Abstract:
This paper presents a new speech enhancement method. This method based on wavelet
filters via multistage convolution with Reverse Biorthogonal Wavelets (RBW) in the
high and low pass band frequency parts of the speech signal. In this method a speech
signal is decomposed into two parts; a high-pass and a low-pass, the noise is then
removed in each band individually in different stages via wavelet filters. The proposed
method provides good results where it does not cut the speech information, which
occurs when utilizing conventional thresholding. The proposed method is tested and the
objective evaluation is used to compare the results with the other used methods. The
new proposed method shows superiority over Donoho and Johnstone thresholding
method and Birge-Massart thresholding strategy method.