首页|基于多模态的线上学习专注度评价

基于多模态的线上学习专注度评价

扫码查看
针对线上学习状态难以有效评估问题,提出一种基于多模态的线上学习专注度评价模型CE-HPE.CE-HPE利用头部姿态与面部表情两种模态特征进行专注度估计,结合 6DRepNet算法预测头部偏航角和俯仰角,利用改进的DAN算法RES-DAN实现表情分类,对两种模态进行量化,并结合权重计算专注度得分.基于RAF-DB数据集实验结果表明,RES-DAN在准确率上优于对比模型,消融研究也验证RES-DAN各模块的有效性.所设计的线上学习专注度评价系统实现了单人和多人的实时或定期专注度量化评估,能有效地对线上学习状态进行检测和评估.
Multimodal Concentration Evaluation in Online Learning
This paper presents a multi-modal concentration evaluation model CE-HPE for the challenge of evaluating the status of online learning effectively.CE-HPE estimates concentration of learners with two modal features such as head posture and facial expression.It combines the 6DRepNet algorithm to predict the yaw angle and the pitch angle of head,and classifies facial expressions with the improved DAN algorithm RES-DAN.CE-HPE quantifies two modal features and calculates the concentration score with their weights.Experiments based on the RAF-DB data set show that the accuracy of RES-DAN is superior to that of comparative models.Furthermore,ablation studies also verify the effectiveness of each module in RES-DAN.We develop an online learning concentration evaluating system which can calculate the quantification score of concentration in online learning for single player mode and multiplayer mode in real time or a period.The results show that our system can effectively detect and evaluate the concentration status in online learning.

concentration evaluationhead posturefacial expressionRES-DANmultimodality

王一可、孙英娟、蒲东兵、崔灿、王敬

展开 >

东北师范大学信息科学与技术学院,吉林 长春 130117

沈阳城市学院智能与工程学院,辽宁 沈阳 110100

长春师范大学计算机科学与技术学院,吉林 长春 130032

专注度评价 头部姿态 面部表情 RES-DAN 多模态

2024

长春师范大学学报
长春师范学院

长春师范大学学报

CHSSCD
影响因子:0.312
ISSN:1008-178X
年,卷(期):2024.43(8)