A Multimodal Dataset of Driving-Related Spontaneous Mind-Wandering

Existing datasets mostly focus on distraction triggered by external tasks and lack high-quality multimodal benchmark data targeting spontaneous mind wandering.This dataset centers on spontaneous cognitive distraction in driving. Adopting a three-stage experimental protocol (3–5 minutes of baseline recording, 3–5 minutes of focused driving, and 30 minutes of free driving), it records data on focused attention and spontaneous mind wandering states of 50 drivers in a monotonous highway simulation scenario.

The dataset integrates four types of modal signals synchronously: electroencephalography (EEG), eye movement, peripheral physiological signals (EDA, PPG), and vehicle control signals, achieving millisecond-level multimodal temporal alignment. For state labeling, a dual subjective-objective verification method combining the Attention Disengagement Index (ADI) and self-reports is proposed to conduct fine-grained dynamic labeling of distracted segments.This dataset can provide standardized benchmark data for research on the identification of driving-induced spontaneous mind wandering, classification of focused/distracted states, and driver status early warning.