Regularized Least Square based Identification for Wiener Systems

This paper presents a regularized least square algorithm for Wiener system identification by using bilinear parameterized formulation. This method is based on the decomposition of model form and assumes the invertibility of the non-linearity involved in the modeling of wiener model. In addition, the output of the linear block is corrupted with noise signal resulting in a model with correlated noise disturbance. The standard least square solution provides unregularized estimates in the presence of correlated noise disturbance. Therefore, an approach based on the decomposition of the wiener model form is formulated under the regularity constraint on the coefficients of basis functions to be used to model the nonlinearity. Simulation examples are given in the presence of noise to show the effectiveness of the decomposition based regularized least square iterative method.