Abstract: Coefficients of continuous wavelet transform (CWT) of the characteristic components of a non-stationary signal can be assembled for high-amplitude energy region in time-scale plane. An improved CWT gray moment with time-scale window is constructed in this paper to represent the information implicated by a CWT coefficient figure and it can be used as a quantitative time-frequency symptom of the signal. Simulation results show that this method captures effectively the time-frequency amplitudes of mutations components and its successful application to vibration analysis on the hydraulic turbines at the Three Gorges hydropower station has produced useful time-frequency information of non-stationary vibrations. Thus, the CWT method provides a new quantitative approach to extraction of hydraulic turbine vibration symptoms.