Acoustic Feedback Cancellation Using Optimal Step-Size Control of the Partition Block Frequency-Domain Adaptive Filter

摘要

Acoustic feedback control is very critical in systems like hearing aids and public address systems. In hearing aid, lack of control of acoustic feedback can lead to instability of the system and causes annoying sound to the user. To control the acoustic feedback adaptive filtering approach is widely used. Limited tracking capability and slow convergence of the adaptive algorithms are also important limitations to be considered. However in hearing aid applications, for improving battery life, avoiding high computational complexity and long delays with appreciable feedback control is important. In this paper we proposed an AFC algorithm combining with decorrela-tion properties, by means of the prediction-error method (PEM), low delay and low computational complexity by means of a partitioned-block and frequency-domain implementation (PBFDAF), and included the optimal step-size (OSS) technique for achieving the fast convergence and low steady-state misalignment. Also we incorporated feedback path change detector (FPCD) in the proposed algorithm to improve convergence in the scenario of non-stationary feedback paths.