On Kernel Design for Regularized Volterra Series Model with Application to Wiener System Identification

In this paper, we propose a kernel design method for the regularized identification of Volterra series models. In particular, we design the kernels based on two types of prior knowledge: one is the structural knowledge for block-oriented models and the other one is the prior knowledge for LTI systems. The proposed method is also applied to the identification of Wiener systems. The efficacy of the proposed kernels is verified in Monte Carlo simulations.