On Error Classification from Physiological Signals within Airborne Environment

2025-04-17

AI Summary

Human error accounts for 70–80% of aviation safety incidents, necessitating robust real-time detection methods applicable in operational flight environments. Method: This study conducts the first systematic validation of physiological signals for real-time error detection under authentic flight conditions—including high-G maneuvers (up to 2G)—using synchronized multimodal acquisition (EEG, eye-tracking, ECG) and machine learning classifiers. Evaluation spans three distinct contexts: laboratory, level-flight, and high-G maneuvering. Contribution/Results: EEG achieves 87.83% classification accuracy—comparable to laboratory performance (89.23%) with no statistically significant degradation—demonstrating exceptional robustness. Eye-tracking attains 82.50% accuracy, whereas ECG yields 51.50%. These findings overcome a critical translational bottleneck in deploying neurophysiological feedback from controlled lab settings to dynamic, safety-critical flight operations. The work establishes the first empirical foundation and methodological framework for neurophysiologically informed, adaptive human–machine collaborative safety monitoring in aviation.

Technology Category

• Humans and AI: Human-in-the-loop Machine Learning
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior
• Natural Language Processing: Safety and Robustness

Application Category

• Security and Privacy: Large-scale security measurements
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI

Abstract

Human error remains a critical concern in aviation safety, contributing to 70-80% of accidents despite technological advancements. While physiological measures show promise for error detection in laboratory settings, their effectiveness in dynamic flight environments remains underexplored. Through live flight trials with nine commercial pilots, we investigated whether established error-detection approaches maintain accuracy during actual flight operations. Participants completed standardized multi-tasking scenarios across conditions ranging from laboratory settings to straight-and-level flight and 2G manoeuvres while we collected synchronized physiological data. Our findings demonstrate that EEG-based classification maintains high accuracy (87.83%) during complex flight manoeuvres, comparable to laboratory performance (89.23%). Eye-tracking showed moderate performance (82.50%), while ECG performed near chance level (51.50%). Classification accuracy remained stable across flight conditions, with minimal degradation during 2G manoeuvres. These results provide the first evidence that physiological error detection can translate effectively to operational aviation environments.

Problem

• Assess error detection accuracy from physiological signals in real flight conditions
• Compare EEG, eye-tracking, and ECG performance across lab and airborne environments
• Validate feasibility of physiological error classification during complex flight manoeuvres

Innovation

• EEG-based classification maintains high accuracy
• Eye-tracking shows moderate performance
• ECG performs near chance level