Title:
SIGNAL PROCESSING METHOD AND ELECTRONIC DEVICE
Document Type and Number:
WIPO Patent Application WO/2022/068608
Kind Code:
A1
Abstract:
A signal processing method and an electronic device (10). By combining a first video obtained by means of a camera (140) (S220) and a first audio signal obtained by means of a microphone array (130) (S210), a target sound source direction of a target user who is interacting with the electronic device (10) by speech is determined (S240), such that the estimation accuracy of the target sound source direction can be greatly increased; speech enhancement is performed on a second audio signal obtained by means of the microphone array (130) (S260) on the basis of a user lip video in the target sound source direction obtained by means of the camera (140) (S250) and a preset speech enhancement model; since the speech enhancement model integrates correspondences between pronunciations and lip shapes, a clean third audio signal can be recovered in combination with the user lip video and the speech enhancement model; finally, speech recognition efficiency can be effectively increased.
Inventors:
BAO GUANGZHAO (CN)
CHEN LIWEN (CN)
HUANG LEI (CN)
CHEN LIWEN (CN)
HUANG LEI (CN)
Application Number:
PCT/CN2021/118948
Publication Date:
April 07, 2022
Filing Date:
September 17, 2021
Export Citation:
Assignee:
HUAWEI TECH CO LTD (CN)
International Classes:
G10L15/26; G06K9/00; G10L21/0216; H04N5/232
Domestic Patent References:
| WO2017129239A1 | 2017-08-03 |
Foreign References:
| CN110691196A | 2020-01-14 | |||
| CN106328156A | 2017-01-11 | |||
| CN106679651A | 2017-05-17 | |||
| CN110858488A | 2020-03-03 | |||
| CN110082723A | 2019-08-02 | |||
| CN110389597A | 2019-10-29 | |||
| CN202011065346A | 2020-09-30 |
Other References:
See also references of EP 4207186A4
Attorney, Agent or Firm:
LONGSUN LEAD IP LTD. (CN)
Download PDF: