Publication Type

Conference Proceeding Article

Version

publishedVersion

Publication Date

7-2018

Abstract

Sensor technologies have gained attention as an effective means to monitor physical and mental wellbeing of elderly. In this study, we examined the possibility of using passive in-home sensors to detect frailty in older adults based on their day-to-day in-home living pattern. The sensor-based elderly monitoring system consists of PIR motion sensors and a door contact sensor attached to the main door. A set of pre-defined features associated with elderly’s day-to-day living patterns were derived based on sensor data of 46 elderly gathered over two different time periods. A series of feature vectors depicting different behavioral aspects were derived to train and test three machine learning algorithms; Logistic Regression, Linear Discriminant Analysis and Naïve Bayes. The best prediction scores yielded by seven features, namely, daytime napping, time in the bedroom, night-time sleep, kitchen activity level, kitchen use duration, in-home transitions and away duration. These features produced an area under the ROC curve of 98%, 79% and 93%, for Logistic Regression, Linear Discriminant Analysis and Naïve Bayes algorithms respectively. The findings of this study provide implications on how a non-intrusive sensor-based monitoring system comprised of a minimum set of sensors coupled with predictive analytics can be used to detect frail elderly.

Keywords

Ageing-in-place, Frailty detection, Non-intrusive in-home sensors

Discipline

Gerontology | Software Engineering

Research Areas

Software and Cyber-Physical Systems

Publication

ITAP 2018: Proceedings of 4th International Conference on Human Aspects of IT for the Aged Population, Las Vegas, July 15-20

Volume

10927

First Page

290

Last Page

302

ISBN

9783319920375

Identifier

10.1007/978-3-319-92037-5_22

Publisher

Springer

City or Country

Cham

Copyright Owner and License

Publisher

Additional URL

https://doi.org/10.1007/978-3-319-92037-5_22

Share

COinS