Driving　fatigue　has　been　considered　as　a　significant　risk　factor　in　transportation　accidents,　and　the　development　of　the　human　cognitive　state　based　on　electroencephalogram　(EEG)　has　become　a　major　focus　in　the　field　of　driving　safety.　However,　it　faces　portable　and　real-time　problems　on　its　practical　application.　This　study　uses　MindWave　to　collect　the　Attention　and　Meditation　EEG　from　the　left　prefrontal　lobe　of　the　subject,　and　uses　the　relation　between　Attention　and　Meditation　EEG　when　the　subject　is　in　the　state　of　concentration,　relaxation,　fatigue　and　sleep　being　measured　first.　As　a　result,　a　new　method　for　driving　fatigue　detection　based　on　the　correlation　coefficient　between　drivers　Attention　and　Meditation　EEG　is　proposed.　Meanwhile,　the　k-Nearest　Neighbors　(k-NN)　algorithm　is　introduced　to　classify　the　correlation　coefficient　between　the　drivers　Attention　and　Meditation　EEG,　so　as　to　detect　driving　fatigue　and　alert.　Lastly,　the　software　running　on　an　Android　smart　device　is　developed　based　on　the　above　technologies,　and　the　experiment　proves　that　it　has　noninvasive　and　real-time　advantages,　while　its　sensitivity　and　specificity　are　68.31%　and　90.43%　respectively.　©　2014　IEEE.