Vibration signal analysis is significant for motor rolling bearing defect diagnosis and online monitoring. To effectively recover fault features of motor rolling bearing vibration signals, a signal characteristic extraction method according to feature mode decomposition (FMD) and dispersion entropy (DE) is presented. First, the vibration signals in various failure states are broken down into a number of intrinsic mode functions (IMFs) using the FMD approach. After this, the defect feature vector of the original vibration signal is created by combining the DE of each IMF component. An SVM fault diagnosis model was developed utilizing the feature vectors recovered by FMD-DE in order to confirm the efficacy of the method proposed in this paper in extracting fault characteristics of motor rolling bearing vibration signals. Simulation experiments show that this method can effectively extract the vibration signal characteristics of motor rolling bearings and correctly classifying fault kinds according to the features that are extracted.