Speech signal enhancement through wavelet domain MMSE filtering

A new speech enhancement system that combine robust signal enhancement and minimum signal distortion is proposed in this paper. The proposed method introduces frequency depended, parametric, MMSE filtering techniques that involve wavelet packets. Voice activity detection (VAD) is used to further distinguish speech from noise and help to adaptively remove noise components from color noise eruptive noisy speech, while perceptual criteria is also taken into account. Experimental results and objective quality measurement test results validate the proposed speech enhancement system and illustrate the benefit of the proposed wavelet domain MMSE filtering as an excellent speech enhancement method to provide sufficient noise reduction and good intelligibility and perceptual quality, without causing considerable signal distortion and musical background noise method.