Multimodal Phenotyping Dataset of Driving Fatigue.

Driving fatigue analysis using multimodal physiological signals has become a growing focus in human factor engineering research. We introduce the Multimodal Phenotyping Dataset of Driving Fatigue (MPD-DF), a public resource comprising data from 50 participants (35 females and 15 males) during a standardized 2-hour simulated driving task. The dataset includes multidimensional subjective and objective metrics for fatigue assessment, comprising multimodal physiological recordings (32-channel electroencephalogram, single-lead electrocardiogram, dual-channel electrooculogram, and thoracic respiratory effort signals), fatigue-associated questionnaire results, and expert physician annotations of fatigue levels. To the best of our knowledge, it contains the largest number of signal modalities and provides the first physician-annotated, EEG-based fatigue labels among existing public datasets. The dataset validity was evaluated across multiple dimensions, including fatigue induction efficacy, statistical analysis of questionnaire outcomes, physiological signal quality, and correlations between physiological signals and annotated fatigue levels. This extensively characterized and validated dataset provides a benchmark resource for developing fatigue detection algorithms, facilitating cross-dataset validation, and advancing mechanistic studies of driving fatigue to enhance transportation safety.