Wavelet-Based Identification of Linear Discrete-Time Systems: Robustness Issue

DISCRETE-TIME SYSTEMS Milo s Doroslova cki H. (Howard) Fan ;1 Lei Yao Department of EECS, The George Washington University, Washington, DC 20052, USA Department of ECECS, University of Cincinnati, Cincinnati, OH 45221-0030, USA Abstract: Discrete-time linear time-varying systems are modeled by discrete-time wavelets. System identi cation using the output error method is studied. The output of the unknown system is in general corrupted by noise. The minimum mean square error, minimum cumulative mean square error and least squares are considered. The optimal system model parameters are found. Conditions are derived that provide consistency to the optimal parameters obtained by the least squares approach. It is shown that due to the good time-frequency localization of wavelets, parameter estimates are robust to narrow-band noise and/or impulse noise.