System Identification in the Frequency Domain

An emerging new technique of system identification is the maximum likelihood estimation of transfer functions in the frequency domain. The basic ideas are described in the book, Johan Schoukens and Rik Pintelon, "Identification of Linear Systems - A Practical Guideline to Accurate Modeling" (Pergamon Press, 1991). In cooperation with the authors, István Kollár has developed the Frequency Domain System Identification Toolbox for Matlab. This toolbox comprises the whole identification procedure from excitation signal design, through data preprocessing, parameter estimation, presentations to model validation. The toolbox is being distributed by Gamax, Inc. Lately its Matlab7 compatible version has also been developed.

There are several ways for further developments, as

• graphical user interface (GUI) which offers simple to use graphical tools for the use of the core functions
• extension of the methods from SISO to MIMO systems
• inclusion of new functionalities and design methods like
• crest factor minimization using Deczky's method
• use of a smaller-bias logarithmic cost function
• use of orthogonal polynomials for identification of high-order systems
• etc.
• exploration of conditioning issues and introduction of new methods with better conditioning

The work is done in cooperation with the Department Algemene Electriciteit (Dienst ELEC) of the Vrije Universiteit Brussel (Belgium).

Related publications