Adaptive period estimation of a class of periodic random processes

Abstract

The problem of period uncertainty when evaluating spectrum estimates for wide sense cyclostationary processes is addressed in this paper. In particular, the extended Kalman filter (EKF) and a parallel bank of Kalman filters are investigated as different methods for adaptive estimation of a time-varying period. An example is given concerning an AR(1) process and a number of time-varying periods are adaptively tracked for different periodic functions. Convergence characteristics are also assessed. Finally, a combined detection-estimation approach is also investigated.