Accurate Speech Enhancement Method using Compressive Sensing with Variable Sparse Level

Abstract

In this paper, we propose an accurate speech enhancement method using Compressive Sensing with Variable Sparse Level. Compressive sensing (CS)-based methods can efficiently enhance speech signal quality but they often require a noise level in advance of speech enhancement processing. We try to estimate the noise level from a given input signal, and introduce some threshold levels for input amplitudes so that speech components are effectively enhanced. Performance of the proposed method is evaluated through computer simulation.