Abstract:The rock acoustic emission signals have transient property and diversity under impact loading, which belong to the typical non-stationary signals. The EMD method was used to decompose the rock acoustic emission signals into a number of IMF components with different characteristics time scale and then the energy characteristic of each stationary IMF component was extracted. The test results showed that the energy of the rock acoustic emission signals were mainly distributed in the first four IMF components, and unevenly distributed; The spectrum of the IMF component with the original signal spectrum were in substantial agreement; If the density of rock is less, the speed of the longitudinal wave is lower, and the elasticity modulus is less, the dominant frequency of acoustic emission signals under impact loading become more concentrated and tend to be low. Compared with wavelet packet analysis, EMD analysis was more suitable for the non-stationary signals.