EUVIMEDEUVIMED entdecken
Crossref

An EEG-based motor imagery intention decoding framework for smart vehicle window control

Yuchen Guo, Tianyao Zhang, Jiaxue Cai, Liyun Deng, Zhenhai Gao

Biomedical Signal Processing and Control · 2027-02 · 130 · 111624

DOI: 10.1016/j.bspc.2026.111624 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Biomedical Signal Processing and Control, 2027-02. Verfasst von Yuchen Guo und 4 weiteren Autor:innen. Die Arbeit nennt 56 Literaturangaben. Gefördert von: Jilin Provincial Science and Technology Department.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Non-invasive electroencephalography (EEG)-based motor imagery (MI) decoding remains a challenging brain-computer interface (BCI) problem due to the low signal-to-noise ratio, high-dimensional spatiotemporal dynamics, inter-trial variability, and non-Euclidean topology of EEG signals. This study proposes Brain2AutoWin (B2AW), an EEG-based MI intention decoding paradigm that employs vehicle window operation as a low-risk, semantically meaningful, and quantifiable testbed for neural intention recognition in intelligent cockpit scenarios. A 32-channel EEG dataset was collected from 20 healthy participants in a high-fidelity driving simulator while they performed five MI tasks corresponding to left-window upward imagery, left-window downward imagery, right-window upward imagery, right-window downward imagery, and neutral state. To achieve fine-grained MI intention decoding, we developed B2AWNet, a topology-aware deep learning framework integrating parallel spatial–temporal representation learning, electrode-topology graph convolution, and lightweight Poolformer-based global feature mixing. The proposed model achieved an average accuracy of 95.05 % ± 3.01 % in the five-class MI-EEG classification task under subject-dependent offline evaluation and consistently outperformed representative EEG decoding baselines in comparative and ablation experiments. Moreover, leave-one-subject-out evaluation demonstrated the capability of B2AWNet to extract discriminative representations across unseen participants. Beyond classification performance, scalp topographic analysis and cortical source localization were conducted to examine neurophysiological plausibility. The results revealed task-related activation patterns involving motor, premotor, parietal, prefrontal, and temporal regions, with consistent lateralized responses for left- and right-side imagery conditions. These findings demonstrate the effectiveness of B2AWNet for offline MI-EEG intention decoding within a vehicle-oriented experimental paradigm and provide a validated signal-processing framework for future EEG-based implicit human–vehicle interaction research.

Förderung

  • Jilin Provincial Science and Technology Department

Zitieren

APA

Guo, Y., Zhang, T., Cai, J., Deng, L., & Gao, Z. (2027). An EEG-based motor imagery intention decoding framework for smart vehicle window control. Biomedical Signal Processing and Control, 130, 111624. https://doi.org/10.1016/j.bspc.2026.111624

Vancouver

Guo Y, Zhang T, Cai J, Deng L, Gao Z. An EEG-based motor imagery intention decoding framework for smart vehicle window control. Biomedical Signal Processing and Control. 2027;130:111624. doi:10.1016/j.bspc.2026.111624

BibTeX
@article{Guo2027,
  author = {Guo, Yuchen and Zhang, Tianyao and Cai, Jiaxue and Deng, Liyun and Gao, Zhenhai},
  title = {An EEG-based motor imagery intention decoding framework for smart vehicle window control},
  journal = {Biomedical Signal Processing and Control},
  year = {2027},
  volume = {130},
  pages = {111624},
  doi = {10.1016/j.bspc.2026.111624}
}
Kostenlos starten →