The Algorithm Of Wavelet-Thresholding Denosing In Speech Encoding

Vast amounts of redundant information contained in speech signal. It is very important to research on how to design the efficient encoding , while at the same time reducing the distortion of speech signal. In a variety of encoding, Continuously Variable Slope Delta Modulation (CVSD) is a speech algorithm which is widely applied in the field of satellite communication system, mobile communication system, military special communication net, and so on. Based on the CVSD, how to design better methods to improve the encoding quality of speech signal and save bandwidth resources needs us to make further study.The wavelet transform in denoising has been achieved some good results. Wavelet can analyze signal at time domain and the frequency domain simultaneously, so it can de-noise effectively. The mother wavelets is a gather, the more suitable mother wavelet we select, the better the result of de-noising. At the same time, the threshold also influences the result of de-noising for a same mother wavelet. Therefore we need to select the conscious threshold.In this thesis, the method of wavelet-thresholding denoising is introduced based on the traditional CVSD coding. We first analyze the soft-thresholding and hard-thresholding denoising methods in wavelet thresholding denoising, and then we present a new thresholding function based on them in order to improve the denoising performance. This method can overcome shortcomings of soft-thresholding and hard-thresholding denoising method. We select four layers of db4 wavelet as mother wavelet, and add a factor to the modified project which can reduce the fixed deviation in the thresholding function. At the same time we can adjust the size of this factor in order to obtain better denoising results. Subsequently, we encode and decode the speech signal based on CVSD. At last, we validate the effectiveness and superority of the improved scheme by Matlab simulation.