An Analysis System of Students' Attendence State in Classroom Based on Human Posture Recognition

Listening state is a value reflecting the efficiency of lectures. As various human gesture recognition algorithms proposed, we could make advantages of these algorithms to construct a system to evaluate the listening state of the students. With the spring of many outstanding algorithm of machine vision, the basic theories have reached matured to complete some complex assignment. The system proposed here is a kind of integration of these technologies, which also need innovation when it's actually implemented. Also, wasted surveillance video resources universally existed in our society, which should be made advantage of but not. Especially in the scene of the classrooms, the schools hold tremendous surveillance video resources. The system is proposed to analyze the state of the students through the human posture recognition. Its basic theory is the recognition of the human body key points, including the rather important points of one's body. We could extract these information from the videos firstly, and make some analysis such as the computation of some specific angles. After these treatment of data, we could get rough results of the state of the students, whose correct rate could be higher through machine learning. There are much efforts to be made before the system could fit the requirement of the specific environment, such as semantic segmentation and super resolution. The system would also have clear visualization when finally faces customers.