Volterra Filter Identification Using Penalized Least Squares
Nowak, Robert David
Volterra filters have been applied to many nonlinear system identification problems. However, obtaining good filter estimates from short and/or noisy data records is a difficult task. We propose a penalized least squares estimation algorithm and derive appropriate penalizing functionals for Volterra filters. An example demonstrates that penalized least squares estimation can provide much more accurate filter estimates than ordinary least squares estimation.