Dynamic Expression Recognition-Based Quantitative Evaluation of Teaching Validity Using Valence-Arousal Emotion Space

Dynamic expression recognition-based quantitative evaluation of teaching validity using V-A emotion space is proposed, in which the students' expressions in class are recognized to obtain their inner evaluation of the teaching validity. The convolutional neural network is used to realize dynamic expression recognition by using the videos of students' listening state in real classroom scenes, and the dropout layer is used to prevent overfitting. The output of Softmax is mapped to the V-A emotion space to obtain the quantification of students' studying status with originality. Finally, Analytic Hierarchy Process method is adopted to evaluate the teaching validity comprehensively from the learning status of students. Experiments on JAFFE database and self-built database show that this method is superior to the most advanced methods. Simulation experiments on JAFFE database show that the emotion recognition rate of the proposed method is 97.57%, 0.97%, 2.26% and 5.04% higher than that of the deep-learning-based system (DLS), feature selection strategy using co-clustering (CCFS) and the exemplar-based SVM (ES-VM) respectively. Experiments on self-built database verify the effectiveness of teaching validity evaluation method.