Model-Based Kalman Filtering Approaches for Frequency Tracking

Extended Kalman Filter is commonly used in the framework of frequency estimation. This approach relies on the definition of a state-space signal model suitable for frequency tracking purposes. Moreover, a typical trade-off involved in the selection of the filter design parameters, regards both the robustness and the performances of the estimation technique. In this paper, a novel third-order signal model is presented, beside the classical harmonic signal representation. The Unscented Kalman Filter approach is then applied in the context of frequency estimation. This technique is compared with the more classical Extended Kalman Filter by means of suitable criteria to probe performance and robustness.