Detection and estimation of frequency-random signals (Corresp.)

This is the authors' abstract. We don't add key points for this paper.

A random frequency-varying signal is approximated by a finite-parameter model. A maximum { a posteriori} estimate of the parameter set leads to an estimation criterion, which when maximized is also the decision statistic obtained with an estimator-correlator detector. An efficient implementation of the frequency-parameter estimate (which corresponds to maximizing the criterion) is presented. The resulting estimate is compared to previous techniques found in the literature. Finally, performance bounds for the resulting detector are evaluated and compared to the asymptotically optimum low-energy coherence detector.