Anomaly Detection of Elderly Patient Activities in Smart Homes using a Graph-Based Approach

摘要

Sensor-based smart home provide the ability to track resident activities without interfering in their daily routine. It is useful to detect and predict the behaviors of an elderly resident in order to improve the safety of the residents' home environment and provide aid for their caregiver. This paper presents a graph-based approach that successfully discover patterns and anomalies in resident activities. We analyze activity graphs constructed from smart home daily activities to detect normative patterns as well as temporal, spatial, and behavioral anomalies. We also present case studies for cognitively impaired participants and discuss how these anomalies can be linked to the decline in their cognitive abilities which will ultimately provide clinicians and care givers important knowledge regarding their patients.