Abstract: To address the issue that the vibration signals of solid rocket motor measured by laser Doppler vibrometer contains high frequency noise, a data processing method based on empirical mode decomposition (EMD) was proposed. By suppressing the endpoint effect of traditional EMD methods and modifying the criterion of the intrinsic mode function (IMF), an improved EMD method more suitable for the vibration signal data processing of solid rocket motors was developed. In addition, the IMFs were screened based on the characteristics of correlation coefficient and autocorrelation function for subsequent data processing. The comparison shows that data processing based on the improved EMD method is 5% faster than that of the traditional EMD method, and the denoising effect is 7% or 17% better compared with the traditional EMD or the wavelet decomposition algorithm, respectively. This indicates that the proposed data processing method is efficient, stable, and has good engineering application value.