A Wavelet Packet Threshold De-Noising Algorithm for Speech Enhancement Based on Maximum Information Entropy
Hongli Xu · Measurement & Control Technology · 2011
De-noising algorithm plays a very important position in the speech enhancement,while the traditional wavelet threshold de-noising algorithm will cause loss of part of useful speech signals inevitably.In order to denoise the speech signals with noise better,a wavelet packet analysis method for voice decomposition is seleted, and a new threshold function is used.The threshold and the weighted factors in the weighted threshold function are determined by the theory of maximum information entropy.The simulation results show that the posed algorithm can suppress noise better,and ensure the quality of the reconstructed signal.The de-noising effect is better not only in the audio effects but also in the signal to noise ratio.