Abstract: Acoustic Emission technology was used to diagnose the typical defect of rolling bearing and identify the size of fault.A new processing way of so-called time domain plus frequency domain was proposed for the complicated AE signals,which were processed by combining short-time absolute average and autocorrelation function to extract bearing characteristic frequency simply and effectively.In addition,through researching on the AE signal of bearing rollers in some particular experiments and analyzing the character of the rolling entry into and depart from the fault area,then the time difference between entry and depart can be calculated and the fault width of bearings can be identified.The experiment results prove the way of defect diagnosis and fault size identification is effective and believable.