Research On Cooperative Learning Evaluation Method Based On Video Processing

In recent years,how to cultivate and evaluate non-technical literacy（ideology,morality,values,etc.）has gradually become a research hotspot of experts and scholars in the field of education at home and abroad.The teaching and research staff found that the cooperative learning environment has a better effect on non-technical literacy education.Since the formation of non-technical literacy cannot be quantified,it is generally limited to the subjective evaluation of teachers.In addition,affected by the new crown epidemic,global education has shifted from 2020 to online teaching integrating "Internet +" and "Intelligence +" technologies on a large scale,creating a broader stage for online cooperative learning.This thesis integrates multiple evaluation index points to comprehensively evaluate the learning status of students in online cooperative learning.An artificial intelligence-assisted algorithm was introduced for the evaluation of the degree of concentration and speech quality in the objective evaluation of the process,and the fusion weight of the evaluation index points was obtained by applying the analytic hierarchy process based on the questionnaire data of 54 college teachers.The main work of this thesis is as follows:1.Data set collection and preprocessing: Based on the non-technical literacy online teaching situation carried out by the university,the classroom video is collected and preprocessed,and the online cooperative learning concentration analysis data set is constructed;Deep Speech text conversion,supplemented by manual screening,obtains the text data of students’ speech in cooperative learning.2.Concentration evaluation: Two evaluation methods were studied.SVM-based focus recognition and evaluation method: By extracting dense trajectories of video clips,forming trajectory pooling descriptors,obtaining spatial global features after space-time aggregation,and applying linear SVM classification to determine focus status（recognition accuracy rate 76.21%）;depth-based Learning focus recognition method: Using the Slow-Fast network for focus analysis,and transfer learning to the self-built data set,the recognition accuracy rate is 86.7%.The concentration evaluation score can be obtained by counting the concentration state by time.3.Evaluation of speech quality: Use jieba word segmentation and stop word removal to determine candidate words in students’ speeches,use TF-IDF to filter out the key words of students’ speeches,and use Jaccard’s related thinking to analyze the relevance of students’ speeches.Combined with the length of students’ speeches,we get the results of the evaluation of the speech quality of students in cooperative learning.4.Multi-dimensional integration of AHP: Analytic hierarchy process is used to analyze the questionnaire data of 54 college teachers,and on the basis of fully considering the authority of different experts,the final contribution weights of nine evaluation indicators are calculated and obtained.The weights are combined with the scores of each evaluation index to obtain the comprehensive evaluation scores of cooperative learning students.The experimental results show that the researched cooperative learning student evaluation method based on video processing can achieve the whole-process evaluation of the learning status of students in cooperative learning to a certain extent,which is more suitable for the evaluation of the effect of non-technical literacy education,and helps teachers to adjust the teaching progress in time.And provide personalized guidance and training for different students.