基于人体骨架信息提取的学生课堂行为自动识别

Automatic Recognition of Student’s Classroom Behaviors based on Human Skeleton Information Extraction

利用计算机视觉技术和机器学习技术对学生课堂行为进行自动识别,是过程性评价的一种新方法,近年来逐渐引起了研究者的关注。文章以监控设备拍摄的实际课堂教学视频为数据源,采集、标注了学生课堂行为数据,提取了学生的人体骨架信息。在此基础上,文章采用Boosting算法和卷积神经网络算法,对基于这两类不同机器学习算法的5种模型进行了学生课堂行为自动识别准确率实验。实验结果表明,在学校教室这种识别比较困难的场景,基于人体骨架信息提取的学生课堂行为自动识别可以达到较高的精度,其中基于Boosting算法的XGBoost模型识别准确率最高。文章的研究推动了计算机视觉技术和机器学习技术的进一步应用,有助于解决学生课堂行为自动识别难题,并助力教师优化教学策略、提高教学效率。

Using computer vision technology and machine learning technology to recognize students’ classroom behaviors automatically is a new method of process evaluation, which has gradually attracted researchers’ attention in recent years. This paper captured the actual teaching videos by monitor equipment as the data source, collected and labelled students’ classroom behavior data, extracted students’ human skeleton information. Based on this, the paper adopted Boosting algorithm and convolutional neural network algorithm to carry out the experiment on the automatic recognition accuracy of students’ classroom behaviors based on five models of the two machine learning algorithms. The experimental results showed that the automatic recognition of students’ classroom behaviors based on human skeleton information extraction could achieve high accuracy in the scene of school classroom where the recognition was difficult, and the XGBoost model based on Boosting algorithm has the highest accuracy. The research of this paper promoted further application of computer vision technology and machine learning technology, contributed to solving the problem of automatic recognition of students’ classroom behaviors and helped teachers to optimize teaching strategies and improve teaching efficiency.