A Wavelet Packet Speech Enhancement Algorithm Based on Energy Entropy Improvement

For the wavelet packet speech enhancement can not accurately strip or even make the processed signal distortion and other cases of random non-stationary noise signals, the entropy threshold estimation method is used to estimate the energy structure, which solves the problem of irrational threshold allocation on the wavelet packet noise reduction; at the same time, the threshold function has been improved, and the entropy coefficient adjustment of the new function is introduced, which solves the selection of the threshold function in the process of traditional wavelet packet speech enhancement. Problem. The problem of selecting the threshold function in the traditional wavelet packet speech enhancement process is solved. In order to verify that the proposed speech enhancement algorithm has obvious advantages in dealing with highly complex random non-smooth signals, the improved wavelet packet noise reduction method and multi-threshold noise reduction method are subjectively and objectively evaluated using noisy speech signals.