For the degradation characteristic of the gear performance parameter,a method of fault pattern recognition and analysis,which combines the Delayed Correlation-Envelope（DCE）with the Hidden Markov Model（HMM）is proposed.The DCE is used to extract features from vibration signals.The HMM is trained respectively by using the data under normal condition,gear root crack condition and gear root breaking condition.Moreover,the trained HMMs are used to pattern recognition and assessment model.Finally,by comparing with standard HMM,the experimental results show that the proposed methodology is feasible and effective.