A Comprehensive Framework for Detecting Behavioural Anomalies in the Elderly

The world is seeing a rapid increase in the population of the aged. This, combined with a shortage of affordable care-giving manpower, leads to a dependence on automated systems for monitoring the well-being of the elderly and detecting abnormalities. There exist techniques based on sensors of various types to detect and recognize the daily activities of the elderly and detect anomalies. While such sensor-based techniques are effective at detecting immediate exigencies, they are unable to comprehend gradual deterioration in the behavior of the elderly indicating conditions like dementia and Alzheimer’s, for example. This aspect is also not properly addressed in the literature. This paper introduces an approach for the comprehensive detection of anomalies in the activities of the elderly using a graph-based approach. The approach employs dynamic activity graphs where anomalies are detected using a dissimilarity score. It is capable of detecting both short-term and long-term anomalies in the daily activities of the elderly.