EEG-XR: A Cross-Reality EEG Dataset for Neuroergonomic Analysis Across Desktop, Virtual, and Mixed Reality
Description
EEG-XR is a controlled cross-reality electroencephalography dataset collected during a functionally matched visual-attention task presented across Desktop, immersive virtual reality, and optical see-through mixed reality. The dataset was designed to support research on cognitive workload, short-term time-on-task dynamics, visual-fatigue-related activity, repeated-session stability, P300 Speller, and EEG signal quality across different display modalities. Seventeen healthy adult participants completed two measurement sessions. In each session, participants performed the same 6 × 6 row-column visual-stimulation task under three presentation conditions: a 24-inch Desktop monitor, an Oculus Quest 2 virtual-reality headset, and a Microsoft HoloLens 2 mixed-reality headset. Each modality included two task blocks using the target strings COMPUTER_SCIENCE and WORLD_UNIVERSITY. The task layout, target structure, stimulation sequence, color configuration, nominal event timing, and participant instructions were maintained across modalities. Continuous EEG was recorded at 500 Hz from 16 channels: F3, Fz, F4, FC1, FC2, C3, Cz, C4, CP1, CP2, P3, Pz, P4, O1, Oz, and O2. The recordings are provided in BrainVision format with synchronized row and column event markers. The dataset contains all 204 expected participant-condition recordings. Of these, 203 are complete, while one Session 1 VR recording for participant S06 contains a documented partial usable interval and is retained with an explicit quality flag. The repository includes anonymized participant folders, continuous EEG recordings, event-marker files, task and condition metadata, participant and recording inventories, channel definitions, recording-quality and eligibility information, file checksums, preprocessing configurations, and data-loading resources. Participants are identified only by codes S01-S17. The dataset supports participant-level regional and temporal EEG analyses, including frontal theta, parietal alpha, occipital alpha, occipital beta, theta-to-alpha measures, beta-to-alpha measures, Early-Middle-Late task phases, task comparisons, modality comparisons, and repeated-session analyses. It provides a reproducible resource for investigating EEG responses across non-immersive, immersive, and mixed-reality display environments.
Files
Steps to reproduce
Recruit adult participants with normal or corrected-to-normal vision and obtain written informed consent under an approved institutional ethics protocol. Implement the same 6 × 6 row-column visual-attention interface in Unity 3D for three display conditions: a 24-inch Desktop monitor, an Oculus Quest 2 virtual-reality headset, and a Microsoft HoloLens 2 optical see-through mixed-reality headset. Maintain the same matrix layout, target structure, colors, instructions, and stimulation procedure across modalities. Conduct two measurement sessions for each participant. Use the modality order Desktop–VR–MR in Session 1 and Desktop–MR–VR in Session 2. Within every modality, present the COMPUTER_SCIENCE block before the WORLD_UNIVERSITY block. For each target character, randomly present all six rows and six columns. Use a flash duration of 0.08 s, an inter-stimulus interval of 0.22 s, and a stimulus-onset interval of 0.30 s. Present five complete 12-event sequences per character. Provide a 5 s preparation interval between target characters and a 60 s delay before each task block. Record continuous EEG at 500 Hz from F3, Fz, F4, FC1, FC2, C3, Cz, C4, CP1, CP2, P3, Pz, P4, O1, Oz, and O2, using the right mastoid as reference and the forehead as ground. Store the recordings in BrainVision format. Record synchronized event markers using codes 4, 8, 12, 16, 20, and 24 for rows and 2, 6, 10, 14, 18, and 22 for columns. Verify the .vhdr, .eeg, and .vmrk files, sampling frequency, channel availability, recording duration, marker coverage, and task boundaries. Retain the first complete readable recording for each participant-condition combination and document partial or replaced recordings. To reproduce the spectral analysis, load the curated recordings using Python, MNE-Python, NumPy, SciPy, and pandas. Retain the first 240 s from the documented analysis start point and preserve the original 500 Hz sampling rate. Re-reference to the common average and apply a zero-phase FIR filter of 1–30 Hz for workload and temporal analyses or 1–35 Hz for visual-fatigue analysis. Divide the signal into 4 s windows with 2 s overlap. Reject windows containing nonfinite values, peak-to-peak amplitudes above 250 µV, or standard deviations below 0.10 µV. Retain conditions in which at least 50% of the candidate windows pass quality control. Estimate power spectral density using Welch’s method with 2 s Hann segments and 50% overlap. Calculate frontal theta at F3, Fz, F4, FC1, and FC2; parietal alpha at P3, Pz, and P4; and occipital alpha and beta at O1, Oz, and O2. Aggregate windows and electrodes to obtain one participant-level value per condition. For temporal analysis, divide each 240 s recording into Early, Middle, and Late 80 s phases. Perform participant-level paired tests and repeated-measures analyses using log-transformed values and Holm correction for multiple comparisons.
Institutions
- Nazarbayev UniversityAstana, Astana