Discover datasets

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

A brain-computer interface dataset implementing a gaze-independent rapid serial visual presentation (RSVP) P300 speller paradigm. Twelve healthy participants performed mental typewriting tasks using visual attention to discriminate target symbols from non-targets presented sequentially at screen center. The dataset includes preprocessed EEG recordings (63 channels, 200 Hz sampling rate after downsampling from 1000 Hz acquisition) with event-related potential markers, achieving 94.8% mean accuracy and 1.43 symbols/minute spelling rate.

Participants

12

Channels

63 (10-10)

Citations

95

HED

v8.4.0

Size

5.70 GB

Version

v1.0.2

Updated

Aug 18, 2026

This dataset comprises 64-channel EEG recordings from 11 healthy participants performing an auditory P300-based brain-computer interface task using selective attention to multiple tone streams. Participants attended to one of three auditory streams of musical tones and counted target stimuli while P300 event-related potentials were recorded. Classification using Riemannian geometry achieved >80% accuracy for 5 subjects and >75% for 9 subjects, demonstrating the feasibility of multi-class auditory BCIs based on auditory stream segregation.

Participants

11

Channels

64 (10-10)

Citations

11

HED

v8.4.0

Size

6.31 GB

Version

v1.0.4

Updated

Aug 18, 2026

This dataset contains EEG recordings from 11 healthy participants performing a multi-streamed musical oddball paradigm for auditory brain-computer interface (BCI) research. Participants selectively attended to one of three concurrent instruments in polyphonic music clips (Synth-Pop or Jazz) while counting deviant patterns, enabling classification of attended instrument from EEG signals including P300 and early auditory ERP components. The dataset is a BIDS-formatted derivative converted using MOABB from the original BNCI Horizon 2020 BNCI2015-006 dataset.

Participants

11

Channels

64 (10-10)

Citations

83

HED

v8.4.0

Size

11.8 GB

Version

v1.0.1

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

Imported from OpenNeuro ds004362

Participants

109

Channels

64 (10-10)

Citations

99

HED

v8.1.0

Size

11.1 GB

Version

v1.0.0

Updated

Jul 10, 2026

This dataset comprises preprocessed EEG recordings from a novel 9-class auditory event-related potential (ERP) paradigm designed for brain-computer interface (BCI) applications. Ten healthy participants performed a spelling task using a predictive text entry system (PASS2D) that exploited spatial auditory stimuli varying in pitch and direction. The study demonstrates competitive performance for auditory ERP-based BCIs, achieving 0.8 characters per minute and 3.4 bits/min information transfer rate, with potential applications for communication in patients with severe motor impairment.

Participants

10

Channels

63 (10-10)

Citations

84

HED

v8.4.0

Size

7.62 GB

Version

v1.0.2

Updated

Aug 18, 2026

An auditory brain-computer interface dataset implementing the AMUSE (Auditory Multi-class Spatial ERP) paradigm for spelling applications. The dataset comprises EEG recordings from 10 healthy subjects performing an auditory oddball task with spatial cues from six speaker locations, achieving online spelling performance of up to 1.41 characters per minute. Data includes preprocessed EEG signals with P300 and N200 event-related potentials, suitable for BCI research and benchmarking.

Participants

10

Channels

8 (10-10)

Citations

81

HED

v8.4.0

Size

74.4 MB

Version

v1.0.3

Updated

Aug 18, 2026

This dataset comprises P300 evoked potentials recorded from 10 healthy female subjects using two complementary paradigms: the P300 Speller (overt attention) and GeoSpell (covert attention) interfaces. Scalp EEG was acquired from 16 electrodes at 256 Hz while subjects focused on one of 36 alphanumeric characters across multiple sessions. The dataset includes preprocessed data with bandpass filtering (0.1-20 Hz) and is designed to investigate P300 latency jitter effects on brain-computer interface performance.

Participants

10

Channels

16 (10-10)

Citations

79

HED

v8.4.0

Size

247 MB

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