knuedge_patents

Our patents and intellectual property create a foundation for the future, with great products that can be built from our innovation and ingenuity.


Patent NumberPatent TitleFirst InventorAbstract
9,208,794PROVIDING SOUND MODELS OF AN INPUT SIGNAL USING CONTINUOUS AND/OR LINEAR FITTINGMassimo MascaroVoice enhancement and/or speech features extraction may be performed on noisy audio signals. An input signal may convey audio comprising a speech component superimposed on a noise component. The input signal may be segmented into discrete successive time windows including a first time window spanning a duration greater than a sampling interval of the input signal. A transform may be performed on individual time windows of the input signal to obtain corresponding sound models of the input signal in the individual time windows. A first sound model may describe a superposition of harmonics sharing a common pitch and chirp in the first time window of the input signal. Linear fits in time of the sound models over individual time windows of the input signal may be obtained. The linear fits may include a first linear fit in time of the first sound model over the first time window.
9,185,057SMART MEMORYDouglas A. PalmerSystems and methods to process packets of information using an on-chip processing system include a memory bank, an interconnect module, a controller, and one or more processing engines. The packets of information include a packet header and a packet payload. The packet header includes one or more operator codes. The transfer of individual packets is guided to a processing engine through the interconnect module and through the controller by operator codes included in the packets.
9,183,850SYSTEM AND METHOD FOR TRACKING SOUND PITCH ACROSS AN AUDIO SIGNALDavid C. BradleyA system and method may be configured to analyze audio information derived from an audio signal. The system and method may track sound pitch across the audio signal. The tracking of pitch across the audio signal may take into account change in pitch by determining at individual time sample windows in the signal duration an estimated pitch and an estimated fractional chirp rate of the harmonics at the estimated pitch. The estimated pitch and the estimated fractional chirp rate may then be implemented to determine an estimated pitch for another time sample window in the signal duration with an enhanced accuracy and/or precision.
9,183,494COMPETITIVE BCM LEARNING RULE FOR IDENTIFYING FEATURESDouglas A. MooreDisclosed are systems, apparatuses, and methods for implementing a competitive BCM learning rule used in a neural network. Such a method includes identifying a maximally responding neuron with respect to a feature of an input signal. The maximally responding neuron is the neuron in a group that has a response with respect to the feature of the input signal that is greater than a response of each other neuron in the group. Such a method also includes applying a learning rule to weaken the response of each other neuron with respect to the feature of the input signal. The learning rule may also strengthen the response of the maximally responding neuron with respect to the feature of the input signal.
9,177,561SYSTEMS AND METHODS FOR RECONSTRUCTING AN AUDIO SIGNAL FROM TRANSFORMED AUDIO INFORMATIONDavid C. BradleyA system and method may be configured to reconstruct an audio signal from transformed audio information. The audio signal may be resynthesized based on individual harmonics and corresponding pitches determined from the transformed audio information. Noise may be subtracted from the transformed audio information by interpolating across peak points and across trough points of harmonic pitch paths through the transformed audio information, and subtracting values associated with the trough point interpolations from values associated with the peak point interpolations. Noise between harmonics of the sound may be suppressed in the transformed audio information by centering functions at individual harmonics in the transformed audio information, the functions serving to suppress noise between the harmonics.
9,177,560SYSTEMS AND METHODS FOR RECONSTRUCTING AN AUDIO SIGNAL FROM TRANSFORMED AUDIO INFORMATIONDavid C. BradleyA system and method may be configured to reconstruct an audio signal from transformed audio information. The audio signal may be resynthesized based on individual harmonics and corresponding pitches determined from the transformed audio information. Noise may be subtracted from the transformed audio information by interpolating across peak points and across trough points of harmonic pitch paths through the transformed audio information, and subtracting values associated with the trough point interpolations from values associated with the peak point interpolations. Noise between harmonics of the sound may be suppressed in the transformed audio information by centering functions at individual harmonics in the transformed audio information, the functions serving to suppress noise between the harmonics.
9,142,220SYSTEMS AND METHODS FOR RECONSTRUCTING AN AUDIO SIGNAL FROM TRANSFORMED AUDIO INFORMATIONDavid C. BradleyA system and method may be configured to reconstruct an audio signal from transformed audio information. The audio signal may be resynthesized based on individual harmonics and corresponding pitches determined from the transformed audio information. Noise may be subtracted from the transformed audio information by interpolating across peak points and across trough points of harmonic pitch paths through the transformed audio information, and subtracting values associated with the trough point interpolations from values associated with the peak point interpolations. Noise between harmonics of the sound may be suppressed in the transformed audio information by centering functions at individual harmonics in the transformed audio information, the functions serving to suppress noise between the harmonics.
9,082,078NEURAL PROCESSING ENGINE AND ARCHITECTURE USING THE SAMEDouglas A. PalmerA neural processing engine may perform processing within a neural processing system and/or artificial neural network. The neural processing engine may be configured to effectively and efficiently perform the type of processing required in implementing a neural processing system and/or an artificial neural network. This configuration may facilitate such processing with neural processing engines having an enhanced computational density and/or processor density with respect to conventional processing units.
9,058,820IDENTIFYING SPEECH PORTIONS OF A SOUND MODEL USING VARIOUS STATISTICS THEREOFMassimo MascaroSpeech portions of a sound model may be identified using various statistics associated with the sound model for voice enhancement of noisy audio signals. A spectral motion transform may be performed on an input signal to obtain a linear fit in time of a sound model of the input signal. Statistics may be extracted from the linear fit of the sound model of the input signal. Speech portions of the linear fit of the sound model of the input signal may be identified by detecting a presence of harmonics as a function of time in the linear fit of the sound model of the input signal based on individual ones of the extracted statistics. An output signal may be provided that conveys audio comprising a reconstructed speech component of the input signal with a noise component of the input signal being suppressed.
8,897,461DENOISING AN AUDIO SIGNAL USING LOCAL FORMANT INFORMATIONEric WiewioraA system, method, and computer program product are provided for cleaning an audio segment. For a given audio segment, an offset amount is calculated where the audio segment is maximally correlated to the audio segment as offset by the offset amount. The audio segment and the audio segment as offset by the offset amount are averaged to produce a cleaned audio segment, which has had noise features reduced while having signal features (such as voiced audio) enhanced.
8,849,663SYSTEMS AND METHODS FOR SEGMENTING AND/OR CLASSIFYING AN AUDIO SIGNAL FROM TRANSFORMED AUDIO INFORMATIONDavid C. BradleyA system and method may be provided to segment and/or classify an audio signal from transformed audio information. Transformed audio information representing a sound may be obtained. The transformed audio information may specify magnitude of a coefficient related to energy amplitude as a function of frequency for the audio signal and time. Features associated with the audio signal may be obtained from the transformed audio information. Individual ones of the features may be associated with a feature score relative to a predetermined speaker model. An aggregate score may be obtained based on the feature scores according to a weighting scheme. The weighting scheme may be associated with a noise and/or SNR estimation. The aggregate score may be used for segmentation to identify portions of the audio signal containing speech of one or more different speakers. For classification, the aggregate score may be used to determine a likely speaker model to identify a source of the sound in the audio signal.
8,848,726I/O DATA INTERFACE FOR PACKET PROCESSORSDouglas A. PalmerSystems and methods to process packets of information use an on-chip information processing system configured to receive, resolve, convert, and/or transmit packets of different packet-types in accordance with different protocols. A first packet-type may use a protocol for wired local-area-networking (LAN) technologies, such as Ethernet. A second packet-type may use a proprietary protocol. The proprietary protocol may be used to exchange information with one or more packet processing engines, such as neural processing engines.
8,767,978SYSTEM AND METHOD FOR PROCESSING SOUND SIGNALS IMPLEMENTING A SPECTRAL MOTION TRANSFORMDavid C. BradleyA system and method are provided for processing sound signals. The processing may include identifying individual harmonic sounds represented in sound signals, determining sound parameters of harmonic sounds, classifying harmonic sounds according to source, and/or other processing. The processing may include transforming the sound signals (or portions thereof) into a space which expresses a transform coefficient as a function of frequency and chirp rate. This may facilitate leveraging of the fact that the individual harmonics of a single harmonic sound may have a common pitch velocity (which is related to the chirp rate) across all of its harmonics in order to distinguish an the harmonic sound from other sounds (harmonic and/or non-harmonic) and/or noise.
8,756,265AUDIO FILTER BANK DESIGNShalom HalevyA system, method, and computer program product are provided for reducing ripple in a filter bank. For a given fixed number of linearly-spaced filters in the filter bank, a monotonically increasing function for a Q-factor is specified for the filter bank. Adjustments to each filter’s Q-factor based on the Q-factor function are made in order to produce a nearly constant filter bank ripple throughout the filter bank’s frequency response range.
8,655,815NEURAL PROCESSING UNITDouglas A. PalmerThe subject matter disclosed herein provides methods, apparatus, and articles of manufacture for neural-based processing. In one aspect, there is provided a method. The method may include reading, from a first memory, context information stored based on at least one connection value; reading, from a second memory, an activation value matching the at least one connection value; sending, by a first processor, the context information and the activation value to at least one of a plurality of microengines to configure the at least one microengine as a neuron; and generating, at the at least one microengine, a value representative of an output of the neuron. Related apparatus, systems, methods, and articles are also described.
8,626,700CONTEXT AWARE DEVICE EXECUTION FOR SIMULATING NEURAL NETWORKS IN COMPUTE UNIFIED DEVICE ARCHITECTUREXavier MONRAZA system, method, and computer program product are provided for efficient allocation of attributes corresponding to neurons or connections of multiple types using a common data structure. A map file is generated by a pre-processor in order to map an attribute of a neuron or connection to a particular location within the common data structure based on a type associated with the neuron or connection, while allowing a neuron or connection of a different type to map its own attribute to that same particular location. Additionally, kernel code can be written to reference attribute names made available by the map file in order to provide reusability of code.
8,620,646SYSTEM AND METHOD FOR TRACKING SOUND PITCH ACROSS AN AUDIO SIGNAL USING HARMONIC ENVELOPEDavid C. BradleyA system and method may be configured to analyze audio information derived from an audio signal. The system and method may track sound pitch across the audio signal. The tracking of pitch across the audio signal may take into account change in pitch by determining at individual time sample windows in the signal duration an estimated pitch and a representation of harmonic envelope at the estimated pitch. The estimated pitch and the representation of harmonic envelope may then be implemented to determine an estimated pitch for another time sample window in the signal duration with an enhanced accuracy and/or precision.
8,548,803SYSTEM AND METHOD OF PROCESSING A SOUND SIGNAL INCLUDING TRANSFORMING THE SOUND SIGNAL INTO A FREQUENCY-CHIRP DOMAINDavid C. BradleyA system and method may be configured to process an audio signal. The system and method may track pitch, chirp rate, and/or harmonic envelope across the audio signal, may reconstruct sound represented in the audio signal, and/or may segment or classify the audio signal. A transform may be performed on the audio signal to place the audio signal in a frequency chirp domain that enhances the sound parameter tracking, reconstruction, and/or classification.
8,543,526SYSTEMS AND METHODS USING NEURAL NETWORKS TO REDUCE NOISE IN AUDIO SIGNALSDouglas A. MooreSystems, methods, and computer program products are provided to provide noise reduction for an input signal using a neural network. A feed-forward set of neuron groups is provided to enhance neuron activity within a particular frequency band based on prior reception of activity within that frequency band, and also to attenuate surrounding frequency bands. A surround-inhibition set of neuron groups further attenuates activity surrounding the stimulated frequency band.
8,543,402SPEAKER SEGMENTATION IN NOISE CONVERSATIONAL SPEECHJiyong MASystem and methods for robust multiple speaker segmentation in noisy conversational speech are presented. Robust voice activity detection is applied to detect temporal speech events. In order to get robust speech features and detect speech events in a noisy environment, a noise reduction algorithm is applied, using noise tracking. After noise reduction and voice activity detection, the incoming audio/speech is initially labeled as speech segments or silence segments. With no prior knowledge of the number of speakers, the system identifies one reliable speech segment near the beginning of the conversational speech and extracts speech features with a short latency, then learns a statistical model from the selected speech segment. This initial statistical model is used to identify the succeeding speech segments in a conversation. The statistical model is also continuously adapted and expanded with newly identified speech segments that match well to the model. The speech segments with low likelihoods are labeled with a second speaker ID, and a statistical model is learned from them. At the same time, these two trained speaker models are also updated/adapted once a reliable speech segment is identified. If a speech segment does not match well to the two speaker models, the speech segment is temporarily labeled as an outlier or as originating from a third speaker. This procedure is then applied recursively as needed when there are more than two speakers in a conversation.
8,521,671NEURAL NETWORK FOR CLUSTERING INPUT DATA BASED ON A GAUSSIAN MIXTURE MODELDouglas A. MooreDisclosed are systems, apparatuses, and methods for clustering data. Such a method includes providing input data to each of a plurality of cluster microcircuits of a neural network, wherein each cluster microcircuit includes a mean neural group and a variance neural group. The method also includes determining a response of each cluster microcircuit with respect to the input data. The method further includes modulating the mean neural group and the variance neural group of each cluster microcircuit responsive to a value system.
8,504,499CONSTANT MEMORY IMPLEMENTATION OF A NETWORK OF PHASE MODELS OF NEURONSJeremy M. LEWIDisclosed are systems, apparatuses, and methods for implementing a phase-model neural network using a fixed amount of memory. Such a phase-model neural network includes a plurality of neurons, wherein each neuron is associated with two parameters—an activity and a phase. Example methods include (i) generating a sequence of variables associated with a probability distribution of phases and (ii) sequentially sampling the probability distribution of phases using a fixed amount of memory, regardless of a number of phases used in the phase-model neural network.
8,473,436NEURAL SEGMENTATION OF AN INPUT SIGNAL AND APPLICATIONS USING SIMULATED NEURONS, AND A PHASE MODULATORDouglas A. MooreDisclosed are systems, methods, and computer-program products for segmenting content of an input signal and applications thereof. In an embodiment, the system includes simulated neurons, a phase modulator, and an entity-identifier module. Each simulated neuron is connected to one or more other simulated neurons and is associated with an activity and a phase. The activity and the phase of each simulated neuron is set based on the activity and the phase of the one or more other simulated neurons connected to each simulated neuron. The phase modulator includes individual modulators, each configured to modulate the activity and the phase of each of the plurality of simulated neurons based on a modulation function. The entity-identifier module is configured to identify one or more distinct entities (e.g., objects, sound sources, etc.) included in the input signal based on the one or more distinct collections of simulated neurons that have substantially distinct phases.
7,627,540ADDRESSING SCHEME FOR NEURAL MODELING AND BRAIN-BASED DEVICES USING SPECIAL PURPOSE PROCESSORJames A. SnookA special purpose processor (SPP) can use a Field Programmable Gate Array (FPGA) to model a large number of neural elements. The FPGAs or similar programmable device can have multiple cores doing presynaptic, postsynaptic, and plasticity calculations in parallel. Each core can implement multiple neural elements of the neural model.
7,533,071NEURAL MODELING AND BRAIN-BASED DEVICES USING SPECIAL PURPOSE PROCESSORJames A. SnookA special purpose processor (SPP) can use a Field Programmable Gate Array (FPGA) or similar programmable device to model a large number of neural elements. The FPGAs can have multiple cores doing presynaptic, postsynaptic, and plasticity calculations in parallel. Each core can implement multiple neural elements of the neural model.
7,519,452MOBILE BRAIN-BASED DEVICE FOR USE IN A REAL WORLD ENVIRONMENTAnil K. SethA mobile brain-based device BBD includes a mobile base equipped with sensors and effectors (Neurally Organized Mobile Adaptive Device or NOMAD), which is guided by a simulated nervous system that is an analogue of cortical and sub-cortical areas of the brain required for visual processing, decision-making, reward, and motor responses. These simulated cortical and sub-cortical areas are reentrantly connected and each area contains neuronal units representing both the mean activity level and the relative timing of the activity of groups of neurons. The brain-based device BBD learns to discriminate among multiple objects with shared visual features, and associated “target” objects with innately preferred auditory cues. Globally distributed neuronal circuits that correspond to distinct objects in the visual field of NOMAD <b>10</b> are activated. These circuits, which are constrained by a reentrant neuroanatomy and modulated by behavior and synaptic plasticity, result in successful discrimination of objects. The brain-based device BBD is moveable, in a rich real-world environment involving continual changes in the size and location of visual stimuli due to self-generated or autonomous, movement, and shows that reentrant connectivity and dynamic synchronization provide an effective mechanism for binding the features of visual objects so as to reorganize object features such as color, shape and motion while distinguishing distinct objects in the environment.
7,467,115MOBILE BRAIN-BASED DEVICE HAVING A SIMULATED NERVOUS SYSTEM BASED ON THE HIPPOCAMPUSGerald M. EdelmanA brain-based device (BBD) having a physical mobile device NOMAD controlling and under control by a simulated nervous system. The simulated nervous system is based on an intricate anatomy and physiology of the hippocampus and its surrounding neuronal regions including the cortex. The BBD integrates spatial signals from numerous objects in time and provides flexible navigation solutions to aid in the exploration of unknown environments. As NOMAD navigates in its real world environment, the hippocampus of the simulated nervous system organizes multi-modal input information received from sensors on NOMAD over timescales and uses this organization for the development of spatial and episodic memories necessary for navigation.