Abstract: Influenced by the accuracy of geophysical prospecting, some small faults in coal mine can not be effectively detected, which poses a serious threat to the safety of underground production. Accurate prediction of hidden faults is an urgent need for safe mining in working face. Based on the analysis of a large number of microseismic events, this paper shows that microseismic monitoring can be used as an auxiliary monitoring method for detecting hidden faults in working face. The main methods are: DDT, DWT and energy density methods are used to transform the time domain, frequency domain and energy domain curves of microseismic events, and the "time-frequency-energy" characteristic analysis method of microseismic events is proposed. Based on this method, the difference of "time-frequency-energy" characteristics between the curves measured before and after faults for the same microseismic event is analyzed. Research shows that: in the time domain, there are many peaks of microseismic events before faults, short duration of vibration, undeveloped coda, and the opposite is true after faults. In the frequency domain, the frequency band of microseismic events before faults is complete, and the low, medium and high frequency bands are distributed. The frequency domain curve shows multi-peak characteristics, while the microseismic events after faults are concentrated in the frequency band of 0-20Hz, showing a sharp single peak characteristic. Faults play a similar role of "low-pass filtering" for microseismic events. In the energy domain, the energy of microseismic events before faults has the characteristics of short-term, full-frequency and continuous propagation, while after faults, the energy of microseismic events is opposite. Based on this, the abrupt change reason of microseismic event waveform curve in 3307 working face of a mine is analyzed. It is inferred that there are hidden faults interfering with the propagation of vibration wave in front of the working face. The F2 hidden faults exposed subsequently in the working face verify this inference.