An estimation of time-varying signal frequency using recent finite observations

For estimating a time-varying signal frequency, an alternative estimator with a finite memory structure is proposed. The proposed estimator is developed under a maximum likelihood criterion using only the most recent finite observations on the window. The proposed estimator is first represented in a batch form, and then in a recursive form for computational advantage. The proposed estimator is shown to have good inherent properties, such as unbiasedness and deadbeat. Finally, via computer simulation and comparison, the proposed approach is shown to outperform remarkably the variable forgetting factor (VFF) Kalman filtering approach.