基于小波变换的语音去噪算法研究与实现-信号与信息处理专业论文.docxVIP

Abs Abstract II II Abstract In the process of acquisition and transmission, the speech signals will be disturbed by many different kinds of noise inevitably. The good time-frequency localization property of the wavelet transform theory solves the problem that the time-frequency unitary of Fourier transform of the traditional speech de-noising methods effectively. Based on the wavelet transform theory, the wavelet de-noising method gives the detailed analysis and processing of non-stationary signal on the time-frequency space. And based on the distribution characteristics in the wavelet domain of the noise and speech signal, to threshold the wavelet coefficients on each layer can eliminate the noise component effectively. The main work in this paper is as follows: In the first place, by introducing the characteristics of speech and noise signal, discuss the basic principles of several common speech de-noising methods, and analyze their advantages and disadvantages by simulation comparison. In the next place, on the basis of basic theory of wavelet analysis, give some research on the basic principles of wavelet de-noising theory, and analyze the impact of threshold function and threshold value to the effect of speech signal de-noising, gives simulation comparison of selecting the optimum decomposition layers and selection of wavelet radix. Based on the study of traditional wavelet threshold algorithm and the analysis of its advantages and disadvantages, propose a new threshold function related on the wavelet decomposition layers, and verify the de-noising effects of the threshold function is better than the traditional function by simulation. Finally, based on the limitation of the wavelet threshold algorithm with fixed parameters, propose the wavelet de-noising method based on empirical mode decomposition. Firstly, process the speech signal with noise by empirical mode decomposition, then use the improved wavelet threshold de-noising to process the components in whic