Research On Methods For Speech Enhancement Based On Particle Filter

In general, speech signals are often corrupted acoustically by ambient noise. The noise degrades both the quality and the intelligibility of the speech signals; even make the speech processing system not work well. So speech enhancement technology is strongly desired to solve the problem. The objective of speech enhancement is to reduce background noise, to improve the quality of voice communication, to increase the intelligibility and to ensure the reliability of speech processing system. Furthermore, speech enhancement can greatly improve the performance of many speech encoding and recognition system in the noisy environment. Speech enhancement technology is a very important part of digital speech signal processing, and it is extensively used in wireless telephone, conference call, scene record, military wiretap and other fields. Consequently the study of speech enhancement technology is very promising.The thesis discusses practical speech enhancement technologies, and focuses on the particle filter based methods. Following is the main work of this thesis:(1) A variety of methods for speech enhancement are investigated, such as model based methods, adaptive filtering methods and spectral estimator etc. Particularly the particle filter based method is studied.(2) By exploring the time-varying autoregressive models and its stochastically evolving parameters, we present an algorithm for the estimation of the TVAR model parameters based on particle filter. By the use of reflection coefficients, the decision of the model stability is realized quickly and the strict stability of the model is ensured.(3) Based on former work, we present a particle filter based speech enhancement (Particle Filter based Speech Enhancement: PFSE) algorithm. The experimental results show that the signal-to-noise ratio and the quality of speech signals have been improved obviously after filtered by the proposed PFSE algorithm.