Speaker verification using probabilistic 2D CLAFIC

Lee Ying Chong, Andrew B.J. Teoh

Research output: Contribution to journalArticlepeer-review

Abstract

User-specific subspace method, such as CLAFIC concatenates the speech matrices side by side leads to a high dimensional feature space before forming the correlation matrix. We propose an enhanced CLAFIC method in speaker verification, coined as Probabilistic 2D CLAFIC which apply a straightforward 2D speech matrix projection technique for feature dimension reduction and GMM is used to construct the speaker model to boost up the performance. Our results show the superior performance and low computation time in the proposed system.

Original languageEnglish
Pages (from-to)179-184
Number of pages6
Journalieice electronics express
Volume4
Issue number5
DOIs
Publication statusPublished - 2007 Mar 10

All Science Journal Classification (ASJC) codes

  • Electronic, Optical and Magnetic Materials
  • Condensed Matter Physics
  • Electrical and Electronic Engineering

Fingerprint Dive into the research topics of 'Speaker verification using probabilistic 2D CLAFIC'. Together they form a unique fingerprint.

Cite this