The perhaps most challenging problem in acoustic feedback cancellation using adaptive filters is the bias problem. It is well-known that a probe noise approach can effectively prevent this problem. However, when the probe noise must be inaudible and the steady-state error of the adaptive filter must be unchanged, this approach causes a significantly decreased convergence rate of the adaptive filter, and might thereby be less useful in practical applications. In this work, we propose a new probe noise approach which significantly increases the convergence rate while maintaining the steady-state error of the adaptive algorithm in a multiple-microphone and single-loudspeaker audio system. This is obtained through a specifically designed probe noise signal and a corresponding probe noise enhancement strategy. We show the effects of the proposed probe noise approach by deriving analytical expressions for its system behavior, and through a simulation experiment in an acoustic feedback cancellation system.
Proceedings of the Ieee Sensor Array and Multichannel Signal Processing Workshop, 2012, p. 497-500
2012 IEEE 7th Sensor Array and Multichannel Signal Processing Workshop, 2012