Adaptive detection with tunable robustness via a linear combination of the test statistics and loss factor

Abstract

In this paper, we propose an adaptive detection algorithm with tunable robustness to the mismatched target. We notice that the test statistic of existing algorithms is the function of the test statistic of the generalized likelihood ratio test (GLRT) and the loss factor, which motivates us to propose the tunable robustness test (TRT) via a linear combination of the above two variables. The probability of false alarm (Pfa) and the probability of detection (Pd) for the mismatched target are presented for the rank one target and the multidimensional subspace target, respectively. Through numerical simulations, we verify that the tuning parameter will influence the ability to distinguish between matched and mismatched targets, and therefore the TRT can achieve robustness that existing algorithms do not have.