Multisensor Data Fusion Based on Modified D-S Evidence Theory

Abstract

Abstract-Dempster-Shafer (D-S) evidence theory has been widely used in multisensor data fusion to deal with uncertain information. But unreasonable results may be produced by using D-S combination rule in the case of that data are conflicting with each other. This paper proposes a modified evidence combination method based on information gain and fuzzy preference relations. This method takes account of both historical data and real-time data by introducing the concepts of historical support and realtime support, so it can obtain more accurate results by using more effective information. In order to evaluate the performance of the proposed evidence combination method, an example of classifying the patient's state by five vital signs is given in this paper. The simulation experiment shows that the proposed modified method achieves higher classification accuracy compared with other three data fusion methods.