An improved belief χ ^2 divergence for Dempster–Shafer theory and its applications in pattern recognition

Xueyuan Gao,Fuyuan Xiao
DOI: https://doi.org/10.1007/s40314-022-01975-3
2022-01-01
Computational and Applied Mathematics
Abstract:Dempster–Shafer (D–S) evidence theory is widely used in the field of uncertain information processing, but because of its defective fusion rules, it often results in visual violations when dealing with highly conflicting evidence. To solve the conflict management problem of D–S theory, an improved belief χ 2 divergence (called ℐℬχ ^2 ) is proposed in this paper. The ℐℬχ ^2 divergence takes into account the amount of all possible hypotheses, which allow it be a more credible and efficient solution to measure the dissimilarity between evidences. Moreover, it has good mathematical properties including symmetry, boundedness and non-degeneracy. Next, we designed a novel multi-resource information fusion algorithm based on ℐℬχ ^2 divergence. Finally, application in automobile system fault recognition and iris feature recognition prove the effectiveness and accuracy of the proposed multi-resource information fusion algorithm.
What problem does this paper attempt to address?