Enhancing speaker identification performance using circular hidden Markov model

Author

Shahin, Ismail

Author_Institution

Sharjah Univ., UAE

fYear

2004

fDate

19-23 April 2004

Firstpage

389

Lastpage

390

Abstract

In this paper, circular hidden Markov model (CHMM) is implemented to improve the recognition performance of isolated-word text-dependent speaker identification systems under the neural talking condition. Our results show that the CHMM improves the speaker recognition performance under such a condition compared to the left-to-right hidden Markov model (LTRHMM). The average speaker recognition performance has been improved from 90% using the LTRHMM to 95% using the CHMM. In this research, the linear predictive coding (LPC) cepstral feature analysis is used to form the observation vector for both LTRHMM and CHMM.

Keywords

cepstral analysis; hidden Markov models; linear predictive coding; neural nets; speaker recognition; speech coding; CHMM; LPC; cepstral feature analysis; circular hidden Markov model; isolated-word text-dependent speaker identification system; linear predictive coding; neural talking condition; speaker recognition; Computational complexity; Hidden Markov models; Markov processes; Power system modeling; Probability density function; Probability distribution; Speaker recognition; Speech processing; Speech recognition; Stochastic processes;

fLanguage

English

Publisher

ieee

Conference_Titel

Information and Communication Technologies: From Theory to Applications, 2004. Proceedings. 2004 International Conference on

Print_ISBN

0-7803-8482-2

Type

conf

DOI

10.1109/ICTTA.2004.1307795

Filename

1307795