Optimization On A Portable Fatigue Driving Detection Method Based On The Fusion Of Eeg And Eog

Fatigue driving accidents caused by drivers pose a serious threat to people’s property and life safety,and more and more researchers are investing in the field of fatigue driving research.In traditional EEG data collection experiments,drivers wearing EEG caps for a long time can cause discomfort and affect normal driving activities.Therefore,designing a portable EEG collection device with a more flexible and comfortable wearing method has important practical value.EEG signals are often regarded as the "gold standard" for fatigue detection,but due to the simplification of portable devices,the amount of EEG data collected is limited,and new features need to be introduced to improve the accuracy of fatigue driving detection.Research has shown that EOG signals are closely related to driver fatigue.Therefore,to improve the discomfort caused by traditional EEG cap-wearing methods and improve the accuracy of fatigue-driving detection,this paper proposes a portable fatigue-driving detection method based on the fusion of EEG and EOG signals.This article is based on the SEED-VIG public dataset from Shanghai Jiao Tong University.Considering the temporal characteristics of EEG signals and the difficulty in finding suitable parameter combinations using neural network algorithms,a novel IPSO-LSTM-Attention network model is proposed for the classification of EEG signals into three categories:alertness,fatigue,and drowsiness in different brain regions.The results show that the proposed model achieves a classification accuracy of 0.9698 and a mean squared error of 0.0326,which outperforms previous methods such as support vector machines(SVM),grid search optimized SVM(GS-SVM),and LSTM-Attention with an attention mechanism.Although the fatigue detection accuracy of different brain regions is lower than that of the whole brain EEG,the decrease is within 0.038,which preliminarily verifies the feasibility of a sparse EEG detection scheme.To further improve fatigue detection accuracy using fewer EEG channels,this paper proposes a fusion detection method of EEG and EOG signals.A maximum relevance and minimum redundancy feature fusion algorithm compares the detection accuracy of full-brain and different brain regions’ EEG and frontal EOG fusion.The results show that the proposed IPSO-LSTM-Attention model achieves the best fatigue detection performance.Specifically,the fusion of full-brain EEG and frontal EOG achieves the highest accuracy of 0.9742 and the lowest mean squared error of only 0.0258,the results show that the fusion algorithm is effective.In the case of EEG and frontal EOG fusion in different brain regions,the frontal EEG and frontal EOG fusion results are only slightly worse than the full-brain EEG and frontal EOG fusion results,with a decrease in accuracy of only 0.0051 and an increase in the mean squared error of only 0.0055,but with significantly fewer channels.The frontal EEG and frontal EOG fusion only requires four electrode channels,which can collect both EEG and EOG signals,while the full-brain EEG and frontal EOG fusion requires separate consideration of the EEG channels for the entire brain and the EOG channels,making the frontal EEG and frontal EOG fusion the optimal choice.Compared with existing fatigue detection algorithms based on the same dataset,the proposed IPSO-LSTM-Attention algorithm based on the frontal EEG and frontal EOG fusion achieves high detection accuracy while implementing portable and low-channel detection.In summary,the proposed low-channel solution and classification detection algorithm are effective,feasible,and have practical application value.