An Improved Wavelet Packet Threshold Function for Speech Enhancement Method

摘要

Speech enhancement method based on the traditional wavelet threshold function had larger distortion and constant deviation. In order to solve the problems, an improved wavelet packet threshold function was presented and used in speech enhancement. In the new method, the nonlinear function was applied to compress the wavelet packet coefficients less than threshold, it ensured that enhanced speech was continuous, and thus removed the distortion. Furthermore, the improved method could change freely hard function into soft function by selecting the different parameters, that flexible adjustment character overcame the flaw of the constant deviation between the true and estimated wavelet packet coefficients, thus enhanced speech quality was modified. Simulation experimental results show that the white Gaussian noise and colored noise are removed largely, the Signal-to-noise Ratio (SNR) of enhanced speech are increased, speech distortion is reduced effectively, compared with the existed speech enhancement algorithms based on wavelet threshold.