A spectrum matching method for accurate frequency estimation of real sinusoids

It is well-known that incoherent sampling is detrimental for frequency estimation of a real sinusoid, and the estimation errors get worse when the signal lengths are very short. In this paper, a spectrum matching based frequency estimator is proposed as well as evaluated against other four two-step methods developed to suppress the effect of incoherent sampling. The spectral interference introduced by incoherent sampling is eliminated via a spectrum matching process including modulation and spectral analysis. A further error correction based on Fourier transform is conducted to generate the fine frequency estimate. Simulation results are carried out to show that the proposed method can closely approach the Cramér-Rao lower bound without any error floor, and it can outperform the other four methods particularly for short signal lengths.