It designed a driver fatigue detection system using the eye's open and close status for decision. It firstly uses an adaptive skin color region for face segmentation. After that, the face region gradient is calculated and adaptively thresholded. It uses a color HOG algorithm to pick out the eye region features. All eye feature vectors are pushed into the linear SVM classifier for training. Two stages classifiers are designed. The first SVM classifier is used to detect the location of eyes, and the second one is used to decide whether the eye is closed or open. In a period of time, if the driver is detected a high eye close rate or a long time without the eye closed, the system will make an alarm.