Modified dataset including the NC instances for the three subjects: K3b, K6b and L1b. (ZIP)

as0305/BCI-Competition-IVa-dataset-3 dataset hosted on Hugging Face and contributed by the HF Datasets community

The goal of the "BCI Competition" is to validate signal processing and classification methods for Brain-Computer Interfaces (BCIs).

Classification accuracies (%) comparison on Dataset IVa of BCI Competition III.

This paper introduces a dataset capturing brain signals generated by the recognition of 100 Malayalam words

This EEG Brain Computer Interface (BCI) dataset was collected as part of the study titled: “Continuous Tracking using Deep Learning-based Decoding for Non-invasive Brain-Computer Interface”. If you use a part of this dataset in your work, please cite the following publication: D. Forenzo, H. Zhu, J. Shanahan, J. Lim, and B. He, “Continuous tracking using deep learning-based decoding for noninvasive brain–computer interface,” PNAS Nexus, vol. 3, no. 4, p. pgae145, Apr. 2024, doi: 10.1093/pnasnexus/pgae145. This dataset was collected under support from the National Institutes of Health via grants AT009263, NS096761, NS127849, EB029354, NS124564, and NS131069 to Dr. Bin He. Correspondence about the dataset: Dr. Bin He, Carnegie Mellon University, Department of Biomedical Engineering, Pittsburgh, PA 15213. E-mail: bhe1@andrew.cmu.edu

BCI competition III dataset 4a classification accuracy (%) with different classifiers.

Context : We share a large database containing electroencephalographic signals from 87 human participants, with more than 20,800 trials in total representing about 70 hours of recording. It was collected during brain-computer interface (BCI) experiments and organized into 3 datasets (A, B, and C) that were all recorded following the same protocol: right and left hand motor imagery (MI) tasks during one single day session. It includes the performance of the associated BCI users, detailed information about the demographics, personality and cognitive user’s profile, and the experimental instructions and codes (executed in the open-source platform OpenViBE). Such database could prove useful for various studies, including but not limited to: 1) studying the relationships between BCI users' profiles and their BCI performances, 2) studying how EEG signals properties varies for different users' profiles and MI tasks, 3) using the large number of participants to design cross-user BCI machine learning algorithms or 4) incorporating users' profile information into the design of EEG signal classification algorithms. Sixty participants (Dataset A) performed the first experiment, designed in order to investigated the impact of experimenters' and users' gender on MI-BCI user training outcomes, i.e., users performance and experience, (Pillette & al). Twenty one participants (Dataset B) performed the second one, designed to examined the relationship between users' online performance (i.e., classification accuracy) and the characteristics of the chosen user-specific Most Discriminant Frequency Band (MDFB) (Benaroch & al). The only difference between the two experiments lies in the algorithm used to select the MDFB. Dataset C contains 6 additional participants who completed one of the two experiments described above. Physiological signals were measured using a g.USBAmp (g.tec, Austria), sampled at 512 Hz, and processed online using OpenViBE 2.1.0 (Dataset A) & OpenVIBE 2.2.0 (Dataset B). For Dataset C, participants C83 and C85 were collected with OpenViBE 2.1.0 and the remaining 4 participants with OpenViBE 2.2.0. Experiments were recorded at Inria Bordeaux sud-ouest, France. Duration : Each participant's folder is composed of approximately 48 minutes EEG recording. Meaning six 7-minutes runs and a 6-minutes baseline. Documents Instructions: checklist read by experimenters during the experiments. Questionnaires: the Mental Rotation test used, the translation of 4 questionnaires, notably the Demographic and Social information, the Pre and Post-session questionnaires, and the Index of Learning style. English and french version Performance: The online OpenViBE BCI classification performances obtained by each participant are provided for each run, as well as answers to all questionnaires Scenarios/scripts : set of OpenViBE scenarios used to perform each of the steps of the MI-BCI protocol, e.g., acquire training data, calibrate the classifier or run the online MI-BCI Database : raw signals Dataset A : N=60 participants Dataset B : N=21 participants Dataset C : N=6 participants The article that expained the database is available here: Dreyer, P., Roc, A., Pillette, L. et al. A large EEG database with users’ profile information for motor imagery brain-computer interface research. Sci Data 10, 580 (2023). https://doi.org/10.1038/s41597-023-02445-z {"references": ["Pillette & al (2021). Experimenters Influence on Mental-Imagery based Brain-Computer Interface User Training. International Journal of Human-Computer Studies, pp.102603.", "Camille Benaroch & al (2022). When should MI-BCI feature optimization include prior knowledge, and which one?. Brain-Computer Interfaces, 9 (2), pp.115-128"]} https://doi.org/10.1038/s41597-023-02445-z

Context : We share a large database containing electroencephalographic signals from 87 human participants, with more than 20,800 trials in total representing about 70 hours of recording. It was collected during brain-computer interface (BCI) experiments and organized into 3 datasets (A, B, and C) that were all recorded following the same protocol: right and left hand motor imagery (MI) tasks during one single day session. It includes the performance of the associated BCI users, detailed information about the demographics, personality and cognitive user’s profile, and the experimental instructions and codes (executed in the open-source platform OpenViBE). Such database could prove useful for various studies, including but not limited to: 1) studying the relationships between BCI users' profiles and their BCI performances, 2) studying how EEG signals properties varies for different users' profiles and MI tasks, 3) using the large number of participants to design cross-user BCI machine learning algorithms or 4) incorporating users' profile information into the design of EEG signal classification algorithms.

Sixty participants (Dataset A) performed the first experiment, designed in order to investigated the impact of experimenters' and users' gender on MI-BCI user training outcomes, i.e., users performance and experience, (Pillette & al). Twenty one participants (Dataset B) performed the second one, designed to examined the relationship between users' online performance (i.e., classification accuracy) and the characteristics of the chosen user-specific Most Discriminant Frequency Band (MDFB) (Benaroch & al). The only difference between the two experiments lies in the algorithm used to select the MDFB. Dataset C contains 6 additional participants who completed one of the two experiments described above. Physiological signals were measured using a g.USBAmp (g.tec, Austria), sampled at 512 Hz, and processed online using OpenViBE 2.1.0 (Dataset A) & OpenVIBE 2.2.0 (Dataset B). For Dataset C, participants C83 and C85 were collected with OpenViBE 2.1.0 and the remaining 4 participants with OpenViBE 2.2.0. Experiments were recorded at Inria Bordeaux sud-ouest, France.

Duration : Each participant's folder is composed of approximately 48 minutes EEG recording. Meaning six 7-minutes runs and a 6-minutes baseline.

Documents Instructions: checklist read by experimenters during the experiments. Questionnaires: the Mental Rotation test used, the translation of 4 questionnaires, notably the Demographic and Social information, the Pre and Post-session questionnaires, and the Index of Learning style. English and french version Performance: The online OpenViBE BCI classification performances obtained by each participant are provided for each run, as well as answers to all questionnaires Scenarios/scripts : set of OpenViBE scenarios used to perform each of the steps of the MI-BCI protocol, e.g., acquire training data, calibrate the classifier or run the online MI-BCI

Database : raw signals Dataset A : N=60 participants Dataset B : N=21 participants Dataset C : N=6 participants

Prediction of Finger Flexion IV Brain-Computer Interface Data Competition The goal of this dataset is to predict the flexion of individual fingers from signals recorded from the surface of the brain (electrocorticography (ECoG)). This data set contains brain signals from three subjects, as well as the time courses of the flexion of each of five fingers. The task in this competition is to use the provided flexion information in order to predict finger flexion for a provided test set. The performance of the classifier will be evaluated by calculating the average correlation coefficient r between actual and predicted finger flexion.

ECoG data during individual flexions of the five fingers; movements acquired with a data glove. [48 - 64 ECoG channels (0.15-200Hz), 1000Hz sampling rate, 3 subjects]

Brain-Computer Interfaces, and especially passive Brain-Computer Interfaces (pBCI), with their ability to estimate and detect mental states, are receiving increasing attention from both the scientific and the research and development communities. Many pBCIs aim to increase the safety of complex work environments such as in the aeronautical domain. Therefore, mental workload, vigilance and decision-making are some of the most commonly examined aspects of cognition within this field of research. A large proportion of pBCIs involve a component of machine learning and signal processing as the data that are collected need to be transformed into a reliable estimate of the users’ current mental state (e.g. mental workload). Improving this component is a major challenge for researchers, requiring large quantities of data. While data sharing is common for the active BCI community, open pBCI datasets are scarcer and generally incomplete with regards to the information they report. This is particularly true for datasets encompassing several tasks or sessions, which are of importance for tackling the challenges of transfer learning. Testing new pipelines, feature extraction algorithms and classifiers are central issues for future advances in research within this domain, as well as for algorithm benchmark and research reproducibility.The COG-BCI database presented here is comprised of the recordings of 29 participants over 3 individual sessions with 4 different tasks designed to elicit different cognitive states. This results in a total of over 100 hours of open electrophysiological (EEG) and electrocardiogram (ECG) data. The project was validated by the local ethical committee of the University of Toulouse (CER number 2021-342). The dataset was validated on a subjective, behavioral and physiological level (i.e. cardiac and cerebral activity), to ensure its usefulness to the pBCI community. This body of work represents a large effort to promote the use of pBCIs, as well as the use of open science.

The data are in the Brain Imaging Data Structure (BIDS) format. For more information, please read the COG-BCI_info.pdf file.

Brain-computer interface (BCI) enables bodyless communication with machines or devices; this is done using the translation of the electrical activity in the brain (EEG) signals into outputs command. Motor Imagery BCI (MI-BCI) employs the amplitude changes elicited voluntarily by the mental rehearsal of physical motor actions (known as event-related de-synchronization and synchronization - ERD/ERS). MI-BCI applications focus on mental representations of jointed limbs following the human anatomy constraints (e.g., two arms in a symmetrical distribution), without any exploration or applications that include non-embodied human limbs in BCI systems, even though Rubber Hand Illusion (RHI) experiments demonstrated the human capabilities to create body transfer illusions. This dataset aims to study the feasibility of including a virtual third arm in an MI-BCI system while comparing the effectiveness of using the conventional arrows and fixation cross as a training step (Graz) against a first-person view using a human avatar (Hands). The dataset was used for two studies: an EEG analysis of the induced brain oscillatory activity elicited by the third arm using Event-Related Spectral Perturbation (ERSP); and an offline exploration of the classification of the third arm task. There were two recording sessions with two runs in each one with a resting time between them. The sessions were conducted on two separate days within one week. Ten right-handed volunteers (four women) participated in the study. We collected the EEG data using an OpenBCI 32-bit board at a sampling rate of 250 Hz. Following the 10-20 EEG placement system, eight passive gold cup electrodes were used and placed at the sensorimotor cortex. The experiment involves the execution of four different tasks in two training conditions (both performed in a VR environment). The subjects were invited to rest (RS), or to move (either imaginary or execute when possible) a specific hand: third hand (TH), left hand (LH), and right hand (RH). The Hands condition involved the presentation of a human-like avatar, whereas Graz the presentation of arrows, both following the usual BCI timing protocol (see the Figure). The users performed 20 trials of each task randomly selected with a duration of seven seconds each.

Performance of the 3-Class BCI.Performance of the 3-Class BCI.

ENCODES a protein that exhibits DNA-binding transcription factor binding (ortholog); histone deacetylase binding (ortholog); protein-macromolecule adaptor activity (ortholog); INVOLVED IN antimicrobial humoral response (ortholog); canonical NF-kappaB signal transduction (ortholog); defense response to bacterium (ortholog); PARTICIPATES IN nuclear factor kappa B signaling pathway; ASSOCIATED WITH atherosclerosis (ortholog); Dyslipidemias (ortholog); prion disease (ortholog); FOUND IN Bcl3-Bcl10 complex (ortholog); Bcl3/NF-kappaB2 complex (ortholog); ciliary basal body (ortholog); INTERACTS WITH 1-naphthyl isothiocyanate; 17beta-estradiol; 17beta-estradiol 3-benzoate

This dataset captures longitudinal brain-computer interface (BCI) training data collected from a total of 27 human participants, including 25 healthy individuals and 2 individuals with spinal cord injury (SCI). The dataset is designed to support research into motor imagery (MI)-based BCI performance, neuromodulation through transcutaneous electrical spinal stimulation (TESS), and neuroplasticity associated with longitudinal BCI use. It contains high-resolution EEG and EOG recordings, performance metrics, stimulation metadata, and detailed session logs across multiple experimental conditions.

1. Participants

• Healthy Cohort (n=25):

  • Age: 18–30 years

  • Gender: 48% female

  • No prior history of neurological disorders

• SCI Participants (n=3):

  • Patient-I: C5 complete lesion, 9 years post-injury, consistent motor rehabilitation

  • Patient-II: C4 complete lesion, 3 months post-injury, stimulation at C5/C6

  • (Note: One female SCI participant excluded due to medical complications)

2. Experimental Design

The study included three major participant groups:

• Main Group (n=20):

  • EEG recordings with 32-channel cap

  • Randomized controlled study to evaluate the effects of TESS on BCI learning

  • Split into:

    • TESS Group (n=10): Received 20 minutes of spinal stimulation before each training session

    • Rest Group (n=10): Completed 20 minutes of rest instead

  • Protocol lasted five days, including offline and online BCI sessions

• Cross-Over Group (n=4):

  • Subset of Rest group participants who failed to reach BCI control

  • Underwent a second round of training with the TESS condition after a 6-month washout

  • Included additional pre/post stimulation resting-state EEG recordings

  • Final follow-up session one week post-training to assess retention

  • EEG recordings with 64-channel cap, including pre/post stimulation resting-state EEG

• Single-Pulse Group (n=5):

  • EEG recordings with 64-channel cap

  • Underwent identical training with a single-pulse stimulation protocol instead of multi-pulse TESS

  • Designed to serve as a placebo/control to isolate TESS-specific effects

  • Included resting-state EEG pre/post stimulation

• SCI Participants (n=2):

  • Each completed the full protocol under both TESS and Rest conditions

  • EEG recordings with 64-channel cap, including pre/post stimulation resting-state EEG

3. EEG and EOG Data Acquisition

• Sampling Rate: 512 Hz

• Electrode Montage:

  • 32-channel 10–10 system for healthy participants

  • 64-channel high-density cap for SCI and subgroup participants

  • Ground: AFz; Reference: CPz

• Recording System: ANTNeuro EEGO with Ag/AgCl-coated electrodes

• Artifact Monitoring: Integrated EOG channels for blink and eye movement detection

4. BCI Task and Feedback

• Modality: Motor imagery (left/right hand)

• Task: Kinesthetic MI with bar feedback interface

• Online BCI Decoder: Riemannian geometry-based, recalibrated after conditioning phase

• Feedback Algorithm:

  • Evidence accumulation over sliding windows

  • Decisions based on dynamic thresholds

• Trial Outcomes: Hit, Miss, Timeout

• Session Format:

  • 3–4 runs/session

  • 20 trials/run

  • EEG time-locked to MI cues, execution, and feedback events

5. Stimulation Conditions

• TESS Protocol: Multi-pulse 5 kHz carrier at 30 Hz

• Single-Pulse Protocol: 200 µs biphasic pulse once per 30 Hz period

• Stimulation Duration: 20 minutes/session (pre-BCI)

• Target Site: Cervical spine (C5/C6)

• Additional EEG Collection: 2-minute resting state EEG recordings pre/post stimulation in relevant groups

6. Metadata and Annotations

• Session Labels: Offline, Online-Baseline, Post-Conditioning, Long-Term Follow-Up

• Condition Tags: TESS, Rest, Single-Pulse

• Performance Scores: Trial-level and session-level accuracy, command latency, hit/miss distribution

• Participant Information: Anonymized IDs, handedness, group assignment, SCI details (where applicable)

• Stimulation Logs: Protocol type, duration, electrode configuration, session timestamps

7. Ethical Compliance

• IRB Approval: University of Texas at Austin (Protocol #2020-03-0073)

• ClinicalTrials.gov Registration: NCT05183152

• Informed Consent: Obtained from all participants

Finland BCI: Services: sa: Expectations for the Next 3 Months data was reported at 9.000 % in Apr 2025. This records a decrease from the previous number of 15.000 % for Mar 2025. Finland BCI: Services: sa: Expectations for the Next 3 Months data is updated monthly, averaging 17.500 % from Jan 2005 (Median) to Apr 2025, with 244 observations. The data reached an all-time high of 42.000 % in Nov 2005 and a record low of -60.000 % in Apr 2020. Finland BCI: Services: sa: Expectations for the Next 3 Months data remains active status in CEIC and is reported by Confederation of Finnish Industries. The data is categorized under Global Database’s Finland – Table FI.S001: Business Confidence Indicator: Seasonally Adjusted.

Map based on aerial photo taken in early June 1927. The original oblique photo was digitally flattened, aligned with the current shoreline of BCI, and georeferenced for further use in a geographic information system (GIS). The rectified photograph was digitally smoothed and each pixel received a value between 0 (black, dense canopy) and 255 (white, cleared). The color values were divided into five classes (posterized): (1) 0–31, (2) 31–95, (3) 95–160, (4) 160–220, and (5) 220–255, depicting different forest ages. Isolated patches (<100 m2) were manually removed. Comparison with other BCI forest-age maps indicates that these categories correspond well to the forest disturbance categories distinguished on BCI: old-growth (classes 1 and 2), tall-secondary (class 3), and low-secondary (classes 4 and 5) forest.This map was first published in Albrecht et a. 2017. References: Albrecht, Larissa, Robert F. Stallard, and Elisabeth K. V. Kalko. 2017. "Land use history and population dynamics of free-standing figs in a maturing forest." Plos One 12.https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0177060

Brain–computer interfaces (BCIs) have wide-ranging applications as solutions for replacing or substituting neural output that has been lost because of severe neuromuscular injury or disease, such as individuals with late-stage amyotrophic lateral sclerosis (ALS). The P300-based BCI is one of the most commonly researched BCI for communication. This BCI dataset is curated from data originally generated from previous visual P300-based BCI speller studies, which include single- and multi-session experiments under a wide range of conditions. The BCI data are provided in an enriched and standardised format with BCI data elements that align with developing IEEE P2731 Working Group standards for BCI data to facilitate reusability. The data files, provided in open European Data Format ‘plus’, contain: i) electroencephalography (EEG) signals; ii) the BCI encoder, target characters and stimulus event markers for P300 event related potential analysis; iii) BCI spelling outcomes and feedback event markers for error related potential analysis; and if available, iv) self-reported demographics (age, sex, race, ethnicity); v) ALS diagnosis and a revised ALS Functional Rating Scale score obtained from medical records; and vi) eye tracker signals.

BNCI 2014-001 Motor Imagery dataset Dataset IIa from BCI Competition 4 1.

Dataset Description This data set consists of EEG data from 9 subjects. The cue-based BCI paradigm consisted of four different motor imagery tasks, namely the imagination of movement of the left hand (class 1), right hand (class 2), both feet (class 3), and tongue (class 4). Two sessions on different days were recorded for each subject. Each session is comprised of 6 runs separated by short breaks. One run consists of 48 trials (12 for each of the four possible classes), yielding a total of 288 trials per session.

The subjects were sitting in a comfortable armchair in front of a computer screen. At the beginning of a trial ( t = 0 s), a fixation cross appeared on the black screen. In addition, a short acoustic warning tone was presented. After two seconds ( t = 2 s), a cue in the form of an arrow pointing either to the left, right, down or up (corresponding to one of the four classes left hand, right hand, foot or tongue) appeared and stayed on the screen for 1.25 s. This prompted the subjects to perform the desired motor imagery task. No feedback was provided. The subjects were ask to carry out the motor imagery task until the fixation cross disappeared from the screen at t = 6 s.

Twenty-two Ag/AgCl electrodes (with inter-electrode distances of 3.5 cm) were used to record the EEG; the montage is shown in Figure 3 left. All signals were recorded monopolarly with the left mastoid serving as reference and the right mastoid as ground. The signals were sampled with. 250 Hz and bandpass-filtered between 0.5 Hz and 100 Hz. The sensitivity of the amplifier was set to 100 μV . An additional 50 Hz notch filter was enabled to suppress line noise.

References 1 Tangermann, M., Müller, K.R., Aertsen, A., Birbaumer, N., Braun, C., Brunner, C., Leeb, R., Mehring, C., Miller, K.J., Mueller-Putz, G. and Nolte, G., 2012. Review of the BCI competition IV. Frontiers in neuroscience, 6, p.55.

Comparison of the performance obtained using different γ and c values.