AI KRegression

The single-axis Sample Entropy (X-axis) only slightly decreased from 0.095 ± 0.054 to 0.080 ± 0.048 (p = 0.025*), and the Y-axis also only decreased from 0.106 ± 0.046 to 0.090 ± 0.043 (p = 0.025*), indicating that the complexity changes in a single direction are limited. In contrast, the Pearson correlation of the dual-axis indicators dropped sharply from 0.049 ± 0.227 to –0.062 ± 0.201 (p = 0.002**) before and after fatigue, with a larger change and statistical significance, fully reflecting the collapse of coordination between the horizontal and vertical axes caused by visual fatigue. Further examining the nonlinear interaction coupling indicators, NLID(X|Y) decreased from 0.736 ± 0.185 to 0.654 ± 0.182 (p = 0.013*), and NLID(Y|X) decreased from 0.747 ± 0.151 to 0.698 ± 0.155 (p = 0.030*), clearly revealing the significant weakening of the nonlinear driving relationship between the two axes under fatigue state. Compared with the uniaxial entropy value, these biaxial synchronization and coupling metrics have obvious advantages in sensitivity and significance, which fully verifies the importance of introducing biaxial information for visual fatigue detection. In addition, the appendix provides complete statistical results under each dimension (m=1–5) so that readers can deeply examine the sensitivity and stability of different parameter combinations for fatigue identification indicators.