Research And Application Of Speech Signals Enhancement Based On Hilbert-Huang Transform

The speech signal is general and very important. The actual speech signal becomes very weak because of background noise and environmental noise interference. Speech signal enhancement is a important way to remove noise. A signal processing method must be used to enhance the speech signal, which suppresses background noise and improves the quality of voice communication and system performance. Therefore, in the actual background noise circumstances, it is of great importance to research speech signal enhancement techniques.Fourier transform is the earliest theory of signal analysis in frequency domain. It is a very effective method in analysis of linear stationary signals. However, the Fourier frequency f is a overall variable of time-independent. Speech signals are typically non-stationary and time-varying. The Hilbert-Huang transform based on empirical mode decomposition is of adaptive, multi-scale features, especially suitable to analyze non-stationary and nonlinear signals.In this paper, Hilbert-Huang Transform is used to enhance speech signal with noise. Speech signal is decomposed into a set of independent intrinsic mode function components by empirical mode decomposition method; analytic signal is constructed by Hilbert transformation to research transient characteristics of the signal; then spectral analysis of each intrinsic mode function is did through Fourier transform. Depending on the spectral characteristics of the speech signal and the noise, the new algorithm of speech signal enhancement based on HHT is proposed to remove the noise of the speech signal. The main contents are as follows:First of all, describing the principles of Fourier transform, time-frequency analysis methods and Hilbert transform, discussing spectral features of Hilbert transform and analytic signal and their application in signal analysis. Then, the paper discusses the theory and the algorithms of Hilbert-Huang transform based on empirical mode decomposition, and implements experimental simulation. Lastly, According to the distribution of intrinsic mode functions and the different Hilbert-Huang spectral features and energy of the speech signal and the noise signal, a new algorithm based on EMD speech signal enhancement is put forward. Speech signal enhancement algorithms based on Hilbert transform and spectral subtraction and wavelet threshold is proposed, simulation results show the superiority of the method.