In order to test the applicability of using facial expressions to detect the participation level of on-line learners in their educational activities,three different convolutional neural network models and one pro-posed convolutional neural network model are studied,including All-CNN,NiN-CNN,and VD-CNN.The new model is raised based on the advantages of three basic models,and the proposed model replaces linear convolutional layers with multi-layer perceptrons and uses small(3×3)convolutional filters to increase the depth of the network,and replaces some max pooling layers with convolutional layers with increased stride.The four models are applied to the Dataset for Affective States in E-Learning Environments(DAiSEE),and their performance in detecting learner participation is analyzed.The results show that the proposed model performs better than the other models.
deep learningConvolutional Neural Networkonline learning environmentengagement de-tectionfacial expressions
万方数据
维普
