Modelagem estocástica dos algoritmos adaptativos LMS-SAS e NP-VSS-NLMS no contexto de identificação de sistema

Abstract

This work deals with the stochastic modeling of the least-mean-square algorithm with symmetric/antisymmetric properties (LMS-SAS) and the nonparametric variable step size normalized leas-mean-square (NP-VSS-NLMS) algorithms, operating in a system identification setup with Gaussian (correlated and uncorrelated) input data. Specifically, model expressions are derived describing the mean weight behavior of the adaptive weight vector, learning curves, and evolution of some correlation-like matrices associated with the weight-error vector. Still, model expressions describing the steady-state behavior of NP-VSS-NLMS algorithm are obtained. Based on such expressions, some interesting characteristics of the considered algorithms are discussed, aiming to provide useful design guidelines. Simulation results are shown and discussed confirming the accuracy of the proposed models, for both transient and steady-state phases, under different operating conditions.