Iterative optimization method of GOB-Volterra filters
| Authors: | Favier Gérard, UNSA/CNRS, France
Kibangou Alain, UNSA/CNRS, France
Hassani Moha, FSSM/UCAM, Morocco |
|---|
| Topic: | 1.1 Modelling, Identification & Signal Processing |
|---|
| Session: | Nonlinear System Identification - Volterra Methods |
|---|
| Keywords: | Volterra filters, Generalized orthonormal basis, Laguerre basis, Basis selection, Identification |
|---|
Abstract
An iterative method is proposed to optimize the Volterra kernels expansions on Generalized Orthonormal Bases (GOB). Each kernel is expanded on an independent GOB. The expansion coefficients, also called Fourier coefficients, are estimated in using an orthogonal formulation of the Least Squares (LS) algorithm. The proposed method allows optimization of both the Fourier coefficients and the GOBs poles. It can be seen as a good compromise between the exhaustive method for GOB poles optimization, costly by nature, and the analytical solution to Laguerre poles optimization that generally furnishes worse performances for system approximation than expansions on GOBs.
