Abstract:Roof is composed of heterogeneous rock. When roof failure happens,a large amount of energy would release in the form of seismic wave. There are many common methods,such as Fourier transformation,filter technique,etc. to identity the abnormal signal of seismic wave,but the identification accuracy is quite low. Multi-resolution wavelet technique is used to identify the first and second variation point,based on the Lipschitz. An example analysis shows that multi-resolution wavelet technique can identify the abnormal signal of seismic wave effectively in different scales,and the omen of roof fall can be grasped accurately. The presented study provides a new idea for the prediction of catastrophe of rock engineering.