为了有效提取轴承的故障特征，避免轴承损伤引起的冲击成分受到离散频率分量和强背景噪声的干扰，该文提出了一种新的基于倒谱编辑(cepstrum editing procedure，cep)信号预白化和奇异值分解(singular value decomposition， SVD)的轴承故障特征提取方法。通过CEP预白化处理增强了轴承故障的冲击特性，去除复杂振动信号中的周期性频率成分，产生了只包含背景噪声和碰撞损伤引起的非平稳冲击成分的白化信号。构造预白化信号的Hankel矩阵，进行奇异值分解，通过差分谱理论选择表征故障冲击成分的奇异值进行矩阵重构恢复信号，去除强背景噪声的干扰，实现对故障特征的提取。试验结果表明，该方法较为理想地提取了轴承滚动体和内圈的故障特征，并且在提取效果和运算效率方面要优于基于小波-SVD差分谱故障特征提取方法。“,”In order to extract fault feature of bearing efficiently, and eliminate discrete frequencies caused by shock and strong background noise, a new method that based on combination of pre-whitening technology using cepstrum editing procedure(CEP) and singular value decomposition (SVD) theory was presented. Signal pre-whitening could enhance the impulsiveness of the bearing fault, and isolate the periodic frequencies in complex vibration signal. The pre-whitening signal only contained non-stationary impact components and background noise. Then a Hankel matrix was constructed for pre-whitening signal, and a group of singular values were obtained by SVD processing, the suitable singular values which represented fault of bearing were selected by difference spectrum, the reconstruction signal did not have strong noise interference and could extract the fault feature. The result shows that this method can efficiently extract the fault feature of defective rolling bearing with ball and inner faults, and has better extraction effect and operation efficiency than wavelet-SVD method.