A new method to determine evidence distance

DS evidence theory is a hot topic in multi-sensor data fusion because of its advantages in multi-source information presentation and processing. Considering the difficulties of sensor reliability priority information acquisition, how to compute reliability of sensor via the evidence distance in multisensor data fusion system and how to select consistent evidence for combination via the evidence distance when many homogeneous sensors exist are the open issues. In this paper, a new evidence distance is presented in order to avoid the disadvantage of Diaz's distance measure in the logical selection of tuning parameter and Jousselme's distance measure in not considering the proximity of reference frame to the frame of discernment. This method firstly defines a fuzzy linguistic term set to express the proximity of reference frame to the frame of discernment, in which the semantics of the terms is given by a triangular membership function and a rewarding or penalizing function. Then similarity measurement matrix is achieved by weighting the rewarding or penalizing function using membership degree. Finally Jousselme's distance framework is used to compute the evidence distance. Two numerical examples illustrate the efficiency of the method proposed.