An efficient feedback canceler for hearing aids based on approximated affine projection

Abstract

In the feedback cancelation systems in hearing aids, signal cancelation and coloration artifacts can occur for a narrow-band input. In this paper, we propose a new adaptive feedback cancelation algorithm that can achieve fast convergence by approximating the affine projection (AP) algorithm and prevent signal cancelation by controlling the step-size. A Gram-Schmidt prediction error filter (GS-PEF) is used for a stable approximation of the AP algorithm, and the step-size is varied using the prediction factor of the input signal. Simulations results are presented to verify efficiency of the proposed algorithm.