When collecting the overpressure signal of the blast wave , a large amount of noise is mixed in the blast shock wave overpressure signal due to the high temperature environment of the monitoring, the error of the pressure transducer and the magnetic field disturbances. In order to accurately obtain the characteristics of the overpressure signal, a denoising algorithm based on the combination of Complete EEMD with Adaptive Noise (CEEMDAN) and Savitzky-Golay (SG) is established. Firstly, CEEMDAN was used to decompose the overpressure signal of the blast wave, and then the energy contribution rate of each intrinsic mode function (IMF) was calculated. The SG filtering algorithm was used to denoise IMFs with energy ratios below 0.1% and greater than 0.05%. The experiment result expresses the signal-to-noise ratio is improved by 0.85dB, 0.71dB,3.09dB and 0.25dB ，which comparison with EMD, modified EEMD, CEEMDAN and CEEMDAN-Wavelet threshold denoising, respectively, showed the smallest mean square error. CEEMDAN-SG and CEEMDAN-Wavelet are more effective in noise removal, and CEEMDAN-SG has the highest similarity with the original signal at 0.16s. The algorithm can not only availably remove the high frequency noise, but also retain the characteristics of the original signal, which is appropriate for the denoising process of the blast shock wave overpressure signal.