基于通道加权的多模态特征融合用于EEG疲劳驾驶检测

程文鑫,闫光辉,常文文,吴佰靖,黄亚宁

Channel-weighted multimodal feature fusion for EEG-based fatigue driving detection

Wenxin CHENG,Guanghui YAN,Wenwen CHANG,Baijing WU,Yaning HUANG

表 4 单被试数据的疲劳驾驶检测准确率对比结果

Tab.4 Comparison of fatigue driving detection accuracy for single-subject data

D1 D2 被试 PCNN GRU PCNN-GRU 被试 PCNN GRU PCNN-GRU 1 0.956 7 0.917 9 0.951 8 1 0.862 5 0.852 5 0.885 8 2 0.947 7 0.932 2 0.971 4 2 0.900 6 0.910 7 0.919 0 3 0.977 4 0.978 9 0.963 9 3 0.880 4 0.899 2 0.908 6 4 0.783 1 0.797 4 0.908 9 4 0.872 5 0.888 1 0.890 2 5 0.865 2 0.882 5 0.912 7 5 0.920 1 0.910 9 0.922 6 6 0.819 3 0.790 7 0.874 3 6 0.862 2 0.854 4 0.865 6 7 0.893 1 0.856 9 0.941 3 7 0.901 4 0.900 9 0.919 4 8 0.910 8 0.923 2 0.942 8 8 0.833 1 0.829 9 0.855 9 9 0.939 8 0.956 3 0.970 6 9 0.921 8 0.919 1 0.930 5 10 0.982 7 0.985 7 0.970 6 10 0.882 5 0.878 9 0.870 6 11 0.878 8 0.881 8 0.936 8 11 0.939 8 0.946 0 0.947 3 12 0.825 0 0.820 0 0.872 0 12 0.860 1 0.855 5 0.866 3 13 0.896 1 0.914 9 0.923 2 13 0.921 7 0.933 9 0.948 8 14 0.920 9 0.925 5 0.931 5 14 0.844 8 0.858 5 0.868 7 15 0.980 0 0.970 0 0.990 0 15 0.931 5 0.933 1 0.945 4 16 0.839 9 0.838 1 0.875 0 16 0.939 9 0.928 7 0.932 6 17 0.817 4 0.832 8 0.908 9 17 0.845 8 0.813 9 0.881 1 18 0.874 3 0.917 9 0.929 2 18 0.877 2 0.892 5 0.904 8 19 0.978 2 0.985 7 0.987 9 19 0.910 5 0.909 1 0.914 5 20 0.951 1 0.930 7 0.869 7 20 0.862 2 0.844 2 0.874 8 21 0.959 0 0.941 3 0.988 0 21 0.939 3 0.940 1 0.949 6 22 0.991 7 0.991 0 0.988 7 22 0.909 9 0.912 9 0.937 8 23 0.952 6 0.939 8 0.866 0 23 0.948 8 0.932 4 0.956 6 平均准确率 0.910 5 0.909 2 0.933 7 24 0.890 1 0.909 3 0.898 6 标准差 0.061 3 0.059 8 0.041 1 25 0.878 7 0.856 7 0.904 4 — — — — 26 0.889 4 0.848 4 0.903 1 — — — — 平均准确率 0.893 3 0.890 8 0.907 8 — — — — 标准差 0.032 4 0.036 8 0.029 9