Abstract: Acoustic emission (AE) signals generated during the failure of rock materials were typically characterized by high randomness, low signal−to−noise ratio (SNR), and significant susceptibility to environmental interference, which posed considerable challenges to accurate feature extraction and quantitative analysis in rock mechanics research. In this study, a robust denoising method based on Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) combined with wavelet transform was developed and validated. Synthetic AE signals with added white noise were generated to simulate noisy laboratory environments, and the CEEMDAN algorithm was applied to adaptively decompose these signals into multiple intrinsic mode functions (IMFs), effectively addressing the modal aliasing problem commonly found in empirical mode decomposition (EMD) and ensemble EMD (EEMD). Each IMF was evaluated for its correlation with the original signal, and noise−dominated components were denoised using wavelet transform; the db4 wavelet basis function was selected after comparative analysis of SNR and root mean square error (RMSE). The method was further applied to AE data collected from uniaxial compression tests on granite samples under controlled laboratory conditions, with signal acquisition performed by a multi−channel system at high sampling rates, and all processing and analysis was carried out in MATLAB. The results show that, compared to conventional EMD− or EEMD−based denoising methods, the CEEMDAN−wavelet approach significantly improves signal quality: for simulated AE signals, the SNR increases from 14.86 dB to 15.31 dB (a 3% improvement) and the RMSE decreases from 0.0971 to 0.0877 (a 9.3% reduction); for experimental AE signals, the SNR increases from 9.06 dB to 9.69 dB (a 6.9% improvement), the RMSE drops from 0.0032 to 0.0011 (a 65.6% reduction), and the Pearson correlation coefficient rises from 0.905 to 0.960. These results confirm that the proposed method achieves superior noise suppression while preserving the essential time−frequency characteristics of AE signals, provides practical value for both laboratory and field applications in geotechnical engineering and structural health monitoring, and future work will focus on integrating adaptive parameter optimization and deep learning approaches to further improve the automation and generalizability of AE signal denoising under complex geological conditions.