An EEG-based Brain Cognitive Dynamic Recognition Network for representations of brain fatigue

Fatigue driving will seriously threaten public safety and health, so monitoring the brain’s cognitive state accurately and exploring the fatigue process is essential. This paper proposes a 5-D Brain Cognitive Dynamic Recognition Network (E-5-D-BCDRNet) that synchronizes the brain’s cognitive changes with global features, local features, and sequence changes. The global feature refers to the 3-D brain global cognitive power maps (3-D-BGCM) constructed based on the new fatigue cognitive indicators. Local features refer to the 1-D brain local cognitive array (1-D-BLCA) created based on fatigue recognition contribution of different rhythms in each brain region. This paper uses a non-invasive wearable device to collect electroencephalogram (EEG) data from 14 subjects to evaluate the model’s performance. Compared with the existing advanced methods, the proposed method achieves the best detection effect (88.5%). In addition, this study finds that the features learned based on different indicators are different and distributed in various brain regions by presenting the learned characteristics. Moreover, by comparing the detection results based on the local features of a single rhythm, this study finds that the frontal and occipital regions of the driver are significantly different before and after. More importantly, only a single region can also achieve significant detection effects, which shows that our method can achieve satisfactory results based on small data. In a word, the research in this paper can lay a theoretical foundation for real-time brain fatigue detection and promote the development of intelligent transportation systems.