Cloud-Based Dempster-Shafer Theory (CDST) for Precision-Centric Activity Recognition in Smarter Environments

Abstract

Smart environments such as smart homes, offices, hotels, and cities are increasingly being furnished with a variety of sensors and actuators to constantly monitor all kinds of accidents and incidents in order to make correct inferences and right decisions. However, the recurring challenge here is that disparate and distributed sensors and their ad hoc networks collectively could produce some wrong sensor data. There are several mathematical techniques to deal with the impurity and the uncertainty issues of heterogeneous sensor data. Bayesian, rough set theory, hidden Markov model, and Dempster–Shafer theory (DST) take the lead. DST is being positioned as one of the most powerful methods to deal with the perpetual problem of inaccuracy and uncertainty. In this paper, a novel framework is proposed for activity recognition such as fire/flame/fall detection for smart environments. The framework comprises a cloud-based execution of Dempster–Shafer theory to increase the overall accuracy of incident monitoring and activity recognition. A fire detection application is implemented using several upcoming technologies like RESTful services, Jess-based rule engine, cloud, and mobility.