In the processing of fault vibration signals, an improved CBI-LMD (cubic B-spline interpolation local mean decomposition) method based on self-adaptive waveform matching and an orthogonality criterion is proposed to combat the low decomposition accuracy of the cubic spline interpolation-based local mean decomposition (CSI-LMD) method. First, the raw vibration signal is extended with a self-adaptive waveform matching technique. Next, instead of CSI, the CBI is used to calculate the local means and envelope functions. Finally, the orthogonality criterion is used to set a stopping criterion for the product function. Simulation and experimental results show that the proposed method can effectively extract more accurate feature information in less time than CSI-LMD.