Study on Wavelet Sym3 Threshold-method in Speech De-noising

Speech signal is a kind of unidimensional non-stationary signal,with noise being easily sneaked during its spreading,and this phenomenon makes its original message difficultly distinguished. In order to get pure speech,speech de-noising is studied and the basic method using wavelets is proposed. And then nosiy speech is processed with wavelets sym3 in matlab. The classical hard threshold-method,soft threshold-method,and the way in combination of the former two methods are adopted to process the noisy speech with different strengths. The exprtiment indicates that better effects could be obtained in combination of hard threshold-method and soft threshold-method,including fairly high output signal-to-noise ratio.