Negation of BPA: a belief interval approach and its application in medical pattern recognition

摘要

Assignment of confidence mass appropriately is still an open issue in information fusion. The most popular rule of combination, Dempster’s rule of combination, has been widely used in various fields. In this paper, a belief interval negation is proposed based on belief interval. The belief interval has a stronger ability to express basic probability assignment (BPA) uncertainty. By establishing belief interval in an exhaustive frame of discernment (FOD), the negation is obtained in the form of a new interval. Belief interval negation as an essential tool for measuring uncertainty builds the relationship among BPA, belief interval and entropy. Furthermore, the new negation is applicable to various belief entropies and entropy increment is verified in negation iterations. Two novel uncertainty measures proposed in this paper are applicable to the newly proposed belief interval negation, too. Finally, convergent mass distribution is discussed. Some numerical examples and its application in medical pattern recognition are exhibited.