《基于小波分析的语音信号去噪处理的研究》.docVIP

单位代码：分 类 号：延安大学西安创新学院本科毕业论文（设计）题 目 基于小波分析的语音信号去噪技术研究专 业： 名：黄 同职 称：二〇一四年六月基于小波分析的语音信号去噪技术研究 摘要：本论文利用小波在语音信号去噪方面的良好应用，首先对小波信号进行分析，并用傅立叶变换与小波变换进行比较，发现他们的相同点与不同点，对小波基分析，发现小波基的选取对去噪效果没有太大的影响；后对阈值法分析，发现阈值法具有计算量小、去噪效果好的特点，并且在阈值法中，阈值的选取直接关系到去噪效果的优劣。本论文重点使用小波阈值法，针对不同的阈值函数、阈值处理方法及小波函数做出了选取；针对阈值法中高频信号失真的缺点，我们对小尺度上的小波系数做谱减法预处理；之后以一个小阈值去除剩余噪声，大尺度上仍然利用阈值法处理。经过仿真实验表明，这种处理方法较传统的小波阈值法，保留了更多有用信号，减小了去噪后语音信号的失真。 关键词：小波分析；阈值去噪；谱减法。Research on speech signal de-noise based on the wavelet analysis Abstract：This thesis takes advantage of the wavelet in speech signal denoising of good application. For one thing, we analyze the wavelet signal ,comparing the Fu Liye transform with wavelet transform to find their same and different points.On wavelet basis analysis,we found that the selection of wavelet base has not enough influence on denoising effect. After we make an analysis on the threshold value method, founding the threshold method has small computation and good de-noising effect characteristics. And in the threshold method, the selection of threshold value is directly related to the denoising effect;This thesis focuses on the use of wavelet thresholding method, according to the different threshold function and threshold processing method and wavelet function to make a choice; Aiming at the shortcoming of the high frequency signal distortion in threshold method, we do to small scale wavelet coefficients of spectral subtraction pretreatment and then to a small threshold to remove residual noise, large scale still using the method of threshold processing. After the simulation experiment indicated that, this method compare to the traditional wavelet threshold method retaining more of the useful signal and reducing the noise after the speech signal distortion. Keywords: Wavelet analysis;Threshold de-noising; Spectral subtraction.目 录 1引言 1 2小波分析 2 2.1傅立叶变换 2 2.2小波变换 3 2.3 傅立叶变换与小波变换的比较 5 3小波去噪基本理论 6 3.1 信号和噪声在小波域的传播特性 6 3.2 小波基的选取 6 3.3 小波阈值去噪及通用阈值法(sqtwolog 原理) 8 3.4小波阈值处