Adaptive speech enhancement method based on hearing wavelet packet transforms

An auditory model based on wavelet packet is established. The relationship between the soft and hard threshold of the traditional wavelet analysis is discussed. A new adaptive dynamic threshold calculation method is proposed according to the relationship among wavelet thresholds that Donoho introduced. Experimental results show that the algorithm deal with the problems of music noise better and the speech enhancement effect is also better than the traditional algorithms regarding to the clarity and the intelligibility.