A signal detection system based on Dempster-Shafer theory and comparison to fuzzy detection

This paper describes a signal detection algorithm based on Dempster-Shafer theory. The detector combines evidence provided by multiple waveform features and explicitly considers uncertainty in the detection decision. The detector classifies waveforms as including a signal, not including a signal, or being uncertain, in which case no conclusion regarding presence or absence of a signal is drawn. The probability numbers required in the Dempster-Shafer formulation are defined as piecewise linear functions that can be described by two parameters, and the effects of these parameters on detector performance, using simulated data, are compared to Bayesian detection and to a fuzzy signal detector that also considers uncertainty. The performance of the Dempster-Shafer and fuzzy detectors shows similar dependence on the parameters, although, if parameters are adjusted so that the number of correctly classified waveforms are equal, the Dempster-Shafer detector has more uncertain classifications and fewer errors than the fuzzy detector, providing superior performance. The Dempster-Shafer detector incorporates a different type of uncertainty than the fuzzy detector, which may contribute to this difference in performance. The difference may also reflect the different mathematical operations used