A Real-Time FBCSP–sLDA Motor-Imagery Brain–Computer Interface for Safety-Oriented Wheelchair Control
DOI:
https://doi.org/10.46522/S.2026.01.30Kulcsszavak:
FBCSP–sLDA system, wheelchair control, motor imagery, electroencephalography, brain–computer interfaceAbsztrakt
This paper presents a real-time EEG motor-imagery brain–computer interface designed for safety-oriented wheelchair control in assistive mobility applications and as a model of embodied human–machine interaction relevant to artificial intelligence, robotics, and technologically mediated performance. The architecture combines Filter-Bank Common Spatial Patterns feature extraction with shrinkage Linear Discriminant Analysis to achieve low-latency decoding, robust multiclass discrimination, and real-time deployment compatibility. EEG signals from a 16-channel wireless platform were processed through a multiband Mu beta pipeline targeting "Left-hand MI", "Right-hand MI", and "Both-feet MI" tasks. Hyperparameter optimization via stratified cross-validation showed that the optimal unpruned configuration achieved a classification accuracy of 78.11%. Feet motor imagery provided the highest-class separability, supporting its assignment to safety-critical stop/forward commands. To minimize hazardous command generation during motion, the decoding pipeline was integrated with a context-aware finite-state mechanism that dynamically gates inadmissible commands. Additional confidence-thresholding, temporal voting, and refractory strategies significantly improved runtime reliability, providing an efficient, deployable control system for assistive neurotechnology. Beyond assistive mobility, such brain–computer interfaces may support future forms of accessible performance and interactive art, enabling performers or spectators with severe motor disabilities to control robotic devices, stage elements, sound, lighting, or virtual environments through motor imagery. The proposed approach therefore points toward broader applications in inclusive human–machine performance and accessible creative participation.
Hivatkozások
AN, Jian, ZHOU, Jun, CHEN, Ling and XU, Guanghua, 2025. Safety-aware control architectures for neurointerface-based assistive mobility platforms. Journal of Neural Engineering and Rehabilitation, vol. 22, nr. 1, p. 45-60.
ANG, Kai Keng, ZHENG, Zhang Yang, PHUA, Kok Soon and GUAN, Cuntai, 2008. Filter Bank Common Spatial Pattern (FBCSP) in Brain-Computer Interface. IEEE International Joint Conference on Neural Networks (IEEE World Congress on Computational Intelligence), p. 2390-2397.
BLANKERTZ, Benjamin, LEMEW, Steven, TREDER, Matthias S., HAUFE, Stefan, SANNECK, Henning, MULLER, Klaus-Robert and CURIO, Gabriel, 2011. The Berlin Brain-Computer Interface: Non-medical uses of BCI technology. Frontiers in Neuroscience, vol. 5, p. 1-17.
LEDOIT, Olivier and WOLF, Michael, 2004. A well-conditioned estimator for large-dimensional covariance matrices. Journal of Multivariate Analysis, vol. 88, nr. 2, p. 365-411.
PFURTSCHELLER, Gert and LOPES DA SILVA, Fernando H., 1999. EEG rhythms and the motor imagery system. Clinical Neurophysiology, vol. 110, nr. 11, p. 1842-1857.
RAMOSER, Herbert, MULLER-GERKING, Johannes and PFURTSCHELLER, Gert, 2000. Optimal spatial filtering of single trial EEG during imagined hand movement. IEEE Transactions on Rehabilitation Engineering, vol. 8, nr. 4, p. 441-446.
----------------------- L R F L 139 29 4 R 41 98 5 F 14 11 134
Downloads
Megjelent
Hogyan kell idézni
Folyóirat szám
Rovat
License

This work is licensed under a Creative Commons Attribution 4.0 International License.
A CC-BY engedélyezi a sokszorosítást, terjesztést, archiválást és idézést, feltéve, hogy a szerzőket feltüntetik. A cikk teljes idézése szükséges, amely az ezen az oldalon található Idézési példák között található. Ha lehetséges, kérjük, adja meg a linket erre a honlapra. A szerzői jog a szerzőket illeti.

