Application of Ensemble Learning in EEG Signal Analysis of Fatigue Driving. Issue 4 (February 2021)

• Record Type: Journal Article
• Authors: Rao, Songhui; Li, Kairui; Wu, Jun; Mu, Zhendong
• Is Part Of: Journal of physics. Volume 1744:Issue 4(2021)
• Publication Date: 2021-02
• Subjects: Fatigue Detection -- EEG Signal -- Ensemble Learning -- Functional Brain Network; Physics -- Congresses; 530.5
• DOI: 10.1088/1742-6596/1744/4/042193
• Languages: English
• ISSNs: 1742-6588

Abstract

Abstract: Fatigue detection has now become an important direction of fatigue driving research. As a reliable physiological signal, EEG signals can be used to detect fatigue driving. This passage uses the idea of ensemble learning to establish an ensemble learning classification model of Bagged Tree, RSM Discrimination and RUSBoosted Tree, and uses the fatigue driving experiment data as the object to divide the fatigue driving object into normal state and fatigue state, using Pearson Correlation Coefficient method build a functional brain network. Six methods of Betweenness Centrality, Edge Betweenness Centrality, Clustering Coefficient, Degree, Local Efficiency, and Rich Club Coefficient are used to process and analyze six kinds of EEG signal characteristics, and then bring them into different in the ensemble learning model, the EEG signal characteristics and the ensemble learning model are matched and selected to improve the accuracy and stability of the detection model. By simulating 24 subjects, the results show that the accuracy and robustness of RSM Discrimination ensemble learning model and Degree method are the best, with an average accuracy rate of 90.35% and a standard deviation of 0.081.