A biaxial servo testing machine is used to simulate rockburst of a granite tunnel

and the acoustic emission（AE） system is applied to acquire the waveform signal. The dominant-frequency characteristics of AE signals from initiation to occurrence of rockburst are analyzed using the spectrum analysis theory. The results show that the dominant frequency interval of AE signal is from 0.4 to 110 k Hz

and more than 99% of the dominant-frequency concentration is in the range of 35-65 k Hz. The distribution of the dominant-frequency of the AE signal can be divided into three stages with the evolution of rockburst. In the early stage of rockburst initiation

the dominant frequency is in the range of 35-65 k Hz and 15-25 k Hz

a few of low frequency AE signal close to 0.4 k Hz occurs

and high frequency around 110 k Hz seldom appear. In the middle period of rockburst initiation

the low frequency around 0.4 k Hz nearly disappears

but the high frequency around 110 k Hz begins to appear. In the final stage of rockburst occurrence

the number of dominant-frequency in the range of 35-65 k Hz plummets

the low frequency approximate to 0.4 k Hz almost disappears

but the high frequency close to 110 k Hz appears intensively. The study provides the fundamental understanding of optimizing the frequency band of AE monitoring

which has important theoretical significance with respect to monitoring and early warning of tunnel rockburst.