US20150379397A1 - Secure voice signature communications system - Google Patents
Secure voice signature communications system Download PDFInfo
- Publication number
- US20150379397A1 US20150379397A1 US14/753,296 US201514753296A US2015379397A1 US 20150379397 A1 US20150379397 A1 US 20150379397A1 US 201514753296 A US201514753296 A US 201514753296A US 2015379397 A1 US2015379397 A1 US 2015379397A1
- Authority
- US
- United States
- Prior art keywords
- intelligent device
- artificial intelligent
- spatio
- temporal
- neural network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Abandoned
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
Definitions
- the present invention generally relates to communications system over a communication network.
- the present invention relates to a system and method to couple two or more parts of a spatio-temporal spiking neural network, such as to provide acoustic signature identification or other cognitive tasks and the related communications between two or more components of a spatio-temporal spiking neural network.
- Neural networks have been around since the mid-1940's. Since then, many different neural models have been conceived with the aim to reproduce the considerable cognitive powers of the brain. These Artificial Neural Networks vary from simple sigmoid networks (perceptron) to complex biologically-accurate Spatio-temporal Spiking Neural Networks. It becomes increasingly difficult to determine the distinctions between these networks. For the purpose of clarity, Neural Networks can be divided into four distinct groups;
- Computer software neural networks generally exist of sigmoid function neurons wherein activated values are added and a non-linear mathematical function, such as a logistics function, is applied to the sum.
- Computer software simulations of biologically accurate neuron models focus on building accurate mathematical models of biological networks.
- Analog hardware networks a model is constructed in VLSI, consisting of transistors that behave more or less like synapses and an Integrate and Fire neural network.
- Digital hardware emulation the same is accomplished using logic gates.
- the above mentioned technologies lack in efficiently capturing the dynamics of a biological neuronal networks, including spike-timing dependent plasticity, the effects of neurotransmitters and neuromodulators and diverse spike timing representations. Further, the present technologies do not provide computational retrieving and storing learned tasks, so that they may be able to be re-learned by another dynamic neural network. Additionally, the technologies lack in fast speeds and stability.
- the present technologies do not describe communications between two or more components of artificial neural network, that may be implemented in various applications, such as communication to a remote neural network irrespective of the location, controlling and analyzing a connected appliance or equipment.
- the present invention relates to a system and a method for providing communication between two or more components of a spatio-temporal spiking neural network, such as to provide acoustic signature identification or other cognitive tasks and the related communications between two or more components of a spatio-temporal spiking neural network.
- An embodiment of the present invention provides an apparatus for identifying and learning acoustic signature of a plurality of auditory signals comprising: an input sensor configured to capture a varying potential produced in the plurality of auditory signals from an auditory source; a series of resonators configured to convert the varying potential into a plurality of streams of electrical pulses with spatio-temporal distribution; an artificial intelligent device identifies one or more features of the pulse streams of acoustic signals, representing the acoustic signature, by association in a dynamic spatio-temporal neural network; and the artificial intelligent device learns to respond to the acoustic signature of the acoustic signals by modifying synaptic strengths in the dynamic spatio-temporal neural network.
- Another embodiment of the present invention provides a communication system comprising: a first artificial intelligent device consisting of a dynamic spatio-temporal neural network configured to receive and learn from a plurality of pulse streams of acoustic signals with spatio-temporal distribution; and at least one remote artificial intelligent device consisting of a dynamic spatio-temporal neural network communicating with the first artificial intelligent device, through a communication channel, configured to receive and respond to the pulse streams.
- a first artificial intelligent device consisting of a dynamic spatio-temporal neural network configured to receive and learn from a plurality of pulse streams of acoustic signals with spatio-temporal distribution
- at least one remote artificial intelligent device consisting of a dynamic spatio-temporal neural network communicating with the first artificial intelligent device, through a communication channel, configured to receive and respond to the pulse streams.
- Yet another embodiment of the invention provides a method for identifying and learning acoustic signature of a plurality of auditory signals comprising: capturing, by an input sensor, a varying potential produced in the plurality of auditory signals from an auditory source; converting the varying potential into a plurality of streams of electrical pulses with spatio-temporal distribution, by a series of resonators; identifying, by an artificial intelligent device, one or more features of the pulse streams of acoustic signals, representing the acoustic signature, through association in a dynamic spatio-temporal neural network; and learning, by the artificial intelligent device, to respond to the acoustic signature of the acoustic signals by modifying synaptic strengths in the dynamic spatio-temporal neural network.
- An embodiment of the present invention discloses communication to a remote spatio-temporal spiking neural network for receiving spatio-temporal pulse streams via a parallel or serial communication such as the Internet.
- Another embodiment of the present invention discloses a communication system in order to analyze and control an appliance irrespective of the location.
- Yet another embodiment of the present invention provides a secure communication system because each spatio-temporal neural network is trained to respond to a unique voice signature.
- FIG. 1 illustrates a system for communicating two or more components of a dynamic spatio-temporal neural network, in accordance with an embodiment of the present invention.
- FIG. 2 illustrates a method of learning and identifying acoustic signature of auditory signals using a dynamic spatio-temporal neural network, in accordance with an embodiment of the present invention.
- FIG. 3 illustrates a method of communicating two or more components of a dynamic spatio-temporal neural network, in accordance with an embodiment of the present invention.
- the Dynamic Spatio-Temporal Neural Network model described here captures the dynamics of biological neuronal networks, including spike-timing dependent plasticity, the effects of neurotransmitters and neuromodulators and diverse spike timing representations, allowing the model to perform functions that are presently beyond the scope of Artificial Neural Networks.
- the Digital neural spiking neural networks and particularly the Dynamic Spatio-Temporal Neural Network offer a number of advantages over the prior art technologies.
- the neural model exhibits higher speed, excellent stability and very low power dissipation due to an asynchronous digital organization, with a further advantage that learned functions can be computationally retrieved and stored.
- Similar network models and libraries are disclosed in U.S. Pat. No. 8,250,011 and US Pub. No. 2013/0297537, both of which were invented by the current inventor incorporated by reference herein.
- the digital circuit is very stable and its behavior is completely repeatable. Its characteristics are not dependent on any process technology.
- the neural model is elaborate, including the effects of spike interval neuro-transmitter reuptake, leakage currents, synaptic plasticity, and threshold plasticity.
- Synaptic variables can be computationally retrieved through a microprocessor interface enabling the storage and subsequent reuse of learned tasks.
- a system comprising an artificial intelligent device, consisting of dynamic spatio-temporal neural network, for identifying cognitive signals and learning the ways to respond to the cognitive signals. Further, the system provides for transmitting the signals to and hence, communicating with a remote artificial intelligent device, consisting of dynamic spatio-temporal neural network, over a communication protocol, such as the Internet. Additionally, the system disclosed in the present invention provides for analyzing and controlling one or more appliances from a remote location. Due to the communication protocol, the Internet, a user may be able to analyze or control his/her appliances from a remote location also by utilizing the artificial intelligent device of dynamic spatio-temporal neural network that recognizes, indentifies and learns to respond in a particular way to particular signals with respect to a particular user.
- FIG. 1 illustrates a system for communicating two or more components of a dynamic spatio-temporal neural network, in accordance with an embodiment of the present invention.
- the present invention provides a system comprising an artificial intelligent device 104 of dynamic spatio-temporal neural network.
- the artificial intelligent device 104 receives a plurality of input pulses in form of stimuli.
- Stimuli may be in form of auditory signals, human speech, animal, or any object, or any other type of stimuli generated from different event such as, object recognition, hand or body movement, and others.
- the system 100 relates to an artificial device 104 that receives input pulses in the form of auditory signals, and recognizes and identifies the input signals by implementing the spatio-temporal neural network comprised in the artificial intelligent device 104 . Subsequently, the artificial intelligent device 104 learns to respond to the input pulses.
- the artificial intelligent device 104 receives plurality of input pulses, in the form of acoustic signals or sound waves.
- One or more acoustic signals are captured by an input sensor from an auditory source such as the human voice, animal or insect activity or mechanical vibration.
- the input sensor is further connected to an artificial cochlear 102 consisting of a series of resonators.
- the sound waves in the acoustic signals produces pressure waves in the artificial cochlear 102 that stimulates sensory neurons in the cochlear 102 .
- the sensory neurons each produce a potential that varies at the same rate as the sound pressure waves of various frequencies.
- the artificial cochlear 102 outputs a plurality of pulse patterns with a spatio-temporal distribution that is indicative of features in the acoustic signals.
- the artificial cochlear 102 connects to a first artificial intelligent device 104 .
- the first artificial intelligent device 104 consists of a spatio-temporal neural network.
- the varying potential produced at the artificial cochlear 102 connects to a series of resonators that approximates the function of hair cells in the biological cochlear, producing a series of spatio-temporal pulses that are equivalent to the action potentials in the biological cochlear nerve. Further, the pulse streams with a spatio-temporal distribution are input to the connected artificial intelligent device 104 that consists of a Spatio-Temporal Neural Network. The Spatio-Temporal Neural Network in the artificial intelligent device 104 is subsequently trained to respond to a limited set of acoustic signatures.
- the method of identifying acoustic signatures is by association in a Dynamic Spatio-Temporal Neural Network.
- Each synapse circuit in the Spatio-Temporal Neural Network of the artificial intelligent device 104 performs a temporal integration function on a single pulse stream.
- the Dendrite is a mechanism that provides spatial integration functions in a biological neuron.
- the values resulting from temporal integration in each artificial synapse circuit, the result of pulse interval and pulse frequency are integrated in an artificial dendrite circuit.
- Multiple synapse circuits are connected to each dendrite circuit, each synapse generating a single value. These values are spatially integrated in the dendrite circuit.
- One or more dendrite circuits connect to a soma circuit.
- the soma In biological neurons, the soma consists of the nucleus of the cell and performs a global integration function.
- the artificial soma circuit integrates the values from a plurality of dendric circuits.
- the soma circuit connects to an n artificial axon circuit generates an output pulse sequence that is proportional to the total of integrated dendric values.
- Synapses receive feedback from the Post Synaptic Neuron.
- the state of synapse circuit is changed according to the timing of input and feedback pulses.
- Synaptic Time Dependent Plasticity the spatio-temporal neural network of the artificial intelligent device 104 is trained by exposing it to acoustic pattern information through the input sensor. Once a particular acoustic signature has been learned, the learning function can be mitigated by applying a value that represents a Neuromodulator.
- the acoustic signature is internally represented as pulses, also known as spikes, with spatio-temporal distribution.
- the system 100 provides an artificial intelligent device 104 consisting of spatio-temporal neural network that recognizes and identifies acoustic signature in one or more auditory signals. Subsequently, the artificial intelligent device 104 learns to respond to the auditory signals.
- the present invention also provides the system 100 to couple two or more parts/components of a spatio-temporal spiking neural network comprised in an artificial intelligent device, such as to provide acoustic signature identification or other cognitive tasks and the related communications between two or more components of a spatio-temporal spiking neural network.
- Each component comprises a plurality of neural processing cores and synaptic memory.
- the system 100 provides connecting two or more parts or components of a distributed and spatio-temporal spiking neural network, comprised in the artificial intelligent devices, by some means of parallel or serial communication protocol 108 , such as the Internet.
- the communication protocol may be Intranet or a fast serial bus.
- a plurality of spike times, originating neuron and destinations, are transmitted as packets of information, and feedback from a remote spiking neural network is received as packets of information.
- the protocol for these packets of information can be any established standard, including but not limited to PCI (Peripheral Component Interconnect), PCIe (PCI express), USB (Universal Serial Bus), or TCP (Transmission Control Protocol).
- PCI Peripheral Component Interconnect
- PCIe PCI express
- USB Universal Serial Bus
- TCP Transmission Control Protocol
- a distributed spiking neural network of the artificial intelligent device 104 is used in the recognition and identification of acoustic signals using acoustic signature recognition by means of a spatio-temporal neural network, in accordance with an embodiment of the present invention.
- the first artificial intelligent device 104 connects to a remote artificial intelligent device 104 A via a Serial Address Event Representation Bus 106 , over the TCP/IP 108 .
- the pulse stream patterns with spatio-temporal distribution generated at the first artificial intelligent device 104 defining the features of the events of the input pulses are transmitted to the remote artificial intelligent device 104 A.
- the pulse stream patterns with spatio-temporal distribution consist of the pulse timing and spatial distribution that transmit key aspects of the acoustic signature in the input auditory signals.
- the remote artificial intelligent device 104 A receives spatio-temporal pulse streams via a Serial Address Event Representation 106 A. After receiving the pulse stream patterns, the remote artificial intelligent device 104 A recognizes, identifies the pulse stream patterns and learns to respond in the same as the first artificial intelligent device 104 , by means of Synaptic Time Dependent Plasticity.
- the pulse streams with spatio-temporal distribution communicate over two linked Address Event Representation (AER) busses ( 106 and 106 A).
- AER Address Event Representation
- the properties of the AER bus are that each activation event of an artificial neuron is represented as the address of that neuron.
- the matrix of parallel processing artificial neurons is indexed, with each artificial neuron assigned an index number. This index number represents its address.
- the internal bus 106 is 16 bits wide and allows for ‘near’ connections, while the external serial bus 106 A contains a 40 bits address field, allowing a maximum of 240 connections.
- the communications protocol 108 is optimized to avoid bus contention.
- the serial AER communication bus 106 is connected to a remote Dynamic Spatio-Temporal Neural network of a remote artificial intelligent device 104 A through a communication link, such as the Internet 108 .
- the remote Dynamic Spatio-Temporal Neural Network of the remote artificial intelligent device 104 A has the same structure as of the Dynamic Spatio-Temporal Neural Network of the first artificial intelligent device 104 .
- the remote artificial intelligent device 104 A connects to a computing device for the purpose of analysis or controlling an appliance.
- the remote artificial intelligent device 104 A connects with a microprocessor 110 that further is connected with one or more appliances that need to be controlled remotely.
- the system 100 provides a communication system for connecting two or more components a distributed spatio-temporal neural network comprised in one or more artificial devices, where the input spikes is communicated in a communication protocol over the Internet 108 , so that physical layers of the artificial neural networks can be separated by miles but still be part of the same network.
- FIG. 2 illustrates a method of learning and identifying acoustic signature of auditory signals using a dynamic spatio-temporal neural network, in accordance with an embodiment of the present invention.
- an artificial intelligent device 104 identifies and recognizes features in a plurality of spatio-temporal pulses received by it, implementing Synaptic Time Dependent Plasticity.
- the artificial intelligent device 104 comprises of a dynamic spatio-temporal neural network.
- the artificial intelligent device 104 receives a plurality of input pulses in form of stimuli.
- Stimuli may be in form of auditory signals, human speech, animal, or any object, or any other type of stimuli generated from different event such as, object recognition, hand or body movement, and others.
- an input sensor captures a plurality of acoustic signals from an auditory source such as the human voice, animal or insect activity or mechanical vibration, at step 202 .
- An artificial cochlear 102 is further connected to the input sensor, where the artificial cochlear 102 consists of a series of resonators.
- the artificial cochlear 102 receives the acoustic signals from the input sensor, at step 204 .
- the sound waves in the acoustic signals produces pressure waves in the artificial cochlear 102 that stimulates sensory neurons in the cochlear 102 , at step 206 .
- the sensory neurons of the artificial cochlear 102 each produce a potential that varies at the same rate as the sound pressure waves of various frequencies.
- the varying potential produced at the artificial cochlear 102 connects to a series of resonators that approximates the function of hair cells in the biological cochlear, producing a series of spatio-temporal pulses that are equivalent to the action potentials in the biological cochlear nerve.
- the artificial cochlear 102 outputs a plurality of pulse patterns with a spatio-temporal distribution that is indicative of features in the acoustic signals.
- the artificial cochlear 102 connects to a first artificial intelligent device 104 .
- the first artificial intelligent device 104 consists of a spatio-temporal neural network.
- the Spatio-Temporal Neural Network in the artificial intelligent device 104 is subsequently trained to respond to a limited set of acoustic signatures.
- the artificial intelligent device 104 of spatio-temporal neural network After receiving a series of spatio-temporal neural network, at step 212 , the artificial intelligent device 104 of spatio-temporal neural network recognizes and identifies the acoustic signatures in the auditory signals by association in a dynamic spatio-temporal neural network.
- Each synapse circuit in the Spatio-Temporal Neural Network of the artificial intelligent device 104 performs a temporal integration function on a single pulse stream.
- the Dendrite is a mechanism that provides spatial integration functions in a biological neuron.
- the values resulting from temporal integration in each artificial synapse circuit, the result of pulse interval and pulse frequency are integrated in an artificial dendrite circuit.
- Each dendrite circuit connects to a soma circuit.
- the soma In biological neurons, the soma consists of the nucleus of the cell and performs a global integration function.
- the artificial soma circuit integrates the values from a plurality of dendric circuits.
- the soma circuit connects to an n artificial axon circuit generates an output pulse sequence that is proportional to the total of integrated dendric values.
- Synapses receive feedback from the Post Synaptic Neuron.
- the state of synapse circuit is changed according to the timing of input and feedback pulses.
- This comprises a learning function, known as Synaptic Time Dependent Plasticity.
- Synaptic Time Dependent Plasticity the spatio-temporal neural network of the artificial intelligent device 104 is trained by exposing it to acoustic pattern information through the input sensor.
- the learning function can be mitigated by applying a value that represents a Neuromodulator.
- the acoustic signature is internally represented as pulses, also known as spikes, with spatio-temporal distribution.
- FIG. 3 illustrates a method of communicating two or more components of a dynamic spatio-temporal neural network, in accordance with an embodiment of the present invention.
- the present invention also provides a method for communication between two or more components a distributed spatio-temporal neural network comprised in one or more artificial devices.
- the communication occurs by a means of parallel or serial communication protocol 108 , such as the Internet.
- a plurality of spike times, originating neuron and destinations, are transmitted as packets of information, and feedback from a remote spiking neural network is received as packets of information.
- the protocol for these packets of information can be any established standard, including but not limited to PCI (Peripheral Component Interconnect), PCIe (PCI express), USB (Universal Serial Bus), or TCP (Transmission Control Protocol).
- PCI Peripheral Component Interconnect
- PCIe PCI express
- USB Universal Serial Bus
- TCP Transmission Control Protocol
- the method 300 describes a step 302 where a first artificial intelligent device 104 recognizes and identifies plurality of acoustic signals using acoustic signature recognition by means of a spatio-temporal neural network.
- the first artificial intelligent device 104 connects to a remote artificial intelligent device 104 A via a Serial Address Event Representation Bus 106 , over the TCP/IP 108 .
- the pulse stream patterns with spatio-temporal distribution generated at the first artificial intelligent device 104 defining the features of the events of the input pulses are transmitted to the Serial Address Event Representation (abbreviated as AER) Bus 106 , over the TCP/IP 108 , at step 304 .
- the pulse stream patterns with spatio-temporal distribution consist of the pulse timing and spatial distribution that transmit key aspects of the acoustic signature in the input auditory signals.
- the AER bus transmits the pulse streams with spatio-temporal neural network to the remote artificial intelligent device 104 A, at step 306 .
- the remote artificial intelligent device 104 A has the same structure as of the Dynamic Spatio-Temporal Neural Network of the first artificial intelligent device 104 .
- the pulse streams with spatio-temporal distribution communicate over two linked Address Event Representation (AER) busses ( 106 and 106 A).
- the AER bus 106 and 106 A represents each activation event of an artificial neuron as the address of that neuron.
- the matrix of parallel processing artificial neurons is indexed, with each artificial neuron assigned an index number. This index number represents its address.
- the internal bus 106 is 16 bits wide and allows for ‘near’ connections, while the external serial bus 106 A contains a 40 bits address field, allowing a maximum of 240 connections.
- the communications protocol 108 is optimized to avoid bus contention.
- the remote artificial intelligent device 104 A After receiving the pulse stream patterns, the remote artificial intelligent device 104 A recognizes, identifies the pulse stream patterns and learns to respond in the same manner as the first artificial intelligent device 104 , by means of Synaptic Time Dependent Plasticity, at step 308 .
- the remote artificial intelligent device 104 A is connected to a computing device for analysis and controlling of an appliance. Therefore, at step 310 , the remote artificial intelligent device 104 A connects with a microprocessor further is connected with one or more appliances that need to be controlled remotely.
- the present invention provides a system 100 comprising artificial intelligent device 104 communicating with a remote artificial intelligent device 104 A, where the remote artificial intelligent device 104 A efficiently controls one or more appliances. Therefore, the system 100 controls the appliances from anywhere in the world using a user's device equipped with the first artificial intelligent device 104 . Also, the system 100 provides secure communication, since each neural network in the artificial intelligent device is trained to a unique voice signature of its owner. Further, the transfer of pulse streams is in AER format, which contains voice signature of the user/owner in timing between pulses distribution.
Abstract
Description
- This application claims benefit to U.S. Provisional Patent Application Ser. No. 62/018,562, filed on Jun. 28, 2014 and claims benefit to U.S. Provisional Patent Application Ser. No. 62/019,399, filed on Jun. 30, 2014, the disclosures of each of which are hereby incorporated by reference in their entirety.
- The present invention generally relates to communications system over a communication network. In particular, the present invention relates to a system and method to couple two or more parts of a spatio-temporal spiking neural network, such as to provide acoustic signature identification or other cognitive tasks and the related communications between two or more components of a spatio-temporal spiking neural network.
- Neural networks have been around since the mid-1940's. Since then, many different neural models have been conceived with the aim to reproduce the considerable cognitive powers of the brain. These Artificial Neural Networks vary from simple sigmoid networks (perceptron) to complex biologically-accurate Spatio-temporal Spiking Neural Networks. It becomes increasingly difficult to determine the distinctions between these networks. For the purpose of clarity, Neural Networks can be divided into four distinct groups;
-
- Computer software neural networks also referred to as ‘Sigmoid Networks’
- Computer software simulation of biologically realistic neural networks
- Analog hardware emulation of a sigmoid networks or spiking neural networks
- Digital hardware logic emulation of sigmoid networks or spiking neural networks
- Within each of these categories a number of variations exist that differ in accuracy and computational efficacy. Computer software neural networks generally exist of sigmoid function neurons wherein activated values are added and a non-linear mathematical function, such as a logistics function, is applied to the sum. Computer software simulations of biologically accurate neuron models, such as the Hodgkin-Huxley model of the giant squid neuron, focus on building accurate mathematical models of biological networks. In Analog hardware networks, a model is constructed in VLSI, consisting of transistors that behave more or less like synapses and an Integrate and Fire neural network. In Digital hardware emulation, the same is accomplished using logic gates. Each variation has its specific advantages and application areas.
- The above mentioned technologies lack in efficiently capturing the dynamics of a biological neuronal networks, including spike-timing dependent plasticity, the effects of neurotransmitters and neuromodulators and diverse spike timing representations. Further, the present technologies do not provide computational retrieving and storing learned tasks, so that they may be able to be re-learned by another dynamic neural network. Additionally, the technologies lack in fast speeds and stability.
- Further, the present technologies do not describe communications between two or more components of artificial neural network, that may be implemented in various applications, such as communication to a remote neural network irrespective of the location, controlling and analyzing a connected appliance or equipment.
- Therefore, there exist a need to provide a system and a method to capture dynamics of biological neuronal networks, including spike-timing dependent plasticity, and couple two or more parts of a spatio-temporal spiking neural network, such as to provide cognitive tasks, and resultantly provide communication between them.
- Therefore, the present invention relates to a system and a method for providing communication between two or more components of a spatio-temporal spiking neural network, such as to provide acoustic signature identification or other cognitive tasks and the related communications between two or more components of a spatio-temporal spiking neural network.
- An embodiment of the present invention provides an apparatus for identifying and learning acoustic signature of a plurality of auditory signals comprising: an input sensor configured to capture a varying potential produced in the plurality of auditory signals from an auditory source; a series of resonators configured to convert the varying potential into a plurality of streams of electrical pulses with spatio-temporal distribution; an artificial intelligent device identifies one or more features of the pulse streams of acoustic signals, representing the acoustic signature, by association in a dynamic spatio-temporal neural network; and the artificial intelligent device learns to respond to the acoustic signature of the acoustic signals by modifying synaptic strengths in the dynamic spatio-temporal neural network.
- Another embodiment of the present invention provides a communication system comprising: a first artificial intelligent device consisting of a dynamic spatio-temporal neural network configured to receive and learn from a plurality of pulse streams of acoustic signals with spatio-temporal distribution; and at least one remote artificial intelligent device consisting of a dynamic spatio-temporal neural network communicating with the first artificial intelligent device, through a communication channel, configured to receive and respond to the pulse streams.
- Yet another embodiment of the invention provides a method for identifying and learning acoustic signature of a plurality of auditory signals comprising: capturing, by an input sensor, a varying potential produced in the plurality of auditory signals from an auditory source; converting the varying potential into a plurality of streams of electrical pulses with spatio-temporal distribution, by a series of resonators; identifying, by an artificial intelligent device, one or more features of the pulse streams of acoustic signals, representing the acoustic signature, through association in a dynamic spatio-temporal neural network; and learning, by the artificial intelligent device, to respond to the acoustic signature of the acoustic signals by modifying synaptic strengths in the dynamic spatio-temporal neural network.
- An embodiment of the present invention discloses communication to a remote spatio-temporal spiking neural network for receiving spatio-temporal pulse streams via a parallel or serial communication such as the Internet.
- Another embodiment of the present invention discloses a communication system in order to analyze and control an appliance irrespective of the location.
- Yet another embodiment of the present invention provides a secure communication system because each spatio-temporal neural network is trained to respond to a unique voice signature.
-
FIG. 1 illustrates a system for communicating two or more components of a dynamic spatio-temporal neural network, in accordance with an embodiment of the present invention. -
FIG. 2 illustrates a method of learning and identifying acoustic signature of auditory signals using a dynamic spatio-temporal neural network, in accordance with an embodiment of the present invention. -
FIG. 3 illustrates a method of communicating two or more components of a dynamic spatio-temporal neural network, in accordance with an embodiment of the present invention. - In the following detailed description of embodiments of the invention, numerous specific details are set forth in order to provide a thorough understanding of the embodiment of invention. However, it will be obvious to a person skilled in art that the embodiments of invention may be practiced with or without these specific details. In other instances well known methods, procedures and components have not been described in details, so as not to unnecessarily obscure aspects of the embodiments of the invention.
- Furthermore, it will be clear that the invention is not limited to these embodiments only. Numerous modifications, changes, variations, substitutions and equivalents will be apparent to those skilled in the art, without parting from the spirit and scope of the invention.
- According to the present invention, the Dynamic Spatio-Temporal Neural Network model described here captures the dynamics of biological neuronal networks, including spike-timing dependent plasticity, the effects of neurotransmitters and neuromodulators and diverse spike timing representations, allowing the model to perform functions that are presently beyond the scope of Artificial Neural Networks.
- The Digital neural spiking neural networks and particularly the Dynamic Spatio-Temporal Neural Network offer a number of advantages over the prior art technologies. The neural model exhibits higher speed, excellent stability and very low power dissipation due to an asynchronous digital organization, with a further advantage that learned functions can be computationally retrieved and stored. Similar network models and libraries are disclosed in U.S. Pat. No. 8,250,011 and US Pub. No. 2013/0297537, both of which were invented by the current inventor incorporated by reference herein.
- Power dissipation is extremely low. The digital circuit is very stable and its behavior is completely repeatable. Its characteristics are not dependent on any process technology. The neural model is elaborate, including the effects of spike interval neuro-transmitter reuptake, leakage currents, synaptic plasticity, and threshold plasticity.
- Learning takes place through the modification of neurotransmitter levels resulting from feedback from the post-synaptic neuron, and the intensity of input spikes. Synaptic variables can be computationally retrieved through a microprocessor interface enabling the storage and subsequent reuse of learned tasks.
- In the present invention, a system is provided that comprises an artificial intelligent device, consisting of dynamic spatio-temporal neural network, for identifying cognitive signals and learning the ways to respond to the cognitive signals. Further, the system provides for transmitting the signals to and hence, communicating with a remote artificial intelligent device, consisting of dynamic spatio-temporal neural network, over a communication protocol, such as the Internet. Additionally, the system disclosed in the present invention provides for analyzing and controlling one or more appliances from a remote location. Due to the communication protocol, the Internet, a user may be able to analyze or control his/her appliances from a remote location also by utilizing the artificial intelligent device of dynamic spatio-temporal neural network that recognizes, indentifies and learns to respond in a particular way to particular signals with respect to a particular user.
-
FIG. 1 illustrates a system for communicating two or more components of a dynamic spatio-temporal neural network, in accordance with an embodiment of the present invention. - The present invention provides a system comprising an artificial
intelligent device 104 of dynamic spatio-temporal neural network. The artificialintelligent device 104 receives a plurality of input pulses in form of stimuli. Stimuli may be in form of auditory signals, human speech, animal, or any object, or any other type of stimuli generated from different event such as, object recognition, hand or body movement, and others. Thesystem 100 relates to anartificial device 104 that receives input pulses in the form of auditory signals, and recognizes and identifies the input signals by implementing the spatio-temporal neural network comprised in the artificialintelligent device 104. Subsequently, the artificialintelligent device 104 learns to respond to the input pulses. - In biological sound sensing system, sound produces pressure waves in a cochlear that stimulates sensory neurons. The sensory neurons each produce a potential that varies at the same rate as the sound pressure waves of various frequencies. Since the artificial neural network approximates the biological neural network, the artificial
intelligent device 104 receives plurality of input pulses, in the form of acoustic signals or sound waves. - One or more acoustic signals are captured by an input sensor from an auditory source such as the human voice, animal or insect activity or mechanical vibration. The input sensor is further connected to an artificial cochlear 102 consisting of a series of resonators. The sound waves in the acoustic signals produces pressure waves in the artificial cochlear 102 that stimulates sensory neurons in the cochlear 102. As mentioned above, the sensory neurons each produce a potential that varies at the same rate as the sound pressure waves of various frequencies. The artificial cochlear 102 outputs a plurality of pulse patterns with a spatio-temporal distribution that is indicative of features in the acoustic signals. The artificial cochlear 102 connects to a first artificial
intelligent device 104. The first artificialintelligent device 104 consists of a spatio-temporal neural network. - The varying potential produced at the artificial cochlear 102 connects to a series of resonators that approximates the function of hair cells in the biological cochlear, producing a series of spatio-temporal pulses that are equivalent to the action potentials in the biological cochlear nerve. Further, the pulse streams with a spatio-temporal distribution are input to the connected artificial
intelligent device 104 that consists of a Spatio-Temporal Neural Network. The Spatio-Temporal Neural Network in the artificialintelligent device 104 is subsequently trained to respond to a limited set of acoustic signatures. - The method of identifying acoustic signatures is by association in a Dynamic Spatio-Temporal Neural Network. Each synapse circuit in the Spatio-Temporal Neural Network of the artificial
intelligent device 104 performs a temporal integration function on a single pulse stream. The Dendrite is a mechanism that provides spatial integration functions in a biological neuron. In the Spatio-Temporal Neural network, the values resulting from temporal integration in each artificial synapse circuit, the result of pulse interval and pulse frequency are integrated in an artificial dendrite circuit. Multiple synapse circuits are connected to each dendrite circuit, each synapse generating a single value. These values are spatially integrated in the dendrite circuit. One or more dendrite circuits connect to a soma circuit. In biological neurons, the soma consists of the nucleus of the cell and performs a global integration function. The artificial soma circuit integrates the values from a plurality of dendric circuits. The soma circuit connects to an n artificial axon circuit generates an output pulse sequence that is proportional to the total of integrated dendric values. - Synapses receive feedback from the Post Synaptic Neuron. The state of synapse circuit is changed according to the timing of input and feedback pulses. This comprises a learning function, known as Synaptic Time Dependent Plasticity. In this way by means of Synaptic Time Dependent Plasticity, the spatio-temporal neural network of the artificial
intelligent device 104 is trained by exposing it to acoustic pattern information through the input sensor. Once a particular acoustic signature has been learned, the learning function can be mitigated by applying a value that represents a Neuromodulator. The acoustic signature is internally represented as pulses, also known as spikes, with spatio-temporal distribution. - Therefore, the
system 100 provides an artificialintelligent device 104 consisting of spatio-temporal neural network that recognizes and identifies acoustic signature in one or more auditory signals. Subsequently, the artificialintelligent device 104 learns to respond to the auditory signals. - Further, the present invention also provides the
system 100 to couple two or more parts/components of a spatio-temporal spiking neural network comprised in an artificial intelligent device, such as to provide acoustic signature identification or other cognitive tasks and the related communications between two or more components of a spatio-temporal spiking neural network. Each component comprises a plurality of neural processing cores and synaptic memory. - The
system 100 provides connecting two or more parts or components of a distributed and spatio-temporal spiking neural network, comprised in the artificial intelligent devices, by some means of parallel orserial communication protocol 108, such as the Internet. In an embodiment, the communication protocol may be Intranet or a fast serial bus. A plurality of spike times, originating neuron and destinations, are transmitted as packets of information, and feedback from a remote spiking neural network is received as packets of information. The protocol for these packets of information can be any established standard, including but not limited to PCI (Peripheral Component Interconnect), PCIe (PCI express), USB (Universal Serial Bus), or TCP (Transmission Control Protocol). - As described above in
FIG. 1 , a distributed spiking neural network of the artificialintelligent device 104 is used in the recognition and identification of acoustic signals using acoustic signature recognition by means of a spatio-temporal neural network, in accordance with an embodiment of the present invention. Further, referring toFIG. 1 , the first artificialintelligent device 104 connects to a remote artificialintelligent device 104A via a Serial AddressEvent Representation Bus 106, over the TCP/IP 108. The pulse stream patterns with spatio-temporal distribution generated at the first artificialintelligent device 104 defining the features of the events of the input pulses are transmitted to the remote artificialintelligent device 104A. The pulse stream patterns with spatio-temporal distribution consist of the pulse timing and spatial distribution that transmit key aspects of the acoustic signature in the input auditory signals. The remote artificialintelligent device 104A receives spatio-temporal pulse streams via a SerialAddress Event Representation 106A. After receiving the pulse stream patterns, the remote artificialintelligent device 104A recognizes, identifies the pulse stream patterns and learns to respond in the same as the first artificialintelligent device 104, by means of Synaptic Time Dependent Plasticity. - The pulse streams with spatio-temporal distribution communicate over two linked Address Event Representation (AER) busses (106 and 106A). The properties of the AER bus are that each activation event of an artificial neuron is represented as the address of that neuron. The matrix of parallel processing artificial neurons is indexed, with each artificial neuron assigned an index number. This index number represents its address. In an embodiment, the
internal bus 106 is 16 bits wide and allows for ‘near’ connections, while the externalserial bus 106A contains a 40 bits address field, allowing a maximum of 240 connections. Thecommunications protocol 108 is optimized to avoid bus contention. In a preferred embodiment, the serialAER communication bus 106 is connected to a remote Dynamic Spatio-Temporal Neural network of a remote artificialintelligent device 104A through a communication link, such as theInternet 108. The remote Dynamic Spatio-Temporal Neural Network of the remote artificialintelligent device 104A has the same structure as of the Dynamic Spatio-Temporal Neural Network of the first artificialintelligent device 104. - In an embodiment, the remote artificial
intelligent device 104A connects to a computing device for the purpose of analysis or controlling an appliance. The remote artificialintelligent device 104A connects with amicroprocessor 110 that further is connected with one or more appliances that need to be controlled remotely. - Therefore, the
system 100 provides a communication system for connecting two or more components a distributed spatio-temporal neural network comprised in one or more artificial devices, where the input spikes is communicated in a communication protocol over theInternet 108, so that physical layers of the artificial neural networks can be separated by miles but still be part of the same network. -
FIG. 2 illustrates a method of learning and identifying acoustic signature of auditory signals using a dynamic spatio-temporal neural network, in accordance with an embodiment of the present invention. According to the present invention, an artificialintelligent device 104 identifies and recognizes features in a plurality of spatio-temporal pulses received by it, implementing Synaptic Time Dependent Plasticity. The artificialintelligent device 104 comprises of a dynamic spatio-temporal neural network. The artificialintelligent device 104 receives a plurality of input pulses in form of stimuli. Stimuli may be in form of auditory signals, human speech, animal, or any object, or any other type of stimuli generated from different event such as, object recognition, hand or body movement, and others. - According to the
method 200, an input sensor captures a plurality of acoustic signals from an auditory source such as the human voice, animal or insect activity or mechanical vibration, atstep 202. An artificial cochlear 102 is further connected to the input sensor, where the artificial cochlear 102 consists of a series of resonators. The artificial cochlear 102 receives the acoustic signals from the input sensor, atstep 204. The sound waves in the acoustic signals produces pressure waves in the artificial cochlear 102 that stimulates sensory neurons in the cochlear 102, atstep 206. Thereafter, atstep 208, the sensory neurons of the artificial cochlear 102 each produce a potential that varies at the same rate as the sound pressure waves of various frequencies. The varying potential produced at the artificial cochlear 102 connects to a series of resonators that approximates the function of hair cells in the biological cochlear, producing a series of spatio-temporal pulses that are equivalent to the action potentials in the biological cochlear nerve. - Further at
step 210, the artificial cochlear 102 outputs a plurality of pulse patterns with a spatio-temporal distribution that is indicative of features in the acoustic signals. The artificial cochlear 102 connects to a first artificialintelligent device 104. The first artificialintelligent device 104 consists of a spatio-temporal neural network. The Spatio-Temporal Neural Network in the artificialintelligent device 104 is subsequently trained to respond to a limited set of acoustic signatures. - After receiving a series of spatio-temporal neural network, at
step 212, the artificialintelligent device 104 of spatio-temporal neural network recognizes and identifies the acoustic signatures in the auditory signals by association in a dynamic spatio-temporal neural network. Each synapse circuit in the Spatio-Temporal Neural Network of the artificialintelligent device 104 performs a temporal integration function on a single pulse stream. The Dendrite is a mechanism that provides spatial integration functions in a biological neuron. In the Spatio-Temporal Neural network, the values resulting from temporal integration in each artificial synapse circuit, the result of pulse interval and pulse frequency are integrated in an artificial dendrite circuit. Multiple synapse circuits are connected to each dendrite circuit, each synapse generating a single value. These values are spatially integrated in the dendrite circuit. One or more dendrite circuits connect to a soma circuit. In biological neurons, the soma consists of the nucleus of the cell and performs a global integration function. The artificial soma circuit integrates the values from a plurality of dendric circuits. The soma circuit connects to an n artificial axon circuit generates an output pulse sequence that is proportional to the total of integrated dendric values. - Synapses receive feedback from the Post Synaptic Neuron. The state of synapse circuit is changed according to the timing of input and feedback pulses. This comprises a learning function, known as Synaptic Time Dependent Plasticity. In this way by means of Synaptic Time Dependent Plasticity, the spatio-temporal neural network of the artificial
intelligent device 104 is trained by exposing it to acoustic pattern information through the input sensor. Once a particular acoustic signature has been learned atstep 212, the learning function can be mitigated by applying a value that represents a Neuromodulator. The acoustic signature is internally represented as pulses, also known as spikes, with spatio-temporal distribution. -
FIG. 3 illustrates a method of communicating two or more components of a dynamic spatio-temporal neural network, in accordance with an embodiment of the present invention. The present invention also provides a method for communication between two or more components a distributed spatio-temporal neural network comprised in one or more artificial devices. The communication occurs by a means of parallel orserial communication protocol 108, such as the Internet. A plurality of spike times, originating neuron and destinations, are transmitted as packets of information, and feedback from a remote spiking neural network is received as packets of information. The protocol for these packets of information can be any established standard, including but not limited to PCI (Peripheral Component Interconnect), PCIe (PCI express), USB (Universal Serial Bus), or TCP (Transmission Control Protocol). - As shown in
FIG. 3 , themethod 300 describes astep 302 where a first artificialintelligent device 104 recognizes and identifies plurality of acoustic signals using acoustic signature recognition by means of a spatio-temporal neural network. The first artificialintelligent device 104 connects to a remote artificialintelligent device 104A via a Serial AddressEvent Representation Bus 106, over the TCP/IP 108. The pulse stream patterns with spatio-temporal distribution generated at the first artificialintelligent device 104 defining the features of the events of the input pulses are transmitted to the Serial Address Event Representation (abbreviated as AER)Bus 106, over the TCP/IP 108, atstep 304. The pulse stream patterns with spatio-temporal distribution consist of the pulse timing and spatial distribution that transmit key aspects of the acoustic signature in the input auditory signals. - Further, the AER bus transmits the pulse streams with spatio-temporal neural network to the remote artificial
intelligent device 104A, atstep 306. The remote artificialintelligent device 104A has the same structure as of the Dynamic Spatio-Temporal Neural Network of the first artificialintelligent device 104. The pulse streams with spatio-temporal distribution communicate over two linked Address Event Representation (AER) busses (106 and 106A). TheAER bus internal bus 106 is 16 bits wide and allows for ‘near’ connections, while the externalserial bus 106A contains a 40 bits address field, allowing a maximum of 240 connections. Thecommunications protocol 108 is optimized to avoid bus contention. - After receiving the pulse stream patterns, the remote artificial
intelligent device 104A recognizes, identifies the pulse stream patterns and learns to respond in the same manner as the first artificialintelligent device 104, by means of Synaptic Time Dependent Plasticity, atstep 308. In a further embodiment, the remote artificialintelligent device 104A is connected to a computing device for analysis and controlling of an appliance. Therefore, atstep 310, the remote artificialintelligent device 104A connects with a microprocessor further is connected with one or more appliances that need to be controlled remotely. - Advantageously, the present invention provides a
system 100 comprising artificialintelligent device 104 communicating with a remote artificialintelligent device 104A, where the remote artificialintelligent device 104A efficiently controls one or more appliances. Therefore, thesystem 100 controls the appliances from anywhere in the world using a user's device equipped with the first artificialintelligent device 104. Also, thesystem 100 provides secure communication, since each neural network in the artificial intelligent device is trained to a unique voice signature of its owner. Further, the transfer of pulse streams is in AER format, which contains voice signature of the user/owner in timing between pulses distribution.
Claims (23)
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US14/753,296 US20150379397A1 (en) | 2014-06-28 | 2015-06-29 | Secure voice signature communications system |
US16/282,550 US11429857B2 (en) | 2014-06-28 | 2019-02-22 | Secure voice signature communications system using local and remote neural network devices |
Applications Claiming Priority (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US201462018562P | 2014-06-28 | 2014-06-28 | |
US201462019399P | 2014-06-30 | 2014-06-30 | |
US14/753,296 US20150379397A1 (en) | 2014-06-28 | 2015-06-29 | Secure voice signature communications system |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US16/282,550 Continuation US11429857B2 (en) | 2014-06-28 | 2019-02-22 | Secure voice signature communications system using local and remote neural network devices |
Publications (1)
Publication Number | Publication Date |
---|---|
US20150379397A1 true US20150379397A1 (en) | 2015-12-31 |
Family
ID=54930913
Family Applications (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US14/753,296 Abandoned US20150379397A1 (en) | 2014-06-28 | 2015-06-29 | Secure voice signature communications system |
US16/282,550 Active 2037-09-18 US11429857B2 (en) | 2014-06-28 | 2019-02-22 | Secure voice signature communications system using local and remote neural network devices |
Family Applications After (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US16/282,550 Active 2037-09-18 US11429857B2 (en) | 2014-06-28 | 2019-02-22 | Secure voice signature communications system using local and remote neural network devices |
Country Status (1)
Country | Link |
---|---|
US (2) | US20150379397A1 (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170229117A1 (en) * | 2016-02-05 | 2017-08-10 | Brainchip Inc. | Low power neuromorphic voice activation system and method |
US20180015878A1 (en) * | 2016-07-18 | 2018-01-18 | Toyota Motor Engineering & Manufacturing North America, Inc. | Audible Notification Systems and Methods for Autonomous Vehhicles |
CN111612136A (en) * | 2020-05-25 | 2020-09-01 | 之江实验室 | Neural morphology visual target classification method and system |
US10977550B2 (en) | 2016-11-02 | 2021-04-13 | Samsung Electronics Co., Ltd. | Method of converting neural network and recognition apparatus using the same |
US11151441B2 (en) * | 2017-02-08 | 2021-10-19 | Brainchip, Inc. | System and method for spontaneous machine learning and feature extraction |
US11157798B2 (en) | 2016-02-12 | 2021-10-26 | Brainchip, Inc. | Intelligent autonomous feature extraction system using two hardware spiking neutral networks with spike timing dependent plasticity |
US11429857B2 (en) | 2014-06-28 | 2022-08-30 | Brainchip, Inc. | Secure voice signature communications system using local and remote neural network devices |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5537394A (en) * | 1993-11-04 | 1996-07-16 | Fujitsu Limited | Routing system for communications network |
US7774440B1 (en) * | 2001-07-25 | 2010-08-10 | Scalable Network Technologies, Inc. | Method and system for enhancing performance of a physical network under real-time control using simulation of a reference model |
US8111174B2 (en) * | 2007-10-03 | 2012-02-07 | University Of Southern California | Acoustic signature recognition of running vehicles using spectro-temporal dynamic neural network |
US20150269481A1 (en) * | 2014-03-24 | 2015-09-24 | Qualcomm Incorporated | Differential encoding in neural networks |
Family Cites Families (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6173074B1 (en) | 1997-09-30 | 2001-01-09 | Lucent Technologies, Inc. | Acoustic signature recognition and identification |
EP1827580A1 (en) | 2004-10-21 | 2007-09-05 | Advanced Neuromodulation Systems, Inc. | New stimulation design for neuromodulation |
WO2007071070A1 (en) | 2005-12-23 | 2007-06-28 | Universite De Sherbrooke | Spatio-temporal pattern recognition using a spiking neural network and processing thereof on a portable and/or distributed computer |
US8250011B2 (en) | 2008-09-21 | 2012-08-21 | Van Der Made Peter A J | Autonomous learning dynamic artificial neural computing device and brain inspired system |
US20130297537A1 (en) | 2012-05-02 | 2013-11-07 | Peter AJ. van der Made | Method and System for creating Dynamic Neural Function Libraries |
US8250010B2 (en) | 2009-05-21 | 2012-08-21 | International Business Machines Corporation | Electronic learning synapse with spike-timing dependent plasticity using unipolar memory-switching elements |
CA2835701C (en) * | 2011-06-06 | 2017-03-21 | Socpra Sciences Et Genie, S.E.C. | Method, system and aggregation engine for providing structural representations of physical entities |
US9098811B2 (en) | 2012-06-04 | 2015-08-04 | Brain Corporation | Spiking neuron network apparatus and methods |
US8977582B2 (en) | 2012-07-12 | 2015-03-10 | Brain Corporation | Spiking neuron network sensory processing apparatus and methods |
US20150379397A1 (en) | 2014-06-28 | 2015-12-31 | Brainchip, Inc. | Secure voice signature communications system |
-
2015
- 2015-06-29 US US14/753,296 patent/US20150379397A1/en not_active Abandoned
-
2019
- 2019-02-22 US US16/282,550 patent/US11429857B2/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5537394A (en) * | 1993-11-04 | 1996-07-16 | Fujitsu Limited | Routing system for communications network |
US7774440B1 (en) * | 2001-07-25 | 2010-08-10 | Scalable Network Technologies, Inc. | Method and system for enhancing performance of a physical network under real-time control using simulation of a reference model |
US8111174B2 (en) * | 2007-10-03 | 2012-02-07 | University Of Southern California | Acoustic signature recognition of running vehicles using spectro-temporal dynamic neural network |
US20150269481A1 (en) * | 2014-03-24 | 2015-09-24 | Qualcomm Incorporated | Differential encoding in neural networks |
Non-Patent Citations (2)
Title |
---|
Chan et al. "AER EAR: A Matched Silicon Cochlea Pair with Address Event Representation Interface", IEEE TCS-I, 2007, pp 48-59 * |
Kasabov et al. âDynamic evolving spiking neural networks for on-line spatio- and spectro-temporal pattern recognitionâ, Neural Networks 41 (2013) PP 188â201 * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11429857B2 (en) | 2014-06-28 | 2022-08-30 | Brainchip, Inc. | Secure voice signature communications system using local and remote neural network devices |
US20170229117A1 (en) * | 2016-02-05 | 2017-08-10 | Brainchip Inc. | Low power neuromorphic voice activation system and method |
US10157629B2 (en) * | 2016-02-05 | 2018-12-18 | Brainchip Inc. | Low power neuromorphic voice activation system and method |
US11157798B2 (en) | 2016-02-12 | 2021-10-26 | Brainchip, Inc. | Intelligent autonomous feature extraction system using two hardware spiking neutral networks with spike timing dependent plasticity |
US20180015878A1 (en) * | 2016-07-18 | 2018-01-18 | Toyota Motor Engineering & Manufacturing North America, Inc. | Audible Notification Systems and Methods for Autonomous Vehhicles |
US9956910B2 (en) * | 2016-07-18 | 2018-05-01 | Toyota Motor Engineering & Manufacturing North America, Inc. | Audible notification systems and methods for autonomous vehicles |
US10977550B2 (en) | 2016-11-02 | 2021-04-13 | Samsung Electronics Co., Ltd. | Method of converting neural network and recognition apparatus using the same |
US11151441B2 (en) * | 2017-02-08 | 2021-10-19 | Brainchip, Inc. | System and method for spontaneous machine learning and feature extraction |
CN111612136A (en) * | 2020-05-25 | 2020-09-01 | 之江实验室 | Neural morphology visual target classification method and system |
Also Published As
Publication number | Publication date |
---|---|
US11429857B2 (en) | 2022-08-30 |
US20190188600A1 (en) | 2019-06-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11429857B2 (en) | Secure voice signature communications system using local and remote neural network devices | |
Yu et al. | Spike timing or rate? Neurons learn to make decisions for both through threshold-driven plasticity | |
Taherkhani et al. | A supervised learning algorithm for learning precise timing of multiple spikes in multilayer spiking neural networks | |
Diehl et al. | Unsupervised learning of digit recognition using spike-timing-dependent plasticity | |
Yu et al. | Rapid feedforward computation by temporal encoding and learning with spiking neurons | |
Kasabov | Neucube evospike architecture for spatio-temporal modelling and pattern recognition of brain signals | |
US11157798B2 (en) | Intelligent autonomous feature extraction system using two hardware spiking neutral networks with spike timing dependent plasticity | |
JP2016538632A (en) | Method and apparatus for tagging a class using supervised learning | |
George et al. | Event-based softcore processor in a biohybrid setup applied to structural plasticity | |
Indiveri et al. | Spike-based learning in VLSI networks of integrate-and-fire neurons | |
Wysoski et al. | Adaptive learning procedure for a network of spiking neurons and visual pattern recognition | |
CN117275568A (en) | Primary auditory cortex neuron cell release rate curve simulation method and device | |
Mohemmed et al. | Evaluating span incremental learning for handwritten digit recognition | |
Ahmed et al. | A review on methods, issues and challenges in neuromorphic engineering | |
US20220269923A1 (en) | System and method for humanoid robot control and cognitive self-improvement without programming | |
KR102535635B1 (en) | Neuromorphic computing device | |
Liu | Deep spiking neural networks | |
Lobov et al. | Spiking neurons as universal building blocks for hybrid systems | |
Hourdakis et al. | Computational modeling of cortical pathways involved in action execution and action observation | |
Yashchenko | Neural-like growing networks in the development of general intelligence. Neural-like growing networks (P. II) | |
Yusoff et al. | Face-voice association towards multimodal-based authentication using modulated spike-time dependent learning | |
González-Nalda et al. | Topos: generalized Braitenberg vehicles that recognize complex real sounds as landmarks | |
Bálya et al. | Basic mammalian retinal effects on the prototype complex cell CNN universal machine | |
Vaila et al. | Spiking CNNs with PYNN and NEURON | |
Jimenez-Romero | A Heterosynaptic Spiking Neural System for the Development of Autonomous Agents |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AS | Assignment |
Owner name: BRAINCHIP, INC., CALIFORNIA Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:VAN DER MADE, PETER AJ;MANKAR, ANIL SHAMRAO;REEL/FRAME:036225/0389 Effective date: 20150626 Owner name: BRAINCHIP, INC., CALIFORNIA Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:VAN DER MADE, PETER AJ;MANKAR, ANIL SHAMRAO;REEL/FRAME:036226/0242 Effective date: 20150626 |
|
STCV | Information on status: appeal procedure |
Free format text: NOTICE OF APPEAL FILED |
|
STCB | Information on status: application discontinuation |
Free format text: ABANDONED -- FAILURE TO RESPOND TO AN OFFICE ACTION |