公开(公告)号：US20220109929A1

公开(公告)日：2022-04-07

申请号：US17553976

申请日：2021-12-17

Applicant: Amazon Technologies, Inc.

Inventor： Robert Ayrapetian ,  Philip Ryan Hilmes ,  Mohamed Mansour ,  Carlo Murgia

IPC: H04R3/00 ,  H04R5/027 ,  G06F3/16 ,  H04R5/04 ,  G10L21/0224 ,  G10L21/0208 ,  G10L21/02 ,  G10L21/0272

Abstract: Techniques for improving adaptive interference cancellation (AIC) using cascaded AIC algorithms are described. To improve an accuracy of detecting speech, a device may perform a first stage of AIC to generate isolated audio data and may generate speech mask data indicating time windows when speech is detected in the isolated audio data. Based on the speech mask data, the device may perform second AIC to generate output audio data, with adaptation of the adaptive filter enabled when the speech is not detected and disabled when the speech is detected. Thus, the first AIC improves the accuracy with which the device detects that speech is present and the second AIC reduces distortion in the output audio data by not updating filter coefficient values when the speech is present. The first AIC may use playback audio data, microphone audio data or beamformed audio data as reference signals.

公开(公告)号：US20210327442A1

公开(公告)日：2021-10-21

申请号：US17201843

申请日：2021-03-15

Applicant: Amazon Technologies, Inc.

Inventor： Yuan-Yen Tai ,  Mohamed Mansour ,  Parind Shah

IPC: G10L19/018 ,  G10L19/16 ,  G10L13/08 ,  G10L15/22 ,  G10L15/05

Abstract: A system may embed audio watermarks in audio data using an Eigenvector matrix. The system may detect audio watermarks in audio data despite the effects of reverberation. For example, the system may embed multiple repetitions of an audio watermark before generating output audio using loudspeaker(s). To detect the audio watermark in audio data generated by a microphone, the system may perform a self-correlation that indicates where the audio watermark is repeated. In some examples, the system may encode the audio watermark using multiple repetitions of a multi-segment Eigenvector. Additionally or alternatively, the system may encode the audio watermark using a binary sequence of positive and negative values, which may be used as a shared key for encoding/decoding the audio watermark. The audio watermark can be embedded in output audio data to enable wakeword suppression (e.g., avoid cross-talk between devices) and/or local signal transmission between devices in proximity to each other.

公开(公告)号：US20210134313A1

公开(公告)日：2021-05-06

申请号：US16934668

申请日：2020-07-21

Applicant: Amazon Technologies, Inc.

Inventor： Mohamed Mansour ,  Shobha Devi Kuruba Buchannagari

IPC: G10L21/0232 ,  G10L25/21 ,  G10L25/51

Abstract: A system configured to perform cascade echo cancellation processing to improve a performance when reference signals are asymmetric (e.g., dominant reference signal(s) overshadow weak reference signal(s)). The system may perform cascade echo cancellation processing to separately adapt filter coefficients between the dominant reference signal(s) and the weak reference signal(s). For example, the system may use a dominant reference signal to process a microphone audio signal and generate a residual audio signal, using the residual audio signal to adapt first filter coefficient values corresponding to the dominant reference signal. Separately, the system may use a weak reference signal to process the residual audio signal and generate an output audio signal, using the output audio signal to adapt second filter coefficient values corresponding to the weak reference signal.

公开(公告)号：US10950249B2

公开(公告)日：2021-03-16

申请号：US16141489

申请日：2018-09-25

Applicant: Amazon Technologies, Inc.

Inventor： Yuan-Yen Tai ,  Mohamed Mansour ,  Parind Shah

IPC: G10L19/018 ,  G10L19/16 ,  G10L13/08 ,  G10L15/22 ,  G10L15/05

Abstract: A system may embed audio watermarks in audio data using an Eigenvector matrix. The system may detect audio watermarks in audio data despite the effects of reverberation. For example, the system may embed multiple repetitions of an audio watermark before generating output audio using loudspeaker(s). To detect the audio watermark in audio data generated by a microphone, the system may perform a self-correlation that indicates where the audio watermark is repeated. In some examples, the system may encode the audio watermark using multiple repetitions of a multi-segment Eigenvector. Additionally or alternatively, the system may encode the audio watermark using a binary sequence of positive and negative values, which may be used as a shared key for encoding/decoding the audio watermark. The audio watermark can be embedded in output audio data to enable wakeword suppression (e.g., avoid cross-talk between devices) and/or local signal transmission between devices in proximity to each other.

公开(公告)号：US10657981B1

公开(公告)日：2020-05-19

申请号：US15982392

申请日：2018-05-17

Applicant: Amazon Technologies, Inc.

Inventor： Mohamed Mansour ,  Robert Ayrapetian

IPC: G10L21/0208 ,  H04R1/08 ,  H04M9/08 ,  G06F17/10 ,  G10L15/22 ,  H04R1/28 ,  G10L15/14 ,  G10L15/08

Abstract: Techniques for improving acoustic echo cancellation to attenuate an echo signal generated by a loudspeaker included in a device are described. A system may determine a loudspeaker canceling beam (LCB) (e.g., fixed beam directed to the loudspeaker) and may use the LCB to generate LCB audio data that corresponds to the echo signal. For example, based on a configuration of the loudspeaker relative to microphone(s) of the device, the system may perform simulation(s) to generate a plurality of filter coefficient values corresponding to the loudspeaker. By subtracting the LCB audio data during acoustic echo cancellation, the system may attenuate the echo signal even when there is distortion or nonlinearity or the like caused by the loudspeaker. In some examples, the system may perform acoustic echo cancellation using the LCB audio data and playback audio data.

公开(公告)号：US20200098379A1

公开(公告)日：2020-03-26

申请号：US16141489

申请日：2018-09-25

Applicant: Amazon Technologies, Inc.

Inventor： Yuan-Yen Tai ,  Mohamed Mansour ,  Parind Shah

IPC: G10L19/018 ,  G10L19/16 ,  G10L15/05 ,  G10L13/08 ,  G10L15/22

Abstract: A system may embed audio watermarks in audio data using an Eigenvector matrix. The system may detect audio watermarks in audio data despite the effects of reverberation. For example, the system may embed multiple repetitions of an audio watermark before generating output audio using loudspeaker(s). To detect the audio watermark in audio data generated by a microphone, the system may perform a self-correlation that indicates where the audio watermark is repeated. In some examples, the system may encode the audio watermark using multiple repetitions of a multi-segment Eigenvector. Additionally or alternatively, the system may encode the audio watermark using a binary sequence of positive and negative values, which may be used as a shared key for encoding/decoding the audio watermark. The audio watermark can be embedded in output audio data to enable wakeword suppression (e.g., avoid cross-talk between devices) and/or local signal transmission between devices in proximity to each other.

公开(公告)号：US10598543B1

公开(公告)日：2020-03-24

申请号：US15830622

申请日：2017-12-04

Applicant: Amazon Technologies, Inc.

Inventor： Mohamed Mansour ,  Srivatsan Kandadai ,  Stefan Vlaski

IPC: G01H7/00 ,  H04R1/32 ,  H04R3/00 ,  G01S15/10 ,  G01S7/527 ,  G10L21/0232 ,  G10L21/0208

Abstract: A system that performs wall detection, range estimation, corner detection and/or angular estimation. The system may determine an aggregate impulse response (e.g., impulse response of all components in a room) and may perform a deconvolution to remove a system impulse response (e.g., impulse response associated with loudspeaker(s) and microphone(s)). Thus, the system may use a sparse deconvolution algorithm to estimate a room impulse response (e.g., determine acoustic characteristics of the room). The system may detect a peak in the room impulse response and determine a distance and/or direction to an acoustically reflective surface based on the peak.

公开(公告)号：US10187721B1

公开(公告)日：2019-01-22

申请号：US15630424

申请日：2017-06-22

Applicant: Amazon Technologies, Inc.

Inventor： Mohamed Mansour

IPC: G10L25/21 ,  H04R1/40 ,  H04R3/00 ,  G10L21/0232 ,  G10L25/60 ,  G10L21/0216

Abstract: A beamformer system that can isolate a desired portion of an audio signal resulting from a microphone array. A fixed beamformer is used to dampen diffuse noise while an adaptive beamformer is used to cancel directional coherent noise. A gain is calculated using a signal quality value such as signal-to-noise ratio, signal-to-null ratio or other value. The adaptive beamformer output is adjusted by the gain prior to combining the fixed beamformer output and the adaptive beamformer output to determine the output audio data.