Intelligent internet of things induced preschool education assistance system

Abstract

With the rapid development of AI technology, how to construct smart learning environment has become a hot topic in education. However, most of existing studies focus on the higher education. It is still an open issue to establish smart learning environment in kindergartens for preschool education. In order to solve this issue, this paper designs an intelligent classroom education perception system to assist preschool education. The system end-edge-cloud collaboration structure. The end node captures videos of children in the classroom and send the collected videos to edge node. The edge node contains a NVIDIA® Jetson™ TX2 in which an AI module is deployed. The AI module adopts end-to-end architecture, which contains four parts: human detector, symmetric spatial transformation network module, non-maximum suppression module, and a spatial temporal graph convolutional network module. Compared with previous works, the proposed scheme considers both spatial information and temporal information of skeletal key points. The experimental results show that the proposed smart preschool education assistance system can help teachers to recognize most of common preschool children activities during classroom.