Abstract: Seismic monitoring during tunneling is a critical technology for ensuring safe mining.In response to the limitations of existing monitoring methods in signal denoising, we propose a denoising method for seismic signals during tunneling based on an improved Empirical Wavelet Transform(EWT) and fully adaptive noise ensemble empirical mode decomposition(CEEMDAN) algorithm.The method utilizes the CEEMDAN algorithm to perform multi-scale decomposition of seismic signals, extracting several intrinsic mode functions(IMFs),and combines the EWT algorithm to eliminate high-frequency noise components.This significantly improves the signal-to-noise ratio(SNR) and preserves the time-frequency characteristics of the signal.Empirical analysis of the seismic data from the Zhao Zhuang No.2 well demonstrates that after applying the proposed denoising method, the SNR is enhanced by approximately 180.8%,effectively removing high-frequency noise while retaining the time-frequency features of the valid signal.Compared with traditional denoising methods, the proposed approach offers significant advantages in improving signal clarity and optimizing the accuracy of subsequent seismic signal processing and analysis.This method greatly enhances the effectiveness and precision of advanced seismic detection during tunneling, providing more reliable data support and scientific evidence for support design and geological disaster early warning in the tunneling process of mine roadway.