Cognitive State Monitoring in Aviation Scenarios: A Multimodal Physiological Fusion Approach Based on Resting-State Embedding

Cross-scenario transfer monitoring of pilots’ cognitive states is a key challenge in aviation human factors, focusing on aligning resting-state brain features with dynamic flight cognitive representations. This study innovatively proposes a transfer learning framework via task-free baseline cognitive embedding, using resting-state EEG (Electroencephalogram) to construct a universal feature space for high-precision classification of distraction, concentration, surprise, and neutral states in LOFT (Line Oriented Flight Training) simulations. Methodologically, it employs CEBRA (Consistent Embeddings of Brains and Behaviors) for transferable representations and a multi-modal physiological fusion model for dynamic coupling analysis. Results show superior classification performance over single-scenario training, verifying its robustness against individual baselines and providing a theoretical breakthrough for generalized aviation cognitive monitoring.