An single speech enhancement based on complexity indicators and EMD

Abstract

Speech enhancement is performed in a wide and varied communication system. The ensemble empirical mode decomposition and noise-assisted multivariate empirical mode decomposition have been proposed to improve the empirical mode decomposition more suitable for processing non-linear and non-stationary signals. In this paper, we applied the ensemble empirical mode decomposition and noise-assisted multivariate empirical mode decomposition algorithms on a single speech enhancement system and selected the significant intrinsic mode functions by the complexity indicator. The results demonstrated that the complexity indicator could choose intrinsic mode functions properly and achieve significantly improved speech quality.