Abstract: Gearbox is usually designed to operate on complex conditions with variable speeds, loading and temperatures, which easily lead to fatigue faults of its key components, such as gears and rolling bearings. At present, the diagnosis result is occasionally unstable when the common fault feature extraction methods are used in actual diagnosis of gearbox. Recently, Alpha stable distribution (ASD) and Multi-fractal analysis (MFA) have been investigated to address this limitation. These two methods have their respective advantages, and can be compensated each other. This article reviews and analyzes the existing achievements, and discusses the three aspects of the fault feature extraction method based on ASD, the fault feature extraction method based on MFA, and the fault feature extraction method based on feature fusion of ASD and MFA, respectively. Finally, the future research directions in feature selection, feature fusion are pointed out.