit, an analyzing unit and a display unit for analyzing energy distribution charts of the dual-isotopes, and using specific equations or artificial neural network methods or independent component analysis to compare the energy distribution charts which are with and without metal thin film the invention uses the metal thin film to remove the energy contamination from dual-isotopes simultaneous acquisition whose photopeak energies are close, the invention effectively separates the energy distribution charts without energy crosstalk, therefore, the system improves diagnostic imaging and relieves patient's discomfort. dated 2018-02-27"
9904871,deep convolutional neural network prediction of image professionalism,"in an example embodiment, a deep convolutional neural network (dcnn) is created to assign a professionalism score to an input image. the professionalism score indicates a perceived professionalism of a subject of the input image. the dcnn is designed to automatically learn features of images relevant to the professionalism through a training process.",2018-02-27,"The title of the patent is deep convolutional neural network prediction of image professionalism and its abstract is in an example embodiment, a deep convolutional neural network (dcnn) is created to assign a professionalism score to an input image. the professionalism score indicates a perceived professionalism of a subject of the input image. the dcnn is designed to automatically learn features of images relevant to the professionalism through a training process. dated 2018-02-27"
9904875,processing images using deep neural networks,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for image processing using deep neural networks. one of the methods includes receiving data characterizing an input image; processing the data characterizing the input image using a deep neural network to generate an alternative representation of the input image, wherein the deep neural network comprises a plurality of subnetworks, wherein the subnetworks are arranged in a sequence from lowest to highest, and wherein processing the data characterizing the input image using the deep neural network comprises processing the data through each of the subnetworks in the sequence; and processing the alternative representation of the input image through an output layer to generate an output from the input image.",2018-02-27,"The title of the patent is processing images using deep neural networks and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for image processing using deep neural networks. one of the methods includes receiving data characterizing an input image; processing the data characterizing the input image using a deep neural network to generate an alternative representation of the input image, wherein the deep neural network comprises a plurality of subnetworks, wherein the subnetworks are arranged in a sequence from lowest to highest, and wherein processing the data characterizing the input image using the deep neural network comprises processing the data through each of the subnetworks in the sequence; and processing the alternative representation of the input image through an output layer to generate an output from the input image. dated 2018-02-27"
9904889,methods and systems for artificial cognition,"methods, systems and apparatus that provide for perceptual, cognitive, and motor behaviors in an integrated system implemented using neural architectures. components of the system communicate using artificial neurons that implement neural networks. the connections between these networks form representations—referred to as semantic pointers—which model the various firing patterns of biological neural network connections. semantic pointers can be thought of as elements of a neural vector space, and can implement a form of abstraction level filtering or compression, in which high-dimensional structures can be abstracted one or more times thereby reducing the number of dimensions needed to represent a particular structure.",2018-02-27,"The title of the patent is methods and systems for artificial cognition and its abstract is methods, systems and apparatus that provide for perceptual, cognitive, and motor behaviors in an integrated system implemented using neural architectures. components of the system communicate using artificial neurons that implement neural networks. the connections between these networks form representations—referred to as semantic pointers—which model the various firing patterns of biological neural network connections. semantic pointers can be thought of as elements of a neural vector space, and can implement a form of abstraction level filtering or compression, in which high-dimensional structures can be abstracted one or more times thereby reducing the number of dimensions needed to represent a particular structure. dated 2018-02-27"
9904976,high performance portable convulational neural network library on gp-gpus,systems and methods are disclosed for speeding up a computer having a graphics processing unit (gpu) and a general purpose processor (gp-gpu) by decoupling a convolution process for a first matrix into a row part and a column part; expanding the row part into a second matrix; performing matrix multiplication using the second matrix and a filter matrix; and performing reduction on an output matrix.,2018-02-27,The title of the patent is high performance portable convulational neural network library on gp-gpus and its abstract is systems and methods are disclosed for speeding up a computer having a graphics processing unit (gpu) and a general purpose processor (gp-gpu) by decoupling a convolution process for a first matrix into a row part and a column part; expanding the row part into a second matrix; performing matrix multiplication using the second matrix and a filter matrix; and performing reduction on an output matrix. dated 2018-02-27
9905220,multilingual prosody generation,"methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for multilingual prosody generation. in some implementations, data indicating a set of linguistic features corresponding to a text is obtained. data indicating the linguistic features and data indicating the language of the text are provided as input to a neural network that has been trained to provide output indicating prosody information for multiple languages. the neural network can be a neural network having been trained using speech in multiple languages. output indicating prosody information for the linguistic features is received from the neural network. audio data representing the text is generated using the output of the neural network.",2018-02-27,"The title of the patent is multilingual prosody generation and its abstract is methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for multilingual prosody generation. in some implementations, data indicating a set of linguistic features corresponding to a text is obtained. data indicating the linguistic features and data indicating the language of the text are provided as input to a neural network that has been trained to provide output indicating prosody information for multiple languages. the neural network can be a neural network having been trained using speech in multiple languages. output indicating prosody information for the linguistic features is received from the neural network. audio data representing the text is generated using the output of the neural network. dated 2018-02-27"
9906551,forecasting and classifying cyber-attacks using crossover neural embeddings,"a first collection including a first feature vector and a q&a feature vector is constructed. a second collection is constructed from the first collection by inserting noise in at least one of the vectors. a third collection is constructed by crossing over at least one of vectors of the second collection with a corresponding vector of a fourth collection. the second and the fourth collections have a property similar to one another. using a forecasting configuration, a vector of the third collection is aged to generate a changed feature vector, the changed feature vector containing feature values expected at a future time. the changed feature vector is input into a trained neural network to predict a probability of the cyber-attack occurring at the future time.",2018-02-27,"The title of the patent is forecasting and classifying cyber-attacks using crossover neural embeddings and its abstract is a first collection including a first feature vector and a q&a feature vector is constructed. a second collection is constructed from the first collection by inserting noise in at least one of the vectors. a third collection is constructed by crossing over at least one of vectors of the second collection with a corresponding vector of a fourth collection. the second and the fourth collections have a property similar to one another. using a forecasting configuration, a vector of the third collection is aged to generate a changed feature vector, the changed feature vector containing feature values expected at a future time. the changed feature vector is input into a trained neural network to predict a probability of the cyber-attack occurring at the future time. dated 2018-02-27"
9906718,biomimetic integrated optical sensor (bios) system,"the subject invention includes a biomimetic integrated optical sensor system, based on the integration of a wide field-of-view (wfov) miniature staring multi-aperture compound eye with a high-speed, low-cost, polarization and spectral selective liquid crystal (lc) filter array, a mwir focal plane array (fpa), and a neural network processor.",2018-02-27,"The title of the patent is biomimetic integrated optical sensor (bios) system and its abstract is the subject invention includes a biomimetic integrated optical sensor system, based on the integration of a wide field-of-view (wfov) miniature staring multi-aperture compound eye with a high-speed, low-cost, polarization and spectral selective liquid crystal (lc) filter array, a mwir focal plane array (fpa), and a neural network processor. dated 2018-02-27"
9911069,processing images using deep neural networks,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for image processing using deep neural networks. one of the methods includes receiving data characterizing an input image; processing the data characterizing the input image using a deep neural network to generate an alternative representation of the input image, wherein the deep neural network comprises a plurality of subnetworks, wherein the subnetworks are arranged in a sequence from lowest to highest, and wherein processing the data characterizing the input image using the deep neural network comprises processing the data through each of the subnetworks in the sequence; and processing the alternative representation of the input image through an output layer to generate an output from the input image.",2018-03-06,"The title of the patent is processing images using deep neural networks and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for image processing using deep neural networks. one of the methods includes receiving data characterizing an input image; processing the data characterizing the input image using a deep neural network to generate an alternative representation of the input image, wherein the deep neural network comprises a plurality of subnetworks, wherein the subnetworks are arranged in a sequence from lowest to highest, and wherein processing the data characterizing the input image using the deep neural network comprises processing the data through each of the subnetworks in the sequence; and processing the alternative representation of the input image through an output layer to generate an output from the input image. dated 2018-03-06"
9911413,neural latent variable model for spoken language understanding,"a linguist classifier, for instance intent or slot classifier, is updated using data with only partial annotation indicating overall correctness rather that specific correct intent or slot values, which are treated as “latent” (i.e., unknown) variables. full annotation of the data is not required. a small amount of fully annotated data may be combined with a substantially larger amount of partially annotated data to update the linguistic classifier. in a specific implementation, the linguistic classifier is a neural network and the weights are trained using a reinforcement learning approach.",2018-03-06,"The title of the patent is neural latent variable model for spoken language understanding and its abstract is a linguist classifier, for instance intent or slot classifier, is updated using data with only partial annotation indicating overall correctness rather that specific correct intent or slot values, which are treated as “latent” (i.e., unknown) variables. full annotation of the data is not required. a small amount of fully annotated data may be combined with a substantially larger amount of partially annotated data to update the linguistic classifier. in a specific implementation, the linguistic classifier is a neural network and the weights are trained using a reinforcement learning approach. dated 2018-03-06"
9916531,accumulator constrained quantization of convolutional neural networks,"an apparatus is described herein. the apparatus comprises an accumulator, a controller, and a convolutional neural network. the accumulator is to accumulate a plurality of values within a predetermined bit width. the controller is to determine a parameter quantization and a data quantization. the convolutional neural network is adapted to the data quantization, wherein a quantization point is selected based on the parameter quantization, data quantization, and accumulator bit width.",2018-03-13,"The title of the patent is accumulator constrained quantization of convolutional neural networks and its abstract is an apparatus is described herein. the apparatus comprises an accumulator, a controller, and a convolutional neural network. the accumulator is to accumulate a plurality of values within a predetermined bit width. the controller is to determine a parameter quantization and a data quantization. the convolutional neural network is adapted to the data quantization, wherein a quantization point is selected based on the parameter quantization, data quantization, and accumulator bit width. dated 2018-03-13"
9916825,method and system for text-to-speech synthesis,"there are disclosed methods and systems for text-to-speech synthesis for outputting a synthetic speech having a selected speech attribute. first, an acoustic space model is trained based on a set of training data of speech attributes, using a deep neural network to determine interdependency factors between the speech attributes in the training data, the dnn generating a single, continuous acoustic space model based on the interdependency factors, the acoustic space model thereby taking into account a plurality of interdependent speech attributes and allowing for modelling of a continuous spectrum of the interdependent speech attributes. next, a text is received; a selection of one or more speech attribute is received, each speech attribute having a selected attribute weight; the text is converted into synthetic speech using the acoustic space model, the synthetic speech having the selected speech attribute; and the synthetic speech is outputted as audio having the selected speech attribute.",2018-03-13,"The title of the patent is method and system for text-to-speech synthesis and its abstract is there are disclosed methods and systems for text-to-speech synthesis for outputting a synthetic speech having a selected speech attribute. first, an acoustic space model is trained based on a set of training data of speech attributes, using a deep neural network to determine interdependency factors between the speech attributes in the training data, the dnn generating a single, continuous acoustic space model based on the interdependency factors, the acoustic space model thereby taking into account a plurality of interdependent speech attributes and allowing for modelling of a continuous spectrum of the interdependent speech attributes. next, a text is received; a selection of one or more speech attribute is received, each speech attribute having a selected attribute weight; the text is converted into synthetic speech using the acoustic space model, the synthetic speech having the selected speech attribute; and the synthetic speech is outputted as audio having the selected speech attribute. dated 2018-03-13"
9917725,automotive neural network,network node modules within a vehicle are arranged to form a reconfigurable automotive neural network. each network node module includes one or more subsystems for performing one or more operations and a local processing module for communicating with the one or more subsystems. a management system enables traffic from the one or more subsystems of a particular network node module to be re-routed to an external processing module upon failure of the local processing module of that particular network node module.,2018-03-13,The title of the patent is automotive neural network and its abstract is network node modules within a vehicle are arranged to form a reconfigurable automotive neural network. each network node module includes one or more subsystems for performing one or more operations and a local processing module for communicating with the one or more subsystems. a management system enables traffic from the one or more subsystems of a particular network node module to be re-routed to an external processing module upon failure of the local processing module of that particular network node module. dated 2018-03-13
9921068,devices and methods to facilitate escape from a venue with a sudden hazard,"a device and associated methods for escaping from a venue when a threat is detected is described. venues can be buildings or outside areas and contain the area where the threat constitutes a hazard to a protected person. threats include fire, terrorists, gunmen, explosion, collapse, loss of critical resources and crowd panic. the device incorporates a machine learning system implemented with a neural network or other pattern matching system and is trained in steps. pre-training is based on general requirements such as edge-detection and audio analysis. principles and data for venue layouts and human behavior can be included. the produced model is further trained from data gathered from sensors and servers after entry into the venue. operation of the model produces warnings of threats and a plan of escape with steps of the plan communicated to the protected person by audio, visual or tactile sensory channels.",2018-03-20,"The title of the patent is devices and methods to facilitate escape from a venue with a sudden hazard and its abstract is a device and associated methods for escaping from a venue when a threat is detected is described. venues can be buildings or outside areas and contain the area where the threat constitutes a hazard to a protected person. threats include fire, terrorists, gunmen, explosion, collapse, loss of critical resources and crowd panic. the device incorporates a machine learning system implemented with a neural network or other pattern matching system and is trained in steps. pre-training is based on general requirements such as edge-detection and audio analysis. principles and data for venue layouts and human behavior can be included. the produced model is further trained from data gathered from sensors and servers after entry into the venue. operation of the model produces warnings of threats and a plan of escape with steps of the plan communicated to the protected person by audio, visual or tactile sensory channels. dated 2018-03-20"
9922272,deep similarity learning for multimodal medical images,"the present embodiments relate to machine learning for multimodal image data. by way of introduction, the present embodiments described below include apparatuses and methods for learning a similarity metric using deep learning based techniques for multimodal medical images. a novel similarity metric for multi-modal images is provided using the corresponding states of pairs of image patches to generate a classification setting for each pair. the classification settings are used to train a deep neural network via supervised learning. a multi-modal stacked denoising auto encoder (sdae) is used to pre-train the neural network. a continuous and smooth similarity metric is constructed based on the output of the neural network before activation in the last layer. the trained similarity metric may be used to improve the results of image fusion.",2018-03-20,"The title of the patent is deep similarity learning for multimodal medical images and its abstract is the present embodiments relate to machine learning for multimodal image data. by way of introduction, the present embodiments described below include apparatuses and methods for learning a similarity metric using deep learning based techniques for multimodal medical images. a novel similarity metric for multi-modal images is provided using the corresponding states of pairs of image patches to generate a classification setting for each pair. the classification settings are used to train a deep neural network via supervised learning. a multi-modal stacked denoising auto encoder (sdae) is used to pre-train the neural network. a continuous and smooth similarity metric is constructed based on the output of the neural network before activation in the last layer. the trained similarity metric may be used to improve the results of image fusion. dated 2018-03-20"
9922432,"systems and methods for providing convolutional neural network based image synthesis using stable and controllable parametric models, a multiscale synthesis framework and novel network architectures","systems and methods for providing convolutional neural network based image synthesis using localized loss functions is disclosed. a first image including desired content and a second image including a desired style are received. the images are analyzed to determine a local loss function. the first and second images are merged using the local loss function to generate an image that includes the desired content presented in the desired style. similar processes can also be utilized to generate image hybrids and to perform on-model texture synthesis. in a number of embodiments, condensed feature extraction networks are also generated using a convolutional neural network previously trained to perform image classification, where the condensed feature extraction networks approximates intermediate neural activations of the convolutional neural network utilized during training.",2018-03-20,"The title of the patent is systems and methods for providing convolutional neural network based image synthesis using stable and controllable parametric models, a multiscale synthesis framework and novel network architectures and its abstract is systems and methods for providing convolutional neural network based image synthesis using localized loss functions is disclosed. a first image including desired content and a second image including a desired style are received. the images are analyzed to determine a local loss function. the first and second images are merged using the local loss function to generate an image that includes the desired content presented in the desired style. similar processes can also be utilized to generate image hybrids and to perform on-model texture synthesis. in a number of embodiments, condensed feature extraction networks are also generated using a convolutional neural network previously trained to perform image classification, where the condensed feature extraction networks approximates intermediate neural activations of the convolutional neural network utilized during training. dated 2018-03-20"
9924927,method and apparatus for video interpretation of carotid intima-media thickness,"a system for automatically determining a thickness of a wall of an artery of a subject includes an ecg monitoring device that captures an electrocardiogram (ecg) signal from the subject, and an ultrasound video imaging device, coupled to the ecg monitoring device, that receives the ecg signal from the ecg monitoring device, and captures a corresponding ultrasound video of the wall of the artery of the subject. the system produces a plurality of frames of video comprising the ultrasound video of the wall of the artery of the subject and an image of the ecg signal. a processor is configured to select a subset of the plurality of frames of the ultrasound video based on the image of the (ecg) signal, locate automatically a region of interest (roi) in each frame of the subset of the plurality of frames of the video using a machine-based artificial neural network and measure automatically a thickness of the wall of the artery in each roi using the machine-based artificial neural network.",2018-03-27,"The title of the patent is method and apparatus for video interpretation of carotid intima-media thickness and its abstract is a system for automatically determining a thickness of a wall of an artery of a subject includes an ecg monitoring device that captures an electrocardiogram (ecg) signal from the subject, and an ultrasound video imaging device, coupled to the ecg monitoring device, that receives the ecg signal from the ecg monitoring device, and captures a corresponding ultrasound video of the wall of the artery of the subject. the system produces a plurality of frames of video comprising the ultrasound video of the wall of the artery of the subject and an image of the ecg signal. a processor is configured to select a subset of the plurality of frames of the ultrasound video based on the image of the (ecg) signal, locate automatically a region of interest (roi) in each frame of the subset of the plurality of frames of the video using a machine-based artificial neural network and measure automatically a thickness of the wall of the artery in each roi using the machine-based artificial neural network. dated 2018-03-27"
9928449,tagging similar images using neural network,"an approach is provided in which a knowledge manager selects an extraction layer from a convolutional neural network that was trained on an initial set of images. the knowledge manager processes subsequent images obtained from crawling a computer network that includes extracting image feature sets of the subsequent images from the selected extraction layer and generating tags from metadata associated with the subsequent images. in turn, the knowledge manager receives a new image, extracts a new image feature set from the selected extraction layer, and assigns one or more of the tags to the new image based upon evaluating the new image feature set to the image features sets of the subsequent images.",2018-03-27,"The title of the patent is tagging similar images using neural network and its abstract is an approach is provided in which a knowledge manager selects an extraction layer from a convolutional neural network that was trained on an initial set of images. the knowledge manager processes subsequent images obtained from crawling a computer network that includes extracting image feature sets of the subsequent images from the selected extraction layer and generating tags from metadata associated with the subsequent images. in turn, the knowledge manager receives a new image, extracts a new image feature set from the selected extraction layer, and assigns one or more of the tags to the new image based upon evaluating the new image feature set to the image features sets of the subsequent images. dated 2018-03-27"
9928460,neural network accelerator tile architecture with three-dimensional stacking,"a three dimensional neural network accelerator that includes a first neural network accelerator tile that includes a first transmission coil, and a second neural network accelerator tile that includes a second transmission coil, wherein the first neural network accelerator tile is adjacent to and aligned vertically with the second neural network accelerator tile, and wherein the first transmission coil is configured to wirelessly communicate with the second transmission coil via inductive coupling.",2018-03-27,"The title of the patent is neural network accelerator tile architecture with three-dimensional stacking and its abstract is a three dimensional neural network accelerator that includes a first neural network accelerator tile that includes a first transmission coil, and a second neural network accelerator tile that includes a second transmission coil, wherein the first neural network accelerator tile is adjacent to and aligned vertically with the second neural network accelerator tile, and wherein the first transmission coil is configured to wirelessly communicate with the second transmission coil via inductive coupling. dated 2018-03-27"
9928461,hyper aware logic to create an agent of consciousness and intent for devices and machines,"a neural logic unit network acting as an agent to achieve machine or device consciousness and intent is disclosed.more specifically, an agent of consciousness and intent (the agent) is disclosed consisting of neuronal logic units upon which are mapped and connected to the individual outputs of the host system's entire sensorium and which neuronal logic units are activated by the simultaneous presentation of the results of the host system's recognition, tracking, analyzes and characterization computations similar to those performed by biological unconscious brains.the embodiment of the assembly of neural logic units is referred to as hyper aware logic.",2018-03-27,"The title of the patent is hyper aware logic to create an agent of consciousness and intent for devices and machines and its abstract is a neural logic unit network acting as an agent to achieve machine or device consciousness and intent is disclosed.more specifically, an agent of consciousness and intent (the agent) is disclosed consisting of neuronal logic units upon which are mapped and connected to the individual outputs of the host system's entire sensorium and which neuronal logic units are activated by the simultaneous presentation of the results of the host system's recognition, tracking, analyzes and characterization computations similar to those performed by biological unconscious brains.the embodiment of the assembly of neural logic units is referred to as hyper aware logic. dated 2018-03-27"
9929933,loading a flow table with neural network determined information,"a flow of packets is communicated through a data center. the data center includes multiple racks, where each rack includes multiple network devices. a group of packets of the flow is received onto an integrated circuit located in one of the network devices. the integrated circuit includes a neural network and a flow table. the neural network analyzes the group of packets and in response determines if it is likely that the flow has a particular characteristic. the neural network outputs a neural network output value that indicates if it is likely that the flow has a particular characteristic. the neural network output value, or a value derived from it, is included in a flow entry in the flow table on the integrated circuit. packets of the flow subsequently received onto the integrated circuit are routed or otherwise processed according to the flow entry associated with the flow.",2018-03-27,"The title of the patent is loading a flow table with neural network determined information and its abstract is a flow of packets is communicated through a data center. the data center includes multiple racks, where each rack includes multiple network devices. a group of packets of the flow is received onto an integrated circuit located in one of the network devices. the integrated circuit includes a neural network and a flow table. the neural network analyzes the group of packets and in response determines if it is likely that the flow has a particular characteristic. the neural network outputs a neural network output value that indicates if it is likely that the flow has a particular characteristic. the neural network output value, or a value derived from it, is included in a flow entry in the flow table on the integrated circuit. packets of the flow subsequently received onto the integrated circuit are routed or otherwise processed according to the flow entry associated with the flow. dated 2018-03-27"
9934364,methods for using artificial neural network analysis on flow cytometry data for cancer diagnosis,the present disclosure provides methods for applying artificial neural networks to flow cytometry data generated from biological samples to diagnose and characterize cancer in a subject.,2018-04-03,The title of the patent is methods for using artificial neural network analysis on flow cytometry data for cancer diagnosis and its abstract is the present disclosure provides methods for applying artificial neural networks to flow cytometry data generated from biological samples to diagnose and characterize cancer in a subject. dated 2018-04-03
9934437,system and method for real-time collision detection,"described is a system for collision detection. the system divides an image in a sequence of images into multiple sub-fields comprising complementary visual sub-fields. for each visual sub-field, motion is detected in a direction corresponding to the visual sub-field using a spiking reichardt detector with a spiking neural network. motion in a direction complementary to the visual sub-field is also detected using the spiking reichardt detector. outputs of the spiking reichardt detector, comprising data corresponding to one direction of movement from two complementary visual sub-fields, are processed using a movement detector. based on the output of the movement detector, an impending collision is signaled.",2018-04-03,"The title of the patent is system and method for real-time collision detection and its abstract is described is a system for collision detection. the system divides an image in a sequence of images into multiple sub-fields comprising complementary visual sub-fields. for each visual sub-field, motion is detected in a direction corresponding to the visual sub-field using a spiking reichardt detector with a spiking neural network. motion in a direction complementary to the visual sub-field is also detected using the spiking reichardt detector. outputs of the spiking reichardt detector, comprising data corresponding to one direction of movement from two complementary visual sub-fields, are processed using a movement detector. based on the output of the movement detector, an impending collision is signaled. dated 2018-04-03"
9934462,visualizing deep neural networks,"deep neural networks can be visualized. for example, first values for a first layer of nodes in a neural network, second values for a second layer of nodes in the neural network, and/or third values for connections between the first layer of nodes and the second layer of nodes can be received. a quilt graph can be output that includes (i) a first set of symbols having visual characteristics representative of the first values and representing the first layer of nodes along a first axis; (ii) a second set of symbols having visual characteristics representative of the second values and representing the second layer of nodes along a second axis; and/or (iii) a matrix of blocks between the first axis and the second axis having visual characteristics representative of the third values and representing the connections between the first layer of nodes and the second layer of nodes.",2018-04-03,"The title of the patent is visualizing deep neural networks and its abstract is deep neural networks can be visualized. for example, first values for a first layer of nodes in a neural network, second values for a second layer of nodes in the neural network, and/or third values for connections between the first layer of nodes and the second layer of nodes can be received. a quilt graph can be output that includes (i) a first set of symbols having visual characteristics representative of the first values and representing the first layer of nodes along a first axis; (ii) a second set of symbols having visual characteristics representative of the second values and representing the second layer of nodes along a second axis; and/or (iii) a matrix of blocks between the first axis and the second axis having visual characteristics representative of the third values and representing the connections between the first layer of nodes and the second layer of nodes. dated 2018-04-03"
9934515,content recommendation system using a neural network language model,"the present disclosure relates to applying techniques similar to those used in neural network language modeling systems to a content recommendation system. for example, by associating consumed media content to words of a language model, the system may provide content predictions based on an ordering. thus, the systems and techniques described herein may produce enhanced prediction results for recommending content (e.g. word) in a given sequence of consumed content. in addition, the system may account for additional user actions by representing particular actions as punctuation in the language model.",2018-04-03,"The title of the patent is content recommendation system using a neural network language model and its abstract is the present disclosure relates to applying techniques similar to those used in neural network language modeling systems to a content recommendation system. for example, by associating consumed media content to words of a language model, the system may provide content predictions based on an ordering. thus, the systems and techniques described herein may produce enhanced prediction results for recommending content (e.g. word) in a given sequence of consumed content. in addition, the system may account for additional user actions by representing particular actions as punctuation in the language model. dated 2018-04-03"
9934826,semiconductor device,"to provide a semiconductor device including a first memory cell for holding first analog data, a second memory cell for holding reference analog data, and an offset circuit. the first memory cell and the second memory cell supply a first current and a second current, respectively, when a reference potential is supplied. the offset circuit supplies a third current corresponding to a differential current between the first current and the second current. the first memory and the second memory supply a fourth current and a fifth current, respectively, when a potential corresponding to second analog data is supplied. by subtracting the third current from a differential current between the fourth current and the fifth current, a current that depends on the sum of products of the first analog data and the second analog data is obtained. by providing a plurality of product-sum operation circuits that can be freely connected, a hierarchical neural network can be formed.",2018-04-03,"The title of the patent is semiconductor device and its abstract is to provide a semiconductor device including a first memory cell for holding first analog data, a second memory cell for holding reference analog data, and an offset circuit. the first memory cell and the second memory cell supply a first current and a second current, respectively, when a reference potential is supplied. the offset circuit supplies a third current corresponding to a differential current between the first current and the second current. the first memory and the second memory supply a fourth current and a fifth current, respectively, when a potential corresponding to second analog data is supplied. by subtracting the third current from a differential current between the fourth current and the fifth current, a current that depends on the sum of products of the first analog data and the second analog data is obtained. by providing a plurality of product-sum operation circuits that can be freely connected, a hierarchical neural network can be formed. dated 2018-04-03"
9939548,"systems, methods, and computer medium to produce efficient, consistent, and high-confidence image-based electrofacies analysis in stratigraphic interpretations across multiple wells","embodiments of systems, computer-implemented methods, and non-transitory computer-readable medium having one or more computer programs stored therein are provided to enhance borehole image analysis associated with a hydrocarbon reservoir. a neural network mapping process can first be performed, responsive to openhole log data and core data, to thereby generate a material-type scheme. then, an image-based petrophysical analysis process can distribute and calibrate borehole image data, responsive to the core data and material-type scheme. consequently, an approximated material type and an approximated grain size can be produced for each borehole image reading. the openhole log data, the core data, the material-type scheme, and the approximated material types and grain sizes, for example, can then be displayed to thereby increase consistency in categorizing subsurface material associated with hydrocarbon wells by material type and enhance interpretation of subsurface material texture, fabric, and features to predict subsurface material composition of the hydrocarbon reservoir.",2018-04-10,"The title of the patent is systems, methods, and computer medium to produce efficient, consistent, and high-confidence image-based electrofacies analysis in stratigraphic interpretations across multiple wells and its abstract is embodiments of systems, computer-implemented methods, and non-transitory computer-readable medium having one or more computer programs stored therein are provided to enhance borehole image analysis associated with a hydrocarbon reservoir. a neural network mapping process can first be performed, responsive to openhole log data and core data, to thereby generate a material-type scheme. then, an image-based petrophysical analysis process can distribute and calibrate borehole image data, responsive to the core data and material-type scheme. consequently, an approximated material type and an approximated grain size can be produced for each borehole image reading. the openhole log data, the core data, the material-type scheme, and the approximated material types and grain sizes, for example, can then be displayed to thereby increase consistency in categorizing subsurface material associated with hydrocarbon wells by material type and enhance interpretation of subsurface material texture, fabric, and features to predict subsurface material composition of the hydrocarbon reservoir. dated 2018-04-10"
9939792,systems and methods to adaptively select execution modes,"methods and systems that facilitate efficient and effective adaptive execution mode selection are described. the adaptive execution mode selection is performed in part on-the-fly and changes to an execution mode (e.g., sequential, parallel, etc.) for a program task can be made. an intelligent adaptive selection can be made between a variety execution modes. the adaptive execution mode selection can also include selecting parameters associated with the execution modes. a controller receives historical information associated with execution mode selection, engages in training regarding execution mode selection, and adaptively selects an execution mode on-the-fly. the training can use an approach similar to an artificial neural network in which automated guided machine learning approach establishes correspondences between execution modes and task/input feature definitions based upon historical information. an adaptive selection is performed on-the-fly based on an initial trial run.",2018-04-10,"The title of the patent is systems and methods to adaptively select execution modes and its abstract is methods and systems that facilitate efficient and effective adaptive execution mode selection are described. the adaptive execution mode selection is performed in part on-the-fly and changes to an execution mode (e.g., sequential, parallel, etc.) for a program task can be made. an intelligent adaptive selection can be made between a variety execution modes. the adaptive execution mode selection can also include selecting parameters associated with the execution modes. a controller receives historical information associated with execution mode selection, engages in training regarding execution mode selection, and adaptively selects an execution mode on-the-fly. the training can use an approach similar to an artificial neural network in which automated guided machine learning approach establishes correspondences between execution modes and task/input feature definitions based upon historical information. an adaptive selection is performed on-the-fly based on an initial trial run. dated 2018-04-10"
9940509,object detection method and object detection apparatus,"an object detection method and an object detection apparatus are provided. the object detection method comprises: mapping at least one image frame in an image sequence into a three dimensional physical space to obtain three dimensional coordinates of each pixel in the at least one image frame; extracting a foreground region in the at least one image frame; segmenting the foreground region into a set of blobs; and detecting, for each blob in the set of blobs, an object in the blob through a neural network based on the three dimensional coordinates of at least one predetermined reference point in the blob, to obtain an object detection result.",2018-04-10,"The title of the patent is object detection method and object detection apparatus and its abstract is an object detection method and an object detection apparatus are provided. the object detection method comprises: mapping at least one image frame in an image sequence into a three dimensional physical space to obtain three dimensional coordinates of each pixel in the at least one image frame; extracting a foreground region in the at least one image frame; segmenting the foreground region into a set of blobs; and detecting, for each blob in the set of blobs, an object in the blob through a neural network based on the three dimensional coordinates of at least one predetermined reference point in the blob, to obtain an object detection result. dated 2018-04-10"
9940520,automatic target recognition system with online machine learning capability,"a method and apparatus for real-time target recognition within a multispectral image includes generating radiance signatures from reflectance signatures, sensor information and environment information and detecting targets in the multispectral image with a sparsity-driven target recognition algorithm utilizing set of parameters tuned with a deep neural network.",2018-04-10,"The title of the patent is automatic target recognition system with online machine learning capability and its abstract is a method and apparatus for real-time target recognition within a multispectral image includes generating radiance signatures from reflectance signatures, sensor information and environment information and detecting targets in the multispectral image with a sparsity-driven target recognition algorithm utilizing set of parameters tuned with a deep neural network. dated 2018-04-10"
9940544,event image curation,"in embodiments of event image curation, a computing device includes memory that stores a collection of digital images associated with a type of event, such as a digital photo album of digital photos associated with the event, or a video of image frames and the video is associated with the event. a curation application implements a convolutional neural network, which receives the digital images and a designation of the type of event. the convolutional neural network can then determine an importance rating of each digital image within the collection of the digital images based on the type of the event. the importance rating of a digital image is representative of an importance of the digital image to a person in context of the type of the event. the convolutional neural network generates an output of representative digital images from the collection based on the importance rating of each digital image.",2018-04-10,"The title of the patent is event image curation and its abstract is in embodiments of event image curation, a computing device includes memory that stores a collection of digital images associated with a type of event, such as a digital photo album of digital photos associated with the event, or a video of image frames and the video is associated with the event. a curation application implements a convolutional neural network, which receives the digital images and a designation of the type of event. the convolutional neural network can then determine an importance rating of each digital image within the collection of the digital images based on the type of the event. the importance rating of a digital image is representative of an importance of the digital image to a person in context of the type of the event. the convolutional neural network generates an output of representative digital images from the collection based on the importance rating of each digital image. dated 2018-04-10"
9940551,image generation using neural networks,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for image generation using neural networks. in one of the methods, an initial image is received. data defining an objective function is received, and the objective function is dependent on processing of a neural network trained to identify features of an image. the initial image is modified to generate a modified image by iteratively performing the following: a current version of the initial image is processed using the neural network to generate a current objective score for the current version of the initial image using the objective function; and the current version of the initial image is modified to increase the current objective score by enhancing a feature detected by the processing.",2018-04-10,"The title of the patent is image generation using neural networks and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for image generation using neural networks. in one of the methods, an initial image is received. data defining an objective function is received, and the objective function is dependent on processing of a neural network trained to identify features of an image. the initial image is modified to generate a modified image by iteratively performing the following: a current version of the initial image is processed using the neural network to generate a current objective score for the current version of the initial image using the objective function; and the current version of the initial image is modified to increase the current objective score by enhancing a feature detected by the processing. dated 2018-04-10"
9940573,superpixel methods for convolutional neural networks,"methods, systems, and apparatus for efficiently performing a computation of a convolutional neural network layer. one of the methods includes transforming a x by y by z input tensor into a x′ by y′ by z′ input tensor, wherein x′ is smaller than or equal to x, y′ is smaller than or equal to y, and z′ is larger than or equal to z; obtaining one or more modified weight matrices, wherein the modified weight matrices operate on the x′ by y′ by z′ input tensor to generate a u′ by v′ by w′ output tensor, and the u′ by v′ by w′ output tensor comprises a transformed u by v by w output tensor, wherein u′ is smaller than or equal to u, v′ is smaller than or equal to v, and w′ is larger than or equal to w; and processing the x′ by y′ by z′ input tensor using the modified weight matrices to generate the u′ by v′ by w′ output tensor, wherein the u′ by v′ by w′ output tensor comprises the u by v by w output tensor.",2018-04-10,"The title of the patent is superpixel methods for convolutional neural networks and its abstract is methods, systems, and apparatus for efficiently performing a computation of a convolutional neural network layer. one of the methods includes transforming a x by y by z input tensor into a x′ by y′ by z′ input tensor, wherein x′ is smaller than or equal to x, y′ is smaller than or equal to y, and z′ is larger than or equal to z; obtaining one or more modified weight matrices, wherein the modified weight matrices operate on the x′ by y′ by z′ input tensor to generate a u′ by v′ by w′ output tensor, and the u′ by v′ by w′ output tensor comprises a transformed u by v by w output tensor, wherein u′ is smaller than or equal to u, v′ is smaller than or equal to v, and w′ is larger than or equal to w; and processing the x′ by y′ by z′ input tensor using the modified weight matrices to generate the u′ by v′ by w′ output tensor, wherein the u′ by v′ by w′ output tensor comprises the u by v by w output tensor. dated 2018-04-10"
9940574,system and method to control a model state of a neuromorphic model of a brain,model-based neural control uses a model of a portion of a brain and provides feedback control to the model that is based on a received output from the model. a neuromorphic model-based control system includes a neuromorphic model that includes a neuromorphic network to model the brain portion. a synaptic time-multiplexed (stm) neural model-based control system includes an stm neural network to the model the brain portion. the control systems further include a feedback controller to receive an output of the neuromorphic model or stm neural network and to provide a feedback control input to control a model state of the neuromorphic model or the stm neural network.,2018-04-10,The title of the patent is system and method to control a model state of a neuromorphic model of a brain and its abstract is model-based neural control uses a model of a portion of a brain and provides feedback control to the model that is based on a received output from the model. a neuromorphic model-based control system includes a neuromorphic model that includes a neuromorphic network to model the brain portion. a synaptic time-multiplexed (stm) neural model-based control system includes an stm neural network to the model the brain portion. the control systems further include a feedback controller to receive an output of the neuromorphic model or stm neural network and to provide a feedback control input to control a model state of the neuromorphic model or the stm neural network. dated 2018-04-10
9940577,finding semantic parts in images,"embodiments of the present invention relate to finding semantic parts in images. in implementation, a convolutional neural network (cnn) is applied to a set of images to extract features for each image. each feature is defined by a feature vector that enables a subset of the set of images to be clustered in accordance with a similarity between feature vectors. normalized cuts may be utilized to help preserve pose within each cluster. the images in the cluster are aligned and part proposals are generated by sampling various regions in various sizes across the aligned images. to determine which part proposal corresponds to a semantic part, a classifier is trained for each part proposal and semantic part to determine which part proposal best fits the correlation pattern given by the true semantic part. in this way, semantic parts in images can be identified without any previous part annotations.",2018-04-10,"The title of the patent is finding semantic parts in images and its abstract is embodiments of the present invention relate to finding semantic parts in images. in implementation, a convolutional neural network (cnn) is applied to a set of images to extract features for each image. each feature is defined by a feature vector that enables a subset of the set of images to be clustered in accordance with a similarity between feature vectors. normalized cuts may be utilized to help preserve pose within each cluster. the images in the cluster are aligned and part proposals are generated by sampling various regions in various sizes across the aligned images. to determine which part proposal corresponds to a semantic part, a classifier is trained for each part proposal and semantic part to determine which part proposal best fits the correlation pattern given by the true semantic part. in this way, semantic parts in images can be identified without any previous part annotations. dated 2018-04-10"
9940729,detection of invariant features for localization,"a first image and a second image are provided to a trained neural network. the first image comprises one or more static features and the second image comprises at least one of the one or more static features. a static feature is identified in both the first and second images by a branch of the trained neural network. a three dimensional image comprising the identified static feature is generated and three dimensional geometric information/data related to the static feature is extracted and stored in association with a tile of a digital map. a set of training images may be used to train the trained neural network comprises training image subsets comprising two or more images that substantially overlap that were (a) captured at different times; (b) captured under different (i) weather conditions, (ii) lighting conditions, or (iii) weather and lighting conditions; or both a and b.",2018-04-10,"The title of the patent is detection of invariant features for localization and its abstract is a first image and a second image are provided to a trained neural network. the first image comprises one or more static features and the second image comprises at least one of the one or more static features. a static feature is identified in both the first and second images by a branch of the trained neural network. a three dimensional image comprising the identified static feature is generated and three dimensional geometric information/data related to the static feature is extracted and stored in association with a tile of a digital map. a set of training images may be used to train the trained neural network comprises training image subsets comprising two or more images that substantially overlap that were (a) captured at different times; (b) captured under different (i) weather conditions, (ii) lighting conditions, or (iii) weather and lighting conditions; or both a and b. dated 2018-04-10"
9940935,method and device for voiceprint recognition,"a method is performed at a device having one or more processors and memory. the device establishes a first-level deep neural network (dnn) model based on unlabeled speech data, the unlabeled speech data containing no speaker labels and the first-level dnn model specifying a plurality of basic voiceprint features for the unlabeled speech data. the device establishes a second-level dnn model by tuning the first-level dnn model based on labeled speech data, the labeled speech data containing speech samples with respective speaker labels, wherein the second-level dnn model specifies a plurality of high-level voiceprint features. using the second-level dnn model, registers a first high-level voiceprint feature sequence for a user based on a registration speech sample received from the user. the device performs speaker verification for the user based on the first high-level voiceprint feature sequence registered for the user.",2018-04-10,"The title of the patent is method and device for voiceprint recognition and its abstract is a method is performed at a device having one or more processors and memory. the device establishes a first-level deep neural network (dnn) model based on unlabeled speech data, the unlabeled speech data containing no speaker labels and the first-level dnn model specifying a plurality of basic voiceprint features for the unlabeled speech data. the device establishes a second-level dnn model by tuning the first-level dnn model based on labeled speech data, the labeled speech data containing speech samples with respective speaker labels, wherein the second-level dnn model specifies a plurality of high-level voiceprint features. using the second-level dnn model, registers a first high-level voiceprint feature sequence for a user based on a registration speech sample received from the user. the device performs speaker verification for the user based on the first high-level voiceprint feature sequence registered for the user. dated 2018-04-10"
9941900,techniques for general-purpose lossless data compression using a recurrent neural network,techniques for general-purpose lossless data compression using a neural network including compressing an original content item to a baseline lossless compressed data format. the baseline lossless compressed data format is binarized to a binarized format. the binarized format is arithmetically coded based on probability estimates from a neural network probability estimator. the neural network probability estimator generates the probability estimates for current symbols of the binarized format to be arithmetically coded based on symbols of the binarized format that have already been arithmetically coded.,2018-04-10,The title of the patent is techniques for general-purpose lossless data compression using a recurrent neural network and its abstract is techniques for general-purpose lossless data compression using a neural network including compressing an original content item to a baseline lossless compressed data format. the baseline lossless compressed data format is binarized to a binarized format. the binarized format is arithmetically coded based on probability estimates from a neural network probability estimator. the neural network probability estimator generates the probability estimates for current symbols of the binarized format to be arithmetically coded based on symbols of the binarized format that have already been arithmetically coded. dated 2018-04-10
9942085,early warning and recommendation system for the proactive management of wireless broadband networks,"the present disclosure relates to an early warning and recommendation system for proactive management of a wireless broadband network. without human intervention, the system processes highly heterogeneous network and non-network data and applies unsupervised machine learning to the data to predict and understand the situations that lead to different network state conditions. more specifically, unsupervised clustering is applied to the data to understand “situations” that can lead to non-normal network state conditions. a deep neural network model of situations is then created to further understand the underlying data relationships between the elements of a situation and network states. the deep neural network model and reinforcement learning is used to provide recommendations as to changes in network configuration parameters to improve the state of a predicted situation associated with non-normal network conditions. the system displays warnings and recommendations regarding predicted non-normal network conditions in a user interface for a network operator.",2018-04-10,"The title of the patent is early warning and recommendation system for the proactive management of wireless broadband networks and its abstract is the present disclosure relates to an early warning and recommendation system for proactive management of a wireless broadband network. without human intervention, the system processes highly heterogeneous network and non-network data and applies unsupervised machine learning to the data to predict and understand the situations that lead to different network state conditions. more specifically, unsupervised clustering is applied to the data to understand “situations” that can lead to non-normal network state conditions. a deep neural network model of situations is then created to further understand the underlying data relationships between the elements of a situation and network states. the deep neural network model and reinforcement learning is used to provide recommendations as to changes in network configuration parameters to improve the state of a predicted situation associated with non-normal network conditions. the system displays warnings and recommendations regarding predicted non-normal network conditions in a user interface for a network operator. dated 2018-04-10"
9946960,method for acquiring bounding box corresponding to an object in an image by using convolutional neural network including tracking network and computing device using the same,"a method for acquiring a bounding box corresponding to an object is provided. the method includes steps of: (a) acquiring proposal boxes; (b) selecting specific proposal box among the proposal boxes by referring to (i) a result of comparing distance between a reference bounding box and the proposal boxes and/or (ii) a result of comparing score which indicates whether the proposal boxes includes the object, and then setting the specific proposal box as a starting area of a tracking box; (c) determining a specific area of the current frame as a target area of the tracking box by using the mean shift tracking algorithm; and (d) allowing a pooling layer to generate a pooled feature map by applying pooling operation to an area corresponding to the specific area and then allowing a fc layer to acquire a bounding box by applying regression operation to the pooled feature map.",2018-04-17,"The title of the patent is method for acquiring bounding box corresponding to an object in an image by using convolutional neural network including tracking network and computing device using the same and its abstract is a method for acquiring a bounding box corresponding to an object is provided. the method includes steps of: (a) acquiring proposal boxes; (b) selecting specific proposal box among the proposal boxes by referring to (i) a result of comparing distance between a reference bounding box and the proposal boxes and/or (ii) a result of comparing score which indicates whether the proposal boxes includes the object, and then setting the specific proposal box as a starting area of a tracking box; (c) determining a specific area of the current frame as a target area of the tracking box by using the mean shift tracking algorithm; and (d) allowing a pooling layer to generate a pooled feature map by applying pooling operation to an area corresponding to the specific area and then allowing a fc layer to acquire a bounding box by applying regression operation to the pooled feature map. dated 2018-04-17"
9946970,neural networks for encrypted data,"embodiments described herein are directed to methods and systems for performing neural network computations on encrypted data. encrypted data is received from a user. the encrypted data is encrypted with an encryption scheme that allows for computations on the ciphertext to generate encrypted results data. neural network computations are performed on the encrypted data, using approximations of neural network functions to generate encrypted neural network results data from encrypted data. the approximations of neural network functions can approximate activation functions, where the activation functions are approximated using polynomial expressions. the encrypted neural network results data are communicated to the user associated with the encrypted data such that the user decrypts the encrypted data based on the encryption scheme. the functionality of the neural network system can be provided using a cloud computing platform that supports restricted access to particular neural networks.",2018-04-17,"The title of the patent is neural networks for encrypted data and its abstract is embodiments described herein are directed to methods and systems for performing neural network computations on encrypted data. encrypted data is received from a user. the encrypted data is encrypted with an encryption scheme that allows for computations on the ciphertext to generate encrypted results data. neural network computations are performed on the encrypted data, using approximations of neural network functions to generate encrypted neural network results data from encrypted data. the approximations of neural network functions can approximate activation functions, where the activation functions are approximated using polynomial expressions. the encrypted neural network results data are communicated to the user associated with the encrypted data such that the user decrypts the encrypted data based on the encryption scheme. the functionality of the neural network system can be provided using a cloud computing platform that supports restricted access to particular neural networks. dated 2018-04-17"
9947102,image segmentation using neural network method,"the present disclosure relates to systems, methods, devices, and non-transitory computer-readable storage medium for segmenting three-dimensional images. in one implementation, a computer-implemented method for segmenting a three-dimensional image is provided. the method may include receiving a three-dimensional image acquired by an imaging device, and selecting a plurality of stacks of adjacent two-dimensional images from the three-dimensional image. the method may further include segmenting, by a processor, each stack of adjacent two-dimensional images using a neural network model. the method may also include determining, by the processor, a label map for the three-dimensional image by aggregating the segmentation results from the plurality of stacks.",2018-04-17,"The title of the patent is image segmentation using neural network method and its abstract is the present disclosure relates to systems, methods, devices, and non-transitory computer-readable storage medium for segmenting three-dimensional images. in one implementation, a computer-implemented method for segmenting a three-dimensional image is provided. the method may include receiving a three-dimensional image acquired by an imaging device, and selecting a plurality of stacks of adjacent two-dimensional images from the three-dimensional image. the method may further include segmenting, by a processor, each stack of adjacent two-dimensional images using a neural network model. the method may also include determining, by the processor, a label map for the three-dimensional image by aggregating the segmentation results from the plurality of stacks. dated 2018-04-17"
9947314,semi-supervised learning of word embeddings,"software that trains an artificial neural network for generating vector representations for natural language text, by performing the following steps: (i) receiving, by one or more processors, a set of natural language text; (ii) generating, by one or more processors, a set of first metadata for the set of natural language text, where the first metadata is generated using supervised learning method(s); (iii) generating, by one or more processors, a set of second metadata for the set of natural language text, where the second metadata is generated using unsupervised learning method(s); and (iv) training, by one or more processors, an artificial neural network adapted to generate vector representations for natural language text, where the training is based, at least in part, on the received natural language text, the generated set of first metadata, and the generated set of second metadata.",2018-04-17,"The title of the patent is semi-supervised learning of word embeddings and its abstract is software that trains an artificial neural network for generating vector representations for natural language text, by performing the following steps: (i) receiving, by one or more processors, a set of natural language text; (ii) generating, by one or more processors, a set of first metadata for the set of natural language text, where the first metadata is generated using supervised learning method(s); (iii) generating, by one or more processors, a set of second metadata for the set of natural language text, where the second metadata is generated using unsupervised learning method(s); and (iv) training, by one or more processors, an artificial neural network adapted to generate vector representations for natural language text, where the training is based, at least in part, on the received natural language text, the generated set of first metadata, and the generated set of second metadata. dated 2018-04-17"
9948460,multivariate cryptography based on clipped hopfield neural network,"the systems and methods disclosed herein, in one aspect thereof, can encrypt and decrypt messages using a multivariate extended clipped hopfield neural network that uses a diffie-hellman like key exchange algorithm. the proposed cryptosystem comprises three stages that are involved in the communication. a first stage, where parameters are initialized and private keys are generated, a second stage where various base matrix pairs and threshold vectors are synchronized between the sender and the recipient, and a third stage, where encryption/decryption is performed.",2018-04-17,"The title of the patent is multivariate cryptography based on clipped hopfield neural network and its abstract is the systems and methods disclosed herein, in one aspect thereof, can encrypt and decrypt messages using a multivariate extended clipped hopfield neural network that uses a diffie-hellman like key exchange algorithm. the proposed cryptosystem comprises three stages that are involved in the communication. a first stage, where parameters are initialized and private keys are generated, a second stage where various base matrix pairs and threshold vectors are synchronized between the sender and the recipient, and a third stage, where encryption/decryption is performed. dated 2018-04-17"
9948666,forecasting and classifying cyber-attacks using analytical data based neural embeddings,"a first collection including an analytical feature vector and a q&a feature vector is constructed. a second collection is constructed from the first collection by inserting noise in at least one of the vectors. a third collection is constructed by crossing over at least one of vectors of the second collection with a corresponding vector of a fourth collection, migrating at least one of the vectors of the second collection with a corresponding vector of a fifth collection. using a forecasting configuration, an analytical feature vector of the third collection is aged to generate a changed analytical feature vector containing analytical feature values expected at a future time. the changed analytical feature vector is input into a trained neural network to predict a probability of the cyber-attack occurring at the future time.",2018-04-17,"The title of the patent is forecasting and classifying cyber-attacks using analytical data based neural embeddings and its abstract is a first collection including an analytical feature vector and a q&a feature vector is constructed. a second collection is constructed from the first collection by inserting noise in at least one of the vectors. a third collection is constructed by crossing over at least one of vectors of the second collection with a corresponding vector of a fourth collection, migrating at least one of the vectors of the second collection with a corresponding vector of a fifth collection. using a forecasting configuration, an analytical feature vector of the third collection is aged to generate a changed analytical feature vector containing analytical feature values expected at a future time. the changed analytical feature vector is input into a trained neural network to predict a probability of the cyber-attack occurring at the future time. dated 2018-04-17"
9952566,method for controlling and/or regulating a technical system in a computer-assisted manner,"a computer-implemented method for controlling and/or regulating a technical system, in which actions to be carried out on the technical system are first of all determined using an action selection rule which was determined through the learning of a data-driven model and, in particular, a neural network. on the basis of these actions a numerical optimization searches for actions which are better than the original actions according to an optimization criterion. if such actions are found, the technical system is regulated or controlled on the basis of these new actions, such that the corresponding actions are applied to the technical system in succession. the method is suitable, in particular, for regulating or controlling a gas turbine, wherein the actions are preferably optimized with respect to the criterion of low pollutant emission or low combustion chamber humming.",2018-04-24,"The title of the patent is method for controlling and/or regulating a technical system in a computer-assisted manner and its abstract is a computer-implemented method for controlling and/or regulating a technical system, in which actions to be carried out on the technical system are first of all determined using an action selection rule which was determined through the learning of a data-driven model and, in particular, a neural network. on the basis of these actions a numerical optimization searches for actions which are better than the original actions according to an optimization criterion. if such actions are found, the technical system is regulated or controlled on the basis of these new actions, such that the corresponding actions are applied to the technical system in succession. the method is suitable, in particular, for regulating or controlling a gas turbine, wherein the actions are preferably optimized with respect to the criterion of low pollutant emission or low combustion chamber humming. dated 2018-04-24"
9953171,system and method for tokenization of data for privacy,the present invention describes a system and method for tokenization of data. the system includes a receiver configured to receive a request for tokenization. the request for tokenization comprises an input data to be tokenized. the system also includes a parser configured to determine one or more datatype from the input data. the system further includes a trained artificial neural network to generate a token for the input data based on a tokenization technique corresponding to the datatype of the input data.,2018-04-24,The title of the patent is system and method for tokenization of data for privacy and its abstract is the present invention describes a system and method for tokenization of data. the system includes a receiver configured to receive a request for tokenization. the request for tokenization comprises an input data to be tokenized. the system also includes a parser configured to determine one or more datatype from the input data. the system further includes a trained artificial neural network to generate a token for the input data based on a tokenization technique corresponding to the datatype of the input data. dated 2018-04-24
9953217,system and method for pose-aware feature learning,"a pose-aware feature learning system includes an object tracker which tracks an object on a subject in a plurality of video frames, a pose estimator which estimates a pose of the subject in a track of the plurality of video frames, an image pair generator which extracts a plurality of image pairs from the track of the plurality of video frames, and labels the plurality of image pairs with the estimated pose and as depicting the same or different object, and a neural network trainer which trains a neural network based on the labeled plurality of image pairs, to predict whether an image pair depicts the same or different object and a pose difference for the image pair.",2018-04-24,"The title of the patent is system and method for pose-aware feature learning and its abstract is a pose-aware feature learning system includes an object tracker which tracks an object on a subject in a plurality of video frames, a pose estimator which estimates a pose of the subject in a track of the plurality of video frames, an image pair generator which extracts a plurality of image pairs from the track of the plurality of video frames, and labels the plurality of image pairs with the estimated pose and as depicting the same or different object, and a neural network trainer which trains a neural network based on the labeled plurality of image pairs, to predict whether an image pair depicts the same or different object and a pose difference for the image pair. dated 2018-04-24"
9953425,learning image categorization using related attributes,"a first set of attributes (e.g., style) is generated through pre-trained single column neural networks and leveraged to regularize the training process of a regularized double-column convolutional neural network (rdcnn). parameters of the first column (e.g., style) of the rdcnn are fixed during rdcnn training. parameters of the second column (e.g., aesthetics) are fine-tuned while training the rdcnn and the learning process is supervised by the label identified by the second column (e.g., aesthetics). thus, features of the images may be leveraged to boost classification accuracy of other features by learning a rdcnn.",2018-04-24,"The title of the patent is learning image categorization using related attributes and its abstract is a first set of attributes (e.g., style) is generated through pre-trained single column neural networks and leveraged to regularize the training process of a regularized double-column convolutional neural network (rdcnn). parameters of the first column (e.g., style) of the rdcnn are fixed during rdcnn training. parameters of the second column (e.g., aesthetics) are fine-tuned while training the rdcnn and the learning process is supervised by the label identified by the second column (e.g., aesthetics). thus, features of the images may be leveraged to boost classification accuracy of other features by learning a rdcnn. dated 2018-04-24"
9953634,passive training for automatic speech recognition,"provided are methods and systems for passive training for automatic speech recognition. an example method includes utilizing a first, speaker-independent model to detect a spoken keyword or a key phrase in spoken utterances. while utilizing the first model, a second model is passively trained to detect the spoken keyword or the key phrase in the spoken utterances using at least partially the spoken utterances. the second, speaker dependent model may utilize deep neural network (dnn) or convolutional neural network (cnn) techniques. in response to completion of the training, a switch is made from utilizing the first model to utilizing the second model to detect the spoken keyword or the key phrase in spoken utterances. while utilizing the second model, parameters associated therewith are updated using the spoken utterances in response to detecting the keyword or the key phrase in the spoken utterances. user authentication functionality may be provided.",2018-04-24,"The title of the patent is passive training for automatic speech recognition and its abstract is provided are methods and systems for passive training for automatic speech recognition. an example method includes utilizing a first, speaker-independent model to detect a spoken keyword or a key phrase in spoken utterances. while utilizing the first model, a second model is passively trained to detect the spoken keyword or the key phrase in the spoken utterances using at least partially the spoken utterances. the second, speaker dependent model may utilize deep neural network (dnn) or convolutional neural network (cnn) techniques. in response to completion of the training, a switch is made from utilizing the first model to utilizing the second model to detect the spoken keyword or the key phrase in spoken utterances. while utilizing the second model, parameters associated therewith are updated using the spoken utterances in response to detecting the keyword or the key phrase in the spoken utterances. user authentication functionality may be provided. dated 2018-04-24"
9953661,neural network voice activity detection employing running range normalization,a “running range normalization” method includes computing running estimates of the range of values of features useful for voice activity detection (vad) and normalizing the features by mapping them to a desired range. running range normalization includes computation of running estimates of the minimum and maximum values of vad features and normalizing the feature values by mapping the original range to a desired range. smoothing coefficients are optionally selected to directionally bias a rate of change of at least one of the running estimates of the minimum and maximum values. the normalized vad feature parameters are used to train a machine learning algorithm to detect voice activity and to use the trained machine learning algorithm to isolate or enhance the speech component of the audio data.,2018-04-24,The title of the patent is neural network voice activity detection employing running range normalization and its abstract is a “running range normalization” method includes computing running estimates of the range of values of features useful for voice activity detection (vad) and normalizing the features by mapping them to a desired range. running range normalization includes computation of running estimates of the minimum and maximum values of vad features and normalizing the feature values by mapping the original range to a desired range. smoothing coefficients are optionally selected to directionally bias a rate of change of at least one of the running estimates of the minimum and maximum values. the normalized vad feature parameters are used to train a machine learning algorithm to detect voice activity and to use the trained machine learning algorithm to isolate or enhance the speech component of the audio data. dated 2018-04-24
9959498,neural network instruction set architecture,"a computer-implemented method that includes receiving, by a processing unit, an instruction that specifies data values for performing a tensor computation. in response to receiving the instruction, the method may include, performing, by the processing unit, the tensor computation by executing a loop nest comprising a plurality of loops, wherein a structure of the loop nest is defined based on one or more of the data values of the instruction. the tensor computation can be at least a portion of a computation of a neural network layer. the data values specified by the instruction may comprise a value that specifies a type of the neural network layer, and the structure of the loop nest can be defined at least in part by the type of the neural network layer.",2018-05-01,"The title of the patent is neural network instruction set architecture and its abstract is a computer-implemented method that includes receiving, by a processing unit, an instruction that specifies data values for performing a tensor computation. in response to receiving the instruction, the method may include, performing, by the processing unit, the tensor computation by executing a loop nest comprising a plurality of loops, wherein a structure of the loop nest is defined based on one or more of the data values of the instruction. the tensor computation can be at least a portion of a computation of a neural network layer. the data values specified by the instruction may comprise a value that specifies a type of the neural network layer, and the structure of the loop nest can be defined at least in part by the type of the neural network layer. dated 2018-05-01"
9959500,embedded spin transfer torque memory for cellular neural network based processing unit,"an integrated circuit processor having a processing unit that includes a logical circuit with multiple transistors and a top metal landing pad, and an embedded stt memory. the stt memory includes a dielectric layer formed on the top metal landing pad, an adhesion and topography planarization (atp) layer formed on the dielectric layer, and an mtj film layer disposed on the atp layer. the memory may also include bit lines formed on the mtj film layer. the atp layer may have multiple layers such as a top layer and a bottom layer. the top layer may act as an etch stop for etching the mtj film layer on the top. the atp layer may have a total thickness of 500 a to 4000 a. the bit lines can be configured to send data to the logic circuit of the processing unit to perform one or more convolution neural network computations.",2018-05-01,"The title of the patent is embedded spin transfer torque memory for cellular neural network based processing unit and its abstract is an integrated circuit processor having a processing unit that includes a logical circuit with multiple transistors and a top metal landing pad, and an embedded stt memory. the stt memory includes a dielectric layer formed on the top metal landing pad, an adhesion and topography planarization (atp) layer formed on the dielectric layer, and an mtj film layer disposed on the atp layer. the memory may also include bit lines formed on the mtj film layer. the atp layer may have multiple layers such as a top layer and a bottom layer. the top layer may act as an etch stop for etching the mtj film layer on the top. the atp layer may have a total thickness of 500 a to 4000 a. the bit lines can be configured to send data to the logic circuit of the processing unit to perform one or more convolution neural network computations. dated 2018-05-01"
9959517,self-organizing neural network approach to the automatic layout of business process diagrams,"a method, system, and/or computer program product generates self-organizing layouts of process diagrams. initial weight vectors are distributed uniformly within boundaries of regions in the process diagram. a spatial input vector is randomly generated within the boundaries of each region. in each region in the process diagram, a closest graphical node is found, and a position of a winning graphical node that is the closest graphical node to the random input vector is adjusted. positions of all non-immutable graphical objects, wi, in a topographical neighborhood n(k) of a closest graphical node wc that can cross a boundary of one or more regions from the multiple regions are adjusted. the spatial input vector is recursively generated, the closest graphical node is recursively located, and the positions of all non-immutable graphical objects, wi, in the topographical neighborhood n(k) are recursively adjusted until a maximum number of iterations, kmax is reached.",2018-05-01,"The title of the patent is self-organizing neural network approach to the automatic layout of business process diagrams and its abstract is a method, system, and/or computer program product generates self-organizing layouts of process diagrams. initial weight vectors are distributed uniformly within boundaries of regions in the process diagram. a spatial input vector is randomly generated within the boundaries of each region. in each region in the process diagram, a closest graphical node is found, and a position of a winning graphical node that is the closest graphical node to the random input vector is adjusted. positions of all non-immutable graphical objects, wi, in a topographical neighborhood n(k) of a closest graphical node wc that can cross a boundary of one or more regions from the multiple regions are adjusted. the spatial input vector is recursively generated, the closest graphical node is recursively located, and the positions of all non-immutable graphical objects, wi, in the topographical neighborhood n(k) are recursively adjusted until a maximum number of iterations, kmax is reached. dated 2018-05-01"
9959518,self-organizing neural network approach to the automatic layout of business process diagrams,"a method, system, and/or computer program product generates self-organizing layouts of process diagrams. initial weight vectors are distributed uniformly within boundaries of regions in the process diagram. a spatial input vector is randomly generated within the boundaries of each region. in each region in the process diagram, a closest graphical node is found, and a position of a winning graphical node that is the closest graphical node to the random input vector is adjusted. positions of all non-immutable graphical objects, wi, in a topographical neighborhood n(k) of a closest graphical node wc that can cross a boundary of one or more regions from the multiple regions are adjusted. the spatial input vector is recursively generated, the closest graphical node is recursively located, and the positions of all non-immutable graphical objects, wi, in the topographical neighborhood n(k) are recursively adjusted until a maximum number of iterations, kmax is reached.",2018-05-01,"The title of the patent is self-organizing neural network approach to the automatic layout of business process diagrams and its abstract is a method, system, and/or computer program product generates self-organizing layouts of process diagrams. initial weight vectors are distributed uniformly within boundaries of regions in the process diagram. a spatial input vector is randomly generated within the boundaries of each region. in each region in the process diagram, a closest graphical node is found, and a position of a winning graphical node that is the closest graphical node to the random input vector is adjusted. positions of all non-immutable graphical objects, wi, in a topographical neighborhood n(k) of a closest graphical node wc that can cross a boundary of one or more regions from the multiple regions are adjusted. the spatial input vector is recursively generated, the closest graphical node is recursively located, and the positions of all non-immutable graphical objects, wi, in the topographical neighborhood n(k) are recursively adjusted until a maximum number of iterations, kmax is reached. dated 2018-05-01"
9959862,apparatus and method for recognizing speech based on a deep-neural-network (dnn) sound model,"a speech recognition apparatus based on a deep-neural-network (dnn) sound model includes a memory and a processor. as the processor executes a program stored in the memory, the processor generates sound-model state sets corresponding to a plurality of pieces of set training speech data included in multi-set training speech data, generates a multi-set state cluster from the sound-model state sets, and sets the multi-set training speech data as an input node and the multi-set state cluster as output nodes so as to learn a dnn structured parameter.",2018-05-01,"The title of the patent is apparatus and method for recognizing speech based on a deep-neural-network (dnn) sound model and its abstract is a speech recognition apparatus based on a deep-neural-network (dnn) sound model includes a memory and a processor. as the processor executes a program stored in the memory, the processor generates sound-model state sets corresponding to a plurality of pieces of set training speech data included in multi-set training speech data, generates a multi-set state cluster from the sound-model state sets, and sets the multi-set training speech data as an input node and the multi-set state cluster as output nodes so as to learn a dnn structured parameter. dated 2018-05-01"
9962466,muscle tissue regeneration using muscle fiber fragments,"the invention is directed to methods and compositions for obtaining uniform sized muscle fiber fragments for transplantation. these muscle fiber fragments are able to reconstitute into long fibers that are oriented along native muscle. the implanted muscle cells integrate with native vascular and neural network, as confirmed by histology and immunohistochemistry. this invention is particularly advantageous because autologous muscle can be harvested from a donor site, processed and injected into target sites in the operating room. the fragmented muscle fibers can be readily integrated within the host.",2018-05-08,"The title of the patent is muscle tissue regeneration using muscle fiber fragments and its abstract is the invention is directed to methods and compositions for obtaining uniform sized muscle fiber fragments for transplantation. these muscle fiber fragments are able to reconstitute into long fibers that are oriented along native muscle. the implanted muscle cells integrate with native vascular and neural network, as confirmed by histology and immunohistochemistry. this invention is particularly advantageous because autologous muscle can be harvested from a donor site, processed and injected into target sites in the operating room. the fragmented muscle fibers can be readily integrated within the host. dated 2018-05-08"
9965705,systems and methods for attention-based configurable convolutional neural networks (abc-cnn) for visual question answering,"described herein are systems and methods for generating and using attention-based deep learning architectures for visual question answering task (vqa) to automatically generate answers for image-related (still or video images) questions. to generate the correct answers, it is important for a model's attention to focus on the relevant regions of an image according to the question because different questions may ask about the attributes of different image regions. in embodiments, such question-guided attention is learned with a configurable convolutional neural network (abc-cnn). embodiments of the abc-cnn models determine the attention maps by convolving image feature map with the configurable convolutional kernels determined by the questions semantics. in embodiments, the question-guided attention maps focus on the question-related regions and filters out noise in the unrelated regions.",2018-05-08,"The title of the patent is systems and methods for attention-based configurable convolutional neural networks (abc-cnn) for visual question answering and its abstract is described herein are systems and methods for generating and using attention-based deep learning architectures for visual question answering task (vqa) to automatically generate answers for image-related (still or video images) questions. to generate the correct answers, it is important for a model's attention to focus on the relevant regions of an image according to the question because different questions may ask about the attributes of different image regions. in embodiments, such question-guided attention is learned with a configurable convolutional neural network (abc-cnn). embodiments of the abc-cnn models determine the attention maps by convolving image feature map with the configurable convolutional kernels determined by the questions semantics. in embodiments, the question-guided attention maps focus on the question-related regions and filters out noise in the unrelated regions. dated 2018-05-08"
9965719,subcategory-aware convolutional neural networks for object detection,"a computer-implemented method for detecting objects by using subcategory-aware convolutional neural networks (cnns) is presented. the method includes generating object region proposals from an image by a region proposal network (rpn) which utilizes subcategory information, and classifying and refining the object region proposals by an object detection network (odn) that simultaneously performs object category classification, subcategory classification, and bounding box regression. the image is an image pyramid used as input to the rpn and the odn. the rpn and the odn each include a feature extrapolating layer to detect object categories with scale variations among the objects.",2018-05-08,"The title of the patent is subcategory-aware convolutional neural networks for object detection and its abstract is a computer-implemented method for detecting objects by using subcategory-aware convolutional neural networks (cnns) is presented. the method includes generating object region proposals from an image by a region proposal network (rpn) which utilizes subcategory information, and classifying and refining the object region proposals by an object detection network (odn) that simultaneously performs object category classification, subcategory classification, and bounding box regression. the image is an image pyramid used as input to the rpn and the odn. the rpn and the odn each include a feature extrapolating layer to detect object categories with scale variations among the objects. dated 2018-05-08"
9965720,neural network applications in resource constrained environments,systems and methods are disclosed for applying neural networks in resource-constrained environments. a system may include a sensor located in a resource-constrained environment configured to generate sensor data of the resource-constrained environment. the system may also include a first computing device not located in the resource-constrained environment configured to produce a neural network structure based on the sensor data. the system may further include a second computing device located in the resource-constrained environment configured to provide the sensor data as input to the neural network structure. the second computing device may be further configured to determine a state of the resource-constrained environment based on the input of the sensor data to the neural network structure.,2018-05-08,The title of the patent is neural network applications in resource constrained environments and its abstract is systems and methods are disclosed for applying neural networks in resource-constrained environments. a system may include a sensor located in a resource-constrained environment configured to generate sensor data of the resource-constrained environment. the system may also include a first computing device not located in the resource-constrained environment configured to produce a neural network structure based on the sensor data. the system may further include a second computing device located in the resource-constrained environment configured to provide the sensor data as input to the neural network structure. the second computing device may be further configured to determine a state of the resource-constrained environment based on the input of the sensor data to the neural network structure. dated 2018-05-08
9965863,system and methods for image segmentation using convolutional neural network,"the present disclosure relates to systems, methods, devices, and non-transitory computer-readable storage medium for segmenting three-dimensional images. in one implementation, a computer-implemented method for segmenting a three-dimensional image is provided. the method may include receiving the three-dimensional image acquired by an imaging device, and creating a first stack of two-dimensional images from a first plane of the three-dimensional image and a second stack of two-dimensional images from a second plane of the three-dimensional image. the method may further include segmenting, by a processor, the first stack and the second stack of two-dimensional images using at least one neural network model. the method may also include determining, by the processor, a label map for the three-dimensional image by aggregating the segmentation results from the first stack and second stack.",2018-05-08,"The title of the patent is system and methods for image segmentation using convolutional neural network and its abstract is the present disclosure relates to systems, methods, devices, and non-transitory computer-readable storage medium for segmenting three-dimensional images. in one implementation, a computer-implemented method for segmenting a three-dimensional image is provided. the method may include receiving the three-dimensional image acquired by an imaging device, and creating a first stack of two-dimensional images from a first plane of the three-dimensional image and a second stack of two-dimensional images from a second plane of the three-dimensional image. the method may further include segmenting, by a processor, the first stack and the second stack of two-dimensional images using at least one neural network model. the method may also include determining, by the processor, a label map for the three-dimensional image by aggregating the segmentation results from the first stack and second stack. dated 2018-05-08"
9966137,low power analog or multi-level memory for neuromorphic computing,"a neuron circuit for use in a neural network is disclosed. the neural network includes a plurality of field effect transistors having confined channels. the sources and drains of the field effect transistors are connected in series. a plurality of input terminals for receiving a plurality of input voltages may be connected to a drain terminal of a corresponding field effect transistor. the threshold voltages of the field effect transistors can be programmed by increasing or decreasing a number of excess minority carriers in the confined channels, thereby programming the resistance presented by the field effect transistor.",2018-05-08,"The title of the patent is low power analog or multi-level memory for neuromorphic computing and its abstract is a neuron circuit for use in a neural network is disclosed. the neural network includes a plurality of field effect transistors having confined channels. the sources and drains of the field effect transistors are connected in series. a plurality of input terminals for receiving a plurality of input voltages may be connected to a drain terminal of a corresponding field effect transistor. the threshold voltages of the field effect transistors can be programmed by increasing or decreasing a number of excess minority carriers in the confined channels, thereby programming the resistance presented by the field effect transistor. dated 2018-05-08"
9967693,advanced binaural sound imaging,"monaurally-recorded mono or stereo recordings may be processed and converted into binaurally-recorded audio recordings. an analog process of performing this involves output of at least subsets of the monaurally-recorded recording, such as isolated instrument/vocal tracks, to be played to a dummy with two microphones. a digital process of performing this includes simulating audio input from simulated locations corresponding to audio sources. a neural network process of performing this includes training a neural network using speakers and microphones and then automating conversion from monaural audio to binaural audio based on the training of the neural network. the neural network can also be trained with output speakers to eliminate or reduce dead zones and/or speaker crosstalk.",2018-05-08,"The title of the patent is advanced binaural sound imaging and its abstract is monaurally-recorded mono or stereo recordings may be processed and converted into binaurally-recorded audio recordings. an analog process of performing this involves output of at least subsets of the monaurally-recorded recording, such as isolated instrument/vocal tracks, to be played to a dummy with two microphones. a digital process of performing this includes simulating audio input from simulated locations corresponding to audio sources. a neural network process of performing this includes training a neural network using speakers and microphones and then automating conversion from monaural audio to binaural audio based on the training of the neural network. the neural network can also be trained with output speakers to eliminate or reduce dead zones and/or speaker crosstalk. dated 2018-05-08"
9969507,method for performing diagnostics of a structure subject to loads and system for implementing said method,"a method for performing diagnostics of a structure subject to loads, in particular an aircraft structure, is implemented by an arrangement of sensors located at relevant points of the structure and corresponding neural networks. the method includes training the neural network in order to establish an associative relationship between the state of the structure in a subset of relevant points and the state of the structure in at least one residual relevant point. the state of the structure is detected in a plurality of relevant points under operating conditions. the state of the structure is estimated in at least one residual relevant point by the associated neural network on the basis of the pre-established associated relationship. the state of the estimated structure is compared with the detected state at the residual relevant point, such that an intact state of the structure is determined if the expected and detected values of the state parameter match, or a defective state of the structure is determined if these values differ.",2018-05-15,"The title of the patent is method for performing diagnostics of a structure subject to loads and system for implementing said method and its abstract is a method for performing diagnostics of a structure subject to loads, in particular an aircraft structure, is implemented by an arrangement of sensors located at relevant points of the structure and corresponding neural networks. the method includes training the neural network in order to establish an associative relationship between the state of the structure in a subset of relevant points and the state of the structure in at least one residual relevant point. the state of the structure is detected in a plurality of relevant points under operating conditions. the state of the structure is estimated in at least one residual relevant point by the associated neural network on the basis of the pre-established associated relationship. the state of the estimated structure is compared with the detected state at the residual relevant point, such that an intact state of the structure is determined if the expected and detected values of the state parameter match, or a defective state of the structure is determined if these values differ. dated 2018-05-15"
9970266,methods and systems for improved drilling operations using real-time and historical drilling data,"methods and systems are described for improved drilling operations through the use of real-time drilling data to predict bit wear, lithology, pore pressure, a rotating friction coefficient, permeability, and cost in real-time and to adjust drilling parameters in real-time based on the predictions. the real-time lithology prediction is made by processing the real-time drilling data through a multilayer neural network. the real-time bit wear prediction is made by using the real-time drilling data to predict a bit efficiency factor and to detect changes in the bit efficiency factor over time. these predictions may be used to adjust drilling parameters in the drilling operation in real-time, subject to override by the operator. the methods and systems may also include determining various downhole hydraulics parameters and a rotary friction factor. historical data may be used in combination with real-time data to provide expert system assistance and to identify safety concerns.",2018-05-15,"The title of the patent is methods and systems for improved drilling operations using real-time and historical drilling data and its abstract is methods and systems are described for improved drilling operations through the use of real-time drilling data to predict bit wear, lithology, pore pressure, a rotating friction coefficient, permeability, and cost in real-time and to adjust drilling parameters in real-time based on the predictions. the real-time lithology prediction is made by processing the real-time drilling data through a multilayer neural network. the real-time bit wear prediction is made by using the real-time drilling data to predict a bit efficiency factor and to detect changes in the bit efficiency factor over time. these predictions may be used to adjust drilling parameters in the drilling operation in real-time, subject to override by the operator. the methods and systems may also include determining various downhole hydraulics parameters and a rotary friction factor. historical data may be used in combination with real-time data to provide expert system assistance and to identify safety concerns. dated 2018-05-15"
9971940,automatic learning of a video matching system,"provided content is determined to contain an asset represented by reference content by comparing digital fingerprints of the provided content and the reference content. the fingerprints of the reference content and the provided content are generated using a convolutional neural network (cnn). the cnn is trained using a plurality of frame triplets including an anchor frame representing the reference content, a positive frame which is a transformation of the anchor frame, and a negative frame representing content that is not the reference content. the provided content is determined to contain the asset represented by the reference content based on a similarity measure between the generated fingerprints. if the provided content is determined to contain the asset represented by the reference content, a policy associated with the asset is enforced on the provided content.",2018-05-15,"The title of the patent is automatic learning of a video matching system and its abstract is provided content is determined to contain an asset represented by reference content by comparing digital fingerprints of the provided content and the reference content. the fingerprints of the reference content and the provided content are generated using a convolutional neural network (cnn). the cnn is trained using a plurality of frame triplets including an anchor frame representing the reference content, a positive frame which is a transformation of the anchor frame, and a negative frame representing content that is not the reference content. the provided content is determined to contain the asset represented by the reference content based on a similarity measure between the generated fingerprints. if the provided content is determined to contain the asset represented by the reference content, a policy associated with the asset is enforced on the provided content. dated 2018-05-15"
9971953,visual recognition using deep learning attributes,"a processing device for performing visual recognition using deep learning attributes and method for performing the same are described. in one embodiment, a processing device comprises: an interface to receive an input image; and a recognition unit coupled to the interface and operable to perform visual object recognition on the input image, where the recognition unit has an extractor to extract region proposals from the input image, a convolutional neural network (cnn) to compute features for each extracted region proposal, the cnn being operable to create a soft-max layer output, a cross region pooling unit operable to perform pooling of the soft-max layer output to create a set of attributes of the input image, and an image classifier operable to perform image classification based on the attributes of the input image.",2018-05-15,"The title of the patent is visual recognition using deep learning attributes and its abstract is a processing device for performing visual recognition using deep learning attributes and method for performing the same are described. in one embodiment, a processing device comprises: an interface to receive an input image; and a recognition unit coupled to the interface and operable to perform visual object recognition on the input image, where the recognition unit has an extractor to extract region proposals from the input image, a convolutional neural network (cnn) to compute features for each extracted region proposal, the cnn being operable to create a soft-max layer output, a cross region pooling unit operable to perform pooling of the soft-max layer output to create a set of attributes of the input image, and an image classifier operable to perform image classification based on the attributes of the input image. dated 2018-05-15"
9971958,method and system for generating multimodal digital images,"a computer-implemented method generates a multimodal digital image by processing a vector with a first neural network to produce a first modality of the digital image and processing the vector with a second neural network to produce a second modality of the digital image. a structure and a number of layers of the first neural network are identical to a structure and a number of layers of the second neural network. also, at least one layer in the first neural network has parameters identical to parameters of a corresponding layer in the second neural network, and at least one layer in the first neural network has parameters different from parameters of a corresponding layer in the second neural network.",2018-05-15,"The title of the patent is method and system for generating multimodal digital images and its abstract is a computer-implemented method generates a multimodal digital image by processing a vector with a first neural network to produce a first modality of the digital image and processing the vector with a second neural network to produce a second modality of the digital image. a structure and a number of layers of the first neural network are identical to a structure and a number of layers of the second neural network. also, at least one layer in the first neural network has parameters identical to parameters of a corresponding layer in the second neural network, and at least one layer in the first neural network has parameters different from parameters of a corresponding layer in the second neural network. dated 2018-05-15"
9971965,implementing a neural network algorithm on a neurosynaptic substrate based on metadata associated with the neural network algorithm,"one embodiment of the invention provides a system for mapping a neural network onto a neurosynaptic substrate. the system comprises a metadata analysis unit for analyzing metadata information associated with one or more portions of an adjacency matrix representation of the neural network, and a mapping unit for mapping the one or more portions of the matrix representation onto the neurosynaptic substrate based on the metadata information.",2018-05-15,"The title of the patent is implementing a neural network algorithm on a neurosynaptic substrate based on metadata associated with the neural network algorithm and its abstract is one embodiment of the invention provides a system for mapping a neural network onto a neurosynaptic substrate. the system comprises a metadata analysis unit for analyzing metadata information associated with one or more portions of an adjacency matrix representation of the neural network, and a mapping unit for mapping the one or more portions of the matrix representation onto the neurosynaptic substrate based on the metadata information. dated 2018-05-15"
9971966,processing cell images using neural networks,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for processing cell images using neural networks. one of the methods includes obtaining data comprising an input image of one or more biological cells illuminated with an optical microscopy technique; processing the data using a stained cell neural network; and processing the one or more stained cell images using a cell characteristic neural network, wherein the cell characteristic neural network has been configured through training to receive the one or more stained cell images and to process the one or more stained cell images to generate a cell characteristic output that characterizes features of the biological cells that are stained in the one or more stained cell images.",2018-05-15,"The title of the patent is processing cell images using neural networks and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for processing cell images using neural networks. one of the methods includes obtaining data comprising an input image of one or more biological cells illuminated with an optical microscopy technique; processing the data using a stained cell neural network; and processing the one or more stained cell images using a cell characteristic neural network, wherein the cell characteristic neural network has been configured through training to receive the one or more stained cell images and to process the one or more stained cell images to generate a cell characteristic output that characterizes features of the biological cells that are stained in the one or more stained cell images. dated 2018-05-15"
9972092,utilizing deep learning for boundary-aware image segmentation,"systems and methods are disclosed for segmenting a digital image to identify an object portrayed in the digital image from background pixels in the digital image. in particular, in one or more embodiments, the disclosed systems and methods use a first neural network and a second neural network to generate image information used to generate a segmentation mask that corresponds to the object portrayed in the digital image. specifically, in one or more embodiments, the disclosed systems and methods optimize a fit between a mask boundary of the segmentation mask to edges of the object portrayed in the digital image to accurately segment the object within the digital image.",2018-05-15,"The title of the patent is utilizing deep learning for boundary-aware image segmentation and its abstract is systems and methods are disclosed for segmenting a digital image to identify an object portrayed in the digital image from background pixels in the digital image. in particular, in one or more embodiments, the disclosed systems and methods use a first neural network and a second neural network to generate image information used to generate a segmentation mask that corresponds to the object portrayed in the digital image. specifically, in one or more embodiments, the disclosed systems and methods optimize a fit between a mask boundary of the segmentation mask to edges of the object portrayed in the digital image to accurately segment the object within the digital image. dated 2018-05-15"
9972310,system and method for neural network based feature extraction for acoustic model development,"a system and method are presented for neural network based feature extraction for acoustic model development. a neural network may be used to extract acoustic features from raw mfccs or the spectrum, which are then used for training acoustic models for speech recognition systems. feature extraction may be performed by optimizing a cost function used in linear discriminant analysis. general non-linear functions generated by the neural network are used for feature extraction. the transformation may be performed using a cost function from linear discriminant analysis methods which perform linear operations on the mfccs and generate lower dimensional features for speech recognition. the extracted acoustic features may then be used for training acoustic models for speech recognition systems.",2018-05-15,"The title of the patent is system and method for neural network based feature extraction for acoustic model development and its abstract is a system and method are presented for neural network based feature extraction for acoustic model development. a neural network may be used to extract acoustic features from raw mfccs or the spectrum, which are then used for training acoustic models for speech recognition systems. feature extraction may be performed by optimizing a cost function used in linear discriminant analysis. general non-linear functions generated by the neural network are used for feature extraction. the transformation may be performed using a cost function from linear discriminant analysis methods which perform linear operations on the mfccs and generate lower dimensional features for speech recognition. the extracted acoustic features may then be used for training acoustic models for speech recognition systems. dated 2018-05-15"
9972339,neural network based beam selection,"a neural network model, such as a deep neural network (dnn), is trained using many speech examples to perform beam selection in a microphone array-based speech processing system. the dnn is trained using many different speech examples that are labeled with position or direction information relative to a training microphone array. the dnn may then be trained to recognize a direction of incoming speech so that at runtime the trained dnn may process input audio data from a microphone array and may output to a beam selector an indicator of the desired beam that may be selected for further processing. the dnn may be configured to output a beam index and/or coordinates (or other position data) corresponding to an estimated location of the detected speech. the dnn may also be configured to output acoustic unit data corresponding to speech units (for example corresponding to phonemes, senons, etc. such as those of a detected wakeword or other word).",2018-05-15,"The title of the patent is neural network based beam selection and its abstract is a neural network model, such as a deep neural network (dnn), is trained using many speech examples to perform beam selection in a microphone array-based speech processing system. the dnn is trained using many different speech examples that are labeled with position or direction information relative to a training microphone array. the dnn may then be trained to recognize a direction of incoming speech so that at runtime the trained dnn may process input audio data from a microphone array and may output to a beam selector an indicator of the desired beam that may be selected for further processing. the dnn may be configured to output a beam index and/or coordinates (or other position data) corresponding to an estimated location of the detected speech. the dnn may also be configured to output acoustic unit data corresponding to speech units (for example corresponding to phonemes, senons, etc. such as those of a detected wakeword or other word). dated 2018-05-15"
9974454,method and system for machine learning based assessment of fractional flow reserve,"a method and system for determining fractional flow reserve (ffr) for a coronary artery stenosis of a patient is disclosed. in one embodiment, medical image data of the patient including the stenosis is received, a set of features for the stenosis is extracted from the medical image data of the patient, and an ffr value for the stenosis is determined based on the extracted set of features using a trained machine-learning based mapping. in another embodiment, a medical image of the patient including the stenosis of interest is received, image patches corresponding to the stenosis of interest and a coronary tree of the patient are detected, an ffr value for the stenosis of interest is determined using a trained deep neural network regressor applied directly to the detected image patches.",2018-05-22,"The title of the patent is method and system for machine learning based assessment of fractional flow reserve and its abstract is a method and system for determining fractional flow reserve (ffr) for a coronary artery stenosis of a patient is disclosed. in one embodiment, medical image data of the patient including the stenosis is received, a set of features for the stenosis is extracted from the medical image data of the patient, and an ffr value for the stenosis is determined based on the extracted set of features using a trained machine-learning based mapping. in another embodiment, a medical image of the patient including the stenosis of interest is received, image patches corresponding to the stenosis of interest and a coronary tree of the patient are detected, an ffr value for the stenosis of interest is determined using a trained deep neural network regressor applied directly to the detected image patches. dated 2018-05-22"
9974474,system and method for analyzing progress of labor and preterm labor,"systems and methods for monitoring uterus contraction activity and progress of labor. the system of the subject invention can comprises (1) a plurality of sensors; (2) an amplifying/filtering means; (3) a computing means; and (4) a graphical user interface. accurate clinical data, which can be extracted and provided to the user in real-time using the system of the invention, include without limitation, progress of labor, prediction and monitoring of preterm labor, and intrauterine pressure prediction. in a preferred embodiment, the system of the invention includes an intelligence means, such as a neural network system, to analyze and interpret clinical data for use in clinical diagnosis as well as delivery strategy.",2018-05-22,"The title of the patent is system and method for analyzing progress of labor and preterm labor and its abstract is systems and methods for monitoring uterus contraction activity and progress of labor. the system of the subject invention can comprises (1) a plurality of sensors; (2) an amplifying/filtering means; (3) a computing means; and (4) a graphical user interface. accurate clinical data, which can be extracted and provided to the user in real-time using the system of the invention, include without limitation, progress of labor, prediction and monitoring of preterm labor, and intrauterine pressure prediction. in a preferred embodiment, the system of the invention includes an intelligence means, such as a neural network system, to analyze and interpret clinical data for use in clinical diagnosis as well as delivery strategy. dated 2018-05-22"
9977729,testing applications with a defined input format,"a system and method are provided for testing the performance of applications. by way of example only, the method may include training a neural network with documents containing text elements that are arranged in accordance with a defined format and using the neural network to determine the predictability of the value of individual text elements within a test document. when the neural network indicates that the value of a text element is unlikely, the value may be modified and the modified document may be used to test an application that processes documents in accordance with the defined format.",2018-05-22,"The title of the patent is testing applications with a defined input format and its abstract is a system and method are provided for testing the performance of applications. by way of example only, the method may include training a neural network with documents containing text elements that are arranged in accordance with a defined format and using the neural network to determine the predictability of the value of individual text elements within a test document. when the neural network indicates that the value of a text element is unlikely, the value may be modified and the modified document may be used to test an application that processes documents in accordance with the defined format. dated 2018-05-22"
9977997,training method and apparatus for convolutional neural network model,"disclosed are a training method and apparatus for a cnn model, which belong to the field of image recognition. the method comprises: performing a convolution operation, maximal pooling operation and horizontal pooling operation on training images, respectively, to obtain second feature images; determining feature vectors according to the second feature images; processing the feature vectors to obtain category probability vectors; according to the category probability vectors and an initial category, calculating a category error; based on the category error, adjusting model parameters; based on the adjusted model parameters, continuing the model parameters adjusting process, and using the model parameters when the number of iteration times reaches a pre-set number of times as the model parameters for the well-trained cnn model. after the convolution operation and maximal pooling operation on the training images on each level of convolution layer, a horizontal pooling operation is performed. since the horizontal pooling operation can extract feature images identifying image horizontal direction features from the feature images, such that the well-trained cnn model can recognize an image of any size, thus expanding the applicable range of the well-trained cnn model in image recognition.",2018-05-22,"The title of the patent is training method and apparatus for convolutional neural network model and its abstract is disclosed are a training method and apparatus for a cnn model, which belong to the field of image recognition. the method comprises: performing a convolution operation, maximal pooling operation and horizontal pooling operation on training images, respectively, to obtain second feature images; determining feature vectors according to the second feature images; processing the feature vectors to obtain category probability vectors; according to the category probability vectors and an initial category, calculating a category error; based on the category error, adjusting model parameters; based on the adjusted model parameters, continuing the model parameters adjusting process, and using the model parameters when the number of iteration times reaches a pre-set number of times as the model parameters for the well-trained cnn model. after the convolution operation and maximal pooling operation on the training images on each level of convolution layer, a horizontal pooling operation is performed. since the horizontal pooling operation can extract feature images identifying image horizontal direction features from the feature images, such that the well-trained cnn model can recognize an image of any size, thus expanding the applicable range of the well-trained cnn model in image recognition. dated 2018-05-22"
9978003,utilizing deep learning for automatic digital image segmentation and stylization,"systems and methods are disclosed for segregating target individuals represented in a probe digital image from background pixels in the probe digital image. in particular, in one or more embodiments, the disclosed systems and methods train a neural network based on two or more of training position channels, training shape input channels, training color channels, or training object data. moreover, in one or more embodiments, the disclosed systems and methods utilize the trained neural network to select a target individual in a probe digital image. specifically, in one or more embodiments, the disclosed systems and methods generate position channels, training shape input channels, and color channels corresponding the probe digital image, and utilize the generated channels in conjunction with the trained neural network to select the target individual.",2018-05-22,"The title of the patent is utilizing deep learning for automatic digital image segmentation and stylization and its abstract is systems and methods are disclosed for segregating target individuals represented in a probe digital image from background pixels in the probe digital image. in particular, in one or more embodiments, the disclosed systems and methods train a neural network based on two or more of training position channels, training shape input channels, training color channels, or training object data. moreover, in one or more embodiments, the disclosed systems and methods utilize the trained neural network to select a target individual in a probe digital image. specifically, in one or more embodiments, the disclosed systems and methods generate position channels, training shape input channels, and color channels corresponding the probe digital image, and utilize the generated channels in conjunction with the trained neural network to select the target individual. dated 2018-05-22"
9978013,systems and methods for recognizing objects in radar imagery,"the present invention is directed to systems and methods for detecting objects in a radar image stream. embodiments of the invention can receive a data stream from radar sensors and use a deep neural network to convert the received data stream into a set of semantic labels, where each semantic label corresponds to an object in the radar data stream that the deep neural network has identified. processing units running the deep neural network may be collocated onboard an airborne vehicle along with the radar sensor(s). the processing units can be configured with powerful, high-speed graphics processing units or field-programmable gate arrays that are low in size, weight, and power requirements. embodiments of the invention are also directed to providing innovative advances to object recognition training systems that utilize a detector and an object recognition cascade to analyze radar image streams in real time. the object recognition cascade can comprise at least one recognizer that receives a non-background stream of image patches from a detector and automatically assigns one or more semantic labels to each non-background image patch. in some embodiments, a separate recognizer for the background analysis of patches may also be incorporated. there may be multiple detectors and multiple recognizers, depending on the design of the cascade. embodiments of the invention also include novel methods to tailor deep neural network algorithms to successfully process radar imagery, utilizing techniques such as normalization, sampling, data augmentation, foveation, cascade architectures, and label harmonization.",2018-05-22,"The title of the patent is systems and methods for recognizing objects in radar imagery and its abstract is the present invention is directed to systems and methods for detecting objects in a radar image stream. embodiments of the invention can receive a data stream from radar sensors and use a deep neural network to convert the received data stream into a set of semantic labels, where each semantic label corresponds to an object in the radar data stream that the deep neural network has identified. processing units running the deep neural network may be collocated onboard an airborne vehicle along with the radar sensor(s). the processing units can be configured with powerful, high-speed graphics processing units or field-programmable gate arrays that are low in size, weight, and power requirements. embodiments of the invention are also directed to providing innovative advances to object recognition training systems that utilize a detector and an object recognition cascade to analyze radar image streams in real time. the object recognition cascade can comprise at least one recognizer that receives a non-background stream of image patches from a detector and automatically assigns one or more semantic labels to each non-background image patch. in some embodiments, a separate recognizer for the background analysis of patches may also be incorporated. there may be multiple detectors and multiple recognizers, depending on the design of the cascade. embodiments of the invention also include novel methods to tailor deep neural network algorithms to successfully process radar imagery, utilizing techniques such as normalization, sampling, data augmentation, foveation, cascade architectures, and label harmonization. dated 2018-05-22"
9978374,neural networks for speaker verification,"this document generally describes systems, methods, devices, and other techniques related to speaker verification, including (i) training a neural network for a speaker verification model, (ii) enrolling users at a client device, and (iii) verifying identities of users based on characteristics of the users' voices. some implementations include a computer-implemented method. the method can include receiving, at a computing device, data that characterizes an utterance of a user of the computing device. a speaker representation can be generated, at the computing device, for the utterance using a neural network on the computing device. the neural network can be trained based on a plurality of training samples that each: (i) include data that characterizes a first utterance and data that characterizes one or more second utterances, and (ii) are labeled as a matching speakers sample or a non-matching speakers sample.",2018-05-22,"The title of the patent is neural networks for speaker verification and its abstract is this document generally describes systems, methods, devices, and other techniques related to speaker verification, including (i) training a neural network for a speaker verification model, (ii) enrolling users at a client device, and (iii) verifying identities of users based on characteristics of the users' voices. some implementations include a computer-implemented method. the method can include receiving, at a computing device, data that characterizes an utterance of a user of the computing device. a speaker representation can be generated, at the computing device, for the utterance using a neural network on the computing device. the neural network can be trained based on a plurality of training samples that each: (i) include data that characterizes a first utterance and data that characterizes one or more second utterances, and (ii) are labeled as a matching speakers sample or a non-matching speakers sample. dated 2018-05-22"
9978388,systems and methods for restoration of speech components,"a method for restoring distorted speech components of an audio signal distorted by a noise reduction or a noise cancellation includes determining distorted frequency regions and undistorted frequency regions in the audio signal. the distorted frequency regions include regions of the audio signal in which a speech distortion is present. iterations are performed using a model to refine predictions of the audio signal at distorted frequency regions. the model is configured to modify the audio signal and may include deep neural network trained using spectral envelopes of clean or undamaged audio signals. before each iteration, the audio signal at the undistorted frequency regions is restored to values of the audio signal prior to the first iteration; while the audio signal at distorted frequency regions is refined starting from zero at the first iteration. iterations are ended when discrepancies of audio signal at undistorted frequency regions meet pre-defined criteria.",2018-05-22,"The title of the patent is systems and methods for restoration of speech components and its abstract is a method for restoring distorted speech components of an audio signal distorted by a noise reduction or a noise cancellation includes determining distorted frequency regions and undistorted frequency regions in the audio signal. the distorted frequency regions include regions of the audio signal in which a speech distortion is present. iterations are performed using a model to refine predictions of the audio signal at distorted frequency regions. the model is configured to modify the audio signal and may include deep neural network trained using spectral envelopes of clean or undamaged audio signals. before each iteration, the audio signal at the undistorted frequency regions is restored to values of the audio signal prior to the first iteration; while the audio signal at distorted frequency regions is refined starting from zero at the first iteration. iterations are ended when discrepancies of audio signal at undistorted frequency regions meet pre-defined criteria. dated 2018-05-22"
9980100,device location based on machine learning classifications,"a venue system of a client device can submit a location request to a server, which returns multiple venues that are near the client device. the client device can use one or more machine learning schemes (e.g., convolutional neural networks) to determine that the client device is located in one of specific venues of the possible venues. the venue system can further select imagery for presentation based on the venue selection. the presentation may be published as ephemeral message on a network platform.",2018-05-22,"The title of the patent is device location based on machine learning classifications and its abstract is a venue system of a client device can submit a location request to a server, which returns multiple venues that are near the client device. the client device can use one or more machine learning schemes (e.g., convolutional neural networks) to determine that the client device is located in one of specific venues of the possible venues. the venue system can further select imagery for presentation based on the venue selection. the presentation may be published as ephemeral message on a network platform. dated 2018-05-22"
9984062,generating author vectors,"methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating author vectors. one of the methods includes obtaining a set of sequences of words, the set of sequences of words comprising a plurality of first sequences of words and, for each first sequence of words, a respective second sequence of words that follows the first sequence of words, wherein each first sequence of words and each second sequence of words has been classified as being authored by a first author; and training a neural network system on the first sequences and the second sequences to determine an author vector for the first author, wherein the author vector characterizes the first author.",2018-05-29,"The title of the patent is generating author vectors and its abstract is methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating author vectors. one of the methods includes obtaining a set of sequences of words, the set of sequences of words comprising a plurality of first sequences of words and, for each first sequence of words, a respective second sequence of words that follows the first sequence of words, wherein each first sequence of words and each second sequence of words has been classified as being authored by a first author; and training a neural network system on the first sequences and the second sequences to determine an author vector for the first author, wherein the author vector characterizes the first author. dated 2018-05-29"
9984312,"image registration device, image registration method, and ultrasonic diagnosis apparatus having image registration device","there is provided an image registration device and an image registration method. the device includes: a feature extractor configured to extract, from a first image, a first feature group and to extract, from a second image, a second feature group; a feature converter configured to convert, using a converted neural network in which a correlation between features is learned, the extracted second feature group to correspond to the extracted first feature group, to obtain a converted group; and a register configured to register the first image and the second image based on the converted group and the extracted first feature group.",2018-05-29,"The title of the patent is image registration device, image registration method, and ultrasonic diagnosis apparatus having image registration device and its abstract is there is provided an image registration device and an image registration method. the device includes: a feature extractor configured to extract, from a first image, a first feature group and to extract, from a second image, a second feature group; a feature converter configured to convert, using a converted neural network in which a correlation between features is learned, the extracted second feature group to correspond to the extracted first feature group, to obtain a converted group; and a register configured to register the first image and the second image based on the converted group and the extracted first feature group. dated 2018-05-29"
9984326,spiking neural network simulator for image and video processing,"described is system for simulating spiking neural networks for image and video processing. the system processes an image with a spiking neural network simulator having a plurality of inter-connected modules. each module comprises a plurality of neuron elements. processing the image further comprises performing a neuron state update for each module, that includes aggregating input spikes and updating neuron membrane potentials, and performing spike propagation for each module, which includes transferring spikes generated in a current time step. finally, an analysis result is output.",2018-05-29,"The title of the patent is spiking neural network simulator for image and video processing and its abstract is described is system for simulating spiking neural networks for image and video processing. the system processes an image with a spiking neural network simulator having a plurality of inter-connected modules. each module comprises a plurality of neuron elements. processing the image further comprises performing a neuron state update for each module, that includes aggregating input spikes and updating neuron membrane potentials, and performing spike propagation for each module, which includes transferring spikes generated in a current time step. finally, an analysis result is output. dated 2018-05-29"
9984330,predictive trending of digital entities,"surfacing relevant and predictively trending digital entities to a user in a content feed is provided. aspects of a predictive trending system use one or more predictive models, such as neural networks or regression models, to generate predictive trending scores of digital entities (e.g., documents, people, electronic communications, meetings, locations, digital images, digital videos, digital audio, etc.) based on historical scores and context. by taking into account trends and context, the predictive trending system calculates future trending scores of digital entities, and determines which digital entities are both relevant to a given user and likely to be trending around the user and the people in the user's network in the future. the predictive trending system curates the digital entities determined to be relevant and predicted to be trending around the user, and presents the digital entities in a content feed.",2018-05-29,"The title of the patent is predictive trending of digital entities and its abstract is surfacing relevant and predictively trending digital entities to a user in a content feed is provided. aspects of a predictive trending system use one or more predictive models, such as neural networks or regression models, to generate predictive trending scores of digital entities (e.g., documents, people, electronic communications, meetings, locations, digital images, digital videos, digital audio, etc.) based on historical scores and context. by taking into account trends and context, the predictive trending system calculates future trending scores of digital entities, and determines which digital entities are both relevant to a given user and likely to be trending around the user and the people in the user's network in the future. the predictive trending system curates the digital entities determined to be relevant and predicted to be trending around the user, and presents the digital entities in a content feed. dated 2018-05-29"
9984682,computer-implemented systems and methods for automatically generating an assessment of oral recitations of assessment items,"provide automatic assessment of oral recitations during computer based language assessments using a trained neural network to automate the scoring and feedback processes without human transcription and scoring input by automatically generating a score of a language assessment. providing an automatic speech recognition (“asr”) scoring system. training multiple scoring reference vectors associated with multiple possible scores of an assessment, and receiving an acoustic language assessment response to an assessment item. based on the acoustic language assessment automatically generating a transcription, and generating an individual word vector from the transcription. generating an input vector by concatenating an individual word vector with a transcription feature vector, and supplying an input vector as input to a neural network. generating an output vector based on weights of a neural network; and generating a score by comparing an output vector with scoring vectors.",2018-05-29,"The title of the patent is computer-implemented systems and methods for automatically generating an assessment of oral recitations of assessment items and its abstract is provide automatic assessment of oral recitations during computer based language assessments using a trained neural network to automate the scoring and feedback processes without human transcription and scoring input by automatically generating a score of a language assessment. providing an automatic speech recognition (“asr”) scoring system. training multiple scoring reference vectors associated with multiple possible scores of an assessment, and receiving an acoustic language assessment response to an assessment item. based on the acoustic language assessment automatically generating a transcription, and generating an individual word vector from the transcription. generating an input vector by concatenating an individual word vector with a transcription feature vector, and supplying an input vector as input to a neural network. generating an output vector based on weights of a neural network; and generating a score by comparing an output vector with scoring vectors. dated 2018-05-29"
9984683,automatic speech recognition using multi-dimensional models,"methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for automatic speech recognition using multi-dimensional models. in some implementations, audio data that describes an utterance is received. a transcription for the utterance is determined using an acoustic model that includes a neural network having first memory blocks for time information and second memory blocks for frequency information. the transcription for the utterance is provided as output of an automated speech recognizer.",2018-05-29,"The title of the patent is automatic speech recognition using multi-dimensional models and its abstract is methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for automatic speech recognition using multi-dimensional models. in some implementations, audio data that describes an utterance is received. a transcription for the utterance is determined using an acoustic model that includes a neural network having first memory blocks for time information and second memory blocks for frequency information. the transcription for the utterance is provided as output of an automated speech recognizer. dated 2018-05-29"
9984772,image analytics question answering,a computer-implemented method for predicting answers to questions concerning medical image analytics reports includes splitting a medical image analytics report into a plurality of sentences and generating a plurality of sentence embedding vectors by applying a natural language processing framework to the plurality of sentences. a question related to subject matter included in the medical image analytics report is received and a question embedding vector is generated by applying the natural language processing framework to the question. a subset of the sentence embedding vectors most similar to the question embedding vector is identified by applying a similarity matching process to the sentence embedding vectors and the question embedding vector. a trained recurrent neural network (rnn) is used to determine a predicted answer to the question based on the subset of the sentence embedding vectors.,2018-05-29,The title of the patent is image analytics question answering and its abstract is a computer-implemented method for predicting answers to questions concerning medical image analytics reports includes splitting a medical image analytics report into a plurality of sentences and generating a plurality of sentence embedding vectors by applying a natural language processing framework to the plurality of sentences. a question related to subject matter included in the medical image analytics report is received and a question embedding vector is generated by applying the natural language processing framework to the question. a subset of the sentence embedding vectors most similar to the question embedding vector is identified by applying a similarity matching process to the sentence embedding vectors and the question embedding vector. a trained recurrent neural network (rnn) is used to determine a predicted answer to the question based on the subset of the sentence embedding vectors. dated 2018-05-29
H0014150,signal processor/analyzer with a neural network coupled to an acoustic charge transport (act) device (act),"the unique neural network signal processor/analyzer (unspa), for real-time analysis of analog signals, involves combining acoustic charge transport (act) device(s) and an artificial neural network processor on a single, monolithic substrate. the act will act as input to the neural network. the unspa will allow non-destructive, high-speed, real-time signal analysis with improved performance and decreased size over conventional methods. the unspa can function as, but is not limited to, signal classification, prediction or error detection. applications for such functions could be: target or speech recognition, signal prediction, sensor fusion, adaptive control, image classification, in-line error detection.",1995-02-07,"The title of the patent is signal processor/analyzer with a neural network coupled to an acoustic charge transport (act) device (act) and its abstract is the unique neural network signal processor/analyzer (unspa), for real-time analysis of analog signals, involves combining acoustic charge transport (act) device(s) and an artificial neural network processor on a single, monolithic substrate. the act will act as input to the neural network. the unspa will allow non-destructive, high-speed, real-time signal analysis with improved performance and decreased size over conventional methods. the unspa can function as, but is not limited to, signal classification, prediction or error detection. applications for such functions could be: target or speech recognition, signal prediction, sensor fusion, adaptive control, image classification, in-line error detection. dated 1995-02-07"
H002215,odor discrimination using binary spiking neural network,an odor discrimination method and device for an electronic nose system including olfactory pattern classification based on a binary spiking neural network with the capability to handle many sensor inputs in a noise environment while recognizing a large number of potential odors. the spiking neural networks process a large number of inputs arriving from a chemical sensor array and implemented with efficient use of chip surface area.,2008-04-01,The title of the patent is odor discrimination using binary spiking neural network and its abstract is an odor discrimination method and device for an electronic nose system including olfactory pattern classification based on a binary spiking neural network with the capability to handle many sensor inputs in a noise environment while recognizing a large number of potential odors. the spiking neural networks process a large number of inputs arriving from a chemical sensor array and implemented with efficient use of chip surface area. dated 2008-04-01
RE36450,method and apparatus for automatically determining somatic state,"an apparatus and a method for automatically determining the present somatic state of a human subject. the characteristic values of the subject (e.g., scalp potential, muscle potential, heart-rate, eye-movement and frequency of eye blinks, or any combination thereof) are detected and output signals corresponding to the detected characteristic values are produced, amplified and digitized. the fourier transformation is performed on the output signals. a set of state variables for each selected frequency sub-band of a selected frequency band for each of the output signals is determined. sets of reference weights and sets of reference biases for a neural network from sets of state reference variables corresponding to known somatic states are formed. each of the sets of state variables, the sets of reference weights and the sets of reference biases are applied to the neural network to determine present somatic state of the subject. the present somatic state of the subject is displayed.",1999-12-21,"The title of the patent is method and apparatus for automatically determining somatic state and its abstract is an apparatus and a method for automatically determining the present somatic state of a human subject. the characteristic values of the subject (e.g., scalp potential, muscle potential, heart-rate, eye-movement and frequency of eye blinks, or any combination thereof) are detected and output signals corresponding to the detected characteristic values are produced, amplified and digitized. the fourier transformation is performed on the output signals. a set of state variables for each selected frequency sub-band of a selected frequency band for each of the output signals is determined. sets of reference weights and sets of reference biases for a neural network from sets of state reference variables corresponding to known somatic states are formed. each of the sets of state variables, the sets of reference weights and the sets of reference biases are applied to the neural network to determine present somatic state of the subject. the present somatic state of the subject is displayed. dated 1999-12-21"
RE36823,inference rule determining method and inference device,""" an inference rule determining process according to the present invention sequentially determines, using a learning function of a neural network model, a membership function representing a degree which the conditions of the if part of each inference rule is satisfied when input data is received to thereby obtain an optimal inference result without using experience rules. the inventive inference device uses an inference rule of the type """"if . . . then . . ."""" and includes a membership value determiner (1) which includes all of if part and has a neural network; individual inference quantity determiners (21)-(2r) which correspond to the respective then parts of the inference rules and determine the corresponding inference quantities for the inference rules; and a final inference quantity determiner which determines these inference quantities synthetically to obtain the final results of the inference. if the individual inference quantity determiners (2) each has a neural network structure, the non-linearity of the neural network models is used to obtain the result of the inference with high inference accuracy even if in object to be inferred is non-linear. """,2000-08-15,"The title of the patent is inference rule determining method and inference device and its abstract is "" an inference rule determining process according to the present invention sequentially determines, using a learning function of a neural network model, a membership function representing a degree which the conditions of the if part of each inference rule is satisfied when input data is received to thereby obtain an optimal inference result without using experience rules. the inventive inference device uses an inference rule of the type """"if . . . then . . ."""" and includes a membership value determiner (1) which includes all of if part and has a neural network; individual inference quantity determiners (21)-(2r) which correspond to the respective then parts of the inference rules and determine the corresponding inference quantities for the inference rules; and a final inference quantity determiner which determines these inference quantities synthetically to obtain the final results of the inference. if the individual inference quantity determiners (2) each has a neural network structure, the non-linearity of the neural network models is used to obtain the result of the inference with high inference accuracy even if in object to be inferred is non-linear. "" dated 2000-08-15"
RE41658,"low-voltage, very-low-power conductance mode neuron","a neural network including a number of synaptic weighting elements, and a neuron stage; each of the synaptic weighting elements having a respective synaptic input connection supplied with a respective input signal; and the neuron stage having inputs connected to the synaptic weighting elements, and being connected to an output of the neural network supplying a digital output signal. the accumulated weighted inputs are represented as conductances, and a conductance-mode neuron is used to apply nonlinearity and produce an output. the synaptic weighting elements are formed by memory cells programmable to different threshold voltage levels, so that each presents a respective programmable conductance; and the neuron stage provides for measuring conductance on the basis of the current through the memory cells, and for generating a binary output signal on the basis of the total conductance of the synaptic elements.",2010-09-07,"The title of the patent is low-voltage, very-low-power conductance mode neuron and its abstract is a neural network including a number of synaptic weighting elements, and a neuron stage; each of the synaptic weighting elements having a respective synaptic input connection supplied with a respective input signal; and the neuron stage having inputs connected to the synaptic weighting elements, and being connected to an output of the neural network supplying a digital output signal. the accumulated weighted inputs are represented as conductances, and a conductance-mode neuron is used to apply nonlinearity and produce an output. the synaptic weighting elements are formed by memory cells programmable to different threshold voltage levels, so that each presents a respective programmable conductance; and the neuron stage provides for measuring conductance on the basis of the current through the memory cells, and for generating a binary output signal on the basis of the total conductance of the synaptic elements. dated 2010-09-07"