To address the issue of carrying rolling element bearing （REB） fault diagnosis where the conventional vibration sensor is difficult to install， an instantaneous angular speed （IAS） signal based REB fault diagnosis method by optimized AR model is pro? posed. The forward differential method is used to calculate and estimate the instantaneous angular speed signal. Then， the biased estimation autocorrelation analysis is used to determine the optimal order p by the maximum autocorrelation kurtosis. Periodic com? ponents in the IAS signal are removed by AR prediction， and the residual components containing rich bearing fault information are remained. The residual components are pre-whitened to equalize the importance of each band and to extract fault characteristics from the envelope. Simulation signal and outer ring data from a test rig validate the effectiveness of the proposed method. The ex? perimental comparative analysis results show that the calculation efficiency is improved significantly when compared to the existing method of fast spectral steepness combined with order analysis based on vibration signal.