Accurate estimation of frequency of a single sinusoid based on downsampling

A new phase-based approach for frequency estimation of a single cisoid in the presence of additive white noise is proposed in this paper. The main idea is to divide the observed data into a number of segments by downsampling and exploit this new structure for parameter estimation. The maximum likelihood estimator for frequency is then developed, which is shown to be superior to conventional phase-based methods in terms of uniform performance. Computer simulations also illustrate that the mean square frequency error of the proposed scheme can attain Cramér-Rao lower bound for sufficiently high signal-to-noise ratio conditions.