Adaptive estimation of third-order frequency domain Volterra kernels

Abstract

Techniques for adaptive estimation of frequency domain Volterra kernels of non-linear systems up to the third order are investigated. A simple adaptive algorithm for identifying the frequency domain Volterra kernels is derived based on the assumption that the input is Gaussian. Compared with the conventional adaptive method based on the recursive least squares (RLS) technique, the proposed method requires substantially less computation and computer memory. The insensitivity of the proposed method to the Gaussianity assumption of the input is also demonstrated via computer simulation and application to experimental data. The results show that the proposed adaptive method is comparable in performance with the conventional RLS-based adaptive method even when the input is non-Gaussian.