Score level fusion algorithm using Differential Evolution and Proportional Conflict Redistribution rule

摘要

In this paper, a new score level fusion approach is proposed. It is based on Differential Evolution (DE) technique and Proportional Conflict Redistribution fusion rule. DE technique is used to find the best confidence factors of the belief assignments of the different modalities. The fusion of the weighted belief assignments is then performed by Proportional Conflict Redistribution combination rule. Experiments are conducted on the scores of BANCA multimodal dataset. A comparative study is achieved using our method, Proportional Conflict Redistribution combination rule and the SVM based fusion. The experimental results show that the proposed approach improves significantly the performance compared to the well established methods.