Activity recognition in smart home using weighted dempster-shafer theory

Abstract

Smart homes are equipped with a variety of sensors to monitor the human activities. The information gathered from the heterogeneous sensors may not be always reliable and have different degrees of uncertainly. One of the most important techniques have been proposed to deal with uncertainty is Dempster-Shafer Theory (DST). In this paper, aims to define more precise sensor reliability and decrease uncertainty in sensor data in the activity recognition process within smart home. In the proposed method, in training step some models are built for per activity according extracted features from training samples and then in the prediction step when a new signal sensor is observed, the features extracted from that signal and applied to models and a weight is calculated for that sensor. These weights are considered as sensor reliability and uses in the decision making process. © 2013 SERSC.