Parameter Estimation Algorithms for Hammerstein–Wiener Systems With Autoregressive Moving Average Noise

Abstract

Hammerstein–Wiener (H–W) systems are a class of typical nonlinear systems. This paper studies the gradientbased parameter estimation algorithms for H–W nonlinear systems based on the multiinnovation identification theory and the data filtering technique. The proposed methods include a generalized extended stochastic gradient (GESG) algorithm, a multiinnovation GESG (MIGESG) algorithm, a data filtering based GESG (FGESG) algorithm and a data filtering based MIGESG algorithm. Finally, the computational efficiency of the proposed algorithms are analyzed and compared. The simulation example verifies the theoretical results.