Research on Fatigue Driving Detection Technology Based on CA-ACGAN.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Traffic Safety

Background:

• Driver fatigue is a major cause of road accidents globally.
• Effective fatigue monitoring systems are crucial for enhancing road safety.

Purpose of the Study:

• To develop a precise method for identifying driver fatigue states using electroencephalography (EEG) signals.
• To introduce a novel deep learning framework, the Convolutional and Attention Generative Adversarial Network (CA-ACGAN), for fatigue detection.

Main Methods:

• Constructed a 4D feature data model for EEG signal analysis, considering frequency, spatial, and temporal dimensions.
• Developed the CA-ACGAN framework integrating attention mechanisms, bottleneck residual blocks, and Transformer elements for refined EEG signal processing.
• Utilized a conditional generative adversarial network with a classification head for effective fatigue state discrimination and synthetic data generation.

Main Results:

• The CA-ACGAN model demonstrated superior performance in fatigue detection on the SEED-VIG public dataset compared to existing methods.
• Empirical results showed the model's effectiveness in generating high-quality synthetic EEG data, outperforming other Generative Adversarial Network (GAN) models.
• The proposed model achieved high accuracy in identifying fatigue driving states.

Conclusions:

• The CA-ACGAN model is a valuable tool for fatigue driving identification.
• This research offers new insights into deep learning applications for time series data generation and processing in the context of driver safety.