Abstract: Three kinds of AE signals were obtained through clustering and analyzing the acoustic emission signals of rock burst experiments and the characteristic parameters including AE energy,ringing count,rise time and frequency. By analyzing the characteristic parameters of three kinds of AE signals deeply,excavating their corresponding physical meanings,the fracture characteristics were revealed. Based on the rate of energy release,the main types of rock failure formed by driving rock burst were determined. Use the corresponding AE signals,the rock burst was predicted. The re- sults show that the first kind of signals has high energy,high counts,low rise time and low frequency features,and the “two highs and two low” feature corresponds to the stable expansion of crack with high strength and large scale. The second kind of signals has high energy,high counts,high rise time and low frequency features,and the “ three highs and one low” feature corresponds to the unstable expansion of crack with high strength and large scale. The third kind of signals has low energy,low ringing count,low rise time and low frequency features,and the “four low” feature corresponds to the unstable expansion of crack with low strength and small scale. Among them,the unstable expansion of crack with high strength and large scale is the main driving force of rock burst. The “three highs and one low” feature can provide early warning information for rock burst. It is the key target of early warning research of rock burst disas- ter.