Two discriminative training schemes of GMM for language identification

Author

Dan, Qu ; Bingxi, Wang ; Qiang, Zhang

Author_Institution

Dept. of Information Sci., Information Eng. Univ., Zhengzhou, China

Volume

1

fYear

2004

fDate

31 Aug.-4 Sept. 2004

Firstpage

630

Abstract

In this paper, two discriminative training procedures for a Gaussian mixture model (GMM) language identification system are described. One is based on maximum mutual information criterion (MMI), the other uses minimum classification error (MCE) criterion. Both the proposals are based on the generalized probabilistic descent (GPD) algorithm formulated to estimate the GMM parameters. The evaluation is conducted using the OGI multi-language telephone speech corpus. The experimental results show such system is very effective in language identification tasks.

Keywords

Gaussian processes; natural languages; speech recognition; Gaussian mixture model; discriminative training scheme; generalized probabilistic descent algorithm; language identification system; maximum mutual information criterion; minimum classification error criterion; multilanguage telephone speech corpus; Chromium; Error analysis; Influenza; Information science; Maximum likelihood estimation; Natural languages; Robustness; Speech analysis; Speech recognition; Wire;

fLanguage

English

Publisher

ieee

Conference_Titel

Signal Processing, 2004. Proceedings. ICSP '04. 2004 7th International Conference on

Print_ISBN

0-7803-8406-7

Type

conf

DOI

10.1109/ICOSP.2004.1452742

Filename

1452742