Use of eigenvector expansions in generalized Wiener system models

The authors consider the estimation of Wiener and Volterra kernels of finite dimension, finite order, finite memory nonlinear systems. The linear filters in a generalized Wiener model are constructed using the parallel cascade identification method. The authors demonstrate that these filters span the subspace occupied by the linear part of the system being identified. This approach is compared with the Laguerre expansion method, in which a complete orthonormal basis of the linear space is created. Simulations are used to illustrate the strengths and weaknesses of each approach