Abstract: The discrete wavelet threshold method was used to denoise the SHPB testing signals containing the high-frequency noise. Aimed at the short-duration and abrupt change of non-stationary SHPB testing signals, the Sym5 wavelet basis was chosen as the optimal one which has the lowest root mean square error (RMSE) among various wavelet basis reconstruction for the processing of SHPB testing signals. The unbiased estimation procedure SURE was used to determine the threshold value for each decomposition level, the signal was decomposed to six levels with the threshold values of 0.119, 0.085, 0.089, 0.102, 0.118, 0.116 from level 1 to level 6 respectively. Signal-noise ratio (SNR) and RMSE of the signals denoised by the discrete wavelet transform were compared with those by the conventional low-pass filters in dynamic strain indicators. SNR of the signal denoised by the discrete wavelet transform is greater than that by the low-pass filter, and RMSE of the signal denoised by the discrete wavelet transform is smaller than that by the low-pass filter. The results show that discrete wavelet transform not only has favorable denoising effects, but also can get more accurate reconstruction signals, it can replace the low-pass filters in dynamic strain indicators for the SHPB testing signal denoising.