I. Introduction
As an important component in manufacturing industry, bearings are widely used in various mechanical equipment, especially in the area of motor [1]-[3], urban rail train [4], gearbox [5]-[6] and so on. With the increasingly high-quality requirements for various types of mechanical equipment, condition monitoring for bearings plays an important role for their safety and reliability operation. It is reported that damages are easily existed in bearings under harsh working environment and further affect their service life [7]. Furthermore, their damage may cause their failure and chain reactions on adjacent machine components and the whole rotating machinery. Therefore, it is critical to put forward an effective and applicable method to identify bearing fault state so that it not only reduces the probability of accidents but also saves maintenance costs.