LAGUERRE FILTER REDUCED-ORDER L(1) IDENTIFICATION

Calculable l(1) upper bounds for the truncation error of Laguerre filt ers are presented. A new solution to the finite-dimensional l(1) appro ximation problem is also detailed. The numerical performance of this a lgorithm is shown to be better than existing solutions. An l(1) identi fication method, which makes use of the finite-dimensional l(1) approx imation algorithm, is developed. The identified models are reduced-ord er fixed-parameter filters, which are approximations of high order non parametric models. Examples of the identification method are produced using Laguerre models. The examples considered strongly suggest that a trade-off is necessary between: truncation errors, errors due to the choice of model order, errors due to the choice of the Laguerre parame ter.