Abstract: Classroom video analysis serves as a crucial pathway for uncovering instructional laws and fostering teacher professional growth, and has become the new normal in teaching quality evaluation. The advancement of intelligent technologies offers novel paradigms for classroom research, positioning technology-enhanced classroom analysis as a significant research direction. Grounded in the Human-Centered Artificial Intelligence (HCAI) application framework, this study systematically examines the analytical elements, potential space, and risk mitigation within the field of multimodal classroom video analysis across three dimensions: human-factor design, human-intelligence embodied technical design, and ethically-aligned design. At the level of analytical elements, the focus is on three types of modal information: teacher-student behavior, emotion, and cognition. Among these, behavioral recognition technology is developing most rapidly, affective analysis is maturing, while intelligent analysis of the cognitive dimension remains relatively underdeveloped. At the level of potential space, future efforts should be directed to promote the evolution of analytical subjects from single entities to long-term teacher-student interaction, the evolution of analytical techniques from multi-source data to integrated decision-making, and the evolution of analytical content from element identification to meaning attribution. At the level of risk mitigation, technological applications face challenges such as privacy leakage, algorithmic bias, value alienation, and emotional defense, which must be addressed through privacy stewardship, algorithmic fairness, value guidance, and emotional inclusivity. Based on these findings, this study provides theoretical references and practical insights for precise classroom analysis, precision teaching research, and teacher professional development supported by intelligent technologies.

Key words: multimodal, classroom video analysis, technology ecosystem, technology ethics