An EEG-Based Pilots' Attention Detection Method

In complex flight environments, pilots are susceptible to external stimuli and internal fatigue, which can lead to excessive concentration or distraction. Such fluctuations in attention may heighten the risk of human error, potentially resulting in catastrophic incidents. Consequently, precise monitoring of pilot attention levels is crucial for mitigating fatigue-related risks. Research indicates a correlation between electroencephalography (EEG) and electrooculogram (EOG) signals and attention levels. Additionally, deep learning techniques have proven effective in extracting features from EEG data for predictive purposes. This study introduces the Multi-Scale Fusion EEGNet (MSF_EEGNet) model, which integrates channel attention modules to enhance predictive accuracy. Four subjects participated in three flight trial sets of varying complexity, during which the MSF_EEGNet model was employed to monitor attention levels based on collected EEG and EOG data. Comparative analyses were conducted against EEGNet, SE_EEGNet, and ECA_EEGNet models. The experimental results indicate that the MSF_EEGNet model effectively monitors attention levels, demonstrating its potential for improving pilot safety in dynamic operational contexts.