Eye movement entropy: Application of information entropy in eye-tracker

Abstract: Eye Movement Entropy (EME) is an objective metric derived from information entropy, designed to assess the complexity and randomness of visual scanning behavior, addressing the limitations of traditional eye movement indicators in capturing intricate scanning patterns. Based on different computational methods, EME can be categorized into gaze entropy and  heat-map entropy. Gaze entropy is calculated from gaze trajectory maps, reflecting an individual’s fixation positions, durations, and sequences during visual stimuli exposure. It can be further divided into (a) stationary gaze entropy (SGE), which measures the overall complexity of fixation distribution; (b) gaze transition entropy (GTE), which describes the randomness of transitions between fixations; and (c) dwell time entropy (fixation-time entropy), which assesses the distribution characteristics of fixation durations. In contrast, heat-map entropy quantifies attention distribution by analyzing heat maps, focusing on the uniformity and concentration of overall visual attention compared to gaze entropy.
Empirical research highlights EME as a valuable tool for studying  mental health, driving safety, aviation security, education, product design, and industrial safety, with applications in (a) detecting abnormalities in visual attention and cognitive patterns in individuals with mental disorders; (b) evaluating drivers’ visual scanning behavior and fatigue levels; (c) analyzing pilots’ cognitive changes; (d) monitoring cognitive load and instructional effectiveness in education; (e) preventing human errors and optimizing workforce training in industrial safety; and (f) uncovering variations in cognitive and psychological states across different scenarios.
However, current research on EME still faces challenges, One major issue is the lack of a standardized computational framework, leading to inconsistencies in calculation methods and terminology used across different studies. This inconsistency makes cross-experimental comparisons difficult, ultimately affecting the robustness and ecological validity of EME. Additionally, most studies analyzing gaze transition entropy rely on first-order Markov chains, which assume that an individual’s next fixation point depends only on the current fixation (i.e., short-term dependency). This assumption overlooks the influence of eye movement patterns over longer time scales, potentially limiting the model’s ability to capture more complex visual scanning behaviors. Future research should focus on enhancing the robustness and ecological validity of EME measurements while refining its indicator system. Advancing from static statistical analysis to dynamic behavioral pattern exploration will enable a more comprehensive understanding of human visual cognition. Further studies should also expand its practical applications in different fields, optimizing methodologies for real-world environments to maximize its potential in human behavior research, cognitive assessment, and safety monitoring.