Occluded Image Interpretation Ability Assessment Under Different Mental Workloads: An Eye-Tracking Study

Image interpretation experts play a crucial role in various domains in daily life such as military, healthcare, and education. Accurately assessing an individual's ability to perform such tasks is of great importance. In this study, we designed an image interpretation task with three levels of occlusion, which better reflected the real-world working environments. Using the Eyelink 1000plus eye-tracking system, we collected both behavioral data and eye-tracking data from 64 participants while they performed the image interpretation tasks. Through confirmatory factor analysis, we extracted the $g$-factor, representing the comprehensive image interpretation ability, from the behavioral performance data of the tasks with different levels of occlusion. We divided all participants into two groups according to their $g$-factors and conducted statistical analyses on their eyetracking data. The results showed that, under all three levels of mental workload, the high-ability group exhibited fewer fixations, saccades, and blinks. Additionally, the high-ability group showed significantly longer fixation duration under the medium and high mental workload conditions. Furthermore, our study also provided a few machine learning algorithms for assessing individuals' image interpretation ability based on eye-tracking features with the highest Pearson correlation coefficient of 0.81.