Abstract: Aiming at the problem of velocity and displacement distortion after integration caused by error trend term in acceleration signal tested in explosion test, the velocity and displacement after integration were reconstructed by removing the trend term of explosion acceleration based on least square fitting, wavelet decomposition and EEMD (Ensemble Empirical Mode Decomposition). The research shows that: The fitting order of the least square method, the number of decomposition layers of the wavelet decomposition method and, the standard deviation of the white noise of EEMD method have great influence on the reconstruction speed and displacement amplitude. A new method is proposed to evaluate the trend term correction results by using the spectral deviation index s² and the mean deviation index S², and the appropriate order, decomposition level and white noise standard deviation can be selected through the joint analysis of the frequency distribution and integral time history characteristics of the deviation S² to accurately remove the trend term in the signal. Compared with the explosion test results, it is proved that the three methods can restore the motion trend and residual displacement characteristics under the impact of explosion, and the wavelet decomposition method has some advantages.