Realization of elderly fall integration monitoring system based on AlphaPose and YOLOV4
Hongtao Zheng, Yan Liu, XiaoLi Wu, Yonghua Zhang · 2022 Asia Conference on Algorithms, Computing and Machine Learning (CACML) · 2022
In order to be in a long-term monitoring state when the elderly are not falling, and to save hardware resources as much as possible, and to quickly identify and give an alarm when the elderly have a tendency to fall, this article chooses to use an improved YOLOV4 laid on Jeston nano for rough detection under daily circumstances, and to transmit real-time images to the PC host under abnormal conditions, it uses an Alphapose-based algorithm laid inside it to monitor the details through the instantaneous state angles of the cervical spine and the spine with the xy axis plane.Compared with the current detection system, the proposed multiple Jeston nano+PC cooperation modes save more operating resources. At the same time, the judgment logic for judging falls will have better stability and higher recognition rate than the judgment logic that has been more respected in recent years. At the same time, the speed has also increased many times, confirming the possibility of fall detection integrated into the monitoring system.