Order statistic filters as postdetection processors

The performance of a r a d a r detection system can be improved through postdetection processing. Two popular postdetection processors are the n-pulse and binary integrators. This paper examines the performance of the binary integrator from the viewpoint of order statistics and describes the binary integrator as a n order statistic (OS) filter. A general analysis is developed for OS filters in detection systems and is used to show the OS filter is a consistent and biased estimator of the quantities of the received signal distributions. The features of the OS filter and n-pulse integrator that are critical to detection performance are compared to determine the proper application of each processor. It is shown that the OS filter can be used to emphasize selective regions of the input distributions where good statistical separ-ability between the classes of input signals exist. The results from this analysis are useful for the development of optimization procedures over a variety of input signals and a r e applicable to any detection system where the sampled signals can be modeled statistically, such as in sonar and ultrasonic detection systems. A computer simulation is performed to illustrate the performance of binary and n-pulse integration for the detection of a white chi distributed target in white Weibull clutter and of a white Rayleigh distributed target in white Rayleigh clutter.