Discrimination of temporal fatigue-related states and inferred brain-function coupling in subway drivers under...

The mental fatigue state of subway drivers directly affects operational safety. At the same time, they are easily influenced by individual emotions and are closely related to the state of brain function.

In this study, multimodal data of subway drivers during driving were collected synchronously in a simulated driving environment, including EEG signals, subjective fatigue perception, emotional states and operational performance data. First, the Toeplitz inverse covariance clustering method was used to integrate the subjective fatigue perception, emotional states and operational performance data of subway drivers to derive four data-driven fatigue states. Subsequently, the EEG signals were decomposed into four frequency bands of θ/α/β/γ, and the time domain, frequency domain and spatial domain feature indicators were extracted as input data. A discrimination model for the four data-driven fatigue states based on a convolutional recurrent neural network was constructed. After multiple sets of feature set combinations and integrated strategy tests, the optimal fatigue levels discrimination model was finally selected. Further, Dynamic Bayesian Inference was used to explore the of inferred brain-functional coupling patterns of the EEG signals corresponding to each data-driven fatigue state.

The results showed that across the four data-driven fatigue states, the prefrontal cortex consistently occupied a central role in task planning and execution, while the other brain regions formed a dynamically coordinated network through bidirectional drive and feedback relations, suggesting systematic differences in inter-regional functional coupling patterns across the identified states.