Intelligent technologies have enabled new paradigms to solve educational issues. As a core component of education, intelligent assessment of classroom teaching quality is crucial to education intelligence. Detecting and analyzing real classroom teacher-student behaviors is key to optimizing teaching evaluation systems. However, existing research often lacks a connection between theoretical reviews and practical algorithm validation. This paper bridges this gap by proposing a unified review-exploratory framework for classroom behavior analysis. From the perspective of technology-enabled education, this study adopts behavior recognition as its entry point. By reviewing the literature from 2019 to 2025, it identifies a unified technical framework centered on object detection algorithms. Following an analysis of scenario-specific challenges, this paper evaluates representative algorithms, and establishes an empirical benchmark to validate theoretical findings. Building on this, we explore specific optimization strategies, including backbone reconstruction, feature learning enhancement, attention mechanism embedding, and contextual feature optimization. Results demonstrate that the most of improved models exhibit potential advantages in handling complex classroom tasks. The enhancements are not only reflected in algorithmic metrics but also directly overcome limitations of baselines, providing the academic community with reproducible technical benchmarks. Furthermore, this paper discusses how improvements in algorithmic metrics directly translate into actionable educational outcomes and the significance of technology-driven classroom behavior recognition for educational evaluation research. Finally, we identify four major challenges and propose future research directions concerning technical optimization and ethical standards. This hybrid exploratory study aims to provide theoretical support and practical guidance. It helps improve classroom behavior recognition accuracy, refine teaching evaluation systems, and ultimately promote the optimization of teaching practice.
Song et al. (Wed,) studied this question.