Research On Student Classroom Behavior Recognition And Analysis System Based On Deep Learning

With the advent of the era of big data and artificial intelligence,more and more fields choose to use emerging technologies to improve work efficiency.Whether in the field of production,technology or education,artificial intelligence is gradually chosen to replace the traditional way of work.As an important part of artificial intelligence,the development of deep learning algorithms has made great contributions to text recognition,image processing,semantic analysis and other aspects.In the process of teaching management,analyze the students’ classroom behavior is the effective measure to improve the teaching level and teaching achievements,traditional classroom behavior research methods generally need teachers focus on classroom playback video to watch record,but use the deep learning method study of students’ classroom behavior can alleviate the pressure of the teachers,so that teachers can put more energy into teaching work,improve teaching methods,and guide students to learn in the classroom in a better state.Based on the above situation,this thesis proposes a student classroom behavior recognition method based on deep learning,which combines object detection network and classification network to detect and classify student targets.Before the research and improvement of the algorithm,the classroom behavior recognition data sets of students are constructed firstly,including 2100 student object detection data sets and 31614 student behavior classification data sets.The behavior categories can be divided into five types: general state,lying down,raising hands,turning around and standing.Students choose YOLOv5 object detection algorithm for target detection,and the recognition accuracy reaches 99.2%.Shuffle Net V2 algorithm was selected for student behavior classification and the algorithm was improved.The convolution kernel of deep convolution in Shuffle Net V2 unit was expanded from 3×3 to5×5,and the accuracy rate reached 98.1%.Finally,the two networks are combined together,and the student targets detected and intercepted by YOLOv5 algorithm are used as the input of the improved Shuffle Net V2 algorithm,and the classification results are displayed on the student target box.Teachers can intuitively understand the classroom behavior status of students through the recognition results.In addition,based on the behavior of college students algorithm proposed in this thesis,using Tkinter tool design has realized the student classroom behavior identification analysis system,the system realize the function of the demographic data including the classroom,classroom behavior recognition,etc.,to show the system and testing,the results show that the system can analyze the behavior of college students,implement the data visualization.