An affine projection algorithm with a data-selective method of using the condition number

Abstract

This paper proposes a novel affine projection (AP) algorithm that offers a data-selective method using the condition number as a criterion for low misalignment. Capable of keeping the condition number of an input data matrix small, the proposed AP algorithm adopts a recently reported technique of E-norm. instead of L-2-norm to reduce the computational burden during the estimation of the condition number. The simulation results show that the proposed algorithm has a lower misalignment than the conventional algorithm in the steady state. (C) 2007 Elsevier B.V. All rights reserved.