Discover datasets

This dataset is Study Q of bigP3BCI, an open, machine-learning-ready collection of P300-based brain-computer interface EEG recordings. It contains 32-channel EEG from 36 participants, each recorded over 3 sessions, while they used a 6x6 character speller with color intensification. Target and non-target events are annotated with HED tags. The dataset is a BIDS-formatted derivative of the PhysioNet bigP3BCI release, prepared with MOABB.

Participants

36

Channels

32 (10-10)

Citations

2

HED

v8.4.0

Size

3.70 GB

Version

v1.0.4

Updated

Oct 8, 2026

Study P of the bigP3BCI collection, an open, machine-learning-ready P300-based brain-computer interface dataset. It contains 32-channel EEG recorded at 256 Hz from 19 participants with ALS, each over two sessions, while they used a 9x8 character grid speller with predictive and non-predictive spelling paradigms. Target and non-target flash events are annotated with HED tags, and the data are reformatted from the original PhysioNet release into BIDS.

Participants

19

Channels

32 (10-10)

Citations

2

HED

v8.4.0

Size

4.96 GB

Version

v1.0.4

Updated

Oct 8, 2026

EEG recordings from 18 healthy participants performing a gaze-independent brain-computer interface task based on covert spatial attention shifts. Participants answered yes/no questions by shifting attention to a green or red cross while maintaining central fixation, and N2pc-related ERP responses were classified with canonical correlation analysis. The data were recorded with 30 EEG channels and 2 EOG channels at 250 Hz.

Participants

18

Channels

30 (10-10)

Citations

26

HED

v8.4.0

Size

1.82 GB

Version

v1.0.3

Updated

Oct 8, 2026

EEG recordings from five healthy participants performing cued motor imagery (right hand versus foot) with 118 electrodes, originally published to demonstrate improved single-trial classification through feature combination and multiclass paradigms. The data were downsampled to 100 Hz and comprise 280 cued trials per subject, split into labeled training and unlabeled test sets. The dataset was part of BCI Competition III and is widely used for benchmarking BCI classification algorithms.

Participants

5

Channels

118 (10-05)

Citations

86

HED

v8.4.0

Size

735 MB

Version

v1.0.3

Updated

Aug 18, 2026

This dataset contains 64-channel EEG recorded from 31 healthy participants performing a two-class (left vs. right hand) motor imagery task. Each participant completed up to three sessions of 40 trials. The recordings are part of a multimodal collection that also includes eye-tracking and high-speed video, intended for analyzing ocular activity across BCI paradigms. Events are annotated with HED, and the data are organized in BIDS format via MOABB.

Participants

31

Channels

62 (10-10)

Citations

13

HED

v8.4.0

Size

33.6 GB

Version

v1.0.3

Updated

Aug 18, 2026

EEG recordings from 38 subjects playing in pairs the multi-user version of Brain Invaders, a visual P300-based brain-computer interface using an oddball paradigm on a grid of 36 symbols. Data were acquired with 32 wet electrodes per subject during three randomized conditions (Solo1, Solo2, Collaboration), supporting research on inter-brain synchrony and collaborative BCI. The experiment was conducted at GIPSA-lab, Grenoble, France, in 2014.

Participants

38

Channels

32 (10-10)

Citations

17

HED

v8.4.0

Size

1.24 GB

Version

v1.0.3

Updated

Oct 8, 2026

EEG recordings from a visual P300 Brain-Computer Interface videogame, Brain Invaders, collected at GIPSA-lab (Grenoble, France) in 2014. Participants played a game using an oddball paradigm on a grid of 36 symbols while 16 dry electrodes recorded their EEG, with up to three game sessions per participant. The dataset was designed to study the viability of a calibration-less P300-based BCI with dry electrodes.

Participants

64

Channels

16 (10-10)

Citations

19

HED

v8.4.0

Size

3.06 GB

Version

v1.0.4

Updated

Oct 8, 2026

EEG recordings from 20 healthy participants who wrote ten letters (a, d, e, f, j, n, o, s, t, v) with the right index finger after a visual cue. The dataset supports classification of handwritten characters, either directly from low-frequency and broadband EEG or with a two-step approach that first decodes continuous hand kinematics and then classifies the letter. It is intended for research on non-invasive brain-computer interfaces for communication.

Participants

20

Channels

60 (10-10)

Citations

13

HED

v8.4.0

Size

27.0 GB

Version

v1.0.2

Updated

Aug 18, 2026

EEG dataset from the Basic Guard Duty study of the Brain-Computer Interaction Technologies (BCIT) collection, designed to measure sustained vigilance in a realistic setting. Participants verified replica ID cards against reference face photographs at varying queuing rates while 256-channel BioSemi EEG and eye tracking were recorded. The dataset is formatted in BIDS with HED annotations and accompanies companion calibration and baseline driving datasets.

Participants

21

Channels

256 (biosemi)

Citations

23

HED

v8.0.0

Size

55.1 GB

Version

v1.0.0

Updated

Jun 23, 2026

This dataset contains EEG recordings from 13 healthy participants in Study I of the bigP3BCI collection. Participants used a P300-based speller with a 9x8 checkerboard character grid and a performance-based paradigm. Data were recorded from 16 channels at 256 Hz and are annotated with target and non-target stimulus events using HED. The dataset is a BIDS-formatted derivative of the PhysioNet bigP3BCI release, prepared for BCI benchmarking.

Participants

13

Channels

16 (10-10)

HED

v8.4.0

Size

1.57 GB

Version

v1.0.3

Updated

Oct 8, 2026