Discover datasets

An EEG dataset of imagined speech from 15 healthy participants comparing traditional cue-based and gamified (Pac-Man maze) paradigms for brain-computer interface applications. It comprises 1,800 trials of four Spanish directional commands (avanzar, retroceder, derecha, izquierda) recorded at 500 Hz from 24 electrodes. This BIDS derivative, distributed via MOABB and NEMAR, supports evaluation of how paradigm design affects imagined speech decoding.

Participants

15

Channels

24 (10-10)

HED

v8.4.0

Size

1.52 GB

Version

v1.0.3

Updated

Aug 18, 2026

Scalp EEG recordings from 27 stroke patients performing a lower-limb motor imagery task under a multi-paradigm protocol with longitudinal, repeated-session training. The dataset supports research on motor-imagery brain-computer interfaces for lower-limb and gait rehabilitation after stroke. It is distributed in BIDS format and was collected at Tianjin University.

Participants

27

Channels

64 (10-10)

Citations

27

HED

v8.4.0

Size

56.7 GB

Version

v1.0.5

Updated

Aug 20, 2026

HEFMI-ICH is a hybrid EEG-fNIRS motor imagery dataset for brain-computer interface research in intracerebral hemorrhage (ICH). It includes 37 participants (17 healthy, 20 ICH patients) who performed two-class left- and right-hand grasping motor imagery over up to three sessions. This MOABB-packaged version contains pre-epoched, preprocessed 32-channel EEG, intended for developing BCI-based rehabilitation approaches.

Participants

37

Channels

32 (10-10)

Citations

6

HED

v8.4.0

Size

6.42 GB

Version

v1.0.3

Updated

Aug 18, 2026

An open-access EEG dataset of a motor imagery brain-computer interface experiment conducted under distraction. Sixteen healthy participants performed left- versus right-hand motor imagery, first in a calibration run and then in feedback runs under six conditions: no distraction, eyes closed, news listening, number search, flickering video, and vibro-tactile stimulation. Recordings comprise 63 EEG channels at 1000 Hz and 504 trials per subject.

Participants

16

Channels

63 (10-05)

Citations

13

HED

v8.4.0

Size

72.1 GB

Version

v1.0.7

Updated

Aug 18, 2026

EEG, EMG and vehicle-dynamics recordings from 15 healthy participants who drove a simulated racing car while following a lead vehicle that occasionally braked abruptly. The dataset supports detection of emergency-braking intention from brain and muscle signals before the behavioral response, with target and non-target events for event-related potential and BCI analyses. It is a preprocessed BIDS derivative of the BNCI Horizon 2016-002 collection, distributed via MOABB.

Participants

15

Channels

59 (10-10)

Citations

82

HED

v8.4.0

Size

14.2 GB

Version

v1.0.3

Updated

Oct 8, 2026

This dataset contains 32-channel EEG recorded from 22 healthy participants performing visual imagery of ten cued items: animals (dog, bird, fish), geometric figures (pentagram, square, circle), and objects (scissor, watch, cup, chair). Each participant completed up to two sessions of three runs, giving 16,800 trials in total. It is intended for developing visual imagery-based brain-computer interfaces and is distributed in BIDS format with HED event annotations.

Participants

22

Channels

32 (10-10)

Citations

3

HED

v8.4.0

Size

75.4 GB

Version

v1.0.3

Updated

Aug 20, 2026

EEG recordings from 21 participants performing an offline auditory spatial oddball task for a multi-class brain-computer interface (AMUSE). Five frontal speakers presented unique auditory stimuli under three conditions that varied inter-stimulus interval and speaker configuration, and participants attended to target directions. The data were used to evaluate P300-based spelling and communication paradigms, and are distributed here in preprocessed form through MOABB.

Participants

21

Channels

60 (10-10)

Citations

91

HED

v8.4.0

Size

16.5 GB

Version

v1.0.3

Updated

Oct 8, 2026

EEG recordings from 16 healthy, BCI-naive participants using a motion visually evoked potential (mVEP) speller for gaze-independent communication. Three speller variants were compared: Overt Cake Speller, Covert Cake Speller, and Motion Center Speller. Each used a two-level selection among 30 symbols. The dataset includes practice, calibration, copy-spelling and free-spelling sessions with target and non-target events, recorded from 63 channels. It is a preprocessed derivative distributed in BIDS format.

Participants

16

Channels

63 (10-10)

Citations

46

HED

v8.4.0

Size

3.79 GB

Version

v1.0.3

Updated

Oct 8, 2026

EEG recordings from 13 participants using a P300-based visual speller that relies on covert spatial attention and non-spatial feature attention (color and form), so users do not need to move their gaze. The study compared three speller variants (Hex-o-Spell, Cake Speller, Center Speller) in a calibration and online-test design. Data were recorded with 63 channels at 250 Hz, and event markers distinguish target from non-target intensifications.

Participants

13

Channels

63 (10-05)

Citations

81

HED

v8.4.0

Size

9.27 GB

Version

v1.0.3

Updated

Oct 8, 2026

EEG recordings from 12 healthy participants operating a gaze-independent rapid serial visual presentation (RSVP) P300 speller. Thirty symbols were shown one at a time at the center of the screen under three conditions (NoColor 116 ms SOA, Color 116 ms SOA, Color 83 ms SOA), each with calibration, copy-spelling and free-spelling phases. Online spelling reached a mean accuracy of 94.8% at 1.43 symbols/min, supporting use by patients with impaired oculo-motor control.

Participants

12

Channels

63 (10-10)

Citations

95

HED

v8.4.0

Size

5.70 GB

Version

v1.0.3

Updated

Oct 8, 2026