Research On Theoretical Algorithm Of Empirical Wavelet Transform And Its Application In Speech Signal Processing

As a powerful tool for analyzing time-varying non-stationary signals, the time-frequency analysis method has become a hotspot in the research of the modern signal processing. The empirical wavelet transform (EWT) time-frequency analysis is a novel kind of analysis method. The algorithm which combines the empirical mode decomposition and the advantages of the traditional wavelet transform breaks the limitations of traditional time-frequency analysis. By empirical wavelet transform, complex signals can be decomposed into a series of models reasonably.This paper mainly studies the principle of this new algorithm. Aiming at the problems of the original empirical wavelet transform, we proposed a new method to achieve optimization. Experiments and validations of the EWT algorithm has shown its value as a kind of new time-frequency analysis method in the practical application. More specifically, when dealing with signals which have complex spectrum, the EWT algorithm meet a problem we called it "spectrum subdivision". To improve the performance of the existing method, we put forward a new method based on the Top-Hat transform wavelet transform algorithm, and by comparison, we verified the effectiveness of the improved algorithm. And then we apply the improved EWT algorithm to the speech signal processing, aiming at the characteristics of Chinese pronunciation, we put forward a complete pitch detection system based on improved EWT creatively, and the results of a great amount of simulation experiments verified that the algorithm can detect the instantaneous pitch of Chinese accurately and effectively. The improved wavelet transform algorithm has a high accuracy and strong robustness, as well as the high time resolution, it can be applied to engineering application.