Aiming at the problem of difficult to separate the fault characteristic signals of rolling bearing in a strong background noise environment, a fault feature extraction method of rolling bearing based on adaptive maximum second-order cyclostationary blind convolution (CYCBD) and 1.5-dimensional spectrum was proposed. Firstly, Fourier transform was performed on the vibration signal to obtain the spectral structure of the signal; taking the proposed weighted harmonic sum as the optimization index, all frequencies in the signal spectrum range were used as candidate frequencies to search, and the frequency at the maximum weighted harmonic sum was determined as optimal cycle frequency. Then, the CYCBD with optimized parameters was used to filter the signal, and the 1.5-dimensional spectrum method was used to process the filtered signal. Finally, in order to further verify the effectiveness of the method in extracting bearing fault characteristics, the measured signal was analyzed by envelope analysis method, and the spectral characteristics of the filtered signal were obtained. The experimental results show that after filtering by the method, the Shannon entropy of the signal is 0.094, significantly lower than CYCBD and the EMD method. In addition, clear fault characteristic frequency and multiplier appear in the envelope spectrum of the signal, which indicate that the method has good noise suppression ability and can effectively extract fault pulse components from bearing vibration signals. [ABSTRACT FROM AUTHOR]