Conference Proceeding Article
Unobtrusive in-home monitoring systems are gaining acceptability and are being deployed to enable relatives and caregivers to remotely monitor and provide timely care to their elderly loved ones or senior clients, respectively, who are living independently. Such systems can provide information about nonmovement or inactivity of the elderly resident. As prolonged inactivity could mean potential danger, several algorithms have been proposed to automatically detect unusually long durations of inactivity. Such schemes, however, suffer from low sensitivity due to their high detection latency. In this paper, we propose Dwell Time-enhanced Dynamic Threshold (DTDT), a scheme for computing adaptive alert thresholds that exploit region-specific dwell time to reduce the detection latency. Using extreme value theory, we obtain a closed form expression for the per-region alert thresholds. We perform simulations using real data to evaluate the performance of DTDT and compare it with state-of-the art schemes AID and the algorithm by Moshtaghi et al. Results show that DTDT shows significantly lower detection latency, 1.5– 3 hours shorter, in regions with short dwell times (bathroom and kitchen) while maintaining the same false alarm rate.
Digital Communications and Networking | Software Engineering
Software and Cyber-Physical Systems
2016 IEEE International Conference on Pervasive Computing and Communication Workshops (PerCom Workshops), Sydney, Australia, 14-18 March
City or Country
VALERA, Alvin C.; Hwee-Pink TAN; and BAI, Liming.
Improving the sensitivity of unobtrusive inactivity detection in sensor-enabled homes for the elderly. (2016). 2016 IEEE International Conference on Pervasive Computing and Communication Workshops (PerCom Workshops), Sydney, Australia, 14-18 March. 1-6. Research Collection School Of Information Systems.
Available at: http://ink.library.smu.edu.sg/sis_research/3325
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.