Rock AE (acoustic emission) signals exhibit significant non-stationary and nonlinear characteristics, posing substantial challenges to traditional denoising methods. To address this issue, a hybrid denoising method combining VMD (variational mode decomposition) and SSA (singular spectrum analysis) was proposed. Firstly, the grid search algorithm was employed to optimize the VMD parameters using the spectral overlap ratio as the loss function. Consequently, the original AE signal was adaptively decomposed into several IMFs (intrinsic mode functions) and one res (residual) component, which is treated as a noise reference. Subsequently, the correlation coefficients between each IMF and the res component were calculated. IMFs with correlation coefficients higher than a predefined threshold were identified as noise and discarded, while those below the threshold are classified as noise-contaminated IMFs. For the noise-contaminated IMFs, the maximum singular value of the res component was utilized to determine the reconstruction order for SSA, thereby achieving secondary denoising. Finally, the denoised AE signal was reconstructed from the processed IMFs. Simulation and experimental results demonstrate that the proposed method effectively removes noise while notably preserving the frequency characteristics of the signal. Compared with wavelet packet denoising, CEEMDAN denoising, VMD-wavelet packet denoising, and CEEMDAN-SSA denoising methods, the proposed VMD-SSA method shows superior performance in terms of SNR (signal-to-noise ratio) and MSE (mean square error), providing a reliable foundation for subsequent analyses such as ring-down counting and energy calculation.