Go To Program
136

A Noise Robust Multichannel Algorithm for Blind Estimation of Room Impulse Responses


Liao Lei 1 Khong Andy W. H. 2
1 Electrical and Electronic Engineering, Nanyang Technological University, Singapore
2 Electrical and Electronic Engineering, Nanyang Technological University, Singapore

Blind system identification~(BSI) algorithms for speech dereverberation have generated much interest in recent years. In this paper, we revisit the misconvergence problem that exists in the normalized multichannel frequency-domain least-mean-square (NMCFLMS) algorithm under a noisy environment. To address this misconvergence and to achieve a high rate of convergence, we propose to introduce a power and direct path constraint to the NMCFLMS algorithm.
Simulations results show an improvement both in convergence rate and noise robustness compared to the direct path constrained only NMCFLMS algorithm.


View pdf