Hybrid Real-Time Fall Detection System Based on Deep Learning and Multi-sensor Fusion
Xiaojie Lv, Zongliang Gao, Changshun Yuan, Meng Li, Chao Chen · 2020
As China's aging population increases, falls have become a common health problem among the elderly. Therefore, fall detection has become an important research direction relating to elderly health problems. In this article, we propose a low-cost smart home fall detecting and monitoring system based on the YCL(YOLO V3 combined with the LiteFlowNet) algorithm and multi-sensor fusion. On the one hand, the behavior of the elderly is monitored by using the YCL algorithm to detect fall activities. On the other hand, a multi-sensor fusion algorithm is applied based on the embedded sensors of a smartphone-i.e., acceleration and air pressure sensors-to detect fall actions. The outputs of these two methods are fused to generate a final result, which is improved in terms of the performance and accuracy of fall detection. Once the fall action is confirmed, the smartphone will immediately send an alert SMS to the emergency contact to seek medical assistance. Our experimental results validate that by combining the two methods, the false alarm rate is reduced, and the fall detection can be achieved even in the blind zone of the camera.