Systems identification with fuzzy spline wavelets

This paper describes the use of a nonlinear modeling procedure-building a fuzzy spline wavelets (FSW) model. FSW models are produced with a two step process: in the first step, the data set is compressed and denoised with the use of the fast wavelet transform; and in the second stage, the results of the first stage are presented as a fuzzy rule base with the use of linguistic variables. The hybrid model enjoys the excellent numerical and computational characteristics of the fast wavelet transform, combined with the ability to describe the accumulated knowledge in a human like way, in the form of simple IF...THEN rules using linguistic variables. The method is demonstrated for a truck backing problem.