Analysis of envelope detected mean level CFAR processors using importance sampling

Importance sampling is applied to the analysis of mean level CFAR processors. It is shown that, in contrast to density modifications such as increased variance employed to estimate tail probabilities in conventional importance sampling, in the CFAR case the variance of the random variables comprising the adaptive threshold need to be reduced. Two schemes for implementing this concept are analysed. The simulation gain afforded by this procedure increases as the CFAR window length decreases. In application it is found that the cell averaging, greatest of, and smallest of CFAR detectors preceded by an envelope detector display robust performance in the presence of Gaussian clutter power transitions and interfering targets as compared to the square law detector. For all three structures, estimates of threshold settings that provide desired false alarm probabilities have been determined for the envelope detector using our IS techniques. It is concluded that the envelope detector is a better choice than the theoretically optimum square law detector