Algoritmo ZA-PNLMS com fatores de ativação individuais e ganhos dos coeficientes limitados superiormente

Abstract

Adaptive filtering algorithms have been widespread in academia. An important challenge in this area is the identification of sparse plants. With the aim of identifying plants with different sparsity degrees, this work proposes an adaptive LMS proportional normalized algorithm with zero attractor (ZA-PNLMS – zero-attracting proportionate normalized LMS algorithm), which combines individual activation factors with earnings limited from above. The proposal presented here superiorly limits the gains of the algorithm's coefficients for identifying plants with a high degree of sparsity, leading to a better distribution of the adaptation energy between the algorithm's coefficients. Computational simulations, considering sparse systems with perturbation and tracking, attest that the proposed algorithm is capable of agglutinating the characteristics of good convergence in transient state and low error in steady state.