Tensor-based methods for Wiener-Hammerstein system identification

In this paper, we propose tensor-based methods for identifying nonlinear Wiener-Hammerstein (W-H) systems. In a first step, the parameters of the linear subsystems are estimated using two different approaches based on the PARAFAC decomposition of the fifth-order Volterra kernel associated with the W-H system to be identified. The first approach consists in applying the iterative ALS algorithm, while the second approach uses the TOMFAC algorithm. In a second step, the coefficients of the nonlinear subsystem modeled as a polynomial, are estimated by means of the RLS algorithm. The proposed identification methods are illustrated by means of simulation results.