Multi-channel Hammerstein identification

Abstract

The Hammerstein systems, consisting of a zero-memory nonlinearity followed by a linear dynamic function, exists universally in industrial, chemical, physical and biological systems. Thus an effective modelling method for Hammerstein systems is critical for both relevant scientific research and engineering applications. We propose a novel Hammerstein identification approach, in which a multi-channel mechanism is used to separate the coefficients of the linear and nonlinear blocks more completely. Compared with traditional single-channel identification algorithms, the present identification method can enhance the approximation accuracy remarkably under the weak condition on the persistent excitation (PE) condition of the inputs.