Abstract: The paper proposed a fault diagnosis method of shearer rocker bearing based on improved EEMD and HMM. The method uses improved extreme points symmetric extension and cosine window function to reduce impact of end effect on decomposition results, so as to improve signal decomposition precision; then extractes energy entropy of each intrinsic mode function as input feature vector of HMM for fault pattern recognition. The experimental results show that bearing fault identification rate of the proposed method is above 90%, which indicates the method achieves accurate fault diagnosis of shearer rocker bearing.