As the number of accidents caused by driver fatigue driving increases, research on driving fatigue detection becomes more and more important. By studying the current situation and development trend of driving fatigue detection, this paper takes the driver's individual as the unit, extracts the features of the driver's eyes through the AdaBoost algorithm.And distinguishes whether the driver wears glasses through the vertical projection method, and then recognizes the eyes. Based on the feature-weighted Bayesian algorithm, the three parameters of eye fatigue parameters Perclos, blink frequency and closed eye duration are combined, and the driver's eye fatigue state is comprehensively judged. Simulation experiments are performed on the MATLAB platform. Analysis can be found that the characteristics of the fusion can better reflect whether fatigue occurs. Once tired, take a voice alarm prompt. Key words: Fatigue driving test, feature weighted Bayes, eye feature fusion