Research On Classroom Teaching Effect Evaluation Based On Behavior Analysis

Educational informatization reshapes the way of teaching evaluation and teaching management,uses the new generation of information technology to track and monitor the whole process of teaching,carry out learning situation analysis and learning diagnosis,accurately evaluate the teaching and learning effect,change the result oriented "single" evaluation into a comprehensive and process-based "multi-dimensional" evaluation,and change from only focusing on knowledge teaching to paying more attention to the cultivation of ability and quality.By identifying students’ classroom behavior in real time,tracking students’ state changes,counting the proportion of students’ different behaviors in the classroom,and constructing an empirical classroom teaching evaluation system,this topic can effectively grasp students’ classroom attention and formulate personalized classroom behavior correction strategies for students.At the same time,by analyzing the distribution law of students’ classroom behavior,it provides a data-driven mechanism for teachers’ classroom teaching quality evaluation and teaching method improvement.The main research work of this thesis is as follows:Firstly,the movement track of students in the classroom is obtained through Deep SORT multi-target tracking,and then the classroom interaction behaviors such as "standing" to answer questions,walking out of the seat "climbing the blackboard" or "demonstration performance" are classified,identified and studied.By counting the frequency of these teacher-student interaction behaviors,we can provide factual basis for the quantitative evaluation of classroom teaching effect.Secondly,it analyzes and studies the deep learning target detection algorithm,tries to use the YOLO series detector,compares the YOLOv3 and the original YOLOv4 target algorithm,and finally obtains better performance based on the YOLOv4 algorithm,and identifies the specific states such as "rise","mobilephone","side","bow","book" and so on under the premise of "sitting".At the same time,aiming at the occlusion in the classroom and the low detection accuracy of the original YOLOv4 model,the activation function and non maximum suppression mechanism of the network are improved to obtain better detection effect.Finally,make statistics of students’ classroom state and specify quantitative criteria,analyze the time sequence diagram of each student’s classroom behavior state,and count the time and duration of various classroom behaviors of students.Based on this,explore and construct the quantitative evaluation standard of students’ classroom attention,score each student’s listening quality in each class,and make students understand their listening quality by visual methods such as classroom behavior state sequence diagram and distribution pie chart.