According to the non-stationary and nonlinear characteristics of the spherical roller bearing fault vibration signals in wind turbine， a bearing fault feature extraction method of wind turbine based on intrinsic time-scale decomposition（ITD） is presented. The ITD method can decompose a complex signal into several proper rotation components and a trend component. It can also reveal the dynamic characteristics of non-stationary signals， has higher decomposition efficiency and frequency resolution. The diagnosis results show that the ITD method can effectively extract the bearing fault characteristics of wind turbine and can be applied to online fault diagnosis.