Research On Students’ Classroom Behavior Recognition Model Based On Deep Learning

In classroom teaching,the teachers pay attention to each student’s emotional change and learning state to adjust teaching and effectively improve teaching quality.However,there are some problems in the current classroom,such as the lack of teachers’ energy and the delay of teaching feedback,which affect the improvement of teaching quality and hinder the development of students to a certain extent.In recent years,with the rapid development and wide application of information technology,new technologies such as image processing and artificial intelligence have brought new ideas and methods to improve teaching quality.This thesis aims to study the division,evaluation and classroom participation analysis of students’ classroom behavior state by using automatic target detection and student behavior recognition technology based on deep learning,and design and develop a classroom participation analysis system based on students’ classroom behavior recognition to analyze and evaluate students’ classroom behavior and classroom participation in the teaching process,The main research contents are as follows:(1)Dataset construction.Collect students’ classroom videos as research data,and define students’ behavior categories in combination with relevant pedagogical theories and video observation rules.The main classroom behaviors studied include: listening,writing,lying on the table,raising hands,standing and looking left and right.(2)Data preprocessing.In order to improve the accuracy of the model,the training data are preprocessed and expanded.Specifically,firstly,the collected video data is processed by framing,and then through Yolo＿v3 target detection method detects the position of students in the image,cuts the image according to the detected position information,and obtains student image data with less irrelevant background.Then some data enhancement methods are used to expand the training set.(3)Students’ classroom behavior recognition model.In the experiment,resnet50 network is selected to classify students’ behavior.Firstly,the network is pretrained on the public data set Imagenet,and then the parameter migration method in migration learning is used to train the depth model of students’ behavior recognition.At the same time,the effects of other classification models and main parameters on recognition performance are discussed through a large number of experimental comparisons.(4)Design and implementation of visualization system.Based on the trained model,the Tkinter Library of Python is mainly used to make the student behavior recognition result analysis system.