M.S. Szewczul, D.R. Halverson, and M.W. Thompson
Detector, robustness, noise, Laplace
Employing the canonical form of the Huber-Strassen robust signal detector, we apply differential geometric methods to model detector robustness for both false alarm rate and detection probability, with the result that robustness can be numerically measured. We then allow the detector censoring level to vary, thus yielding a variety of possible robustness and performance options. In order to analyze the possible tradeoff between performance and robustness, we employ a normalized cost function, which allows the designer to select different combinations of performance and robustness by choice of the censoring level. We also compare results for the nominally Laplace noise case to that in which the noise is nominally Gaussian.
Important Links:
Go Back
