Title:
音声認識方法、装置、デバイス、コンピュータ可読記憶媒体及びプログラム
Document Type and Number:
Japanese Patent JP6914236
Kind Code:
B2
Abstract:
A speech recognition method, device, apparatus and a computer-readable storage medium are provided. The method includes: starting a part of a microphone array to acquire a first speech signal; performing an echo cancellation process for the first speech signal to obtain a second speech signal; performing wakeup recognition for the second speech signal, to determine whether a wakeup word is included in the second speech signal; starting the microphone array to acquire a third speech signal, in the case that the wakeup word is included in the second speech signal; performing noise reduction for the third speech signal; and performing speech recognition for a noise-reduced signal. Computational load and power consumption during the speech recognition can be greatly reduced, because most processes of the front-end noise reduction algorithm are not started before the wakeup state and only a part of a microphone array is started.
Inventors:
Gen, Ray
Application Number:
JP2018233967A
Publication Date:
August 04, 2021
Filing Date:
December 14, 2018
Export Citation:
Assignee:
Baidu Online Network Technology (Beijing) Company Limited
International Classes:
G10L15/28; G10K11/34; G10L15/10; G10L15/20; G10L15/32; G10L21/0208; G10L25/51; H04R1/40; H04R3/00; H04R3/02
Domestic Patent References:
| JP2010268324A | ||||
| JP2003330490A | ||||
| JP2016167645A | ||||
| JP2013037174A | ||||
| JP10276491A |
Foreign References:
| WO2017098775A1 |
Attorney, Agent or Firm:
Yoshiyuki Inaba
Toshifumi Onuki
Akihiko Eguchi
Kazuhiko Naito
Toyotaka Abe
Toshifumi Onuki
Akihiko Eguchi
Kazuhiko Naito
Toyotaka Abe