Published March 11, 2026 | Version v1

Tavakolan2017 — Three-class upper extremity motor imagery EEG dataset (re-hosted from Dryad)

  • 1. Simon Fraser University

Description

Re-hosted EEG data from Tavakolan et al. (2017), originally deposited on Dryad (DOI: 10.5061/dryad.6qs86) under a CC0 public domain dedication. Repackaged into per-subject ZIP archives for easier programmatic access.

Paper

M. Tavakolan, Z. Frehlick, X. Yong, and C. Menon, “Classifying three imaginary states of the same upper extremity using time-domain features,” PLoS ONE, vol. 12, no. 3, e0174161, 2017.
DOI: 10.1371/journal.pone.0174161

Paradigm

Three-class motor imagery of the same upper extremity (right arm):

ClassStimulusCodeDescription
REST1Relax without any movement imagery
MI-GRASP2Imagine opening and closing all fingers to grab an object
MI-ELBOW3Imagine moving the forearm up and down (elbow flexion/extension)

A fourth class (Reach-Hold the Glass, StimulusCode 4) was also recorded but excluded from the paper’s analysis. It is included in the raw data.

Each trial: 3 s visual cue (imagery period), followed by a 4–6 s rest interval. 20 trials per class per session.

Participants

  • 12 healthy right-handed subjects
  • 4 sessions per subject on separate days
  • Total: 48 recordings (12 subjects × 4 sessions)

Recording setup

ParameterValue
SystemEGI Geodesic Net Amps 400 series
Cap32-channel GSN-HydroCel Geodesic Sensor Net
Channels32 EEG (E1–E32), Cz as online reference
Sampling rate1000 Hz
Bandpass0.1–100 Hz
Impedance< 50 kΩ
Line frequency60 Hz (Canada)
File formatBCI2000 (.DAT)

File structure

12 per-subject ZIP files (P01.zip–P12.zip). Each contains 4 session ZIPs, each holding one BCI2000 .DAT file:

P01.zip
  P01_Se01.zip  →  BCI2000 .DAT (session 1)
  P01_Se02.zip  →  BCI2000 .DAT (session 2)
  P01_Se03.zip  →  BCI2000 .DAT (session 3)
  P01_Se04.zip  →  BCI2000 .DAT (session 4)
...
P12.zip

The BCI2000 .DAT files contain 280 source channels; only the first 32 are EEG. Channel gain: 0.0238419 μV per raw ADC unit. Events are in the StimulusCode state variable.

Reading the data

pip install BCI2kReader
from moabb.datasets import Tavakolan2017
from moabb.paradigms import MotorImagery

dataset = Tavakolan2017()
paradigm = MotorImagery(
    events=["rest", "right_hand", "right_elbow_flexion"],
    n_classes=3,
)
X, y, metadata = paradigm.get_data(dataset=dataset)

Re-hosting rationale

The original Dryad deposit requires OAuth API credentials for programmatic download, which creates a barrier for automated benchmarking pipelines. This Zenodo re-host provides direct URL access. The data is identical to the Dryad deposit — no modifications were made.

License

CC0 1.0 Universal (Public Domain Dedication), as per the original Dryad deposit.

Notes

This is a re-hosted copy of the dataset originally deposited on Dryad (DOI: 10.5061/dryad.6qs86) by Tavakolan et al. The data is unchanged; only the archive format has been reorganized into per-subject ZIPs for integration with the MOABB benchmarking framework.

Files

P01.zip

Files (5.2 GB)

Name Size
md5:acdad83ae831bef1e3801c6a8a770722
430.9 MB Preview Download
md5:4654aedbed3fbe692714f4afd74e87d6
437.3 MB Preview Download
md5:67f8235cb044e1018ada99fd31a74e0e
431.6 MB Preview Download
md5:fdbac11e3e4086babd7564100e6b1895
436.4 MB Preview Download
md5:8203f4dec361d1ece13a72d867f7604b
424.4 MB Preview Download
md5:6f7ef375cc1e7061846bac90a5bf8b67
427.6 MB Preview Download
md5:3c8084dcda1ae9841a9cd9f636c3d71e
435.6 MB Preview Download
md5:b5e6b173a5ad5150e0c10d1f7b684957
435.0 MB Preview Download
md5:0898d24116735130d340cc28415f2f87
431.3 MB Preview Download
md5:7f881ea973671a1fca15dd83f2b68d9c
429.9 MB Preview Download
md5:3d6f99f9fc49ba90e3eff9f0858629e7
432.7 MB Preview Download
md5:fcfa225b965f35766151ba0dbe0cef07
438.1 MB Preview Download

Additional details

Related works

Is derived from
10.5061/dryad.6qs86 (DOI)
Is part of
https://github.com/NeuroTechX/moabb (URL)
Is supplement to
10.1371/journal.pone.0174161 (DOI)