公开(公告)号：US12190875B1

公开(公告)日：2025-01-07

申请号：US17490572

申请日：2021-09-30

Applicant: Amazon Technologies, Inc.

Inventor： Eli Joshua Fidler ,  Aaron Challenner ,  Zoe Adams ,  Sree Hari Krishnan Parthasarathi ,  Gengshen Fu

IPC: G10L15/00 ,  G10L15/02 ,  G10L15/22 ,  G10L15/08 ,  G10L15/187

Abstract: Systems and methods for preemptive wakeword detection are disclosed. For example, a first part of a wakeword is detected from audio data representing a user utterance. When this occurs, on-device speech processing is initiated prior to when the entire wakeword is detected. When the entire wakeword is detected, results from the on-device speech processing and/or the audio data is sent to a speech processing system to determine a responsive action to be performed by the device. When the entire wakeword is not detected, on-device processing is canceled and the device refrains from sending the audio data to the speech processing system.

公开(公告)号：US20170270919A1

公开(公告)日：2017-09-21

申请号：US15196228

申请日：2016-06-29

Applicant: Amazon Technologies, Inc.

Inventor： Sree Hari Krishnan Parthasarathi ,  Bjorn Hoffmeister ,  Brian King ,  Roland Maas

IPC: G10L15/20 ,  G10L15/16 ,  G10L25/87 ,  G10L15/08 ,  G10L15/02 ,  G10L17/06

CPC classification number: G10L15/20 ,  G10L15/02 ,  G10L15/08 ,  G10L15/16 ,  G10L17/02 ,  G10L17/06 ,  G10L17/18 ,  G10L25/87 ,  G10L2015/088 ,  G10L2025/783

Abstract: A system configured to process speech commands may classify incoming audio as desired speech, undesired speech, or non-speech. Desired speech is speech that is from a same speaker as reference speech. The reference speech may be obtained from a configuration session or from a first portion of input speech that includes a wakeword. The reference speech may be encoded using a recurrent neural network (RNN) encoder to create a reference feature vector. The reference feature vector and incoming audio data may be processed by a trained neural network classifier to label the incoming audio data (for example, frame-by-frame) as to whether each frame is spoken by the same speaker as the reference speech. The labels may be passed to an automatic speech recognition (ASR) component which may allow the ASR component to focus its processing on the desired speech.

公开(公告)号：US20250149036A1

公开(公告)日：2025-05-08

申请号：US18966827

申请日：2024-12-03

Applicant: Amazon Technologies, Inc.

Inventor： Eli Joshua Fidler ,  Aaron Challenner ,  Zoe Adams ,  Sree Hari Krishnan Parthasarathi ,  Gengshen Fu

IPC: G10L15/22 ,  G10L15/02 ,  G10L15/08 ,  G10L15/187

Abstract: Systems and methods for preemptive wakeword detection are disclosed. For example, a first part of a wakeword is detected from audio data representing a user utterance. When this occurs, on-device speech processing is initiated prior to when the entire wakeword is detected. When the entire wakeword is detected, results from the on-device speech processing and/or the audio data is sent to a speech processing system to determine a responsive action to be performed by the device. When the entire wakeword is not detected, on-device processing is canceled and the device refrains from sending the audio data to the speech processing system.

公开(公告)号：US20200035231A1

公开(公告)日：2020-01-30

申请号：US16437763

申请日：2019-06-11

Applicant: Amazon Technologies, Inc.

Inventor： Sree Hari Krishnan Parthasarathi ,  Bjorn Hoffmeister ,  Brian King ,  Roland Maas

IPC: G10L15/20 ,  G10L15/02 ,  G10L17/06 ,  G10L25/87 ,  G10L15/08 ,  G10L15/16 ,  G10L17/18

Abstract: A system configured to process speech commands may classify incoming audio as desired speech, undesired speech, or non-speech. Desired speech is speech that is from a same speaker as reference speech. The reference speech may be obtained from a configuration session or from a first portion of input speech that includes a wakeword. The reference speech may be encoded using a recurrent neural network (RNN) encoder to create a reference feature vector. The reference feature vector and incoming audio data may be processed by a trained neural network classifier to label the incoming audio data (for example, frame-by-frame) as to whether each frame is spoken by the same speaker as the reference speech. The labels may be passed to an automatic speech recognition (ASR) component which may allow the ASR component to focus its processing on the desired speech.

公开(公告)号：US10373612B2

公开(公告)日：2019-08-06

申请号：US15196228

申请日：2016-06-29

Applicant: Amazon Technologies, Inc.

Inventor： Sree Hari Krishnan Parthasarathi ,  Bjorn Hoffmeister ,  Brian King ,  Roland Maas

IPC: G10L15/02 ,  G10L15/08 ,  G10L15/16 ,  G10L15/20 ,  G10L17/02 ,  G10L17/06 ,  G10L17/18 ,  G10L25/78 ,  G10L25/87

Abstract: A system configured to process speech commands may classify incoming audio as desired speech, undesired speech, or non-speech. Desired speech is speech that is from a same speaker as reference speech. The reference speech may be obtained from a configuration session or from a first portion of input speech that includes a wakeword. The reference speech may be encoded using a recurrent neural network (RNN) encoder to create a reference feature vector. The reference feature vector and incoming audio data may be processed by a trained neural network classifier to label the incoming audio data (for example, frame-by-frame) as to whether each frame is spoken by the same speaker as the reference speech. The labels may be passed to an automatic speech recognition (ASR) component which may allow the ASR component to focus its processing on the desired speech.

公开(公告)号：US11853391B1

公开(公告)日：2023-12-26

申请号：US16139607

申请日：2018-09-24

Applicant: Amazon Technologies, Inc.

Inventor： Pranav Prashant Ladkat ,  Oleg Rybakov ,  Nikko Strom ,  Sri Venkata Surya Siva Rama Krishna Garimella ,  Sree Hari Krishnan Parthasarathi

IPC: G06F18/214 ,  G06N20/00

CPC classification number: G06F18/2148 ,  G06N20/00

Abstract: Exemplary embodiments provide distributed parallel training of a machine learning model. Multiple processors may be used to train a machine learning model to reduce training time. To synchronize trained model data between the processors, data is communicated between the processors after some number of training cycles. To improve the communication efficiency, exemplary embodiments synchronize data among a set of processors after a predetermined number of training cycles, and synchronize data between one or more processors of each set of the processors after a predetermined number of training cycles. During the first synchronization among a set of processors, compressed model gradient data generated after performing the training cycles may be communicated. During the second synchronization between the set of processors, trained models or full model gradient data generated after performing the training cycles may be communicated.

公开(公告)号：US11514901B2

公开(公告)日：2022-11-29

申请号：US16437763

申请日：2019-06-11

Applicant: Amazon Technologies, Inc.

Inventor： Sree Hari Krishnan Parthasarathi ,  Bjorn Hoffmeister ,  Brian King ,  Roland Maas

IPC: G10L15/20 ,  G10L15/02 ,  G10L17/06 ,  G10L25/87 ,  G10L15/08 ,  G10L15/16 ,  G10L17/18 ,  G10L25/78 ,  G10L17/02

Abstract: A system configured to process speech commands may classify incoming audio as desired speech, undesired speech, or non-speech. Desired speech is speech that is from a same speaker as reference speech. The reference speech may be obtained from a configuration session or from a first portion of input speech that includes a wakeword. The reference speech may be encoded using a recurrent neural network (RNN) encoder to create a reference feature vector. The reference feature vector and incoming audio data may be processed by a trained neural network classifier to label the incoming audio data (for example, frame-by-frame) as to whether each frame is spoken by the same speaker as the reference speech. The labels may be passed to an automatic speech recognition (ASR) component which may allow the ASR component to focus its processing on the desired speech.