eduzhai > Applied Sciences > Engineering >

A Speech Enhancement Algorithm based on Non-negative Hidden Markov Model and Kullback-Leibler Divergence

  • Save

... pages left unread,continue reading

Document pages: 13 pages

Abstract: In this paper, we propose a novel supervised single-channel speechenhancement method combing the the Kullback-Leibler divergence-basednon-negative matrix factorization (NMF) and hidden Markov model (NMF-HMM). Withthe application of HMM, the temporal dynamics information of speech signals canbe taken into account. In the training stage, the sum of Poisson, leading tothe KL divergence measure, is used as the observation model for each state ofHMM. This ensures that a computationally efficient multiplicative update can beused for the parameter update of the proposed model. In the online enhancementstage, we propose a novel minimum mean-square error (MMSE) estimator for theproposed NMF-HMM. This estimator can be implemented using parallel computing,saving the time complexity. The performance of the proposed algorithm isverified by objective measures. The experimental results show that the proposedstrategy achieves better speech enhancement performance than state-of-the-artspeech enhancement methods. More specifically, compared with the traditionalNMF-based speech enhancement methods, our proposed algorithm achieves a 5 improvement for short-time objective intelligibility (STOI) and 0.18improvement for perceptual evaluation of speech quality (PESQ).

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×