eduzhai > Applied Sciences > Engineering >

Version Control of Speaker Recognition Systems

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 7 pages

Abstract: This paper discusses one of the most challenging practical engineeringproblems in speaker recognition systems - the version control of models anduser profiles. A typical speaker recognition system consists of two stages: theenrollment stage, where a profile is generated from user-provided enrollmentaudio; and the runtime stage, where the voice identity of the runtime audio iscompared against the stored profiles. As technology advances, the speakerrecognition system needs to be updated for better performance. However, if thestored user profiles are not updated accordingly, version mismatch will resultin meaningless recognition results. In this paper, we describe differentversion control strategies for different types of speaker recognition systems,according to how they are deployed in the production environment.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...