A novel single channel speech enhancement using time frequency mask

Abstract

For speech with Gaussian noise, a novel single channel speech enhancement algorithm is proposed in this paper. The algorithm applies time frequency mask to delete noise. It firstly splits the time frequency domain into two areas: Joint Areas (JA) and Disjoints Area (DA), and then the signals are thrown away in the JA, and the noise on the “cross” in the DA is deleted. Furthermore it applies the smooth filter to compensate the distortion. Finally, several speech signal experiments demonstrate that the algorithm performance and practical.