Efficient speech recognition using subvector quantization and discrete-mixture HMMs

Author

Tsakalidis, S. ; Digalakis, V. ; Neumeyer, L.

Author_Institution

Dept. of Electron. & Comput. Eng., Tech. Univ. of Crete, Chania, Greece

Volume

2

fYear

1999

fDate

15-19 Mar 1999

Firstpage

569

Abstract

This paper introduces a new form of observation distributions for hidden Markov models (HMMs), combining subvector quantization and mixtures of discrete distributions. We present efficient training and decoding algorithms for the discrete-mixture HMMs (DMHMMs). Our experimental results in the air-travel information domain show that the high-level of recognition accuracy of continuous mixture-density HMMs (CDHMMs) can be maintained at significantly faster decoding speeds. Moreover, we show that when the same number of mixture components is used in DMHMMs and CDHMMs, the new models exhibit superior recognition performance

Keywords

cepstral analysis; decoding; hidden Markov models; speech coding; speech recognition; vector quantisation; CDHMM; DMHMM; air-travel information; continuous mixture-density HMM; decoding speed; discrete distributions; discrete-mixture HMM; efficient decoding algorithms; efficient training algorithms; experimental results; hidden Markov models; mel-warped cepstral coefficients; mixture components; observation distributions; probabilities; recognition accuracy; recognition performance; speech recognition; subvector quantization; Cepstral analysis; Decoding; Distributed computing; Hidden Markov models; Quantization; Service oriented architecture; Speech recognition; Web sites; Working environment noise; World Wide Web;

fLanguage

English

Publisher

ieee

Conference_Titel

Acoustics, Speech, and Signal Processing, 1999. Proceedings., 1999 IEEE International Conference on

Conference_Location

Phoenix, AZ

ISSN

1520-6149

Print_ISBN

0-7803-5041-3

Type

conf

DOI

10.1109/ICASSP.1999.759730

Filename

759730