The Research Of Speech Enhancement Algorithm

As the development of the technology of speech signal processing in our modern life, the robust of the technology becomes more and more important. How to eliminate the noise in the real environments to obtain the useful information has become an important research subject. Speech enhancement is a useful means to eliminate the interference from the received speech signals which are inevitably interfered by noise.Some speech enhancement algorithms including their advantages and disadvantages are introduced in this paper. Multimicrophone-based speech enhancement algorithms incorporate both spatial and spectral information typically. Hence, they have the potential to improve the traditional one or two microphone solutions that utilize only spectral information.With the study of the speech signal in the Time-Frequency domain, a new speech enhancement algorithm is proposed which is based on that the signal has been decomposed with Matching pursuits in the Time-Frequency domain. The Time-Frequency energy of the signal which had been decomposed with Matching pursuits had a clear distribution in Time-frequency domain. Using this characteristic, the noise can be differentiated by setting a threshold value. The result has shown that using this character to cancel noise, the noise can be eliminated, and the speech can be understood easily. The disadvantage of this method is using much time to finish the decomposed processing.The theory of time-frequency analysis is introduced before this, and a new algorithm which can eliminate the cross-term in Wigner Distribution is proposed.At last, we summarized the result of this thesis, and tabled some proposal methods about future study.