An improved fall detection approach for elderly people based on feature weight and Bayesian classification
Hanqing Wang, Min Li, Jie Li, Jinge Cao, Zhongya Wang · 2016
Aging population and empty-nesters are two big challenges in modern healthcare. Fall incidents can cause various physical injuries and serious consequence without receiving timely assistance. Therefore, fall detection and movement classification have very high research value and application significance. This paper aims to study the optimum feature subset of falls and put forward an improved approach to detect falls. A set of twelve motion features of eleven kinds of activities are extracted from different parts of body. Then, an improved classifier is proposed based on feature weight and Bayesian framework for fall detection. The optimal features will be selected to reduce the number of features required for the classification problem. Finally, the activity types of unknown samples are predicted using the optimal features and the classifier gained above, and the accuracy of classification will be analyzed. It has been verified through experiments that the improved fall detection approach can get higher accuracy (sensitivity 95.75% and specificity 1.24%) and better robustness (AUC 0.993).