基于深度学习的学生课堂行为识别
|
摘要
随着教育步入智能时代,智能化的学生课堂行为识别变得越来越重要。但由于学生的课堂行为复杂多变,使得智能化的学生课堂行为识别存在困难。为了提高智能化学生行为识别的精度,文章首先采集了300名学生的7种典型课堂行为图像,并进行了数据预处理。随后,文章将已在ImageNet数据集上训练好的经典深度网络模型——VGG16网络模型迁移到学生课堂行为识别任务中。最后,通过与其它深度学习模型进行实验对比,文章验证了VGG16网络模型对学生课堂行为具有较高的识别精度。上述研究表明,基于深度学习的学生课堂行为识别能及时、精准地反馈学生的课堂学习情况,有利于教师改进教学方法、优化课堂教学与管理,从而提高教与学的效率,助力教学改革。
As education enters the intelligence era, the recognition of intelligent students' classroom behaviors becomes more and more important. However, it is difficult to recognize intelligent students' classroom behaviors, due to the high complexity and variance of students' classroom behaviors. In order to improve the accuracy of recognizing intelligent students' behaviors, this paper firstly collected images of seven kinds of typical classroom behaviors from 300 students,and further carried out data preprocessing. Then, the VGG16 network model, a classic deep network model trained on the ImageNet dataset, was transferred into the recognition task of students' classroom behaviors. Finally, VGG16 network model was verified to have higher recognition accuracy for students' classroom behaviors compared with other deep learning models. The research showed that the recognition of students' classroom behaviors based on deep learning could give feedback on students' classroom learning situations timely and accurately, and help teachers improve teaching method and optimize classroom teaching and management, thus improving the efficiency of teaching and learning and facilitating teaching reform.
As education enters the intelligence era, the recognition of intelligent students' classroom behaviors becomes more and more important. However, it is difficult to recognize intelligent students' classroom behaviors, due to the high complexity and variance of students' classroom behaviors. In order to improve the accuracy of recognizing intelligent students' behaviors, this paper firstly collected images of seven kinds of typical classroom behaviors from 300 students,and further carried out data preprocessing. Then, the VGG16 network model, a classic deep network model trained on the ImageNet dataset, was transferred into the recognition task of students' classroom behaviors. Finally, VGG16 network model was verified to have higher recognition accuracy for students' classroom behaviors compared with other deep learning models. The research showed that the recognition of students' classroom behaviors based on deep learning could give feedback on students' classroom learning situations timely and accurately, and help teachers improve teaching method and optimize classroom teaching and management, thus improving the efficiency of teaching and learning and facilitating teaching reform.
引文
[1]国务院.新一代人工智能发展规划[OL].
[2]王琦.基于卷积神经网络的学生疲劳状态检测关键技术研究与实现[D].武汉:华中师范大学,2016:1-6.
[3]高巍,刘瑞.基于教育大数据的学习行为分析研究——教师评学与导学的角度[J].教师教育论坛,2018,(7):38-42.
[4]王亚飞,刘邦奇.智能教育应用研究概述[J].现代教育技术,2018,(1):5-11.
[5]周建国,唐东明,彭争,等.基于卷积神经网络的课堂表情分析软件研究与实现[J].成都信息工程大学学报,2017,(5):508-512.
[6]廖鹏,刘宸铭,苏航,等.基于深度学习的学生课堂异常行为检测与分析系统[J].电子世界,2018,(8):97-98.
[7]李森林,彭小宁.基于深度神经网络CNN的学生听课状态应用研究[J].电脑与电信,2017,(10):35-37.
[8]韩敏捷.基于深度学习的动作识别方法研究[D].南京:南京理工大学,2016:16-34.
[9]Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[OL].
[2]王琦.基于卷积神经网络的学生疲劳状态检测关键技术研究与实现[D].武汉:华中师范大学,2016:1-6.
[3]高巍,刘瑞.基于教育大数据的学习行为分析研究——教师评学与导学的角度[J].教师教育论坛,2018,(7):38-42.
[4]王亚飞,刘邦奇.智能教育应用研究概述[J].现代教育技术,2018,(1):5-11.
[5]周建国,唐东明,彭争,等.基于卷积神经网络的课堂表情分析软件研究与实现[J].成都信息工程大学学报,2017,(5):508-512.
[6]廖鹏,刘宸铭,苏航,等.基于深度学习的学生课堂异常行为检测与分析系统[J].电子世界,2018,(8):97-98.
[7]李森林,彭小宁.基于深度神经网络CNN的学生听课状态应用研究[J].电脑与电信,2017,(10):35-37.
[8]韩敏捷.基于深度学习的动作识别方法研究[D].南京:南京理工大学,2016:16-34.
[9]Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[OL].