A Modified Adaptive Line Enhancer for Noisy Speech Signals

Abstract

Adaptive filter is a method for enhancing noisy speech signals. In this work, a refinement for the adaptive line enhancer proposed by Widrow and Sambur is introduced. The new algorithm is based on both the exact value of gradient and variable step-size. These values are computed and used for coefficient updating. The algorithm is applied on different cases of noisy speech signals as well as GSM signals. The new algorithm gave good results even when noise is greater than signal which is considered a novel achievement in the field. It avoids the drawbacks of LMS algorithm introduced by widrow and others. The adaptive line enhancer (ALE) design method updates the FIR filter coefficients in terms of the exact value of gradient and a variable step size.