To solve the problem that there exist no clear and prominent components related to rolling bearing rotational frequency in the time-frequency representation, the instantaneous fault characteristic frequency from the envelope TFR is used as the substitute of the traditional rotational frequency to fulfill the resampling process and then a fault diagnosis algorithm for the rolling bearing based on the fault characteristic order template is proposed to accomplish the final bearing diagnosis. In specific, the proposed algorithm can be summarized as four parts: (1) The spectral kurtosis based filtering, together with the short time Fourier transform (STFT) can lead to the envelope TFR, (2) a revised peak searching algorithm is then used to extract the IFCF trend from the envelope TFR, (3) the resampling step is then carried on using the IFCF trend and the fault characteristic order (FCO) spectrum in the fault phase angle domain is obtained, (4) The fault characteristic order template based on the fault characteristic coefficient (FCC) is finally constructed to recognize the operating state of the rolling bearing under time-varying rotational speed and identifying its corresponding fault type. The effectiveness of the proposed method has been validated by both simulated and experimental bearing vibration signals.