Entropy-based abnormal activity detection fusing RGB-D and domotic sensors


The automatic detection of anomalies in Activeand Assisted Living (AAL) environments is important formonitoring the wellbeing and safety of the elderly at home.The integration of smart domotic sensors (e.g. presencedetectors) and those ones equipping modern mobile robots (e.g.RGB-D cameras) provides new opportunities for addressing thischallenge. In this paper, we propose a novel solution to combinelocal activity levels detected by a single RGB-D camera withthe global activity perceived by a network of domotic sensors.Our approach relies on a new method for computing sucha global activity using various presence detectors, based onthe concept of entropy from information theory. This entropyeffectively shows how active a particular room or environment’sarea is. The solution includes also a new application ofHybrid Markov Logic Networks (HMLNs) to merge differentinformation sources for local and global anomaly detection.The system has been tested with a comprehensive dataset ofRGB-D and domotic data containing data entries from 37different domotic sensors (presence, temperature, light, energyconsumption, door contact), which is made publicly available.The experimental results show the effectiveness of our approachand its potential for complex anomaly detection in AAL settings.

2017 IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems (MFI)