Yael Ehrlich

Identfication of Arbitrarily Time-Variant Systems

August 7, 1997
11:00 AM
206 Egan

Abstract

Many systems encountered in engineering applications are time-varying in nature. In the past these were often modeled as time-invariant or slowly-varying systems, resulting in model mismatch and a growing interest in adequate processing for identification of such systems. Appropriate schemes for the processing of data from such systems are of interest in the modeling of communication channels, power systems, and others, and can be applied to control mechanisms, echo cancellation, time-delay estimation, weak signals measurements, biological signals, array processing, and more.

In this dissertation we introduce a technique for identification of systems with arbitrarily time-variant responses from measurements of their (stationary) input and (nonstationary) output signals, and without any prior assumptions about the unknown system response. This is in contrast to currently available procedures, all of which rely on prior statistical characterization of the time-variant system response. We present several examples to demonstrate the utility of our approach and the quality of the resulting estimates. Finally, a way to implement the provided solution in a lattice (order-recursive) form is also presented and analyzed.

Thesis Committee:
Prof. H. Lev-Ari (advisor)
Prof. D. Brady
Prof. J. Preisig