基于小波变换的语音信号去噪研究

发布时间：2018-03-07 00:18

本文选题：小波变换　切入点：阈值函数　出处：《燕山大学》2006年硕士论文　论文类型：学位论文

【摘要】：语音对话是人们相互通讯和交流最方便快捷的手段。但是人们在语音通讯过程中不可避免的会受到来自周围环境、传输介质的干扰,人们引入了噪声,影响了人们的听辨。在过去,我们一般使用短时傅立叶变换(SFFT)在频域内对语音信号进行分析去噪,但是对于白噪声,这种方法的效果往往不尽人意。小波变换(WT)是一种当今在信号处理领域中十分活跃的理论。本文主要基于小波变换对语音去噪进行了研究,首先介绍了小波变换的基础理论、基于小波变换的信号去噪方法以及在语音去噪中的应用。目前在信号去噪中,基于小波变换的方法得到了广泛地应用,这些方法主要是基于传统的硬阈值和软阈值方法,本文提出了一种新的双变量阈值函数,能有效地弥补硬、软阈值方法的不足,是硬、软阈值方法很好的一个改进方案。克服了采用硬阈值法去噪效果不佳和软阈值法过度光滑使信号失真的缺点。当噪声和信号对应的小波系数在临界点大小相差比较明显时,阈值的选取可以有较大的裕度,因此选取就比较容易。基于上述考虑本文提出了基于能量元的小波阈值语音去噪算法,其中运用了双变量阈值函数,并通过实验验证了该算法的有效性和优异性。
[Abstract]:Dialogue is the mutual communication and exchange of the most convenient means. But it is inevitable in the voice communication process will suffer from the surrounding environment, the interference of the transmission medium, we introduce noise, affect people's perception. In the past, we generally use short when Fu Liye transform (SFFT) in the frequency domain of speech signal analysis of denoising, but for white noise, this method is not effective. Wavelet transform (WT) is one of the very active in the field of signal processing theory.
This paper is mainly based on the wavelet transform of speech denoising is studied, firstly introduced the basic theory of wavelet transform, signal denoising method based on wavelet transform and application for speech signal denoising. The signal denoising method based on wavelet transform, have been widely used, these methods are mainly traditional hard threshold and based on the soft threshold method, this paper proposes a new bivariate threshold function, can effectively compensate for the lack of hard, soft threshold method is hard, an improved scheme soft threshold method very well. To overcome the hard threshold denoising effect and soft threshold method over smooth signal distortion the disadvantages of wavelet coefficients. When the noise and signal corresponding to the difference is obvious at the critical point size, threshold selection can have a larger margin, so choose more easily. Based on the above considerations is proposed in this paper. The wavelet threshold speech denoising algorithm is used in the energy element, in which the bivariate threshold function is used, and the effectiveness and the advantages of the algorithm are verified by experiments.

【学位授予单位】：燕山大学
【学位级别】：硕士
【学位授予年份】：2006
【分类号】：TN912.3

【引证文献】