基于小波包分析的声音特性提取大学本科毕业论文.docVIP

成都信息工程学院 学位论文 基于小波包分析的声音特性提取 论文作者姓名： 申请学位专业： 申请学位类别： 指导教师姓名（职称）： 论文提交日期： 基于小波包分析的声音特性提取 摘 要 说话人识别就是利用说话人的语音特征对说话人身份进行辨识或确认。与语音识别一样，说话人识别也是在提取原始语音信号某些特征参数的基础上，建立相应的模板和模型，然后按照一定的判决规则进行识别。其中前期特征参数提取的效果直接影响到后期识别的准确性。本设计完成了一个基于小波包分析的声音特征参数的提取模型。论文首先介绍小波分析理论的基本知识，阐述了小波理论、小波分析在声音特征提取方面的应用，然后在分析MFCC提取原理的基础上，结合小波包分析理论，得到特征参数。用特征参数提取方法通过对两个人的数字音频样本提取的特征参数对比来观察各个参数之间的区别。仿真实验证明该提取方法可以作为区别说话人的声音特征提取方法，配合识别算法可以达到较高的识别率。 关键词：小波包；梅尔倒谱系数；特征参数 Extraction of Feature Coefficient Based on Wavelet Packet Analysis Abstract Speaker Recognition is to use the speaker feature on the identity of the speaker identification or confirmation. As with speech recognition, speaker recognition is to extract some characteristic parameters of the original speech signal based on the establishment of appropriate templates and models, and according to some decision rules for recognition. One early feature extraction directly affects the effect of late the recognition accuracy. The design is completed, a wavelet packet analysis based on a simple model of the voice feature parameter extraction. Paper first introduces the basics of wavelet analysis theory，Described the wavelet theory and wavelet analysis in the application of sound feature extraction, and then we can get two characteristic parameters based on the analysis of MFCC extraction and the theory of wavelet packet analysis. Feature extraction method used by digital audio samples of two characteristic parameters extracted compared to observe the distinction between the various parameters. Simulation results show that the extraction method can distinguish the speakers voice as a feature extraction method, with the recognition algorithm can achieve higher recognition rate. Keywords:Wavelet packet;Mel cepstral;Parameter 目 录 论文总页数：21页 1 引 言 1 1.1 课题背景 1 1.2 国内外研究现状 1 1.3 本课题研究的意义 1 1.4 本课题研究的方法 1 2 小波理论 2 2.1 小波分析背景 2 2.2 小波变换简介 2 2.3从傅里叶变换到小波分析 2 2.4 提升小波变换概述 3 2.5 小波包分析 3 2.6 小波信号分析在声音处理中的应用 4 3 常见声音特征参数提取算法 4