Title:
UPDATING METHOD AND DEVICE FOR SELF-LEARNING VOICEPRINT RECOGNITION
Document Type and Number:
WIPO Patent Application WO/2019/127897
Kind Code:
A1
Abstract:
Disclosed in the present invention is an updating method and device for self-learning voiceprint recognition. In the present invention, when a user selects verification, voice to be verified is received and voiceprint features to be verified are extracted. When determined that a matching value between said voiceprint features to be verified and the most up-to-date fusion voice features complies with the requirements of preset voiceprint evaluation standards, the voiceprint verification is passed. After determining that said voiceprint features to be verified comply with the fusion conditions, said voiceprint features to be verified are used as source material voiceprint features to ensure that source material voiceprint features which meet fusion conditions are a verification operation initiated by the user. The most up-to-date source material voiceprint features are selected for fusion so as to obtain new fusion voiceprint features. The most up-to-date source material voiceprint features are selected during the whole process to obtain new fusion voiceprint features, and at the same time, the most up-to-date fusion voiceprint features are selected to verify voiceprint features to be verified, thus solving the technical problem of the rate of correct recognition in existing voiceprint recognition technology being reduced because of voiceprint drifting.
Inventors:
CHEN SHUDONG (CN)
Application Number:
PCT/CN2018/077535
Publication Date:
July 04, 2019
Filing Date:
February 28, 2018
Export Citation:
Assignee:
SPEAKIN TECH CO LTD (CN)
International Classes:
G10L17/02; G10L17/00
Foreign References:
| CN106506524A | 2017-03-15 | |||
| CN107331400A | 2017-11-07 | |||
| CN107424614A | 2017-12-01 | |||
| US20170140760A1 | 2017-05-18 |
Attorney, Agent or Firm:
UNITALEN ATTORNEYS AT LAW (CN)
Download PDF: