Speech Enhancement Based On TVAR And Particle Filter

Speech enhancement is an important branch of speech signal processing, also is an effective solution to noise pollution. The goal of speech enhancement is either to improve the perceived quality of the speech, or to increase its intelligibility by getting as clean speech as possible from the noisy one. It has been successfully applied in many areas such as the daily life and military in recent years, also has a significant value in speech communication and recognition.The approaches of speech enhancement are numerous. The paper firstly analysis some methods which are not based on speech model, such as spectrum subtraction,wiener filter,subspace method. These algorithms do not use any prior information of the speech, so the enhancement is limited. This paper discusses the speech enhancement algorithm based on TVAR and particle filter, then analyze the simulation results. At last, use Regularized Particle Filter (RPF) algorithm which can avoid the loss of diversity during re-sampling process combining TVAR model to deal with background noise from noise speech. Simulation result shows performance of the modified approach is better than the original one. The output SNR and quality both have been greatly significantly improved.The experimental results indicate that the algorithm proposed in this paper has such advantages:(1) Regularized Particle Filter algorithm can effectively avoid particle diversity issues in the resampling step and the effect is remarkable; (2) It can adapt the low SNR(Signal to Noise Ratio)condition; (3) It can get rid of most noise in noisy speech, perceived quality of the speech.