Challenges in Studying Falls of Community-Dwelling Older Adults in the Real World
Xin Yu Hu, Rahav Dor, Steven L. Bosch, Anita Khoong, Jing Li, Susan Lynn Stark, Chenyang Lu · 2017
Despite over a decade of research and development in fall detection systems, accurate and reliable systems in use are few. The existing fall detection approaches leave three major challenges unsolved: (1) insufficient fall data for model training process, (2) unreliable labeling of ground truth, and (3) resorting to artificial falls to model falls. In this paper we highlight these challenges in a clinical study with community-dwelling adults. The data collected from the real world reveal significant differences between artificial falls and actual falls, and also to illuminate the limitations of existing algorithms. We further make recommendations for future work, based on the challenges, experience, and lessons we learned from this study.