Classification and detection of EEG data for tired driving based on Mamba framework

Fatigue driving is an important factor that causes frequent traffic accidents. The fatigue driving pre-detection system is a research direction at present. Compared to the driving habits of different people, EEG signals can show physiological consistency, and it is a major challenge to mine the characteristics of driver fatigue from the multi-channel EEG data. In this paper, based on Mamba framework, CNN feature extraction module is selected to realize coupling, and a CNN-Mamba model is constructed and applied to EEG fatigue driving detection task. In this paper, we selected a fatigue driving dataset published in 2019, which contains 30 channels of EEG data from 11 subjects. The results show that the average accuracy of the model is 72.73% in the cross-subject EEG classification of 11 subjects, which is significantly improved compared with other advanced methods, and the recognition accuracy performance is improved by 9.65% compared with EEGNet.