Classroom behavior analysis and digital teaching quality evaluation based on spatiotemporal graph neural network

作者：Yang Kong, Rongwei Dong, Hui Zhang · 发表于：Discover Artificial Intelligence · 年份：2025 · DOI：10.1007/s44163-025-00623-z · 被引用次数：2 · 研究领域：Advanced Technologies in Various Fields、Technology-Enhanced Education Studies、Human Pose and Action Recognition

Modern smart classrooms are unable to replicate traditional classroom observation methods, which are biased and resource-intensive. AI-based behavior recognition systems often disregard spatiotemporal dynamics. The goal of this effort is to develop an intelligent, scalable, and objective system for evaluating classroom behavior analysis and teaching quality. This research proposes that EduSpatioNet utilizes a YOLOv8-based entity detection algorithm and a spatiotemporal graph neural network with multi-head attention, adaptive sampling, and hierarchical pooling. Classroom video and sensor data are used to create interaction graphs for student–teacher interactions. Behavior recognition experiments indicate 92% accuracy and 87% agreement with expert instruction assessments. A statistical analysis shows significant improvements over baseline models. The results show that EduSpatioNet is efficient, interpretable, and scalable, making it a reliable real-time educational intelligence solution.