ON THE ESTIMATION OF PARAMETERS FOR FREQUENCY-DOMAIN MODELS

The problem of estimating parameters for frequency domain models is considered. Past approaches have most commonly based estimation criteria upon modulus and phase transformations of the model and sample frequency response functions. As an alternative, a complex vector estimation criterion is proposed and is implemented in a nonlinear, Gauss-Marquardt optimization algorithm. When compared to the previous methods of fitting modulus and phase transformations, the complex vector estimation methodology has less bias and variance and is more robust.