In this paper, a new pattern recognition method for machine working condition based on gravity center of Hilbert time-frequency spectrum and support vector machine (SVM) is demonstrated in detail based on vibration signal analysis. Firstly, synchronous average is used on monitored signal for preprocessing, which is used to remove effect from cyclostationary characteristics. Vibration signal data information changes from time domain to angle domain, which is more suitable for rotary machine pattern recognition. Then, Hilbert time-frequency spectrum is obtained according to Empirical Mode Decomposition (EMD) and Hilbert transform. Thirdly, gravity center of Hilbert spectrum is calculated to construct a feature vector for machine working condition pattern recognition. In the end, different working condition can be classified by using support vector machine (SVM). Rolling bearing pattern recognition is as an example to verify the effectiveness of this method in the research. According to the result analysis, it can be concluded that this method will be helpful for the development of machine preventative maintenance.