基于机器学习的课堂行为识别预测系统研究  

Research on Classroom Behavior Recognition Prediction System Based on Machine Learning

作　　者：赵浩[1] ZHAO Hao(Jinshan College,Fujian Agriculture and Forestry University,Fuzhou 350002,China)

机构地区：[1]福建农林大学金山学院,福州350002

出　　处：《长春大学学报》2025年第2期8-12,共5页Journal of Changchun University

基　　金：福建省教育厅项目(JAT201001,jx210301);福建省教育科学“十四五”规划项目(FJJKBK22-203)。

摘　　要：现有的学生行为识别方法存在识别准确率低等问题,研究提出了基于改进You Only Look Once version 5模型的学生行为识别方法。该方法引入增强交并比和Varifocal Loss函数,半自动标注优化数据集,构建行为识别预测系统。实验表明,改进模型收敛速度最快,最小损失值比其他模型平均低0.018,看黑板行为识别准确率提升4.7%。看手机行为置信度提升3.3%。由此可得,改进模型能够提升行为识别准确率,为教学工作提供帮助。There are problems in existing behavior recognition methods such as low recognition accuracy and so on.A student behavior recognition method based on the improved You Only Look Once version 5 model is proposed.The method introduces Enhanced Inersection over Union Loss(EIoU)and Varifocal Loss functions,semi-automatically annotates the optimized data set,and constructs a behavior recognition prediction system.The experimental results show that the improved model has the fastest convergence rate,the minimum loss value is 0.018 lower than that of other models on average,and the recognition accuracy of watching blackboard behavior is improved by 4.7%.The confidence degree of watching mobile phone behavior is improved by 3.3%.It can be concluded that the improved model can improve the accuracy of behavior recognition and can provide help for teaching work.

关 键 词：新教改 学生行为识别 机器学习 YOLOv5 EIoU损失函数 

分 类 号：H319[语言文字—英语]