Extracting features of the faults is one of the most important processes in the aircraft engine incipient faults diagnosis.We study the aircraft engine rotor faults and present two methods for extracting feature of faults based on empirical mode decomposition(EMD) and wavelet transformation respectively.Through the numerical data magnitude analyzing and emulation,we compare the characteristics between two methods.Under the environment of Matlab7.0,a soft system of extracting feature of the fault is developed.The experiment result shows that the time-frequency method based on EMD can effectively extract the feature of unbalanced fault signal and is proper for non-frequency modulation signal procession.