On Fatigue Detection for Air Traffic Controllers Based on Fuzzy Fusion of Multiple Features

Area of Science:

• Aviation Safety
• Human Factors Engineering
• Machine Learning

Background:

• Fatigue detection for air traffic controllers is crucial for aviation safety.
• Existing methods primarily rely on facial features, limiting detection accuracy.
• A multifeature approach is needed to enhance fatigue monitoring.

Purpose of the Study:

• To propose and evaluate a novel ensemble learning model for air traffic controller fatigue detection.
• To integrate facial and voice features for improved accuracy.
• To introduce the Facial and Voice Stacking (FV-Stacking) method.

Main Methods:

• Facial features extracted using OpenCV and Dlib, analyzed with M-CNN and E-CNN for eye and mouth states (PERCLOS, FOM).
• Voice features extracted using Mel-Frequency Cepstral Coefficients (MFCC).
• Ensemble learning via a stacking model to fuse facial and voice features.

Main Results:

• The FV-Stacking model achieved a detection accuracy of 97%.
• This surpasses single-model accuracy (92%) and state-of-the-art methods (88%).
• Experiments conducted on 14 air traffic controllers.

Conclusions:

• The proposed FV-Stacking method effectively enhances fatigue detection accuracy for air traffic controllers.
• Multifeature fusion significantly outperforms single-feature approaches.
• This method offers a promising advancement in aviation safety research.