Research on key parameters of speech denoising algorithm based on wavelet packet transform

Wavelet packet transform is an efficient method in speech denoising processing. In this paper, we research on various wavelet packet basis, decomposition layers, values of the threshold and threshold functions which are key parameters in wavelet packet denoising. Furthermore, we adopt three methods to evaluate the effects of denoised speech, including signal-noise-ratio(SNR), wavelet spectrum distortion measure(WSDM), and mean opinion score(MOS) algorithms. The experimental results show that if we choose the key parameters referred in this paper, we can obtain an ideal effect for speech denoising by the nonlinear filtering based on threshold in wavelet domain.