Real-time frequency estimation of sinusoids with low-frequency disturbances

This paper mainly concentrates on extracting dominant frequency estimates from sinusoidal signals corrupted with low frequency disturbances perturbation by means of a synthetic identifier in a real-time manner. It is assumed that the frequencies of such signals are unknown, and the magnitudes of the dominant signals are sufficiently large compared with those of disturbances. A state space based model for such sinusoidal signal is presented. The proposed frequency identifier consists of a high-gain observer and a gradient based estimator. The utilization of the observer is for effectively attenuating the low frequency disturbances so that the dominant frequency components can be isolated. The gradient estimators can then operate on the estimated states from the observer to generate frequency estimation.