Abstract: Embodiments of this application disclose method and apparatus for positioning a target audio signal by an audio interaction device, and an audio interaction device The method includes: obtaining audio signals in a plurality of directions in a space, and performing echo cancellation on the audio signal, the audio signal including a target-audio direct signal; obtaining weights of a plurality of time-frequency points in the audio signals, a weight of each time-frequency point indicating, at the time-frequency point, a relative proportion of the target-audio direct signal in the audio signals; weighting time-frequency components of the audio signal at the plurality of time-frequency points separately for each of the plurality of directions by using the weights of the plurality of time-frequency points, to obtain a weighted audio signal energy distribution; and obtaining a sound source azimuth corresponding to the target-audio direct signal in the audio signals accordingly.

Abstract: Embodiments of this application discloses an azimuth estimation method performed at a computing device, the method including: obtaining, in real time, multi-channel sampling signals and buffering the multi-channel sampling signals; performing wakeup word detection on one or more sampling signals of the multi-channel sampling signals, and determining a wakeup word detection score for each channel of the one or more sampling signals; performing a spatial spectrum estimation on the buffered multi-channel sampling signals to obtain a spatial spectrum estimation result, when the wakeup word detection scores of the one or more sampling signals indicates that a wakeup word exists in the one or more sampling signals; and determining an azimuth of a target voice associated with the multi-channel sampling signals according to the spatial spectrum estimation result and a highest wakeup word detection score, thereby improving the accuracy of the azimuth estimation in a voice interaction process.

Abstract: A keyword detection method includes: obtaining an enhanced speech signal of a to-be-detected speech signal, the enhanced speech signal corresponding to a target speech speed; performing speed adjustment on the enhanced speech signal to obtain a first speed-adjusted speech signal having a first speech speed, the first speech speed being different from the target speech speed; obtaining a first speech feature signal according to the first speed-adjusted speech signal; obtaining a detection result according to a first keyword detection result corresponding to the first speech feature signal, the detection result indicating whether a target keyword exists in the to-be-detected speech signal; and performing an operation corresponding to the target keyword in response to determining that the target keyword exists according to the detection result.

Abstract: A system configured to perform microphone occlusion event detection. When a device detects a microphone occlusion event, the device will modify audio processing performed prior to speech processing, such as by disabling spatial processing and only processing audio data from a single microphone. The device detects the microphone occlusion event by determining inter-level difference (ILD) values between two microphone signals and using the ILD values as input features to a classifier. For example, when a far-end reference signal is inactive, the classifier may process a first ILD value within a high frequency band. However, when the far-end reference signal is active, the classifier may process the first ILD value and a second ILD value within a low frequency band.

Abstract: A keyword detection method includes: obtaining an enhanced speech signal of a to-be-detected speech signal, the enhanced speech signal corresponding to a target speech speed; performing speed adjustment on the enhanced speech signal to obtain a first speed-adjusted speech signal having a first speech speed, the first speech speed being different from the target speech speed; obtaining a first speech feature signal according to the first speed-adjusted speech signal; obtaining a detection result according to a first keyword detection result corresponding to the first speech feature signal, the detection result indicating whether a target keyword exists in the to-be-detected speech signal; and performing an operation corresponding to the target keyword in response to determining that the target keyword exists according to the detection result.

Abstract: Embodiments of this application disclose method and apparatus for positioning a target audio signal by an audio interaction device, and an audio interaction device The method includes: obtaining audio signals in a plurality of directions in a space, and performing echo cancellation on the audio signal, the audio signal including a target-audio direct signal; obtaining weights of a plurality of time-frequency points in the audio signals, a weight of each time-frequency point indicating, at the time-frequency point, a relative proportion of the target-audio direct signal in the audio signals; weighting time-frequency components of the audio signal at the plurality of time-frequency points separately for each of the plurality of directions by using the weights of the plurality of time-frequency points, to obtain a weighted audio signal energy distribution; and obtaining a sound source azimuth corresponding to the target-audio direct signal in the audio signals accordingly.

Abstract: An apparatus includes multiple microphones to generate audio signals based on sound of a far-field acoustic environment. The apparatus also includes a signal processing system to process the audio signals to generate at least one processed audio signal. The signal processing system is configured to update one or more processing parameters while operating in a first operational mode and is configured to use a static version of the one or more processing parameters while operating in the second operational mode. The apparatus further includes a keyword detection system to perform keyword detection based on the at least one processed audio signal to determine whether the sound includes an utterance corresponding to a keyword and, based on a result of the keyword detection, to send a control signal to the signal processing system to change an operational mode of the signal processing system.

Abstract: A method for multi-channel echo cancellation and noise suppression is described. One of multiple echo estimates is selected for non-linear echo cancellation. Echo notch masking is performed on a noise-suppressed signal based on an echo direction of arrival (DOA) to produce an echo-suppressed signal. Non-linear echo cancellation is performed on the echo-suppressed signal based, at least in part, on the selected echo estimate.

Abstract: An apparatus includes multiple microphones to generate audio signals based on sound of a far-field acoustic environment. The apparatus also includes a signal processing system to process the audio signals to generate at least one processed audio signal. The signal processing system is configured to update one or more processing parameters while operating in a first operational mode and is configured to use a static version of the one or more processing parameters while operating in the second operational mode. The apparatus further includes a keyword detection system to perform keyword detection based on the at least one processed audio signal to determine whether the sound includes an utterance corresponding to a keyword and, based on a result of the keyword detection, to send a control signal to the signal processing system to change an operational mode of the signal processing system.

Abstract: A method for audio signal processing is described. The method includes decomposing a recorded auditory scene into a first category of localizable sources and a second category of ambient sound. The method also includes recording an indication of the directions of each of the localizable sources. The method may be performed with a device having a microphone array.

Abstract: A method includes receiving, at a processor, a first data frame at a first time from a first microphone. The method also includes receiving a second data frame at the first time from a second microphone. The method further includes calculating a power ratio of the first microphone and the second microphone based on the first data frame and the second data frame in response to determining that the first data frame and the second data frame are noise data frames.

Abstract: Implementations and applications are disclosed for detection of a transition in a voice activity state of an audio signal, based on a change in energy that is consistent in time across a range of frequencies of the signal. For example, such detection may be based on a time derivative of energy for each of a number of different frequency components of the signal.

Abstract: A method of orientation-sensitive recording control includes indicating, within a portable device and at a first time, that the portable device has a first orientation relative to a gravitational axis and, based on the indication, selecting a first pair among at least three microphone channels of the portable device. This method also includes indicating, within the portable device and at a second time that is different than the first time, that the portable device has a second orientation relative to the gravitational axis that is different than the first orientation and, based on the indication, selecting a second pair among the at least three microphone channels that is different than the first pair. In this method, each the at least three microphone channels is based on a signal produced by a corresponding one of at least three microphones of the portable device.

Abstract: A multi-microphone system performs location-selective processing of an acoustic signal, wherein source location is indicated by directions of arrival relative to microphone pairs at opposite sides of a midsagittal plane of a user's head.

Abstract: Systems, methods, apparatus, and machine-readable media for voice activity detection in a single-channel or multichannel audio signal are disclosed.

Abstract: A disclosed method selects a plurality of fewer than all of the channels of a multichannel signal, based on information relating to the direction of arrival of at least one frequency component of the multichannel signal.

Abstract: A method includes receiving, at a processor, a first data frame at a first time from a first microphone. The method also includes receiving a second data frame at the first time from a second microphone. The method further includes calculating a power ratio of the first microphone and the second microphone based on the first data frame and the second data frame in response to determining that the first data frame and the second data frame are noise data frames.

Abstract: A method for echo cancellation and noise suppression is disclosed. Linear echo cancellation (LEC) is performed for a primary microphone channel on an entire frequency band or in a range of frequencies where echo is audible. LEC is performed on one or more secondary microphone channels only on a lower frequency range over which spatial processing is effective. The microphone channels are spatially processed over the lower frequency range after LEC. Non-linear noise suppression post-processing is performed on the entire frequency band. Echo post-processing is performed on the entire frequency band.

Abstract: Based on phase differences between corresponding frequency components of different channels of a multichannel signal, a measure of directional coherency is calculated. Application of such a measure to voice activity detection and noise reduction are also disclosed.

Abstract: Phase-based processing of a multichannel signal, and applications including proximity detection, are disclosed.