Toward passive BCI: asynchronous decoding of neural responses to direction- and angle-specific perturbations during a simulated cockpit scenario

Jalilpour, Shayan; Müller-Putz, Gernot · 2022 · Crossref

DOI: 10.1038/s41598-022-10906-5

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study investigates the feasibility of using passive brain-computer interfaces (pBCIs) to detect and classify balance perturbations in real-time, aiming to enhance aviation and driving safety. Sudden loss of balance, caused by factors like turbulence or mechanical faults, can lead to catastrophic events if pilots cannot react quickly enough. The authors propose a system that decodes Perturbation Evoked Potentials (PEPs)—specific neural responses to balance loss—from ongoing electroencephalography (EEG) signals. By identifying these neural signatures asynchronously, a pBCI could assist in stabilizing vehicles before the pilot consciously reacts, thereby improving human-machine interaction. The experimental design involved fifteen healthy participants seated in a simulated glider cockpit. A robot induced postural changes by tilting the glider left or right at angles of 5° and 10°, embedded within an oddball paradigm where frequent small movements (1.5°) masked the timing of larger perturbations. EEG data were recorded using a 63-channel system at 512 Hz, synchronized with acceleration data from a Myo armband to mark perturbation onsets. The researchers employed a hierarchical classification approach: a binary classifier first detected the presence of a perturbation versus rest, followed by a multiclass classifier to distinguish the specific direction and angle. Features were extracted using Bilinear Common Spatial Pattern (BCSP) for binary classification and Fisher algorithm for multiclass classification, with Radial Basis Function Support Vector Machines (RBF-SVM) serving as the classifier. The model was trained on calibration data and validated in an asynchronous, online simulation mode. The results demonstrated high performance in detecting balance disturbances. The binary classifier achieved an average accuracy of 89.83% and an F1 score of 0.93, indicating robust detection of perturbations amidst ongoing EEG activity. The multiclass classifier, which distinguished between four conditions (left/right, 5°/10°), achieved an average accuracy of 73.64% and an F1 score of 0.60. Specifically, direction classification yielded an average accuracy of 86.09%, while angle discrimination reached 79.75% (73.5% for 5° and 86% for 10°). The average detection time for perturbations was 800 ± 200 ms after onset. Electrophysiological analysis revealed that while early PEP components (N1, P2) were similar across conditions, significant differences emerged in later time windows (320–620 ms), particularly in central and frontocentral regions, allowing for the discrimination of direction and angle. These findings confirm the practicality of pBCIs for monitoring balance disturbances in realistic, dynamic scenarios. The ability to asynchronously decode the existence and specific expression of perturbations from EEG signals suggests that such systems could be integrated into aviation or driving assistant technologies. By providing rapid, automated responses to balance loss, pBCIs could mitigate risks associated with human reaction delays, offering a promising avenue for enhancing safety in high-stakes operational environments.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-nvidia on 2026-08-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• neuro workload indices