Conventional spectral analysis fails in identifying systems operating with feedback from noisy input-output data. It is shown that polyspectral and cyclic methods can identify closed-loop systems with data corrupted by a class of stationary and cyclostationary noise. The class is specified, and nonparametric transfer function estimators are developed and simulated.
Bibliographical noteFunding Information:
Manuscript received April 1, 1994. This work was supported in part by ONR Grant NOOO14-93-1-0485, The author is with the Department of Electrical Engineering, University of Viginia, Charlottesville, VA 22903-2442 USA. IEEE Log Number 9408790.