Recursive identification of continuous-time Wiener systems (Q4485274)

Recursive algorithms for identification of component subsystems of a continuous-time Wiener system, i.e. cascade connection of a continuous-time linear dynamic subsystem and a nonlinear static element, are presented. The system is driven by a Gaussian white random signal and disturbed by a Gaussian white random noise. The impulse response of the linear dynamic part is recovered with a correlation method. As to the static element, it is shown that the inverse of the nonlinear characteristic is a regression function. Then, to recover this inverse, two regression estimates are presented. Convergence conditions of the algorithms, respectively, to the unknown impulse response and the inverse of the nonlinear characteristic, are given and convergence rates are evaluated. Illustrative simulation examples are included.