moabb.datasets.BNCI2020_001#

class moabb.datasets.BNCI2020_001[source]#

BNCI 2020-001 Reach-and-Grasp Electrode Comparison dataset.

Dataset summary

#Subj

15

#Chan

varies (11-64)

#Classes

3

#Trials / class

80

Trials length

5 s

Freq

256 Hz

#Sessions

3

#Runs

4

Total_trials

7200

Participants

  • Population: healthy

  • Handedness: right-handed

Equipment

  • Amplifier: g.tec

  • Electrodes: Gel-based electrodes

  • Montage: 10-20

  • Reference: Car

Preprocessing

  • Data state: raw

  • Bandpass filter: 0.3-60 Hz

  • Steps: zero-phase 4th order Butterworth bandpass filter (0.3-60 Hz), extended infomax ICA for eye artifact removal (gel and water-based only), artifact rejection by amplitude threshold (>125 µV), artifact rejection by abnormal joint probability (4 SD threshold), artifact rejection by abnormal kurtosis (4 SD threshold)

  • Re-reference: Car

Data Access

Experimental Protocol

  • Paradigm: imagery

  • Feedback: visual (screen showing number of completed grasps)

  • Stimulus: avatar

Dataset from [1].

Dataset Description

This dataset contains EEG data from 45 subjects (15 per electrode type) performing natural reach-and-grasp movements with different electrode systems. Three electrode types were compared:

  • Gel-based electrodes (g.tec g.USBamp system): 58 EEG + 6 EOG channels

  • Water-based electrodes (BitBrain EEG-Versatile): 32 EEG + 6 EOG channels

  • Dry electrodes (BitBrain EEG-Hero): 11 EEG channels (no EOG)

The study investigates the feasibility of decoding natural reach-and-grasp movements from EEG signals recorded with different electrode technologies, including mobile systems suitable for real-world applications.

Participants

  • 45 healthy able-bodied subjects (15 per electrode type)

  • All subjects performed the same experimental protocol

  • Each subject used only one electrode type

  • Location: Graz University of Technology, Austria (in collaboration with BitBrain, Spain)

Recording Details

  • Sampling rate: 256 Hz (all systems)

  • Reference: Earlobe (right for gel, left for water/dry)

  • Ground: AFz (gel/water), left earlobe (dry)

  • Filters: 0.3-100 Hz bandpass (3rd-4th order Butterworth)

Experimental Procedure

  • Self-paced reaching and grasping actions toward objects on a table

  • Two grasp types: palmar grasp (empty jar) and lateral grasp (spoon in jar)

  • Rest condition: Quiet sitting with fixation

  • 80 trials per grasp type distributed across 4 runs

  • Window of interest: [-2, 3] seconds relative to movement onset

Event Codes

  • palmar_grasp: Movement onset for palmar grasp (reaching to empty jar)

  • lateral_grasp: Movement onset for lateral grasp (reaching to jar with spoon)

  • rest: Onset of rest period

Electrode Types

Subjects are grouped by electrode type (15 per type). The subject index maps to:

  • 1-15: Gel-based electrode recording

  • 16-30: Water-based electrode recording

  • 31-45: Dry electrode recording

Classification Results (from original paper)

Grand average peak accuracy on unseen test data:

  • Gel-based: 61.3% (8.6% STD)

  • Water-based: 62.3% (9.2% STD)

  • Dry electrodes: 56.4% (8.0% STD)

References

[1]

Schwarz, A., Escolano, C., Montesano, L., & Muller-Putz, G. R. (2020). Analyzing and Decoding Natural Reach-and-Grasp Actions Using Gel, Water and Dry EEG Systems. Frontiers in Neuroscience, 14, 849. https://doi.org/10.3389/fnins.2020.00849

Notes

Added in version 1.3.0.

This dataset is valuable for comparing electrode technologies in naturalistic movement paradigms. Data is available under CC BY 4.0 license.

On this page