Abstract: Identification of driving state based on driving behavior is presented in this paper. The driving behavior data was collected from the driving simulation chamber under the experimental conditions. The operating characteristics in fatigue driving state were analyzed including the steering wheel and throttle amplitude changes. Then the duty ratio of no steering and the mean value of throttle amplitude were extracted as the discrimination index for fatigue state. The no steering time was optimized using variance analysis. Finally the Fisher linear discrimination algorithm was used to identify the driving state with high recognition accuracy.