BLIND IDENTIFICATION OF NONMINIMUM-PHASE FIR SYSTEMS - CUMULANTS MATCHING VIA GENETIC ALGORITHMS

In this paper, a new method for estimating the parameters of a nonmini mum phase, linear, time-invariant (NMP-LTI) system, from only its outp ut measurements, is presented. The method makes use of genetic algorit hms (GA) to minimize a cost function defined in terms of system's outp ut cumulants. It is demonstrated by computer simulations that the new method is insensitive to model order mismatch and is capable of identi fying the parameters of an NMP system using output data which may be c orrupted by additive Gaussian noise. (C) 1998 Elsevier Science B.V. Al l rights reserved.