Robust Noise Canceller Algorithm with SNR-Based Stepsize Control and Gain Adjustment

Abstract

This paper proposes a robust noise canceller algorithm with SNR-based stepsize control and gain adjustment. Use of estimated SNRs for stepsize control reduces interference by the target signal in adaptation. A second SNR estimate, which is the output over an adjusted reference input, initially controls the stepsize to promote coefficient growth, followed by a first SNR estimate which is the output over the noise replica. Changeover from the second to the first SNR estimate takes place when the coefficient growth is saturated. The power gap between the reference input and the noise to be cancelled is adjusted by a factor estimated during an initial period. Evaluations with clean speech and noise recorded at a busy station demonstrate that the coefficient error by the proposed algorithm is as much as 8dB smaller than that without gain adjustment whereas conventional algorithms exhibit initial increase in the coefficient error and never reach the switchover status at a high SNR.