Abstract: Through FSWT （Frequency Slice Wavelet Transform）, we carried on fine time-frequency feature extraction for frozen shaft blasting vibration signal, and using the characteristic that its inverse transformation can cut any frequency range, six reconstructing sub-band signals are refined, and we get all the sub-band energy percentage and the correlation coefficient of the original signals. The results show that: frozen vertical shaft blasting vibration signal belongs to the more high broadband non-stationary random vibration, and the energy is mainly distributed below the 500 Hz, energy ratio is 99.02%; peak vibration amplitude and energy ratio decrease with the increase of frequency, sub-band signal energy percentage and its correlation coefficient has positive correlation relation.