AN IMPROVED DYNAMIC TIME WARPING METHOD COMBINING DISTANCE DENSITY CLUSTERING FOR EYE MOVEMENT ANALYSIS

Analyzing eye movement data to evaluate learning status has become crucial in intelligent education. The eye movement scanning path can directly or indirectly reflect changes in thinking patterns and psychological states. By analyzing the scanning path, we can explore the commonality and differences in learners' eye movement behaviors and provide essential references for improving visual content and giving guidance. This paper first studies the time series representation and clustering of the learner's scanning path under the same task. Then, the three learning states of concentration, mind-wandering, and information wandering are evaluated through the clustering results. Specifically, the improved DBA algorithm (iDBA) is proposed to extract group eye movement patterns, combined with the dynamic time warping (DTW) algorithm to calculate the similarity of scanning paths and determine the clustering seeds, while the distance density clustering (DDC) algorithm is used for clustering. Experiments show that time series-based eye movement pattern mining can identify group viewing behaviors. Meanwhile, clustering reveals different reading strategies and provides the ability to assess learning status.