Abstract: A sound generation apparatus includes sound collection means configured to collect a sound of a sound source in a space, image capture means configured to capture an image of the sound source, estimation means configured to estimate an attribute of the sound source from the image captured by the image capture means, sound generation means configured to obtain an acoustic characteristic of a target sound included in the sound collected by the sound collection means and to generate multiple masking sounds on the basis of the acoustic characteristic and the attribute of the sound source estimated by the estimation means, display means configured to display the attribute of the sound source estimated by the estimation means, sound selection means configured to receive selection of a masking sound from the masking sounds generated by the sound generation means, and sound output means configured to output the selected masking sound.

Abstract: An infant calming/sleep-aid device that includes a moving platform and a sound generator, the sound and motion adapted to calm a fussy baby, induce sleep, and maintain sleep under normal conditions. The device makes a determination as to whether sound signals represent sound coming from inside the device or outside the device. If the sound signals are coming from the inside the device, then the signals are evaluated in a specified frequency band to determine whether the sound is a baby cry. If a determination is made that there is a baby cry, then a threshold analysis is performed to quantify the cry and compare it to a threshold value. If the cry is above a specified threshold, the device moves the platform and/or generates sound.

Abstract: In some implementations, a user device may detect a communication associated with the user device and a user. The user device may analyze, using a machine learning model, a context of the communication to determine a risk score associated with the communication, the risk score indicating a likelihood that the communication is associated with a fraudulent actor attempting to solicit sensitive information from the user. The user device may monitor the communication to identify a trigger in the communication that indicates that sensitive information is to be shared, wherein monitoring the communication includes performing one or more operations to secure the communication based on the risk score satisfying one or more thresholds. The user device may detect, based on monitoring the communication, a portion of the communication that includes the sensitive information. The user device may obfuscate the portion of the communication.

Abstract: Methods and systems for determining periods of excessive noise for smart speaker voice commands. An electronic timeline of volume levels of currently playing content is made available to a smart speaker. From this timeline, periods of high content volume are determined, and the smart speaker alerts users during periods of high volume, requesting that they wait until the high-volume period has passed before issuing voice commands. In this manner, the smart speaker helps prevent voice commands that may not be detected, or may be detected inaccurately, due to the noise of the content currently being played.

Abstract: A video encoding system in which pixel data is decomposed into frequency bands prior to encoding. The frequency bands are organized into blocks that are provided to a block-based encoder that encodes the blocks and passes the encoded blocks to a wireless interface that packetizes the blocks for transmittal over a wireless connection. The encoder may categorize the encoded frequency bands into multiple priority levels, and may tag each frequency block with metadata indicating the frequency band represented in the block, the priority of the frequency band, and timing information. The wireless interface may then transmit or drop packets according to the priority levels of the encoded frequency blocks in the packets and/or according to the timing information of the frequency blocks in the packets.

Abstract: A method of training an audio separation network is provided. The method includes obtaining a first separation sample set, the first separation sample set including at least two types of audio with dummy labels, obtaining a first sample set by performing interpolation on the first separation sample set based on perturbation data, obtaining a second separation sample set by separating the first sample set using an unsupervised network, determining losses of second separation samples in the second separation sample set, and adjusting network parameters of the unsupervised network based on the losses of the second separation samples, such that a first loss of a first separation result outputted by an adjusted unsupervised network meets a convergence condition.

Abstract: Methods and systems for advanced stereo processing of an audio signal are disclosed. The methods and systems include selecting a coding mode of either transform coding or linear predictive coding and performing advanced stereo processing when in the selected coding mode. Both encoding and decoding operations are provided.

Abstract: Methods, systems, and media for processing queries relating to presented media content are provided.

Abstract: An example apparatus for reducing to reduce noise in audio includes a preprocessor to receive audio input from a microphone and preprocess the audio input to generate preprocessed audio. The apparatus also includes an acoustic event detector to detect an acoustic event corresponding to a disturbance in the preprocessed audio. The apparatus further includes a noise reduction model selector to select a specific disturbance model based on the detected acoustic event. The apparatus further includes a noise suppressor to attenuate components related to the disturbance in the preprocessed audio using the selected specific disturbance model to generate enhanced audio with suppressed noise.

Abstract: An action identification device acquires sound data from a microphone, calculates a feature amount of the sound data, determines whether or not a user is present in a space in which the microphone is installed, calculates a noise feature amount indicating a feature amount of noise based on the calculated feature amount and stores the calculated noise feature amount in a noise feature amount storage unit in a case where the user is not present in the space, subtracts the noise feature amount stored in the noise feature amount storage unit from the calculated feature amount to extract an action sound feature amount indicating a feature amount of an action sound generated by an action of the user in a case where the user is present in the space, and identifies an action of the user by using the action sound feature amount.

Abstract: A bitstream encoding or decoding job is broken up into a plurality of segments, each of which is independent from subsequent segments and corresponds to a respective fence identifier. The segments are individually processed and progress is indicated using the fence identifiers. In some cases, a first segment is encoded, transmitted, decoded, and processed before a second segment is encoded. As a result, in some cases, segment statuses are more easily tracked, hardware is used more efficiently, end-to-end processing time is reduced, and less communication network bandwidth is used.

Abstract: An audio infusion system and method are disclosed. A source audio track is separated into a plurality of audio tracks (e.g., instrumental, vocal, or mixes thereof) and the audio tracks are individually processed to generate a plurality of binaural beat tracks. At least one spatialized track is also generated by filtering the source audio track to provide a filtered track, generating one or more spatialization trajectories based on certain audio feature(s) of the source audio track (e.g., tempo) and a target end-state effect, and spatializing the filtered track using the spatialization trajectories. Other tracks may also be generated, such as one or more infrasonic tracks, ultrasonic tracks, enhanced bass tracks, and/or subharmonic tracks. The tracks may be played simultaneously or mixed for delivery to an end user device.

Abstract: To filter unwanted sounds from a conference call, a first voice signal is captured by a first device during a conference call and converted into corresponding text, which is then analyzed to determine that a first portion of the text was spoken by a first user and a second portion of the text was spoken by a second user. If the first user is relevant to the conference call while the second user is not, the first voice signal is prevented from being transmitted into the conference call, the first portion of text is converted into a second voice signal using a voice profile of the first user to synthesize the voice of the first user, and the second voice signal is then transmitted into the conference call. The second portion of text is not converted into a voice signal, as the second user is determined not to be relevant.

Abstract: There is provided mechanisms for frame loss concealment. A method is performed by a receiving entity. The method comprises adding, in association with constructing a substitution frame for a lost frame, a noise component to the substitution frame. The noise component has a frequency characteristic corresponding to a low-resolution spectral representation of a signal in a previously received frame.

Abstract: A computer-implemented method for suppressing noise from audio signal uses both statistical noise estimation and neural network noise estimation to achieve more desirable noise reduction. The method is performed by a noise suppression computer software application running on an electronic device. The noise suppression computer software application first transforms the speech signal in time domain into frequency domain before determining a statistical noise estimate and a neural network noise estimate. The noise suppression computer software application merges the two noise estimates to derive a final noise estimate, and determines and refines a noise suppression filter. The filter is applied to the speech signal in frequency domain to obtain an enhanced signal. The enhanced signal is transformed back into time domain.

Abstract: A method for multi-channel voice activity detection includes receiving a sequence of input frames characterizing streaming multi-channel audio captured by an array of microphones. Each channel of the streaming multi-channel audio includes respective audio features captured by a separate dedicated microphone. The method also includes determining, using a location fingerprint model, a location fingerprint indicating a location of a source of the multi-channel audio relative to the user device based on the respective audio features of each channel of the multi-channel audio. The method also includes generating an output from an application-specific classifier. The first score indicates a likelihood that the multi-channel audio corresponds to a particular audio type that the particular application is configured to process.

Abstract: A method for speech enhancement, the method may include receiving or generating sound samples that represent sound signals that were received during a given time period by an array of microphones; frequency transforming the sound samples to provide frequency-transformed samples; clustering the frequency-transformed samples to speakers to provide speaker related clusters, wherein the clustering is based on (i) spatial cues related to the received sound signals and (ii) acoustic cues related to the speakers; determining a relative transfer function for each speaker of the speakers to provide speakers related relative transfer functions; applying a multiple multiple output (MIMO) beamforming operation on the speakers related relative transfer functions to provide beamformed signals; and inverse-frequency transforming the beamformed signals to provide speech signals.

Abstract: A segment detecting device according to an embodiment includes at least one memory; and at least one processor. The at least one processor receives at least one of (i) an input signal including a first signal and a second signal or (ii) feature data representing one or a plurality of features of the input signal, estimates a level of the second signal by inputting the input signal or the feature data into a neural network, and determines a segment including the second signal in the input signal based on the level of the second signal.

Abstract: Systems and methods for creating a view of an environment are disclosed. Exemplary implementations may: receive parameters and measurements from at least two of one or more microphones, one or more imaging devices, a radar sensor, a lidar sensor, and/or one or more infrared imaging devices located in a computing device; analyze the parameters and measurements received from the multimodal input; generate a world map of the environment around the computing device; and repeat the receiving of parameters and measurements from the input devices and the analyzing steps on a periodic basis to maintain a persistent world map of the environment.

Abstract: A hearing aid includes an input providing an input signal representing sound in an environment, the input signal including no speech signal, or one or more speech signals from one or more speech sound sources and additional signal components, termed noise signal, from one or more other sound sources, an own voice detector, a voice activity detector, and a talker extraction unit to determine and/or receive one or more speech signals as separated one or more speech signals from speech sound sources other than the hearing aid user and to detect the speech signal originating from the voice of the user. The talker extraction unit provides separate signals, each including, or indicating presence of, one of the one or more speech signals. A noise reduction system determines speech overlap and/or gap between the speech signal originating from the user's voice and each of the separated one or more speech signals.

Abstract: Examples of the disclosure relate to apparatus, methods and computer programs for controlling noise reduction in audio signals including audio captured by a plurality of microphones. The apparatus includes circuitry for obtaining one or more audio signals wherein the one or more audio signals include audio captured by a plurality of microphones and dividing the obtained one or more audio signals into a plurality of intervals. The circuitry may also be configured for determining one or more parameters relating to one or more noise characteristics for different intervals and controlling noise reduction applied to the different intervals based on the determined one or more parameters within the different intervals.

Abstract: The present disclosure provides a method and apparatus for generating equivalent nonlinear noise. The method includes randomly generating white noise obeying zero-mean Chi-square distribution; and filtering the input white noise by using an equivalent nonlinear noise spectrum as an equivalent nonlinear filter to obtain the equivalent nonlinear noise. According to the embodiments of the present application, equivalent nonlinear noise may be generated based on zero-mean Chi-square distribution random variables combined with equivalent nonlinear noise spectrum filtering, and more accurate equivalent nonlinear noise than the related art may be generated, thus accurately evaluating an influence of nonlinearity of devices on Bit Error Rate (BER).

Abstract: A method, an apparatus, and logic to post-process raw gains determined by input processing to generate post-processed gains, comprising using one or both of delta gain smoothing and decision-directed gain smoothing. The delta gain smoothing comprises applying a smoothing filter to the raw gain with a smoothing factor that depends on the gain delta: the absolute value of the difference between the raw gain for the current frame and the post-processed gain for a previous frame. The decision-directed gain smoothing comprises converting the raw gain to a signal-to-noise ratio, applying a smoothing filter with a smoothing factor to the signal-to-noise ratio to calculate a smoothed signal-to-noise ratio, and converting the smoothed signal-to-noise ratio to determine the second smoothed gain, with smoothing factor possibly dependent on the gain delta.

Abstract: A speech signal processing method and apparatus. The method includes preprocessing a speech signal that is in a first frequency band and that is collected by an ear canal speech collector, to obtain a first speech signal; preprocessing a speech signal that is in a second frequency band and that is collected by at least one external speech collector, to obtain an external speech signal, where frequency ranges of the first frequency band and the second frequency band are different; performing correlation processing on the first speech signal and the external speech signal to obtain a second speech signal; and outputting a target speech signal, where the target speech signal includes the first speech signal and the second speech signal.

Abstract: An electronic device has one or more microphones that pick up a sound. At least one feature extractor processes the audio signals from the microphones, that contain the picked up the sound, to determine several features for the sound. The electronic device also includes a classifier that has a machine learning model which is configured to determine a sound classification, such as artificial versus natural for the sound, based upon at least one of the determined features. Other aspects are also described and claimed.

Abstract: An encoder for providing an audio stream on the basis of a transform-domain representation of an input audio signal includes a quantization error calculator configured to determine a multi-band quantization error over a plurality of frequency bands of the input audio signal for which separate band gain information is available. The encoder also includes an audio stream provider for providing the audio stream such that the audio stream includes information describing an audio content of the frequency bands and information describing the multi-band quantization error. A decoder for providing a decoded representation of an audio signal on the basis of an encoded audio stream representing spectral components of frequency bands of the audio signal includes a noise filler for introducing noise into spectral components of a plurality of frequency bands to which separate frequency band gain information is associated on the basis of a common multi-band noise intensity value.

Abstract: Example embodiments of the present disclosure relate to devices, methods, apparatuses and computer readable storage media of extended bandwidth allocation. In example embodiments, a first device receives, from a second device, an indication of a reference bandwidth. The first device receives, from the second device, an indication of a bandwidth extended from the reference bandwidth and receives, from the second device, an indication of at least one frequency offset of a first set of narrowbands with respect to a second set of narrowbands associated with the reference bandwidth. Then, the first device determines, based on the reference bandwidth, the extended bandwidth and the at least one frequency offset, the first set of narrowbands for communication with the second device.

Abstract: An encoder for providing an audio stream on the basis of a transform- domain representation of an input audio signal includes a quantization error calculator configured to determine a multi-band quantization error over a plurality of frequency bands of the input audio signal for which separate band gain information is available. The encoder also includes an audio stream provider for providing the audio stream such that the audio stream includes information describing an audio content of the frequency bands and information describing the multi-band quantization error. A decoder for providing a decoded representation of an audio signal on the basis of an encoded audio stream representing spectral components of frequency bands of the audio signal includes a noise filler for introducing noise into spectral components of a plurality of frequency bands to which separate frequency band gain information is associated on the basis of a common multi-band noise intensity value.

Abstract: A sound field control apparatus includes a microphone configured to receive an utterance of a user, an output interface configured to output at least one of a sound signal and image data, and one or more processors configured to cancel a sound signal in a specific area around the microphone, obtain room impulse response information based on a user utterance position when the utterance of the user is received, and output a sound signal for providing an independent sound field to the user based on the room impulse response information.

Abstract: A speech noise reduction processing method, an apparatus, a computer device and a storage medium. The method comprises: responsive to a distance between a speech collection device and a target object being detected to reach a preset value, acquiring a noisy speech signal collected by the speech collection device and performing frequency division processing on the noisy speech signal to obtain a low-frequency band signal; acquiring an amplitude spectrum and a phase spectrum of the low-frequency band signal; acquiring a modulation domain signal corresponding to the amplitude spectrum; performing spectral subtraction to obtain a noise-reduced modulation domain amplitude spectrum; compensating the modulation domain phase spectrum to obtain the compensated modulation domain phase spectrum; and obtaining a noise-reduced low-frequency band signal based on the compensated modulation domain phase spectrum, the noise-reduced modulation domain amplitude spectrum and the phase spectrum of the low-frequency band signal.

Abstract: A takeoff location and a landing location are received for an autonomous vertical takeoff and landing (VTOL) vehicle that includes a plurality of rotors. An autonomous and noise-reduced flight trajectory for the autonomous VTOL vehicle is determined based at least in part on the takeoff location, the landing location, a jerk function, and a noise function, including by minimizing the jerk function and minimizing the noise function. A set of one or more desired forces or moments is determined for the autonomous VTOL vehicle based at least in part on autonomous and noise-reduced flight trajectory. A plurality of motor control signals is determined for the plurality of rotors based at least in part on the set of one or more desired forces or moments.

Abstract: A system, method and computer readable medium for dynamic noise reduction in a voice call. The system includes an encoder having a short-time Fourier transform module to determine a magnitude spectrum and a phase spectrum of an input audio signal, including speech and dynamic noise. A separator coupled to the encoder comprises a temporal convolution network (TCN) used to develop a separation mask using the magnitude spectrum as input. The TCN is trained using a frequency SNR function used to calculate loss during training. A mixer is coupled to the separator to multiply the separation mask with the magnitude spectrum to separate the speech from the dynamic noise to obtain a denoise magnitude spectrum. A decoder coupled to the mixer and the encoder includes an inverse short-time Fourier transform module to reconstruct the input audio signal without the dynamic noise using the denoise magnitude spectrum and the phase spectrum.

Abstract: Disclosed is a method, an apparatus, and a device for transient noise detection. The method includes: obtaining an audio frame signal having a preset duration; performing wavelet decomposition on a first audio frame signal to obtain a first wavelet decomposition signal corresponding to the first audio frame signal; determining a first reference audio intensity value of a first sub-wavelet decomposition signal according to reference audio intensity values of all samples in the first sub-wavelet decomposition signal; determining energy distribution information of the first wavelet decomposition signal according to first reference audio intensity values of all sub-wavelet decomposition signals in the first wavelet decomposition signal; and determining a probability that the first audio frame signal is transient noise according to the energy distribution information of the first wavelet decomposition signal.

Abstract: This application provides an information play control method and apparatus, an electronic device, a computer-readable storage medium and a computer program product, and relates to the artificial intelligence technology. The method includes displaying play rate options in response to a triggering operation for a rate control identifier of a play interface, the play rate options comprising a smart adjustment option, the smart adjustment option being used for triggering play rate adjustment for multimedia information; determining a target rate ratio of the multimedia information based on content of the multimedia information in response to a triggering operation for the smart adjustment option; adjusting a play rate of the multimedia information based on the target rate ratio to obtain adjusted multimedia information; and playing the adjusted multimedia information.

Abstract: Provided is an information processing apparatus including a control section that controls operations of operation bodies in accordance with a result of a voice recognition process. In accordance with a result of a voice recognition process that is based on a voice collected by one of the operation bodies or a voice recognition environment recognized from sensor information collected by one of the operation bodies, the control section controls an operation of another one of the operation bodies.

Abstract: A transmission apparatus recognition apparatus includes a storage unit that stores K sets of template feature groups for estimating K (an integer of 2 or more) kinds of information indicative of a transmission apparatus, a degree-of-similarity calculation unit that generates an i (an integer of 1 to K)-th sample feature from a radio feature, and calculates an i-th degree-of-similarity group, based on the i-th sample feature and an i-th set of the template feature group, a summed degree-of-similarity calculation unit that calculates a summed degree of similarity by summing K degrees of similarity by using an i-th weighting factor with respect to 1 to K of i, and an estimation unit that estimates that K information pieces, which are correlated in advance with calculation sources of K degrees of similarity having the summed degree of similarity that is highest, are information indicative of the transmission apparatus.

Abstract: Provided are an electronic apparatus and a controlling method thereof. The electronic apparatus includes an inputter and a processor configured to, based on receiving an audio signal through the inputter, obtain a speech intelligibility for the audio signal, and modify the audio signal so that the speech intelligibility becomes a target intelligibility that is set based on scene information regarding a type of audio included in the audio signal, and the type of audio includes at least one of a sound effect, shouting, music, or a speech.

Abstract: A method to combine contact and acoustic microphones in a headset for voice wake and voice processing in immersive reality applications is provided. The method includes receiving, from a contact microphone, a first acoustic signal, determining a fidelity and a quality of the first acoustic signal, receiving, from an acoustic microphone, a second acoustic signal, and when the fidelity and quality of the first acoustic signal exceeds a pre-selected threshold, combining the first acoustic signal and the second acoustic signal to provide an enhanced acoustic signal to a smart glass user. A non-transitory, computer-readable medium storing instructions to cause a headset to perform the above method, and the headset, are also provided.

Abstract: A speech processing device includes a processor. The processor performs operations including: detecting a single-talk state based on a speech signal collected by each of microphones, the single-talk state in which any one of persons speaks; estimating a mixing rate indicating a ratio of a speech signal of the main speaking person to a speech signal of another person based on a sound pressure ratio of the speech signals collected by the microphones in the single-talk state of the main speaking person and a sound pressure ratio of the speech signals collected by the plurality of microphones in the single-talk state of the another person; and determining whether suppression of a crosstalk component due to speaking of the another person contained in the speech signal of the main speaking person is necessary based on an estimation result of the mixing rate.

Abstract: The present disclosure discloses an audio encoding and decoding method and an audio encoder and decoder. The audio encoding method includes: obtaining a current frame of an audio signal, where the current frame includes a high frequency band signal and a low frequency band signal; obtaining a first encoding parameter based on the high frequency band signal and the low frequency band signal; obtaining a second encoding parameter of the current frame based on the high frequency band signal, where the second encoding parameter includes tone component information; and performing bitstream multiplexing on the first encoding parameter and the second encoding parameter, to obtain an encoded bitstream.

Abstract: Methods, apparatus, systems, and articles of manufacture are disclosed to fingerprint an audio signal via exponential normalization. An example apparatus includes an audio segmenter to divide an audio signal into a plurality of audio segments including a first audio segment and a second audio segment, the first audio segment including a first time-frequency bin, the second audio segment including a second time-frequency bin, a mean calculator to determine a first exponential mean value associated with the first time frequency bin based on a first magnitude of the audio signal associated with the first time frequency bin and a second exponential mean value associated with the second time frequency bin based on a second magnitude of the audio signal associated with the second time frequency bin and the first exponential mean value.

Abstract: An audio modification system includes one or more processors configured to receive audio data indicative of communication of a user while the user is positioned in a first portion of a control area, determine that the communication of the user is confidential based on the audio data, determine generated audio data configured to render the communication of the user at least partially inaudible in a second portion of the control area, and output a control signal indicative of instructions to provide the generated audio data.

Abstract: Provided is an echo cancellation apparatus capable of calculating an acoustic coupling amount with high accuracy regardless of the magnitude of the near-end speaker component and without using a double talk detector. The echo cancellation apparatus cancels an echo included in a sound pickup signal picked up by a microphone placed at a near-end and includes an acoustic coupling amount calculation unit that updates and calculates an acoustic coupling amount estimated value of a component of a reproduction signal, which is a signal picked up by a microphone placed at a far-end included in the sound pickup signal, such that an update amount is decreased the greater a magnitude of a component other than an echo component is in the sound pickup signal; a gain calculation unit that calculates a gain coefficient on the basis of the acoustic coupling amount estimated value; and an integration unit that integrates the gain coefficient with the sound pickup signal and generates an echo cancellation signal.

Abstract: An apparatus for encoding audio information is provided. The apparatus for encoding audio information includes a selector for selecting a comfort noise generation mode from two or more comfort noise generation modes depending on a background noise characteristic of an audio input signal, and an encoding unit for encoding the audio information, wherein the audio information includes mode information indicating the selected comfort noise generation mode.

Abstract: Embodiments of this disclosure provide a speech enhancement method and apparatus, an electronic device, and a computer-readable storage medium. The method includes: obtaining a clean speech sample; decomposing the clean speech sample to obtain a first sparse matrix and m base matrices, values in the first sparse matrix being all positive numbers, and m being a positive integer greater than 1; obtaining, according to the first sparse matrix and a weight matrix of a target neural network, state vectors of neurons in a visible layer of the target neural network; and updating the weight matrix according to the state vectors of the neurons in the visible layer and the clean speech sample, to obtain a deep dictionary used for speech enhancement.

Abstract: The present technology relates to an information processing device, an information processing method, and a program that enable easier visualization of a sound field. The information processing device includes: a calculation unit configured to calculate, on the basis of positional information indicating a plurality of observation positions in a space and an audio signal of sound observed at each of the plurality of observation positions, an amplitude or phase of the sound at each of a plurality of positions in the space at a first time interval; an event detection unit configured to detect an event; and an adjustment unit configured to perform, in a case where the event is detected, control such that the amplitude or the phase is calculated at a second time interval shorter than the first time interval. The present technology can be applied to a sound-field visualization system.

Abstract: A method and a device for audio repair and a readable storage medium are provided. The method includes the following. Multiple audio frames are sequentially inputted into a cache module, where the cache module is sequentially composed of multiple processing units, and a processing unit located at a center of the multiple processing units is a center processing unit (201). At least one audio frame contained in the center processing unit is assigned as a target frame (202). A noise point presented as a short-term high-energy pulse in the target frame is detected according to audio characteristics of the multiple audio frames in the cache module (203). The target frame is repaired to remove the noise point in the target frame (204).

Abstract: A noise suppression device transforms observation signals to spectral components of multiple channels, calculates an arrival time difference, calculates weight coefficients based on the arrival time difference, estimates whether each of the spectral components of the plurality of frames is a spectral component of target sound or not, estimates a weighted S/N ratio of each of the spectral components of the plurality of frames based on the result of the estimation and the weight coefficients, calculates gains of the spectral components of the plurality of frames by using the weighted S/N ratios, outputs spectral components of an output signal by suppressing spectral components of observation signals of sounds other than the target sound in the spectral components of the plurality of frames by using the gains, and transforms the spectral components of the output signal to an output signal in a time domain.

Abstract: An audio signal processing method implemented by an audio system with an audio sensor and a speaker unit includes: measuring, by the audio sensor, acoustic signals reaching the audio sensor, producing a sensor audio signal; retrieving a speaker audio signal corresponding to a speaker acoustic signal from the speaker unit while measuring the acoustic signals reaching the audio sensor to produce the sensor audio signal; converting the speaker and sensor audio signals to speaker and sensor audio spectra; estimating, based on the speaker audio spectrum, an echo audio spectrum of an echo audio signal caused by the speaker acoustic signal in the sensor audio signal; computing, based on the echo audio spectrum and the sensor audio spectrum, echo suppression gains to be applied to the sensor audio spectrum, by using a MMSE-LSA estimator; and applying the echo suppression gains to the sensor audio spectrum.

Abstract: Techniques are described for selectively adapting and/or selectively utilizing a noise reduction technique in detection of one or more features of a stream of audio data frames. For example, various techniques are directed to selectively adapting and/or utilizing a noise reduction technique in detection of an invocation phrase in a stream of audio data frames, detection of voice characteristics in a stream of audio data frames (e.g., for speaker identification), etc. Utilization of described techniques can result in more robust and/or more accurate detections of features of a stream of audio data frames in various situations, such as in environments with strong background noise. In various implementations, described techniques are implemented in combination with an automated assistant, and feature(s) detected utilizing techniques described herein are utilized to adapt the functionality of the automated assistant.