公开(公告)号：US20220262345A1

公开(公告)日：2022-08-18

申请号：US17662021

申请日：2022-05-04

申请人： Google LLC

发明人： Matthew Sharifi ,  Kevin Kilgour ,  Dominik Roblek ,  James Lin

IPC分类号： G10L15/06 ,  G06N3/04 ,  G06N3/08 ,  G10L13/00 ,  G10L15/16 ,  G10L15/22

摘要： A method of training a custom hotword model includes receiving a first set of training audio samples. The method also includes generating, using a speech embedding model configured to receive the first set of training audio samples as input, a corresponding hotword embedding representative of a custom hotword for each training audio sample of the first set of training audio samples. The speech embedding model is pre-trained on a different set of training audio samples with a greater number of training audio samples than the first set of training audio samples The method further includes training the custom hotword model to detect a presence of the custom hotword in audio data. The custom hotword model is configured to receive, as input, each corresponding hotword embedding and to classify, as output, each corresponding hotword embedding as corresponding to the custom hotword.

公开(公告)号：US11341954B2

公开(公告)日：2022-05-24

申请号：US16717518

申请日：2019-12-17

申请人： Google LLC

发明人： Matthew Sharifi ,  Kevin Kilgour ,  Dominik Roblek ,  James Lin

IPC分类号： G10L15/06 ,  G06N3/04 ,  G06N3/08 ,  G10L13/00 ,  G10L15/16 ,  G10L15/22 ,  G10L15/08

摘要： A method of training a custom hotword model includes receiving a first set of training audio samples. The method also includes generating, using a speech embedding model configured to receive the first set of training audio samples as input, a corresponding hotword embedding representative of a custom hotword for each training audio sample of the first set of training audio samples. The speech embedding model is pre-trained on a different set of training audio samples with a greater number of training audio samples than the first set of training audio samples. The method further includes training the custom hotword model to detect a presence of the custom hotword in audio data. The custom hotword model is configured to receive, as input, each corresponding hotword embedding and to classify, as output, each corresponding hotword embedding as corresponding to the custom hotword.

公开(公告)号：US20210183367A1

公开(公告)日：2021-06-17

申请号：US16717518

申请日：2019-12-17

申请人： Google LLC

发明人： Matthew Sharifi ,  Kevin Kilgour ,  Dominik Roblek ,  James Lin

IPC分类号： G10L15/06 ,  G10L15/22 ,  G10L15/16 ,  G10L13/04 ,  G06N3/04 ,  G06N3/08

摘要： A method of training a custom hotword model includes receiving a first set of training audio samples. The method also includes generating, using a speech embedding model configured to receive the first set of training audio samples as input, a corresponding hotword embedding representative of a custom hotword for each training audio sample of the first set of training audio samples. The speech embedding model is pre-trained on a different set of training audio samples with a greater number of training audio samples than the first set of training audio samples. The method further includes training the custom hotword model to detect a presence of the custom hotword in audio data. The custom hotword model is configured to receive, as input, each corresponding hotword embedding and to classify, as output, each corresponding hotword embedding as corresponding to the custom hotword.