Application of Human Body Posture Recognition Technology in Robot Platform for Nursing Empty-Nesters

Abstract

In order to solve the problem of missing the optimal treatment time for empty-nesters after falling, a mobile robot is designed that can follow the elderly autonomously and send GSM SMS to their children or community hospital when discovering their fall. The robot takes industrial PC as the main control module and STM32F1 as the bottom drive module. Kinect V1 depth camera was used to obtain the point cloud image and RGB image of the elderly, and the ROS system analyzed the image data to achieve target following. A simplified DeeperCut human posture estimation model and MPII data set was used to train deep neural network ResNet to detect the body posture coordinates through the changes of its data. Several tests have shown that the robot can follow the elderly in real time, accurately identify their falling posture and send alarm messages timely.