Feature extraction is the key technology in the field of image and voice recognition or fault diagnosis. Having deeply studied the advantages and disadvantages of the feature extraction based on the wavelet transform and Empirical Mode Decomposition(EMD), a method combining advantages of the two methods is proposed to extract the feature information. Firstly, the proposed method acquires stabilized single mode state components by EMD, and then Wavelet Packet(WP) analysis is performed to single mode state components. Finally, by comparing the method with other methods through simulation and example test, it is proved that the proposed method not only features higher feasibility, but also can extract fault information more accurately.