Bearing acoustic fault signals are easily disturbed by background noise, making weak fault impulsive features difficult to identify directly. To address this problem, this paper proposes an FMD-RFCA-based fault feature enhancement method for bearing acoustic signals. First, Feature Mode Decomposition (FMD) is used to decompose the original acoustic signal into several candidate modes with different frequency-band characteristics and impulsive responses. Then, the RFCA criterion is introduced to evaluate the relative contribution of the fault characteristic frequency and its harmonics in the envelope spectrum of each mode, and the optimal mode with the strongest fault relevance is selected. Experiments on nine single-fault bearing conditions show that FMD-RFCA provides better fault frequency enhancement and impulsive feature preservation than VMD.