A Censored Sample Mean Approach to Nonparametric Identification of Nonlinearities in Wiener Systems

A new, censored sample mean nonparametric identification algorithm for estimation of a nonlinear characteristic in Wiener system using properly preselected input-output data is proposed. Conditions imposed on the unknown characteristic are weak. In particular, its invertibility and global continuity are not required. The algorithm is based on computation of local sample-mean of proper output measurements. The mean square consistency of the estimate is proved for each continuity point of the unknown characteristic and the issue of the asymptotic convergence rate is discussed. Computer simulations are included to illustrate efficiency of the method also for small and moderate number of data.