moabb.datasets.BNCI2025_002#

class moabb.datasets.BNCI2025_002[source]#

BNCI 2025-002 Continuous 2D Trajectory Decoding dataset.

Dataset summary

#Subj

20

#Chan

64

#Classes

3

#Trials / class

varies

Trials length

8 s

Freq

200 Hz

#Sessions

3

#Runs

1

Total_trials

varies

Participants

  • Population: healthy

  • Age: 24 years

  • Handedness: right-handed (Edinburgh Handedness Inventory)

  • BCI experience: naive BCI users in terms of motor decoding (4 had previous EEG experience)

Equipment

  • Amplifier: BrainVision

  • Electrodes: active electrodes (actiCAP)

  • Montage: channels_fyrxyz

  • Reference: Car

Preprocessing

  • Data state: continuous signals with synchronized decoded control signals, paradigm targets, visual feedback trajectories, and error metrics

  • Bandpass filter: 0.18-3 Hz

  • Steps: resampling and alignment to EEG timeline, GND channel removal, channel location assignment

  • Re-reference: common average

Data Access

Experimental Protocol

  • Paradigm: imagery

  • Feedback: none

  • Stimulus: avatar

Dataset from [1].

Dataset Description

This dataset contains EEG recordings from participants performing a continuous 2D trajectory decoding task using attempted movement. The study investigates continuous decoding of hand movement trajectories from EEG signals, with participants tracking a moving target on screen while their dominant arm is strapped to restrict actual motor output (simulating attempted movement conditions similar to paralyzed individuals).

The experimental paradigm includes both calibration and online decoding phases, with varying levels of EEG feedback (0%, 50%, 100%) to evaluate the impact of feedback on decoding performance.

Note: Only 2 of the original 20 participants’ data is currently available on the BNCI server.

Participants

  • 10 able-bodied subjects (5 male, 5 female)

  • Mean age 24 +/- 5 years, all right-handed

  • 4 had prior EEG experience

  • Location: Institute of Neural Engineering, Graz University of Technology, Austria

Recording Details

  • Equipment: 64-channel actiCAP system (Brain Products GmbH)

  • Channels: 60 EEG + 4 EOG electrodes

  • Original sampling rate: 200 Hz

  • Electrode positions: 10-10 system with modifications (Fp1, Fp2, FT9, FT10 used as EOG; TP9, TP10 relocated to PPO1h, PPO2h)

  • Reference: Common average

  • Data synchronized using Lab Streaming Layer (LSL)

Experimental Procedure

Each session consists of:

  • Calibration phase: 2 eye runs (38 trials, 8s each) + 4 snake runs (48 trials, 23s each)

  • Online phase with 3 perception conditions: - perc0: No EEG feedback (baseline) - perc50: 50% EEG feedback - perc100: 100% EEG feedback

Trial types:

  • Snake runs: Tracking a moving white target with decorrelated x/y coordinates

  • Free runs: Tracing static shapes (diagonal/circle) at self-paced speed

Data Organization

  • 3 sessions per subject (recorded over 5 days)

  • 3 perception levels per session (perc0, perc50, perc100)

  • Files named: {subject_id}_ses{session}_perc{level}.mat

References

[1]

Kobler, R. J., Almeida, I., Sburlea, A. I., & Muller-Putz, G. R. (2022). Continuous 2D trajectory decoding from attempted movement: across-session performance in able-bodied and feasibility in a spinal cord injured participant. Journal of Neural Engineering, 19(3), 036005. https://doi.org/10.1088/1741-2552/ac689f

Notes

Added in version 1.3.0.

This dataset is designed for continuous decoding research, specifically for predicting 2D hand movement trajectories from EEG. Unlike classification-based motor imagery datasets, this dataset contains continuous trajectory labels suitable for regression-based decoders.

The paradigm “imagery” is used for compatibility with MOABB’s motor imagery processing pipelines, though the actual task involves attempted (rather than imagined) movements.

See also

BNCI2014_001

4-class motor imagery dataset

BNCI2014_004

2-class motor imagery dataset

On this page