A Classroom Teaching Video Evaluation Algorithm Based on Deep Learning and Emotion Recognition

Currently, there are issues with the evaluation of classroom teaching quality that are not intelligent and subjective. Deep learning algorithms are capable of intelligent teaching evaluation. This article proposes a formative evaluation algorithm for classroom teaching quality based on Local Binary Patterns (LBP), emotion recognition algorithm, and teacher-student scoring algorithm. When processing teaching videos generated in smart classrooms, this algorithm first extracts short video clips containing characters through Local Binary Patterns. Secondly, these segments are divided into segments including only teachers or students using a facial recognition model. Once again, input these fragments into the emotion recognition model (dual stream fusion network) to extract emotional properties. Finally, the deep learning model is utilized to forecast emotional traits and obtain teaching quality evaluation outcomes such as teacher-student interaction. Experiments have demonstrated that this algorithm can effectively simulate the results of manual evaluation of students’ classroom performance while significantly cutting down on evaluation time, providing the groundwork for analyzing other types of data in the future.