Abstract
Rapid evolution of intelligent flight cockpits necessitates real-time monitoring of pilots’ cognitive load to ensure safety. This study introduces CLNet, a lightweight neural network for accurate classification of cognitive load states. Utilizing multi-scale spatiotemporal convolution, CLNet enhances feature extraction from physiological signals, significantly boosting accuracy. Ablation studies highlight the roles of the squeeze-and-excitation and temporal gate convolution modules. To evaluate the effectiveness of CLNet, we developed the Airfield Traffic Pattern Cognitive Load (ATPCL) dataset, including electroencephalogram (EEG), electrocardiogram (ECG), and electromyography (EMG) signals recorded during key flight phases. On the ATPCL dataset, CLNet achieved an accuracy of 95.1%. We also built an integrated online monitoring system for real-time data collection, processing, and visualization. This system employs the CLNet for rapid cognitive load assessment and updating within milliseconds. Our system provides a valuable tool for real-time pilot cognitive load evaluation, supporting advancements in aviation research and applications.
Authors
D., & L.Y.L.K.Z.J.W.S.&.W.
https://doi.org/10.1080/10447318.2025.2541298