Title:
VOICEPRINT RECOGNITION METHOD AND DEVICE EMPLOYING DEEP LEARNING, AND APPARATUS
Document Type and Number:
WIPO Patent Application WO/2021/051608
Kind Code:
A1
Abstract:
A voiceprint recognition method and device employing deep learning, and an apparatus. The method comprises: acquiring a verified voice of a target person, and using MFCCs to perform feature extraction on the verified voice so as to obtain a verified voice feature (101); inputting the verified voice feature into a neural network model, and performing verification processing (102); adjusting, according to a verification processing result, a parameter of a function in the neural network model, so as to obtain a target neural network model capable of recognizing a voice of the target person (103); using MFCCs to perform feature extraction on an acquired voice to undergo recognition, so as to obtain a voice feature to undergo recognition (104); and inputting the voice feature to undergo recognition into the target neural network model, performing voice recognition processing, and determining whether or not the voice to undergo recognition belongs to the target person (105). A trained neural network model is used to recognize a voice and to determine whether or not the voice belongs to the target person, thus improving the speed and accuracy of the voice recognition process, and effectively improving the efficiency of recognition.
Inventors:
WANG JIANZONG (CN)
ZHAO FENG (CN)
ZHAO FENG (CN)
Application Number:
PCT/CN2019/118402
Publication Date:
March 25, 2021
Filing Date:
November 14, 2019
Export Citation:
Assignee:
PING AN TECH SHENZHEN CO LTD (CN)
International Classes:
G10L17/18
Domestic Patent References:
| WO2018107810A1 | 2018-06-21 |
Foreign References:
| CN110010133A | 2019-07-12 | |||
| CN109074822A | 2018-12-21 | |||
| CN108564954A | 2018-09-21 |
Attorney, Agent or Firm:
BEIJING STRONG WISDOM INTELLECTUAL PROPERTY AGENCY LTD. (CN)
Download PDF: