Sixth International Conference on Spoken Language Processing
(ICSLP 2000)

Beijing, China
October 16-20, 2000

Speaker Adaptation in Noisy Environments Based on Parameter Estimation Using Uncertain Data

Naoto Iwahashi (1), Akihiko Kawasaki (1,2)

(1) Sony Computer Science Labs Inc., Shinagawa-ku, Tokyo, 141-0022 Japan
(2) Dept. Electrical, Electronics and Computer Engineering, Waseda University, Shinjuku-ku, Tokyo, Japan

This paper describes new method for the speaker adaptation of HMM parameters in environments with background noise. This method is based on Bayesian estimation, and calculates the a posteriori distribution of clean-speech HMM parameters from their a priori distribution by using noisy speech observations. The advantage of the method is that the distribution of the noise can be taken into account in adapting clean-speech HMMs to a target speakerís speech without noise. The results of the experiments using noninformative prior show that the recognition performance in a noise-free environment was improved by this method even when the SNR of the noisy speech data used for the adaptation was -6 dB.


Full Paper

Bibliographic reference.  Iwahashi, Naoto / Kawasaki, Akihiko (2000): "Speaker adaptation in noisy environments based on parameter estimation using uncertain data", In ICSLP-2000, vol.4, 528-531.