EEG Motor Movement/Imagery Dataset v1.0.0 – 30 Subjects

A powerful, high‑resolution EEG collection featuring 64‑channel recordings sampled at 160 Hz, captured from 30 healthy adults performing a comprehensive battery of motor execution and imagery tasks. Derived from PhysioNet's EEGMMIDB, this dataset offers rich, well‑annotated data across 14 runs per subject, each lasting 1–2 minutes:

1. Baseline (Eyes Open)
2. Baseline (Eyes Closed)
3. Left/Right Fist Movement
4. Left/Right Fist Imagery
5. Both Fists/Feet Movement
6. Both Fists/Feet Imagery …and each of the above tasks repeated twice (physionet.org).

Annotations include detailed timing channels with event codes (T0 = rest, T1 = motion onset, T2 = right‑fist or feet onset), exported both in EDF+ and BCI2000 formats with accompanying .event files for precise epoch alignment (physionet.org).

Electrodes adhere to the 10‑10 international montage (64 channels), excluding peripheral electrodes like Nz, F9/F10, FT9/10, A1/A2, TP9/10, P9/10—ensuring full scalp coverage with high signal quality (physionet.org).

🧠 Ideal For

• BCI researchers exploring motor imagery paradigms
• Signal processing enthusiasts working with CSP, ERD/ERS, wavelet, fractal or spectral features
• Machine learning practitioners training LDA/SVM baselines or developing CNN/LSTM deep-learning architectures
• Transfer learning & cross-subject studies for evaluating generalization

⚙️ File Details

All recordings are provided in EDF+ format (~3.4 GB total for all subjects), with paired .event annotation files for convenient epoch extraction. Compatible with PhysioToolkit, BCI2000, MNE-Python, EEGLAB, and MATLAB.

🛡️ Source & License

Originally recorded by Schalk et al. under the BCI2000 framework and contributed by Schalk, McFarland, Wolpaw and collaborators—this dataset is published by PhysioNet under the Open Data Commons Attribution License v1.0 (physionet.org, mne.tools).

With its blend of high‑density EEG, rigorous task design, and robust annotation, this dataset is an exceptional resource for advancing BCI methods, neuroscience exploration, and intelligent signal decoding.

This dataset is a cleansed, transformed, and structured version of Physionet EEG Motor Movement/Imagery Dataset (https://physionet.org/content/eegmmidb/1.0.0/) The original dataset is licensed under Open Data Commons Attribution (https://physionet.org/content/eegmmidb/view-license/1.0.0/)

The dataset is offered in two format for versatility and convenience; A MATLAB structure and a collection of CSV files.

Dataset

This dataset was created by 6541816

Contents

Acknowledgements

This data set was originally created and contributed to PhysioBank by Gerwin Schalk (schalk at wadsworth dot org) and his colleagues at the BCI R&D Program, Wadsworth Center, New York State Department of Health, Albany, NY. W.A. Sarnacki collected the data. Aditya Joshi compiled the dataset and prepared the documentation. D.J. McFarland and J.R. Wolpaw were responsible for experimental design and project oversight, respectively. This work was supported by grants from NIH/NIBIB ((EB006356 (GS) and EB00856 (JRW and GS)).

To access the initial publication of this dataset, please visit this link to PhysioBank: https://physionet.org/content/eegmmidb/1.0.0/

Experiment Protocol

This data set consists of over 1500 one- and two-minute EEG recordings, obtained from 109 volunteers, as described below.

Subjects performed different motor/imagery tasks while 64-channel EEG were recorded using the BCI2000 system (http://www.bci2000.org). Each subject performed 14 experimental runs: two one-minute baseline runs (one with eyes open, one with eyes closed), and three two-minute runs of each of the four following tasks:

[Task 1] A target appears on either the left or the right side of the screen. The subject opens and closes the corresponding fist until the target disappears. Then the subject relaxes.
[Task 2] A target appears on either the left or the right side of the screen. The subject imagines opening and closing the corresponding fist until the target disappears. Then the subject relaxes.
[Task 3] A target appears on either the top or the bottom of the screen. The subject opens and closes either both fists (if the target is on top) or both feet (if the target is on the bottom) until the target disappears. Then the subject relaxes.
[Task 4] A target appears on either the top or the bottom of the screen. The subject imagines opening and closing either both fists (if the target is on top) or both feet (if the target is on the bottom) until the target disappears. Then the subject relaxes.

In summary, the experimental runs were:
1. Baseline, eyes open
2. Baseline, eyes closed
3. Task 1 (open and close left or right fist)
4. Task 2 (imagine opening and closing left or right fist)
5. Task 3 (open and close both fists or both feet)
6. Task 4 (imagine opening and closing both fists or both feet)
7. Task 1
8. Task 2
9. Task 3
10. Task 4
11. Task 1
12. Task 2
13. Task 3
14. Task 4

Each event code includes an event type indicator (T0, T1, or T2) that is concatenated to the Task # it belongs with (i.e TASK1T2). The event type indicators change definition depending on the Task # it is associated with. For example, TASK1T2 would correspond to the onset of real motion in the right fist, while TASK3T2 would correspond to onset of real motion in both feet:

[T0] corresponds to rest

[T1] corresponds to onset of motion (real or imagined) of:

• the left fist (in runs 3, 4, 7, 8, 11, and 12; for Task 1 (real) and Task 2 (imagined))
• both fists (in runs 5, 6, 9, 10, 13, and 14; for Task 3 (real) and Task 4 (imagined))

[T2] corresponds to onset of motion (real or imagined) of:

• the right fist (in runs 3, 4, 7, 8, 11, and 12; Task 1 (real) and Task 2 (imagined))
• both feet (in runs 5, 6, 9, 10, 13, and 14; for Task 3 (real) and Task 4 (imagined))

Note: The data files in this dataset were converted into the .set format for EEGLAB. The event codes in the .set files of this dataset will contain the concatenated event codes above for all event files for clarity purposes. The non-converted .edf files along with the accompanying PhysioBank-compatible annotation files for all the runs of each subject can be found in the sourcedata folder. In the non-converted .edf files the event codes will only be shown as T0, T1, and T2 regardless of task type. All the Matlab scripts used for the .set conversion and renaming of event codes of the PhysioBank .edf files can be found in the code folder.

Montage

The EEGs were recorded from 64 electrodes as per

Dataset Attribution and Description Dataset Title: EEG Motor Movement/Imagery Dataset Original Source: PhysioNet EEG Motor Movement/Imagery Dataset https://www.physionet.org/content/eegmmidb/1.0.0/ Original Authors: Gerwin Schalk, Dennis J. McFarland, Thilo Hinterberger, Niels Birbaumer, Jonathan R. Wolpaw

This dataset comprises over 1,500 one- and two-minute EEG recordings obtained from 109 volunteers. Subjects performed various motor and imagery tasks while 64-channel EEG was recorded using the BCI2000 system. Each subject participated in 14 experimental runs, including baseline and task-specific sessions.

The dataset is licensed under the Open Data Commons Attribution License v1.0 (ODC-By), which permits users to freely share, modify, and use the database, provided that appropriate credit is given to the original authors. physionet.org

As a student using this dataset for educational purposes, I would like to express my sincere gratitude to the original authors for their invaluable contribution to the field of brain-computer interfaces (BCI) research. In sharing this dataset on Kaggle, I aim to facilitate access for the research community and support further advancements in BCI research.

When using this resource, please cite the original publication: Schalk, G., McFarland, D.J., Hinterberger, T., Birbaumer, N., Wolpaw, J.R. BCI2000: A General-Purpose Brain-Computer Interface (BCI) System. IEEE Transactions on Biomedical Engineering 51(6):1034-1043, 2004.

Please include the standard citation for PhysioNet: Goldberger, A., Amaral, L., Glass, L., Hausdorff, J., Ivanov, P. C., Mark, R., ... & Stanley, H. E. (2000). PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals. Circulation [Online]. 101 (23), pp. e215–e220. RRID:SCR_007345.

This dataset was obtained from: https://physionet.org/content/eegmmidb/1.0.0/

Each Sx_Tasky_z.csv file corresponds to an EEG recording with 64 channels, lasting 4 seconds, sampled at a rate of 160 Hz. Sx represents the subject number x performing a task named y, and the final z indicates the repetition number.

It can be summarized as follows:

Sampling rate: 160 Hz Recording duration: 4 seconds Number channels: 64 Number subjects: 109

Each subject performs multiple times one of the following tasks for 4 seconds

Task0: Rest;             
Task1: Open and close the right fist; 
Task2: Open and close the left fist; 
Task3: Open and close both fists;   
Task4: Open and close both feet;   
Task5: Imagine doing Task1;     
Task6: Imagine doing Task2;      
Task7: Imagine doing Task3;      
Task8: Imagine doing Task4;

Dataset

This dataset was created by Hina Tariq

Released under MIT

Contents

Context

This dataset was created by the creators of the BCI2000 instrumentation system. It contains EEG recordings from 109 subjects performing various motor execution (moving) and motor imagery (imagining) tasks. It is one of the most widely used benchmarks for Brain-Computer Interface (BCI) research.

The original dataset is hosted on PhysioNet. This version is uploaded to Kaggle to facilitate easy experimentation and kernel creation.

Content

• Subjects: 109 volunteers.
• Channels: 64 EEG channels (International 10-10 system).
• Sampling Rate: 160 Hz.
• Format: EDF+ (European Data Format). https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F23408205%2F5fa88e8357e91d361ac3d63744faea55%2F64_channel_sharbrough.png?generation=1763816230126421&alt=media" alt="64 Channel EEG Montage"> ### Experimental Protocol (The "Runs") Each subject performed 14 experimental runs. To use this data correctly, you must map the Run Number to the specific task.

Understanding Annotations (Events)

The EDF+ files contain annotations (events) indicating when a specific action started. * T0: Rest / Baseline (Corresponds to code 1 in many loaders) * T1 & T2: The meaning of these codes changes depending on the Run.

File Naming Convention

The files are named S[subject_id]R[run_id].edf. * S001R01.edf: Subject 1, Run 1 (Baseline eyes open) * S050R04.edf: Subject 50, Run 4 (Motor Imagery: Left vs Right)

Usage Information

This dataset is perfect for: 1. BCI Classification: distinguishing between imagined movements (e.g., Left vs. Right). 2. Event-Related Desynchronization/Synchronization (ERD/ERS) analysis. 3. Deep Learning: Training CNNs (like EEGNet) or Transformers on raw EEG data.

Loading Data (Python/MNE)

This dataset is natively supported by the mne library. ```python import mne

To read a specific file

raw = mne.io.read_raw_edf("S001R04.edf", preload=True)

• This data set was taken from: https://physionet.org/content/eegmmidb/1.0.0/
• Sampling rate: 160 Hz
• Number channels (electrodes): 64 Channels (10-20 system)
• Number of subjects: 109
• Classes: Open eyes, Closed eyes
• Each CSV file contains an EEG recording of 61 seconds, both with eyes open or closed.