Discover datasets

This dataset comprises event-related potential (ERP) recordings from 13 healthy subjects performing a visual matrix speller task using a calibrationless brain-computer interface approach. The study introduces learning from label proportions (LLP), an unsupervised classification method that exploits known target/non-target stimulus ratios to enable online BCI operation without prior calibration. Subjects performed copy-spelling tasks using a 6×7 character grid across three sessions, achieving 84.5% character accuracy without labeled training data.

Participants

13

Channels

31 (10-10)

Citations

46

HED

v8.4.0

Size

5.15 GB

Version

v1.0.2

Updated

Aug 18, 2026

This derivative dataset comprises EEG recordings from 21 ALS subjects participating in the BigP3BCI Study M, a P300-based brain-computer interface experiment using a 9x8 adaptive/checkerboard visual speller paradigm. The data were acquired at 256 Hz using 16-channel EEG montage and represent a subset of the largest public P300 BCI dataset containing recordings from approximately 267 subjects across 20 studies. This dataset is processed and organized according to BIDS standards for facilitating reproducible BCI research and benchmarking.

Participants

21

Channels

16 (10-10)

HED

v8.4.0

Size

2.38 GB

Version

v1.0.2

Updated

Aug 18, 2026

This dataset comprises EEG recordings from a P300 visual matrix speller study comparing three unsupervised learning methods (Expectation-Maximization, Learning from Label Proportions, and their combination MIX) for brain-computer interface decoding. Twelve healthy participants performed a copy-spelling task using a modified 6×6 character grid extended with 10 # symbols as visual blanks (46 total symbols), recorded at 1000 Hz from 31 EEG channels. The study demonstrates that unsupervised learning methods can achieve performance comparable to supervised approaches without requiring calibration data.

Participants

12

Channels

31 (10-10)

Citations

26

HED

v8.4.0

Size

4.81 GB

Version

v1.0.2

Updated

Aug 18, 2026

BigP3BCI Study F is a derivative EEG dataset comprising P300-based brain-computer interface recordings from 10 ALS patients across 3 sessions using a 6x6 character grid speller paradigm. This dataset is extracted from the larger BigP3BCI benchmark (DOI: 10.13026/0byy-ry86) and processed using MOABB 1.5.0. The dataset contains 16-channel EEG data sampled at 256 Hz with visual stimulus presentations classified as target and non-target events, designed for evaluating P300 detection algorithms and BCI speller applications.

Participants

10

Channels

16 (10-10)

HED

v8.4.0

Size

1.91 GB

Version

v1.0.2

Updated

Aug 18, 2026

This dataset comprises EEG recordings from 8 ALS patients participating in a P300-based brain-computer interface speller paradigm, comparing 9x8 dry and wet electrode configurations. Subjects completed 2 sessions each with visual stimulus presentation and target/non-target classification tasks. The data were acquired at 256 Hz using 16-channel g.USBamp systems and represent a subset of the larger BigP3BCI dataset, the most comprehensive public P300 BCI resource.

Participants

8

Channels

16 (10-10)

HED

v8.4.0

Size

1.22 GB

Version

v1.0.3

Updated

Aug 18, 2026

This dataset comprises EEG recordings from 8 healthy subjects performing a P300-based brain-computer interface task using a 6x6 checkerboard visual stimulus paradigm. Recorded at 256 Hz with 16 channels using g.USBamp hardware, the data includes target and non-target event classifications for speller application development. This derivative dataset is part of the BigP3BCI study, the largest public P300 BCI dataset containing recordings from approximately 267 subjects across multiple stimulus paradigms.

Participants

8

Channels

16 (10-10)

HED

v8.4.0

Size

518 MB

Version

v1.0.2

Updated

Aug 18, 2026

This dataset comprises EEG recordings from 5 healthy subjects performing a P300-based brain-computer interface (BCI) speller task using a 9x8 adaptive/checkerboard stimulus paradigm across 2 sessions. The data were acquired at 256 Hz using 16 EEG channels and represent a derivative subset of the BigP3BCI study, the largest public P300 BCI dataset containing recordings from approximately 267 subjects across 20 studies. This dataset is suitable for P300 detection, feature extraction, and within-subject BCI performance evaluation.

Participants

5

Channels

16 (10-10)

HED

v8.4.0

Size

790 MB

Version

v1.0.3

Updated

Aug 18, 2026

This dataset comprises EEG recordings from 15 healthy subjects performing six different upper limb movements (elbow flexion/extension, forearm supination/pronation, hand open/close) and rest conditions in both movement execution and motor imagery modalities. The study investigates neural encoding of individual upper limb movements using low-frequency EEG signals (0.3-3 Hz) and achieves classification accuracies of 55-87% for executed movements and 27-73% for imagined movements. Source localization analysis identifies discriminative movement information in premotor areas, primary motor cortex, somatosensory cortex, and posterior parietal cortex, with applications toward non-invasive control of motor neuroprostheses and robotic arms.

Participants

15

Channels

61 (10-05)

Citations

94

HED

v8.4.0

Size

34.0 GB

Version

v1.0.3

Updated

Aug 18, 2026

A high-gamma EEG dataset comprising 14 healthy subjects performing motor imagery tasks (left hand, right hand, feet, and rest) recorded at 500 Hz with 128 channels. This is a BIDS-formatted derivative of the original dataset described in Schirrmeister et al. 2017, which was used to develop and validate deep convolutional neural networks for end-to-end EEG decoding. The derivative demonstrates that deep learning approaches can match or exceed traditional feature-based methods (FBCSP) while learning interpretable spectral power modulations in alpha, beta, and high-gamma frequency bands.

Participants

14

Channels

128 (10-05)

Citations

76

HED

v8.4.0

Size

31.3 GB

Version

v1.0.3

Updated

Aug 18, 2026

This dataset comprises continuous 2D trajectory decoding from attempted arm movements recorded via 60-channel EEG from 10 participants (9 able-bodied, 1 with spinal cord injury) across 3 sessions. Participants performed motor imagery tasks including target-tracking (snakerun) and self-paced shape tracing (freerun) with visual feedback. The preprocessed data, sampled at 20 Hz after extensive artifact removal and filtering, is designed for regression-based continuous decoding research and investigation of learning effects across sessions.

Participants

10

Channels

60 (10-05)

Citations

20

HED

v8.4.0

Size

7.38 GB

Version

v1.0.3

Updated

Aug 18, 2026