Abstract: To solve the problem that it is difficult to extract the early weak fault features of in-wheel motor bearings, an original feature extraction method based on optimized singular spectrum decomposition (OSSD) and enhance multipoint optimal minimum entropy deconvolution adjusted (EMOMEDA) is proposed to realize the fault detection and extraction, then timely grasp the operation safety of the in-wheel motor. Firstly, an OSSD method that adaptively optimizes the number of components by a new time-frequency composite index (TCI) is proposed which is used to pre-process the original signal, and the sensitive singular spectrum component (SSC) is selected by the envelope spectral peak index (ESPI). Then the EMOMEDA method is proposed. An improved waveform continuation strategy is designed to restore the length of the deconvolution signal which overcomes the edge effect of the MOMEDA algorithm, and the optimal deconvolution signal is obtained through the second deconvolution operation. Finally, the envelope analysis is performed on the optimal deconvolution signal to realize the enhanced extraction of fault features. The feasibility of the proposed method is verified by simulation and test signals respectively, and its superiority is proved by comparing it with a variety of fault feature extraction methods. The results show that the proposed method can effectively extract weak fault features and has considerable advantages in feature enhancement.