Amazon Technologies, Inc.

Google Inc.

Navin Chatlani

Amit Singh Chhetri

Robert Ayrapetian

Trausti Thor Kristjansson

Krishna Kamath Koteshwara

Srivatsan Kandadai

Chandan Karadagur Ananda Reddy

Ananth Raghavendra

Inseok Heo

Implementations described herein relate to providing noise suppression for speech data with reduced power consumption. In some implementations, a computer-implemented method includes receiving a current time frame of speech data, e.g., after receiving a previous time frame associated with a previous noise suppression mask. The current time frame is transformed to a current frequency frame in the frequency domain. A noise classifier is used to determine whether to create a current noise suppression mask for the current frame. If it is determined to create the mask, the mask is created and multiplied by the current frequency frame to obtain a noise-suppressed frequency frame. If it is determined to not create the current mask, the previous noise suppression mask is multiplied with the current frequency frame to obtain the noise-suppressed frequency frame, without creating a mask. The noise-suppressed frequency frame is transformed to a time frame and output.

Noise suppression for speech data with reduced power consumption

A device capable of autonomous motion may move in an environment and may receive audio data from a microphone. A model may be trained to process the audio data to suppress noise from the audio data. The model may include an encoder that includes one or more convolutional layers, one or more recurrent layers, and a decoder that includes one or more convolutional layers.

Autonomously motile device with noise suppression

A device capable of autonomous motion may move in an environment and may receive audio data from a microphone. A model may be trained to process the audio data to determine mask data, which may be used to mask noise in the audio data. Training data for the model may be normalized before training, and different loss functions may be used for different types of training data.

A computing device may receive audio data from a microphone representing audio in an environment of the device, which may correspond to an utterance and noise. A model may be trained to process the audio data to cancel noise from the audio data. The model may include an encoder that includes one or more dense layers, one or more recurrent layers, and a decoder that includes one or more dense layers.

Systems and methods for noise cancellation

A device capable of motion includes a beamformer for determining audio data corresponding to one or more directions. The beamformer includes a target beamformer that boosts audio from a target direction and a null beamformer that suppresses audio from that direction. When the device outputs sound while moving, the target and null beamformers capture and compensate for Doppler effects in output audio that reflects from nearby surfaces back to the device.

Autonomously motile device with adaptive beamforming

A device capable of moving a component of the device motion is capable of determining audio data corresponding to a direction relative to the device ('beamforming'). When the component moves, the device determines a new position of the component and selects one of a set of beamforming filter coefficients corresponding to the position. Using the filter coefficients, the device determines the directional audio data.

Autonomously motile device with beamforming

A system configured to improve echo cancellation for nonlinear systems. The system generate reference audio data by isolating portions of microphone audio data that correspond to playback audio data. For example, the system may determine a correlation between the playback audio data and the microphone audio data in individual time-frequency bands in a frequency domain. In some examples, the system may substitute microphone audio data associated with output audio for the playback audio data. The system may generate the reference audio data based on portions of the microphone audio data that have a strong correlation with the playback audio data. The system may generate the reference audio data by selecting these portions of the microphone audio data or by performing beamforming. This results in precise time alignment between the reference audio data and the microphone audio data, improving performance of the echo cancellation.

Growing community of inventors

Palo Alto, CA, United States of America

Navin Chatlani