Abstract: The mine seismic signals in the production of metal mines is complex, it is difficult to identify and extract appears.By spectrum analysis, component efficient and frequency domain filtering are applied to composite signals, the separated composite signal is obtained.In this paper, two typical signals for seismic and blasting are taking as examples, which are calculated by spectrum characteristic analysis, filtering decomposition and windowing of boundary signal.This method provides data and method support for the classification and identification of mine seismic signals, the statistics of signal sample bank characteristics, and the evaluation of signal trigger rationality in seismic safety monitoring of metal mines.At the same time, it is helpful to improve the accuracy of identifying microseismic events.