The　characteristic　of　EEG　signal　in　drowsy　driving　was　researched.　A　method　based　on　power　spectrum　analysis　and　FastICA　algorithm　was　proposed　to　determining　the　fatigue　degree.　In　a　driving　simulation　system,　the　EEG　signals　of　subjects　were　captured　by　instrument　NT-9200　in　two　states,　one　state　was　sober,　and　the　other　was　drowsy.　The　multi　channel　signals　were　analyzed　with　FastICA　algorithm,　to　remove　ocular　electric,　myoelectric　and　power　frequency　interferences.　Power　spectral　densities　were　calculated　after　FFT,　and　the　fatigue　index　F　was　gotten　finally.　Experimental　results　show　that　the　method　presented　in　this　paper　can　be　used　to　determine　the　drowsiness　degree　of　EEG　signal　effectually.　©2010　IEEE.