An Improved Adaptive Exponential Functional Link based NAEC

Abstract

Nonlinear distortion is an inevitable phenomena in hands-free devices. It is of particular concern in acoustic echo cancellation systems, as the distortion will result in poor echo cancellation performance. Functional link (FL) algorithms which fall under the category of linear-in-parameter model have shown promising results in literature under the influence of distortion. These FL algorithms depend on trigonometric functions in modeling distortion. Recently, adaptive exponential FL algorithm has been proposed for nonlinear system identification. Motivated by the improvement brought by this approach, in this work, a modified adaptive exponential algorithm is proposed to model sparse RIR under the influence of distortion. Simulations carried out validates the improvement in performance.