In order to improve the fault diagnosis rate of rolling bearings, a fault diagnosis method based on the combination of high-order spectrum(HOS) and Tamura texture features is proposed. First, the impact of the fault vibration signal of the rolling bearing is extracted by the high-order spectrum method. Then, the high-order spectrum is processed to obtain a two-dimensional contour map. According to the characteristics that the contour maps are similar when the bearing faults are the same and different at the same time, the Tamura texture description method based on human visual perception is used to extract features. The parameters are then input into a multi-class support vector machine(SVM) for classification. The results show that the high-order spectrum combined with the Tamura texture feature of the rolling bearing fault diagnosis method can achieve high fault identification accuracy with fewer characteristic parameters, and the identification accuracy of mixed vibration signals with different fault sizes is stable and the diagnosis effect is good.