A new class of gradient adaptive step-size LMS algorithms

Authors:
  • BFB. Farhang-Boroujeny
    B. Farhang-Boroujeny

Abstract
The gradient adaptive step-size least-mean-square (LMS) algorithms [an important family of variable step-size LMS (VSLMS) algorithms] are revisited. We propose a simplification to a class of the studied algorithms and show that this leads to a new class of VSLMS algorithms with reduced complexity but with no observable loss in performance.

This publication has 7 references indexed in Scilit: