Evaluation of Accelerometer-Based Fall Detection Algorithms on Real-World Falls

• DOI: 10.1371/journal.pone.0037062
• Publication Date: 2012-05-17T13:16:43Z
• Authors: Fabio Bagalà; Clemens Becker; Angelo Cappello; Lorenzo Chiari; Kamiar Aminian; Jeffrey M. Hausdorff; Wiebren Zijlstra; Jochen Klenk
• Keywords: Fall prevention

Despite extensive preventive efforts, falls continue to be a major source of morbidity and mortality among elderly. Real-time detection their urgent communication telecare center may enable rapid medical assistance, thus increasing the sense security elderly reducing some negative consequences falls. Many different approaches have been explored automatically detect fall using inertial sensors. Although previously published algorithms report high sensitivity (SE) specificity (SP), they usually tested on simulated performed by healthy volunteers. We recently collected acceleration data during number real-world patient population with high-fall-risk as part SensAction-AAL European project. The aim present study is benchmark performance thirteen fall-detection when are applied database 29 To best our knowledge, this first systematic comparison found that SP average algorithms, was (mean±std) 83.0%±30.3% (maximum value = 98%). SE considerably lower (SE 57.0%±27.3%, maximum 82.8%), much than values obtained false alarms generated 1-day monitoring three representative fallers ranged from 3 85. factors affect falls, also discussed. These findings indicate importance testing in real-life conditions order produce more effective automated alarm systems higher acceptance. Further, results support idea large, shared could, potentially, provide an enhanced understanding process information needed design evaluate high-performance detector.