Implementação Eficiente de Filtros de Hammerstein Adaptativos Usando o Algoritmo NLMS

This paper presents a novel strategy to implement adaptive Hammerstein filters aiming to circumvent numerical problems arising from the cascade adaptive process inherent to these filters. Such problems are often overcome maintaining one of the filter coefficients fixed, which may significantly impair the convergence speed of the adaptive algorithm. The proposed strategy uses a coefficient normalization procedure, which avoids maintaining one of the coefficients fixed and thus permits to obtain a faster convergence speed. As a result, an adaptive algorithm with very good performance and acceptable computational cost is obtained. Simulation results are presented attesting the effectiveness of the proposed strategy