Motion Recognition System of Physical Aggression Behavior in Classroom
Ting-Hua Lin, Ming-Te Wu · 2021
In order to detect the physical aggressive behavior of students in the classroom early, a motion recognition system is proposed in this paper. This motion recognition system can detect some physical aggression to provide teachers with early resolution of disputes to reduce greater harm. The system first uses monitors to monitor the movements of the students in the classroom and obtains the image of the body by background removal and saliency map technologies. Secondly, the velocity of the body movement is detected, and the velocity value is used as a judgment to determine whether it is physical aggressive behavior. Finally, a machine learning based confusion matrix is used to verify the accuracy of the proposed method. The experimental results confirm that the system can accurately detect the physical aggressive behavior of the collected clips.