Grinding machine condition monitoring is very important during the manufacturing process. Vibration analysis is usually used to its pattern recognition. But traditional signal analysis method limits the accuracy of recognition because of non-stationary and nonlinear characteristics. In this paper, a novel approach is presented in detail for grinding machine fault diagnosis. The method is based on the new developed Hilbert Marginal Spectrum and wavelet transform, named as wavelet-Hilbert marginal spectrum (WHMS). A rolling bearing’s pattern recognition of grinding machine is used to testify the effectiveness of this method, which can accurately detect flaw of the rolling bearing in early stage. Thus, it can be concluded that this promising method will contribute the development of grinding machine condition monitoring.