NLMS algorithm with variable step-size using set-membership identification

Abstract

In this paper, set-membership identification is used to derive a simple algorithm which is a sign version of the normalized least mean square algorithm. Convergence analysis is carried out. With some simulation examples, the performance of the algorithm, in the cases of slow and fast variations of a parameter, is compared with the modified Dasgupta-Huang optimal bounding ellipsoid algorithm. These examples show the performance of the proposed algorithm.