Design of Indoor Fall Detection System for the Elderly Based on ZYNQ

摘要

Aiming at the problems of prevention of elderly fall damage, a kind of intelligent video surveillance system is designed in this research. CMOS camera transmits the collected video to the video processing module based on ZYNQ(ARM+FPGA). CNN (Convolutional Neural Network) image classification algorithm and hardware acceleration technology are used to process images in real time and identify whether the elderly have fallen. When the elderly falls, the system immediately transmit message about the elderly to the guardian via 5th generation mobile networks(5G). The results of experiments demonstrate the accuracy of this fall detection system is about 92%, response time for alarming is about 0.413s, system satisfices the requirement of real-time.