A newfamily of robust non Gaussian detectors based on a geometric heuristic

This paper addresses the detection problem in moderately non-Gaussian environments. We first propose an analysis of the classical Gaussian and SIRV detectors from a new point of view. We define a robustness criteria with respect to signal mismatch and we demonstrate that the non-Gaussian detector is robust only for small mismatches. We then propose a new family of detectors based on a geometric heuristic that exploits advantages of both Gaussian and non-Gaussian detectors. This new family is robust to signal mismatch and presents very interesting detection performance.