Mining the Patterns of Teachers’ Nonverbal Behavior: Automated Recognition and Systematic Exploration

While the majority of scholarly investigations concerning pedagogical behavior predominantly underscore verbal communication, the significance of nonverbal behavior remains inadequately examined. This research endeavors to propose a theoretical framework pertaining to educators’ nonverbal communication and introduces two artificial intelligence-based recognition models: GestureNet, which focuses on hand gestures, and JointBDOE, which pertains to body orientation. Utilizing video data derived from 32 high-caliber classroom sessions, the research meticulously analyzes gesture patterns through the application of descriptive statistics, sequence analysis, and variance testing. The results indicate a notable prevalence of descriptive and pointing gestures alongside a predominance of frontal orientations. Marked disparities in gesture utilization were identified across various educational stages and subjects, with primary school educators exhibiting more frequent gesture sequences. These findings contribute to the enhancement of automated behavior analysis and provide practical implications for professional development and the design of classroom interactions.