The present disclosure relates to a method for performing inference on input data using a neural network and a processing device employing the aforementioned method. The method comprises the steps of obtaining and storing input data, obtaining parameter data indicating the parameters of the first layer and storing the parameter data in a parameter data storage location and processing the input data using the first layer parameter data, to form first layer output data. The method further comprises storing the first layer output data, obtaining parameter data of the second layer and storing the second layer parameter data by replacing the first layer parameter data with the second layer parameter data, processing the first layer output data using the stored second layer parameter data to form second layer output data; and storing the second layer output data.

A system configured to perform deep adaptive acoustic echo cancellation (AEC) to improve audio processing. Due to mechanical noise and continuous echo path changes caused by movement of a device, echo signals are nonlinear and time-varying and not fully canceled by linear AEC processing alone. To improve echo cancellation, deep adaptive AEC processing integrates a deep neural network (DNN) and linear adaptive filtering to perform echo and/or noise removal. The DNN is configured to generate a nonlinear reference signal and step-size data, which the linear adaptive filtering uses to generate output audio data representing local speech. The DNN may generate the nonlinear reference signal by generating mask data that is applied to a microphone signal, such that the reference signal corresponds to a portion of the microphone signal that does not include near-end speech.

An acoustic device includes: an imaging device configured to take a sample image of a space as a sound field and create an image data on the space based on the taken sample image; a sound collector configured to collect a sound generated in the space or to collect a previously-collected acoustic data therein; and a computation part configured to previously compute a plurality of parameters relevant to a coefficient of spatial acoustic filter corresponding to the sample image of the space and previously learn a sound field model of the space shown in the sample image. The computation part is configured to construct a sound field model of the sample image taken by the imaging device or of a previously-taken sample image, from the acoustic data collected by the sound collector, using the coefficient of spatial acoustic filter.

Implementations of the subject technology provide systems and methods for determining whether to interrupt a user of an audio device that is operating in a noise-cancelling mode of operation. For example, the user may desire to be interrupted by one or more pre-designated contacts that are identified at an associated electronic device as interrupt-authorized contacts, or by a person who speaks a designated keyword to the user.

The present disclosure provides an acoustic device including a microphone array, a processor, and at least one speaker. The microphone array may be configured to acquire an environmental noise. The processor may be configured to estimate a sound field at a target spatial position using the microphone array. The target spatial position may be closer to an ear canal of a user than each microphone in the microphone array. The processor may be configured to generate a noise reduction signal based on the environmental noise and the sound field estimation of the target spatial position. The at least one speaker may be configured to output a target signal based on the noise reduction signal. The target signal may be used to reduce the environmental noise. The microphone array may be arranged in a target area to minimize an interference signal from the at least one speaker to the microphone array.

An active noise reduction system includes a reference signal generator configured to generate a reference signal, a canceling sound generator configured to generate a canceling sound, an error detector configured to detect an error between a noise and the canceling sound and generate an error signal corresponding to the error, and a controller configured to control the canceling sound generator based on the reference signal and the error signal, wherein the controller is configured to update an estimation value of acoustic characteristics in an internal space of a mobile body based on the reference signal and the error signal, estimate a head position of an occupant in the internal space based on the updated estimation value of the acoustic characteristics, and update a control filter based on the estimated head position of the occupant, the control filter being a filter for controlling the canceling sound generator.

A method for hybrid noise suppression includes receiving a processed audio signal from an audio device. The processed audio signal results from a partial audio processing performed on a noisy audio input signal. The method further includes predicting a noise suppression parameter using a neural network model operating on the processed audio signal and generating a noise-suppressed audio signal from the processed audio signal, using the noise suppression parameter. The method further includes generating a noise-suppressed audio output signal from the noise-suppressed audio signal using an additional audio processing and outputting the noise-suppressed audio output signal.

Various implementations include systems for processing audio signals to remove artifacts introduced by a machine learning system in challenging environments. In particular implementations, a method includes generating a processed audio signal for a hearing assistance device in which the processed audio signal is intended to perceptually dominate a user auditory experience, including: processing an unprocessed audio signal received by the hearing assistance device, wherein the processing includes utilizing a machine learning (ML) system to generate an ML enhanced audio signal; determining a mixing coefficient from an environmental noise assessment; mixing the ML enhanced audio signal with the unprocessed audio signal using the mixing coefficient to generate the processed audio signal; and outputting the processed audio signal.

A method for hybrid noise suppression includes receiving a processed audio signal from an audio device. The processed audio signal results from a partial audio processing performed on a noisy audio input signal. The method further includes predicting a noise suppression parameter using a neural network model operating on the processed audio signal and generating a noise-suppressed audio signal from the processed audio signal, using the noise suppression parameter. The method further includes generating a noise-suppressed audio output signal from the noise-suppressed audio signal using an additional audio processing and outputting the noise-suppressed audio output signal.

The present disclosure provides an acoustic device including a microphone array, a processor, and at least one speaker. The microphone array may be configured to acquire an environmental noise. The processor may be configured to estimate a sound field at a target spatial position using the microphone array. The target spatial position may be closer to an ear canal of a user than each microphone in the microphone array. The processor may be configured to generate a noise reduction signal based on the environmental noise and the sound field estimation of the target spatial position. The at least one speaker may be configured to output a target signal based on the noise reduction signal. The target signal may be used to reduce the environmental noise. The microphone array may be arranged in a target area to minimize an interference signal from the at least one speaker to the microphone array.