ERROR ANALYSIS AND OPTIMAL ESTIMATION PROCEDURES IN IDENTIFICATION OF NON-LINEAR VOLTERRA SYSTEMS

In the problem of nonlinear system identification, the white-noise (crosscorrelation) method has become increasingly popular during the last five years. Since the ideal white noise is physically unrealizable, the method employs quasi-white test signals which approximate its statistical properties to a satisfactory and determinable degree. A family of such signals (CSRS) which has been recently introduced appears to be most advantageous in nonlinear system identification. Its principal advantage is that it allows the thorough and effective error analysis and, consequently, the optimization of the identification procedure. This paper presents that error analysis and outlines the optimal procedures which ought to be followed for the optimal application of the method to nonlinear Volterra systems identification. © 1979.