A multimodal visual fatigue assessment model based on back propagation neural network and XGBoost

An experiment was conducted using a subjective questionnaire, ophthalmological parameters, electroencephalogram (EEG) signals, electrocardiogram (ECG) signals, and eye-tracking parameters. The goal was to investigate the impact of display modes (2D, normal 3D, and enhanced 3D) on visual fatigue. The results of paired samples ttests for both subjective and objective parameters indicated a significant influence of the display mode on visual fatigue. A visual fatigue assessment model employing multimodal parameters and based on a backpropagation neural network (BPNN) and XGBoost was proposed. The results suggested that the model was effective in predicting visual fatigue states, which the F1 and Accuracy value were both 0.94.