Abstract

Emotions can influence and regulate learners’ attention, memory, thinking, and other cognitive activities. The similarities and differences between English and non-English majors in terms of English classroom learning engagement were compared, and the significant factors affecting the emotional, cognitive, and behavioral engagement of the two groups of students in the English classroom were different. English majors’ affective engagement in the classroom was not significant, which was largely related to their time and frequency of English learning. Traditional methods of learner emotion recognition suffer from low recognition rate, complex algorithms, poor robustness, and easy to lose key information about facial expression features. The paper proposes a convolutional neural network-based learner emotion recognition method, which includes three convolutional layers, three pooling layers, and one fully connected layer. In the future, the method will can be applied to the construction of smart learning environments, providing technical support for improving learner models, realizing emotional interactions, and mining learning behaviors, etc.