Target Localization using Machine Learning and Belief Functions: Application for Elderly People in Indoor Environments
Résumé
Target localization is an important issue in indoor environments. It is vital to assist elderly people in distress. In this paper, we propose a target localization method using machine learning and the belief functions theory. At first, the target area is partitioned into cells and a database of received signal strengths is collected in each cell. In order to consider ambiguity of data, subsets of cells are created. Machine learning is used to assign masses for these subsets. These masses are then used as inputs for the belief functions theory. The latter combines evidence and uses decision-making criteria to associate each cell with a confidence level of having the target, thus the person, residing at each cell. Experimental results show the efficiency of the proposed method in real case scenario, as compared with other state-of-the-art techniques.