Improved empirical wavelet transform (IEWT) is a new kind of self-adaptation method of signal analysis. Combine this method with fast spectral kurtosis(FSK) filtering, and the gear and rolling bearing fault diagnosis can be achieved. Firstly, the signal is decomposed using the IEWT method, and the two components which have the most obvious fault characteristics are extracted and reconstructed. Then, the reconstructed signal is filtered using the FSK filtering method. Lastly, the filtered signal is analyzed using spectral envelope method and the fault characteristics of signal are extracted. By analyzing the broken gear teeth and rolling bearing fault signals, it is indicated that the method based on IEWT and FSK of envelope spectrum analysis is more distinctive than spectral envelope method and envelope spectrum analysis based on EMD and FSK. It can effectively identify the fault characteristics of gear and rolling bearings.