A Method for Video Learning Analysis Based on Gaze Velocity Motifs

With the increasing prevalence of video learning resources, eye tracking technology is becoming increasingly important to investigate students' eye movement patterns during video learning, in order to better understand their cognitive characteristics. This study employs a gaze velocity motif approach to systematically analyze learners' eye movement behaviors across different functional areas of video content, including the instructional software area, teacher area, and non-instructional area. The results indicate that the distribution of gaze velocity motifs in all three areas significantly deviates from that of random sequences, demonstrating the regularity of eye movement behavior; eye movement patterns are highly similar across instructional areas, whereas the non-instructional area exhibits significant differences. Motif transition analysis and transition entropy results show that eye movement patterns in the non-instructional area are the most difficult to predict, followed by the teacher area, while those in the instructional software area are the easiest to predict, reflecting the impact of task relevance on the dynamic changes of eye movement behavior. Compared with traditional gaze metrics (e.g., fixation count, fixation duration, and average gaze velocity), gaze velocity motif metrics exhibit higher sensitivity in revealing the relationship between learners' attention distribution and learning performance, particularly with specific motif proportions in the instructional software area showing significant correlations with learning outcomes. The data analysis validates that the proposed method provides a feasible approach for understanding gaze characteristics in video learning and opens new avenues for exploring learners' attention and search behaviors during video learning.