经验小波变换的理论算法研究及其在语音信号处理中的应用

发布时间：2018-04-20 09:38

本文选题：时频分析 + 经验小波变换　；参考：《南京理工大学》2016年硕士论文

【摘要】：时频分析方法作为分析时变非平稳信号的有力工具,成为现代信号处理研究的一个热点,经验小波变换算法是一种新型的时频分析方法。该算法打破了传统时频分析算法在自适应方面的局限性,结合了经验模式分解和传统小波变换的优势,可以将复杂信号分解为更具有物理意义的模式。本论文深入研究了经验小波变换算法的原理,对算法中存在的问题进行改进和优化,验证了经验小波变换算法作为一种新的时频分析方法在实际应用中的价值。具体而言,针对该算法在处理具有复杂频谱的信号时出现的频谱划分问题,提出了改进方法-—基于Top-Hat变换的经验小波变换算法,通过比较,验证了新方法的有效性；而后,我们把改进的经验小波变换算法应用到语音信号处理中来,针对汉语发音的特点,创新性地提出了一套完整的基于改进的经验小波变换的基频检测算法,通过大量仿真实验,验证该算法能够准确、有效地测出复杂信号的瞬时基频。与传统的基频检测算法相比,新算法具有高准确度,较好的鲁棒性以及高时间分辨率的特点,可以应用到工程领域中。
[Abstract]:As a powerful tool for analyzing time-varying non-stationary signals, time-frequency analysis has become a hot topic in modern signal processing. Empirical wavelet transform (EWT) algorithm is a new time-frequency analysis method. This algorithm breaks the limitation of the traditional time-frequency analysis algorithm in the aspect of adaptation and combines the advantages of empirical mode decomposition and traditional wavelet transform to decompose complex signals into more physically meaningful patterns. In this paper, the principle of empirical wavelet transform (EWT) algorithm is deeply studied, and the problems in the algorithm are improved and optimized, and the value of EWT algorithm as a new time-frequency analysis method is verified. Specifically, aiming at the spectrum partition problem of the algorithm when processing signals with complex spectrum, an improved method, empirical wavelet transform algorithm based on Top-Hat transform, is proposed. The validity of the new method is verified by comparison. We apply the improved empirical wavelet transform algorithm to speech signal processing. According to the characteristics of Chinese pronunciation, we creatively propose a complete fundamental frequency detection algorithm based on improved empirical wavelet transform. It is proved that the algorithm can accurately and effectively detect the instantaneous fundamental frequency of complex signals. Compared with the traditional fundamental frequency detection algorithm, the new algorithm has the characteristics of high accuracy, good robustness and high time resolution, so it can be applied to the engineering field.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TN912.3

【参考文献】