刘超1，邵知宇2，董翠香3
(1.扬州大学 体育学院，江苏 扬州 225127；2.扬州大学 电气与能源动力工程学院，江苏 扬州 225127；3.华东师范大学 体育与健康学院，上海 200241)
摘 要：依托先进的MMAction2模型，首次将其应用于体育课堂教学行为评价，设计一套涵盖多元智能算法分析和可视化反馈功能的完整系统，实现从数据采集到行为分析的自动化与智能化。研究表明，该系统由4个核心模块组成：感知层(负责数据采集与输入)、平台层(进行数据处理与存储)、模型层(完成行为识别与分析)以及应用层(提供数据可视化与结果反馈)。这些模块高效协同且构建了一个完整的教学行为评价体系。实际测试表明，该系统在篮球课堂教学中达到92%的行为识别准确率，分析结果与人工标注一致性高达95%，可显著提升教学评价的效率与准确性。
关 键 词：体育课堂教学行为；人工智能；评价系统；神经网络；MMAction2模型
中图分类号：G807 文献标志码：A 文章编号：1006-7116(2025)05-0127-09
Design and application of evaluation system for teaching behavior in physical education classroom based on the MMAction2 model
LIU Chao1，SHAO Zhiyu2，DONG Cuixiang3
(1.School of Physical Education，Yangzhou University，Yangzhou 225127，China；2.School of Electrical，Energy and Power Engineering，Yangzhou University，Yangzhou 225127，China；3.School of Physical Education and Health，East China Normal University，Shanghai 200241，China)
Abstract: Leveraging the advanced MMAction2 model, the study applies it for the first time to the evaluation of teaching behavior in physical education classroom, and designs a complete system encompassing multi-intelligence algorithm analysis and visual feedback functions, so as to achieve automation and intelligence from data collection to behavior analysis. This research shows that this system consists of four core modules such as the perception layer (responsible for data collection and input), the platform layer (for data processing and storage), the model layer (for behavior recognition and analysis), and the application layer (providing data visualization and result feedback), and these modules efficiently collaborate to construct a comprehensive evaluation system for teaching behavior. And practical tests demonstrate that this system achieves a 92% accuracy rate in behavior recognition and a 95% consistency between analysis results and manual annotations in basketball classroom teaching, which significantly enhances the efficiency and accuracy of teaching and evaluation.
Keywords: teaching behavior in physical education classroom；artificial intelligence；evaluation system；neural network；MMAction2 model