Abstract: In order to expend the means of charge induction comprehensive analysis,improve the accuracy of rock burst forecast based on charge induction method,combining the physical test and the theoretical analysis,a test of charge monitoring in coal failure process was carried out,the characteristic and law of charge time-frequency domain signal in coal different failure processes were investigated based on the charge induction mechanism and Fourier transform meth- od,and noise reduction was carried out by a digital filter designed. The results show that there was a good consistency between the amplitude of charge time-frequency domain signal and the loading variations. It was fixed that the main frequency spectrum information of white noise,which produced by the non-contact type inductive charge signal monito- ring device,the amplitude of white noise time-frequency domain increased with sampling frequency improved. There had different signal characteristics at different loading stages,the late of elastic stage was minor amplitude intermitten- cy type, nonlinear stage was major amplitude intermittency type, post-peak phase was major amplitude continuous type. The charge signal was a kind of very low frequency dispersed signal that main vibration frequency in the 1-100 Hz range and its frequency amplitude obviously increased with loading,this feature could be used as a major indi- cator of distinguishing between charge signal and electromagnetic radiation,the white noise was removed and the useful signal reserved with no distortion used the digital filter designed. The time-frequency domain characteristics of charge signal could distinguish between beneficial signals and interference signals,that could improve monitoring accuracy ef- fectively,achieve the purpose of forecast and early warning rock burst.