Some aspects of classifier fusion based on fuzzy integrals

Fuzzy integral is a valid method for combining multiple classifiers. In classifier fusion system based on fuzzy integral, the fuzzy measure will have much impact on system's performance. Also many authors have done researches on how to determine the fuzzy measure. Our paper presents some new opinions about classifier fusion based on fuzzy integral and gives the lower bound of error rate of fusion system. This paper points the condition that the systems must give wrong classification and the condition that the systems give possibly correct classification. It will be helpful for improving classifier fusion system and designing classifiers in application.