Reflections on the effectiveness of a high density ambient sensor deployment for monitoring healthy aging

Abstract

The percentage of the world’s population aged over 65 is growing at unprecedented rates. Many countries face the challenge of supporting an aging population despite increasing healthcare costs, and an insufficient number of caregivers. Emerging technologies, like the Internet of Things (IoT) and Artificial Intelligence (AI), can help. Activities of daily living (ADLs) are a good indicator of healthy aging and provide a baseline for detecting an elder’s changes in physical and cognitive state. Monitoring and accurately classifying elders’ activities helps prevent and mitigate some common risks faced by elders when they age-in-place. In our study, we partnered with a senior care provider to add sensors to five apartments in an independent living facility and provided the requisite 24/7 monitoring. Ambient sensors were deployed in each of the apartments and collected high-density IoT sensor data for six months for each of the study participants, who were all over the age of 70. While we successfully created a model to classify ADLs through the recognition and observation of patterns based on high-density ambient sensor placement, sensor data and semantics, and characteristics of activities, we discovered challenges in capturing every ADL for each participant. Often, the ADLs captured for each participant offered unique and personalized indicators of healthy aging. This paper explores the challenges of deploying consumer-grade, IoT, sensors, and the application of AI technology to learn and model elders’ ADLs. We also share results of our exploration to classify ADLs in the data using manual crowd-sourcing, rule-based reasoning and machine learned analytics.