Empirical evaluation of Affine Projection algorithm for acoustic feedback cancellation in hearing AIDS

Abstract

A hearing aid equipment is used for amplifying the audio signal in order to enhance the hearing efficiency of people with hearing impairment. One major problem that hearing aids users usually encounter is screeching sound, which results from acoustic feedback in hearing aids. One effective and widely used solution for solving this problem is to employ an adaptive filter to produce signal for canceling out the acoustic feedback signal. Affine Projection (AP) algorithm is a new adaptive filter. This algorithm has properties that lie between those of the normalized LMS algorithm and RLS algorithm. It has less complexity than RLS and much faster convergence than normalized LMS, which depends on the projection order. This thesis evaluates AP algorithm in fixed point arithmetic for determining number of bits, and also the use of Levinson Durbin Recursion for inverting the autocorrelation matrix.