DBN based multi-stream models for speech

Author

Zhang, Yimin ; Diao, Qian ; Huang, Shan ; Hu, Wei ; Bartels, Chris ; Bilmes, Jeff

Author_Institution

Intel China Res. Center, Beijing, China

Volume

1

fYear

2003

fDate

6-10 April 2003

Abstract

We propose dynamic Bayesian network (DBN) based synchronous and asynchronous multi-stream models for noise-robust automatic speech recognition. In these models, multiple noise-robust features are combined into a single DBN to obtain better performance than any single feature system alone. Results on the Aurora 2.0 noisy speech task show significant improvements of our synchronous model over both single stream models and over a ROVER based fusion method.

Keywords

belief networks; noise; speech recognition; Aurora 2.0 noisy speech task; DBN based multistream speech models; HMM; ROVER based fusion method; asynchronous multistream model; automatic speech recognition; dynamic Bayesian network; hidden Markov models; multiple noise-robust features; single stream models; synchronous multistream model; Automatic speech recognition; Bayesian methods; Concatenated codes; Data mining; Feature extraction; Hidden Markov models; Noise robustness; Speech recognition; Standards development; Streaming media;

fLanguage

English

Publisher

ieee

Conference_Titel

Acoustics, Speech, and Signal Processing, 2003. Proceedings. (ICASSP '03). 2003 IEEE International Conference on

ISSN

1520-6149

Print_ISBN

0-7803-7663-3

Type

conf

DOI

10.1109/ICASSP.2003.1198911

Filename

1198911