Parametric Hammerstein-Wiener model estimation via dual Hammerstein identification

In this paper, we propose an approach to identify parametric Hammerstein-Wiener models. The approach identifies two Hammerstein models alternately, recovering the intermediate signal and parameters in both linear dynamic blocks and static nonlinear blocks. Identification of Hammerstein models can be implemented by using the iterative method or the two-stage over-parametrization method, both leading to efficient computations. Simulation results show that our approach converges fast, and is robust when the input or output blocks are highly nonlinear.