基于头部姿态识别的学习状态检测系统的实现

Implementation of a learning state detection system based on head posture recognition

教室一直以来都是教师进行教学活动的重要场所,为了充分利用课堂监控系统并加强对学生课堂状态的监测,设计了基于头部姿态识别的学生学习状态检测系统。首先,对SSD模型算法的检测后处理中非极大值抑制(NMS)算法进行优化,精准去除冗余候选框;其次,结合系统的应用场景教室,对SSD算法模型的预测特征图进行选取,在保证模型检测精度的同时提高模型的检测效率;最后,采用基于模型的头部姿态识别方式对学生头部姿态进行15 s内动态识别。该系统可以识别2种头部动作、4种头部姿态。经过对教室中真实课堂教学视频的测试,结果表明,该系统可以实现对教室中学生的头部姿态进行有效识别,识别准确率较高,可以应用到课堂学习状态监测中。

The classroom has always been an important place for teachers to conduct teaching activities.In order to make full use of the classroom monitoring system and strengthen the monitoring of students'classroom status,a student learning status detection system based on head posture recognition is designed.Firstly,the Non-Maximum Suppression(NMS)algorithm in the detection post-processing is optimized to accurately remove redundant candidate frames.Secondly,combined with the application scenario of the system—classroom,the feature map of the SSD algorithm model is selected to improve the detection efficiency of the model while ensuring the accuracy of the model detection.Finally,a model-based head posture recognition method is used to dynamically recognize the student's head posture within 15 seconds.The system can recognize two head movements and four head postures.After testing the real classroom teaching videos in the classroom,the results show that the system can effectively recognize the head posture of students in the classroom.With relatively high recognition accuracy,it can be applied to classroom learning status monitoring.