EEG-based analysis for pilots' at-risk cognitive competency identification using RF-CNN algorithm.

Area of Science:

• Cognitive Science
• Neuroscience
• Marine Engineering

Background:

• Current ship pilot screening systems lack focus on cognitive competency.
• Situation awareness (SA) is crucial for piloting performance and is linked to unsafe behaviors.
• Low SA levels can negatively impact ship piloting operations.

Purpose of the Study:

• To develop an identification model for detecting at-risk cognitive competency (low SA) in ship pilots.
• To utilize wearable electroencephalogram (EEG) signal acquisition technology for cognitive assessment.
• To enhance the existing ship pilot screening system with a focus on cognitive evaluation.

Main Methods:

• Developed a Random Forest-Convolutional Neural Network (RF-CNN) model for SA level detection.
• Collected EEG signals from pilots in poor visibility conditions.
• Extracted 12 correlation features from frontal and central EEG regions using a 5s sliding window, followed by PCA-based RF feature combination for CNN input.

Main Results:

• The proposed RF-CNN model achieved an accuracy of 84.8% in identifying at-risk cognitive competency.
• This accuracy represents a 6.7% improvement over individual RF (78.1%) and CNN (81.6%) models.
• Significant correlations were found between SA levels and specific EEG frequency metrics (α/β, θ/(α+θ), (α+θ)/β) in frontal and central regions.

Conclusions:

• The RF-CNN model effectively identifies at-risk cognitive competency in ship pilots using EEG data.
• This technology provides crucial support for developing intelligent, adaptive pilot cognitive evaluation systems.
• The study lays a foundation for monitoring cognitive processes in ship piloting operations.