Detecting stressful older adults-environment interactions to improve neighbourhood mobility: A multimodal physiological sensing, machine learning, and risk hotspot analysis-based approach. (October 2022)

• Record Type: Journal Article
• Authors: Torku, Alex; Chan, Albert P.C.; Yung, Esther H.K.; Seo, JoonOh
• Is Part Of: Building and environment. Volume 224(2022)
• Publication Date: 2022-10
• Subjects: Older adult -- Person-environment interaction -- Environmental stress -- Physiological sensing -- Machine learning -- Risk hotspot analysis; Buildings -- Environmental engineering -- Periodicals; Building -- Research -- Periodicals; Constructions -- Technique de l'environnement -- Périodiques; Electronic journals; 696
• DOI: 10.1016/j.buildenv.2022.109533
• Languages: English
• ISSNs: 0360-1323

Abstract

Abstract: Not only is the global population ageing, but also the built environment infrastructure in many cities and communities are approaching their design life or showing significant deterioration. Such built environment conditions often become an environmental barrier that can either cause stress and/or limit the mobility of older adults in their neighbourhood. Current approaches to detecting stressful environmental interactions are less effective in terms of time, cost, labour, and individual stress detection. This study harnesses the recent advances in wearable sensing technologies, machine learning intelligence and hotspot analysis to develop and test a more efficient approach to detecting older adults' stressful interactions with the environment. Specifically, this study monitored older adults' physiological reactions (Photoplethysmogram and electrodermal activity) and global positioning system (GPS) trajectory using wearable sensors during an outdoor walk. Machine learning algorithms, including Gaussian Support Vector Machine, Ensemble bagged tree, and deep belief network were trained and tested to detect older adults' stressful interactions from their physiological signals, location and environmental data. The Ensemble bagged tree achieved the best performance (98.25% accuracy). The detected stressful interactions were geospatially referenced to the GPS data, and locations with high-risk clusters of stressful interactions were detected as risk stress hotspots for older adults. The detected risk stress hotspot locations corresponded to the places the older adults encountered environmental barriers, supported by site inspections, interviews and video records. The findings of this study will facilitate a near real-time assessment of the outdoor neighbourhood environment, hence improving the age-friendliness of cities and communities. Highlights: Older adults' physiological, GPS, and environmental data were collected during a walk. ML algorithms were trained to detect stressful older adults-environment interactions. Ensemble bagged tree algorithm achieved the best performance with 98% accuracy. Detected high-risk stress hotspots matched the locations of environmental barriers. This study will facilitate near real-time built environment assessment in communities.