pp. 1657-1679
S&M4011 Technical Paper of Special Issue https://doi.org/10.18494/SAM5420 Published: April 30, 2025 Comparative Analysis of Lightweight OpenPose and MoveNet AI Models for Real-time Fall Detection and Alert Systems [PDF] Yuh-Shihng Chang, Yi-Xiang Zheng, and Zheng-Yu Ku (Received October 28, 2024; Accepted April 4, 2025) Keywords: AI, human pose estimation, fall detection, human activity recognition and behavior understanding
Falls are the leading cause of injury-related deaths among adults aged 65 and older. The age-adjusted fall mortality rate increased by 41%, from 0.0553% in 2012 to 0.0780% in 2021. The models OpenPose and MoveNet can be used to detect human movements associated with falls in dynamic images. In this study, we utilized these AI models to independently identify falls among elderly individuals living alone and compared the efficiency of these two AI fall detection models in capturing dynamic images. In this study, we employed a web camera as a sensor and integrated it with the two systems mentioned above to detect human limb movements, determine whether the monitored individual has fallen, and send emergency notifications to family members, caregivers, and nursing staff via communication software. AI-based posture detection technology is vital for elderly individuals who live alone, have poor health, or have limited mobility. Our evaluation of the detection efficiency of different AI fall detection models provides valuable references for applications in healthcare systems.
Corresponding author: Yuh-Shihng Chang![]() ![]() This work is licensed under a Creative Commons Attribution 4.0 International License. Cite this article Yuh-Shihng Chang, Yi-Xiang Zheng, and Zheng-Yu Ku , Comparative Analysis of Lightweight OpenPose and MoveNet AI Models for Real-time Fall Detection and Alert Systems , Sens. Mater., Vol. 37, No. 4, 2025, p. 1657-1679. |