Research of speech enhancement method based on Hilbert-Huang Transform and wavelet transform

摘要

Hilbert-Huang Transform(HHT) is a new and self-adaptable method for time-frequency analysis. Using empirical mode decomposition(EMD), nonlinear and non-stationary signals can be decomposed into several intrinsic mode functions(IMF), and each IMF has its own physical meaning. The theory of HHT is studied and a speech enhancement method based on EMD and wavelet transform is brought forward. Firstly, the noisy signal is decomposed into six IMFs and a residual signal with the EMD; secondly, each signal is applied wavelet threshold filter and gets seven new signals; finally the speech is reconstructed by the seven new signals. Experiments show that this method can improve the SNR, speech articulation and intelligibility, it is more effective than using direct wavelet threshold filter and the spectral subtraction.