Using SlowFast Network to Analyze the Relationship between Teacher-Guided Action Rhythm and Teaching Efficiency in Classroom Interactive Videos

To address the difficulty of quantitatively characterizing teachers’ nonverbal behaviors and clarifying their relationship with teaching effectiveness in classroom interaction videos, this study employs a SlowFast dual-pathway neural network to analyze the rhythm of teacher-guided actions, providing an effective framework for intelligent video understanding and spatiotemporal signal analysis. Considering the increasing demand for real-time visual information processing and multimodal perception in advanced wireless sensing and communication systems, the proposed approach models teacher rhythm parameters by combining the Slow pathway for subtle facial expressions and gesture analysis with the Fast pathway for large-scale body motion extraction. Furthermore, a classroom behavior coding framework is integrated with student attention, interaction response time, and in-class assessment performance to evaluate teaching efficiency. Experimental results show that the SlowFast model achieves an accuracy of 89.7% in recognizing periodic teacher-guided movements, representing an 11.2% improvement over a single-pathway 3D-CNN model. Correlation analysis further demonstrates that the variance of teacher gesture rhythm is significantly negatively correlated with student attention (r = -0.83, p < 0.01), while moderate movement frequency exhibits a positive correlation with classroom interaction response speed (r = 0.76). The proposed framework confirms a significant relationship between teacher-guided action rhythm and teaching efficiency, providing a quantitative basis for optimizing instructional behavior and offering valuable references for intelligent visual sensing, spatiotemporal signal processing, and multimodal information analysis in advanced engineering applications.