The Research Of Speech De-Nosing Based On Wavelet Transform

Speech signal is the most convenient and shortcut way of intercommunion.However, in the course of the intercommunion, speech signal is disturbed andpolluted inevitable by surroundings and transmission mediums. As a result, weare unable to catch on the meanings of speech signal. In the past, we used toanalyze speech signal and filter the noise by short-time Fourier transform in thefrequent domain. However, this method is not effective to the white noise.Wavelet transform is a very popular method in the processing of digital signalnowadays.This paper mainly presents the research of the speech denosing based onwavelet transform. Firstly, we introduce the basic theory of wavelet, methods ofsignal denosing based wavelet transform and application of de-nosing methods inspeech. Recently in the signal denosing domain more and more researchers areused to reducing noised based wavelet. In these algorithms hard threshold methodand soft threshold method are mostly included. But both of these two methodshave limitations and are not effective. Now we present a new method that we cancall it two variable threshold function. This method can supply a gap.When thedifference of the wavelet coefficients of signal and noise in critical point isgreat ,we can choose threshold easily. So, we present the algorithm of noisedreduction in speech based energy member wavelet thresholding. Lastly,Experiment results demonstrate that this method is effective and excellent.