基于人脸姿态识别的课堂学习状态反馈系统的设计与实现  

Design and implementation of classroom learning state feedback system based on face pose recognition

作　　者：吴丽娟[1] 任海清 关贵明 梁岱立 黄尧 WU Lijuan;REN Haiqing;GUAN Guiming;LIANG Daili;HUANG Yao(College of Physical Science and Technology,Shenyang Normal University,Shenyang 110034,China;Troops 31441of Northern Theater of the Chinese Peoplel's Liberation Army,Shenyang 110001,China)

机构地区：[1]沈阳师范大学物理科学与技术学院,沈阳110034 [2]北部战区31441部队,沈阳110001

出　　处：《沈阳师范大学学报（自然科学版）》2022年第2期127-132,共6页Journal of Shenyang Normal University:Natural Science Edition

基　　金：辽宁省教育厅科学研究经费项目(LFW202003)

摘　　要：人脸识别技术是目前计算机视觉中的热门研究方向之一,已经被应用于很多领域。“抬头率”已成为判断学生的专注程度、检验课堂教学效果的重要因素之一。设计开发了基于人脸姿态识别的学生课堂学习状态反馈系统,完成了人脸数据集的采集、制作与训练,并增加了YOLOv5s训练模型的深度和宽度,在保证其速度的前提下提升了识别的准确度。在此基础上,将识别参数Conf与IOU的阈值调整到适当值,使识别结果更加清晰。实验结果表明:改进之后的识别系统在不影响速度的同时提高了识别的准确度;较暗环境下的人脸识别准确度可以达到0.887,能快速识别出学生课堂的学习状态,符合课堂教学实时监测要求;对于有遮挡的人脸识别,其准确度也可以达到0.5以上,满足特殊情况下的人脸识别需求。Face recognition technology is one of the most popular research fields in computer vision,and has been applied in many fields.“Head-up rate”has become one of the important factors to judge student’s attentiveness and to test the effect of classroom teaching.Therefore,in this paper we design and develop the feedback system of student’s classroom learning state based on face pose recognition,complete the collection,production and training of the face data set,and increased the depth and width of YOLOv5straining model.The accuracy of recognition is improved while the speed is guaranteed.On this basis,the threshold value of the identification parameters Conf and IOU are adjusted to the appropriate value to make the identification results more clearly.The experimental results show that the improved system can improve the recognition accuracy without affecting the speed,and the average accuracy of face detection in dark environment can reach 0.887,it can quickly identify the learning state of students in class,meet the requirements of real-time monitoring in class teaching,and the confidence level of face recognition for occluded students can reach 0.5 or above,which meets the needs of face recognition under special circumstances.

关 键 词：人脸识别 人脸数据集 模型深度 模型宽度 抬头率 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] G434[自动化与计算机技术—计算机科学与技术]