Title :
A new NLMS algorithm for slow noise magnitude variation
Author :
Gazor, Saeed ; Shahtalebi, Kamal
Author_Institution :
Dept. of Electr. & Comput. Eng., Queen´´s Univ., Canada
Abstract :
A set-membership (SM) normalized least-mean-square (NLMS) (SMNLMS) algorithm is developed using SM theory in the class of optimal bounding ellipsoid (OBE) algorithms. This signed version of NLMS algorithm requires a priori knowledge of a bound for the error magnitude, which is unknown in most applications. A very simple algorithm is proposed for the case in which the unknown magnitude of the measurement noise is slowly time-varying. The proposed algorithm is able to extract the noise magnitude information and exploit this magnitude to enhance or accelerate the learning process without risk of overbounding or performance loss due to underbounding. The performance of the proposed algorithm is compared with that of SMNLMS using some simulation examples.
Keywords :
adaptive equalisers; decision feedback equalisers; interference suppression; least mean squares methods; parameter estimation; signal processing; NLMS algorithm; OBE algorithm; SMNLMS algorithm; adaptive decision-feedback equalizer; error magnitude; interference cancellation; measurement noise; normalized least-mean-square algorithm; optimal bounding ellipsoid algorithms; overbounding; set-membership theory; signal processing; slow noise magnitude variation; underbounding; Acceleration; Convergence; Data mining; Decision feedback equalizers; Ellipsoids; Noise measurement; Parameter estimation; Parametric statistics; Performance loss; Samarium;
Journal_Title :
Signal Processing Letters, IEEE
DOI :
10.1109/LSP.2002.805312