A comparison between hidden Markov models and vector quantization for speech independent speaker recognition

Author

Weber, D.M. ; Du Preez, J.X.

Author_Institution

Dept. of Electr. & Electron. Eng., Stellenbosch Univ., South Africa

fYear

1993

fDate

34187

Firstpage

139

Lastpage

144

Abstract

We compare Vector Quantization and Hidden Markov Models for speaker recognition for real time recognition. A scheme to reject speakers not known to the system is described and tested. Results show that the HMM algorithm outperforms the VQ algorithm. Using a 64 state HMM, a speaker recognition accuracy of 96.1% was achieved. The rejection option generated 25.7% false rejections for a 95% confidence of a correct decision. VQ best results were 93.1% with a 61% false rejection rate for codebooks of size 128

Keywords

hidden Markov models; speaker recognition; speech coding; vector quantisation; HMM algorithm; VQ algorithm; codebooks; false rejections; hidden Markov models; real time recognition; speaker recognition accuracy; speech independent speaker recognition; vector quantization; Africa; Electronic equipment testing; Feature extraction; Hidden Markov models; Loudspeakers; Real time systems; Speaker recognition; Speech processing; System testing; Vector quantization;

fLanguage

English

Publisher

ieee

Conference_Titel

Communications and Signal Processing, 1993., Proceedings of the 1993 IEEE South African Symposium on

Conference_Location

Jan Smuts Airport

Print_ISBN

0-7803-1292-9

Type

conf

DOI

10.1109/COMSIG.1993.365856

Filename

365856