Research on Fall Detection Based on Millimeter-Wave Radar Fusion of Human Key Points

With the acceleration of global aging, the traditional elderly care model is confronted with pain points such as a shortage of caregivers, an increased probability of death among the elderly due to falls, and a lagging response to fall rescue. This paper proposes a smart fall detection system for the elderly based on the fusion of millimeter-wave radar and human key points. The target human body information is collected through millimeter-wave radar to generate three-dimensional spatial point cloud data. The point cloud data is processed in two dimensions. The MoveNet pose detection model is used to identify the key point information of human bones, capture the visual pose, combine the dynamic and static processes of falls, determine whether the human body falls according to the threshold and respond in time. It achieved an accuracy rate of 97.6% in the test of the Ti-MIMO Dataset, which is a significant improvement compared with the single vision scheme. The system effectively solves the problem of privacy leakage for the elderly and shows significant advantages in complex scenarios such as low light and occlusion.