On the Accuracy of Generalized Hammerstein Models for Nonlinear Active Noise Control

In this paper we extend a filtered-X affine projection (AP) algorithm for active noise control to controllers exploiting the recently proposed functional link artificial neural network (FLANN) based on trigonometric expansions. In fact, these structures belong to the class of filters characterized by a linear relationship between their output and the filter coefficients. Such a class includes linear filters as well as Volterra filters, for which filtered-X AP algorithms have already been derived. We show that the mentioned FLANN structures can be efficiently replaced by novel schemes using simple polynomial terms. These structures constitute a particular class of Volterra filters that we call generalized Hammerstein models