Adaptive filter for perceptual speech enhancement

In this paper, we present a new perceptual speech enhancement method. It is based on an improved spectral subtraction filter and a spectral attenuation filter. In the classic noise reduction techniques, both musical noise and distortions are introduced in the enhanced speech signal. Consequently, the perceptual appearance is affected. Within this context, our method consists in making the musical noise, produced by the classical methods, inaudible. Evaluation tests prove the effectiveness of the proposed method in reducing noise and also in minimizing distortions in the proceeded speech signal.