NOISY SPEECH RECOGNITION BY USING VARIANCE ADAPTED HIDDEN MARKOV-MODELS

被引:4
|
作者
CHIEN, JT
LEE, LM
WANG, HC
机构
[1] Department of Electrical Engineering, National Tsing Hua University, Hsinchu
关键词
HIDDEN MARKOV MODELS; SPEECH RECOGNITION;
D O I
10.1049/el:19951066
中图分类号
TM [电工技术]; TN [电子技术、通信技术];
学科分类号
0808 ; 0809 ;
摘要
Shrinkage of variances in HMM due to additive noise is an factor in speech recognition process. This variance shrinkage can be nonparametrically estimated from a training database. In the Letter we take this factor into account in the projection measure for noisy speech recognition. Experiments show that recognition rates can be improved significantly.
引用
收藏
页码:1555 / 1556
页数:2
相关论文
共 50 条