Earth Mover's divergence of belief function

Divergence is used to measure the difference of two systems, and it is widely applied in many fields. To solve this problem more efficiently, Dempster-Shafer evidence theory has been proposed, different from the traditional probability distribution, and because of its processing advantages of uncertainty, has been widely used in many aspects of reality. In this paper, a new method of belief divergence measure of mass functions is proposed, named as Earth Mover's divergence of belief function, which is a generalization of Earth Mover's distance (Wasserstein distance). Compared with other existing methods of divergence measuring, the EM divergence can show good performance in the presence of higher degrees of uncertainty and more conflicts. Numerical examples help have a better understanding of the Earth Mover's divergence of belief function. Based on the new method of belief divergence measure, there is a combination model proposed to address the problem of data fusion. Application in target recognition is used to show the efficiency of the proposed method of divergence measure.

Automated summary

This article proposes a new method for measuring divergence in belief distribution, called Earth Mover's divergence of belief function, which shows good performance in the presence of higher degrees of uncertainty and conflicts. Numerical examples and application in target recognition demonstrate the efficiency of the proposed divergence measure method.