Time-Frequency Thresholding: A New Algorithm In Wavelet Package Speech Enhancement

In speech enhancement based on wavelet package, a new concept of time-frequency thresholding and the corresponding algorithm are presented. Based on the masking properties of human auditory system to wavelet denoising, the threshold is modulated to vary with the time-frequency characteristic of the speech´s pre-estimate, but not merely the time-variant information as has been adopted in former researches. And via the proposed thresholding, the wavelet package coefficients can be adjusted adaptively and a better denoised speech can be achieved. Experimental results show that compared with the former algorithms the proposed method and algorithm has better performance with less distortion and residual noise.