Speech Enhancement Based on Nonparametric Bayesian Method
Jiawen WU,Qinting LIU,Delu ZENG,Xinghao DING,Lin LI
DOI: https://doi.org/10.6043/j.issn.0438-0479.201702026
2017-01-01
Abstract:A new speech enhancement strategy is proposed by utilizing a nonparametric Bayesian method with Spike-Slab priori (NBSP).As a sparse representation framework,the dictionary learning,sparse coefficients representation and noise variance estimation are replaced by a single procedure of Bayesian posterior estimation.First,the noisy speech is divided into clean speech,Gaussian noise and rest noise.Then,each part is modeled with a certain priori distribution.Finally,upon the adoption of Markov Chain Monte Carlo sampling algorithm,the posterior distribution can be obtained,as the clean speech and all other parameters.Without knowing the noise variance,NBSPcould be performed directly on the noisy speech to infer the sparsity of the speech.Experiments were executed on NOIZEUS database.Experiments are executed on noisy speeches from NOIZEUS database with SNR ranging from 0 dB to 10 dB,which contain three types of noise (white,train and street).And the subjective and objective measures like PESQ score and the output SegSNR are implemented to evaluate the performance of NBSP and the other state-of-the-art methods.Corresponding results show that NBSP achieves better performances,especially in conditions of non-stationary noise with low input SNR.