Dynamic Bayesian Network Inversion for Robust Speech Recognition(Speech and Hearing)
スポンサーリンク
概要
- 論文の詳細を見る
This paper presents an inversion algorithm for dynamic Bayesian networks towards robust speech recognition, namely DBNI, which is a generalization of hidden Markov model inversion (HMMI). As a dual procedure of expectation maximization (EM)-based model reestimation, DBNI finds the 'uncontaminated' speech by moving the input noisy speech to the Gaussian means under the maximum likelihood (ML) sense given the DBN models trained on clean speech. This algorithm can provide both the expressive advantage from DBN and the noise-removal feature from model inversion. Experiments on the Aurora 2.0 database show that the hidden feature model (a typical DBN for speech recognition) with the DBNI algorithm achieves superior performance in terms of word error rate reduction.
- 社団法人電子情報通信学会の論文
- 2007-07-01
著者
-
Xie Lei
School Of Computer Science Northwestern Polytechnical University
-
YANG Hongwu
Dept. of Computer Science & Technology, Tsinghua University
-
Yang Hongwu
Dept. Of Computer Science & Technology Tsinghua University:college Of Physics And Electronics Engineer Northwest Normal University