Identification of Wiener models with data pre-filtering

Abstract

Data pre-filtering, i.e. filtering both the input and output data with an identical scalar filter, is a useful tool in the identification of linear time-invariant systems but is generally not applicable to nonlinear system identification. We show that, if the correlation approach is used to identify Wiener models, which are nonlinear, pre-filtering can be used if the input excitation is Gaussian.