Frequency estimation via multiple lags of correlations in the presence of broadband colored noise

This paper provides the performance analysis of frequency estimation using multiple lags of correlations in the presence of additive white Gaussian noise and broadband colored noise. For the analysis, the colored noise is modeled as a moving average process. The threshold performance is investigated and different phase unwrapping techniques useful for correlation-based frequency estimation are compared. It is shown that the phase unwrapping obtained via a DFT coarse search has the best performance in the presence of additive white Gaussian noise. However, the DFT peak search is shown to be biased in the presence of colored noise. As such, a new phase unwrapping technique, using the method of iterative filtering, is proposed. It is shown that using this phase unwrapping, an estimator based on multiple lags of correlation achieves the CRB while maintaining a low SNR threshold, in the presence of colored noise.