A low complexity noise reduction approach for stationary noise

Abstract

A novel multi-microphone speech enhancement method with low computational complexity using the generalized sidelobe canceller (GSC) structure based on LMS (least mean square) algorithm is proposed under stationary noise condition. In adaptive module of the presented method, speech signals in one frame share the same weight and update simultaneously instead of calculating weights respectively. Experimental results show that, in stationary noise environment, the proposed method which employs modified LMS algorithm outperforms conventional LMS algorithms in noise reduction, while the computational cost is lower, compared with conventional ones.