Third International Conference on Spoken Language Processing (ICSLP 94)

Yokohama, Japan
September 18-22, 1994

An Initial Study on Speaker Adaptation for Mandarin Syllable Recognition With Minimum Error Discriminative Training

Chih-Heng Lin (1,2), Pao-Chung Chang (1), Chien-Hsing Wu (1)

(1) Telecommunication Laboratories, Ministry of Communications, Taiwan
(2) Department of Electrical Engineering, National Taiwan University

This paper presents a method of speaker adaptation for Mandarin syllable recognition. Based on a minimum error classification (MEC) criterion, we use the generalized probabilistic decent (GPD) algorithm to adjust iteratively the parameters of the hidden Markov models (HMM). The experiments on the multi-speaker Mandarin syllable database of Telecommunication Laboratories (T.L.) yield the following results: 1) Efficient speaker adaptation can be achieved through discriminative training using the MEC criterion and the GPD algorithm. 2) The computations required can be reduced through the use of the confusion sets in Mandarin base syllables. 3) For the discriminative training, the adjustment on the mean values of the Gaussian mixtures has the most prominent effect on speaker adaptation.

Full Paper

Bibliographic reference.  Lin, Chih-Heng / Chang, Pao-Chung / Wu, Chien-Hsing (1994): "An initial study on speaker adaptation for Mandarin syllable recognition with minimum error discriminative training", In ICSLP-1994, 307-310.