Adaptive detection algorithms for multiple-target spectral data

The problem of detecting narrowband signal components in broadband noise is considered. The optimum two-sample detector for the single-target situation is shown to be seriously degraded by the presence of other target lines in the (assumed) noise-only sample. Modified two-sample statistics, based on ranking and censoring techniques, are then proposed and analyzed. These statistics maintain acceptable performance when fewer than a specified number of interfering targets are present, and require only a minor increase-in signal-to-noise ratio to equal the single-target performance of the optimum two-sample statistic.