EEG dataset for Drowsiness Detection

We present a publicly available EEG dataset for the detection of driver drowsiness, collected with the Upside Down Labs BioAmp EXG pill. Data were acquired from healthy adult volunteers, each seated in a stationary chair while wearing a custom EEG helmet. Electrode contacts were embedded in the helmet lining at Fp1 and Fp2 (international 10\u201320 system) with an earlobe reference (A1), leveraging the hairless prefrontal region for optimal signal quality .Each subject underwent two 20\u2011minute sessions (totaling 40 minutes):Active condition: participants performed a simulated driving task with sustained alertness.Drowsy condition: participants relaxed or engaged in minimal activity to induce fatigue.Signals were sampled at 512\u202fHz, generating approximately 1.2\u202fmillion raw EEG samples per subject. Raw data were band\u2011pass filtered, and artifacts were removed using standard rejection criteria. We then segmented the cleaned signals into overlapping 1\u2011second windows (512 samples, 0.5\u2011second step), yielding 4,769 labeled segments classified as \u201cactive\u201d or \u201cdrowsy.\u201dThe dataset comprises two CSV files\u2014one containing raw EEG signals for the active state and the other containing raw EEG signals for the Inactive state, facilitating reproducible research in drowsiness detection, feature extraction, and classification.  We aim to support the development of robust driver-monitoring systems and advance the study of EEG\u2011based vigilance assessment.