Introducing the complex wavelet transform into the fault feature extraction of automobile main reducer, a fault feature extraction method for automobile main reducer based on complex Morlet wavelet transform was proposed. Aimed at the characteristics of main reducer vibrating signal, two groups of compounding information for the extraction of mechanical fault feature were constructed according to the magnitudes and phases of complex wavelet transform. The simulation results show that the phases of complex wavelet transform and compounding information are more sensitive to singular points of signal than magnitudes, which can extract singularity of signal efficiently and position the singular points of signal accurately. Subsequently, the fault signal of automobile main reducer was analyzed by complex wavelet transform with its compounding information and real wavelet transform respectively. The results show that main reducer fault feature points can be positioned accurately by compounding information, and the decomposition just needs scale 1 calculation, which reduces the calculation greatly for fault feature extraction.