On the Choice of Filter Bank Parameters for Wavelet-Packet Identification of Dynamic Systems

摘要

This paper is concerned with a recently proposed technique for linear system identification in frequency subbands using wavelet-packet filter banks. More specifically, the effect of using different mother wavelets and resolution levels is investigated. The study is based on simulated examples involving the identification of a servomechanism model. The results reveal that the identification outcome can be improved by using wavelet niters with better frequency selectivity, as well as by increasing the number of resolution levels in the filter bank. In this context, the advantages of using wavelet packets instead of standard wavelet decompositions are also discussed.