Design and Development of University Teacher Behavior Analytical System in English Classroom based on Hybrid Deep Learning

Accurate analysis of university teacher behavior is necessary to enhance teaching quality and increase students' involvement in college English education. Traditional observation processes are time-consuming and prone to subjectivity, disrupting comprehensive and consistent evaluation. This study introduces a Deep Learning (DL) model that combines ResNet50 and Long Short-Term Memory (LSTM) networks to facilitate efficient analysis of teacher behavior. ResNet50 captures spatial information from classroom images, which are frames from labeled video recordings. LSTM captures temporal relationships between these sequential images to detect dynamic behavioral patterns. The suggested model utilizes the SCB-Dataset5, which includes labeled image sequences of different teaching activities, to train and evaluate the algorithm’s effectiveness. Experimental findings indicate that the suggested model attains an accuracy of 99.2%, precision of 98.9%, recall of 99.0%, specificity of 98.5%, and an F1-score of 98.95%, outperforming other DL architectures. The hybrid method offers a robust and flexible foundation for automated instructor behavior analysis that supports real-time observation and feedback in learning environments. The research demonstrates the potential of integrating spatial and temporal learning approaches to enhance intelligent classroom analytics, facilitating data-based teaching strategy enhancement and the development of smart education systems.