Comparison of techniques for combination of evidence in a signal detection task

Abstract

Signal decision models based on Dempster-Shafer and fuzzy set theories were developed to facilitate the implementation of a rule-based interpretation system for sensory evoked potential waveforms. The models do not require a priori probabilities, and they provide measures of uncertainty. The two models were applied to a set of brainstem auditory evoked potential waveforms. Both had lower absolute error rates than discriminant analysis, although they yielded large numbers of uncertain waveforms. The fuzzy set approach may be less sensitive to the exact form of the membership functions.<>