An ARMA model identification algorithm

R. Moddemeijer

University of Twente, Department of Electrical Engineering,
P.O. Box 217, NL-7500 AE Enschede, The Netherlands,

present addres

University of Groningen, Department of Computing Science,
P.O. Box 800, NL-9700 AV Groningen, The Netherlands,
phone: +31.50.363 3940 - fax: +31.50.363 38005 - e-mail: rudy@cs.rug.nl

Abstract

To identify from electroencephalogram (EEG) signals the mechanism, which causes the spreading of epileptic seizures in the brain, we use a model identification algorithm. We have choosen for autoregressive and moving average (ARMA) modeling. We present an off-line maximum likelihood (ML) of least squares algorithm, based on an iterative Gauss-Newton minimization. A sytematic parameter search is integrated in the inversion of the Hessian-matrix. The optimal model is selcted with the Akaike criterion. We are able to identify a model in a large parameter space unsing only a few active parameters. Finally we present some promising results.

Publication

Seventh Symposium on Information Theory in the Benelux, May 22-23, 1986, Noorwijkerhout, The Netherlands, Ed. Boekee, D.E., pp. 151-159, Werkgemeenschap Informatie- en Communicatietheorie, Enschede, ISBN 90-6275-272-1, BibTeX

Related work


other publications