The interplay between the time and frequency domains for linear systems is well known and most useful. In the case of identifying linear models from observed da ta, this interplay manifests itself in two ways.

The observed data is of course primarily recorded in the time domain (even though the recording equipment may deliver them in the frequency domain). To build linear models from the data we can either do the fitting in the time domain and evaluate the resulting modle's properties in the frequency domain. We can also transfer the data themselves to the frequency domain and fit models directly there. In this contribution we shall consider some aspects of these two possibilities.

Back to Adaptive Control, Filtering, and Signal Processing Table of Contents