A Novel Speech Enhancement Method Based on the Constraints of Temporal Continuity

Abstract

The temporal continuity of speech is a vital feature, whose utilizing makes speech enhancement better. By adding the feature in this paper, we present a novel speech enhancement method based on temporal continuity constrained low-rank sparse matrix decomposition (TCCLSMD). This approach makes up for the deficiencies of the constrained low-rank and sparse matrix decomposition (CLSMD) by leading into the temporal continuity constraints. The proposed approach based on the sparse matrix obtained by singular value decomposition, and the discrete sparse matrix is reduced by adding temporal continuity to reduce discrete isolated points, retaining more speech information and reducing the speech distortion. Under various kinds of noise settings, compared with the CLSMD method, the experimental results show that the proposed method reduces speech distortion, makes the residual noise less, and raises speech intelligibility.