In　order　to　acquire　an　objective　and　accurate　driving　fatigue　threshold,　electroencephalography　(EEG)　signals　of　drivers　were　collected　from　driving　simulator,　and　the　time-domain　characteristics　of　EGG　signals　of　drivers　in　sober　and　mental　fatigue　states　were　comparatively　analyzed.　Considering　the　different　complexity　of　EEG　signals　in　sober　and　fatigue　states,　the　sample　entropy　of　EEG　signals　were　calculated　to　characterize　the　complexity　of　signals,　and　used　as　the　index　for　identifying　driving　fatigue.　Based　on　the　obtained　EGG　sample　entropy,　the　receiver　operating　characteristic　(ROC)　curve　analysis　was　introduced　to　obtain　the　discriminating　threshold　of　driving　fatigue.　The　results　indicate　that　when　the　EEG　sample　entropy　value　is　between　(0.32,　0.71),　the　driver　is　in　the　transitional　period　of　fatigue,　may　be　in　a　fatigue　state;　the　sample　entropy　of　less　than　0.605　can　be　identified　as　the　threshold　of　driving　fatigue,　and　the　accuracy　is　0.95.