Identification of non-linear systems by using wavelet expansions

The paper addresses the problem of identification of nonlinear characteristics of systems from a class of complex block-oriented dynamic systems. Non-linearities are recovered from the noisy input-output measurements. Wavelet functions with compact supports (Daubechies wavelets) are used in the identification algorithms. Convergence of the algorithms is shown and the asymptotic convergence rates (true for a number of measurements tending to infinity) are given. These theoretical results are supplemented by a set of numerical experiments in which performance of the algorithms is additionally tested for small and moderate number of measurements.