On determining the radar threshold for non-Gaussian processes from experimental data

The problem of detecting radar signals embedded in clutter is an area of great interest. In many radar applications, it is important to set thresholds to achieve a false alarm probability (P/sub F/) of 10/sup -5/ or lower. Using conventional Monte Carlo techniques, where thresholds are set based on raw percentiles, an extremely large number of samples is required. We use the generalized Pareto distribution to approximate the extreme tail of the distributions and propose the ordered sample least squares (OSLS) method for estimating its parameters.