Abstract: The trend removing algorithm of blasting vibration monitoring signals were studied, based on empirical mode decomposition (EMD), wavelet transform, and the least square method. Effects of trend removing by the three methods were compared and the time-frequency feature of reconstructed signals were analyzed. The study indicates that wavelet and least square methods need a predefine decomposition function basis in the trend removing of monitoring signals. EMD is adaptive, more practical and flexible in processing blasting vibration signals which are short and nonstationary. Thusly, the EMD method can remove trend more effectively and improve the resolution of a time-frequency domain in the preprocessing analysis of blasting vibration monitoring signals. It has an important reference value for the extracting time-frequency feature of blasting vibration signals accurately.