Text-independent speaker identification based on explicit exploitation of stochastic characteristics of test utterance

W. H. Tsai*, Wen-Whei Chang, C. S. Huang

*Corresponding author for this work

Research output: Contribution to journalArticle

Abstract

The statistics characteristics of test utterance for speaker identification via cross liklihood ratio (CLR) and Bayesian information criterion (BIC) was discussed. A large CLR signifies a higher similarity between test speaker's voice and the voice of a client speaker. The CLR-based approach and BIC-based approach were found to be superior to the conventional approach based on maximum liklihood decisions.

Original languageEnglish
Pages (from-to)1556-1557
Number of pages2
JournalElectronics Letters
Volume37
Issue number25
DOIs
StatePublished - 6 Dec 2001

Fingerprint Dive into the research topics of 'Text-independent speaker identification based on explicit exploitation of stochastic characteristics of test utterance'. Together they form a unique fingerprint.

  • Cite this