Effectively controlling the vibration amplitude in the rotor system startup process under different speeds is an important measure to ensure the safe operation of rotor systems, whereas the traditional rotor balancing method is based on steady-state response of the rotor with lower frequency of balancing. In light of the shortage of current rotor balancing methods for processing non-stationary data, we proposed a new rotor balancing method under all working conditions that is based on parameterized time-frequency analysis. According to the method, we constructed an appropriate rotation operator based on the rotating frequency and rotated the time-frequency characteristics of the startup signal, which accurately extracted the rotating frequency component. By adding the trial weight and obtaining the transfer matrix under various speeds by means of holobalancing, we achieved rotor balancing under all operating conditions, thus avoiding the shortcoming of the traditional rotor balancing method. Experimental results showed that this method easily and quickly determined the unbalanced weight and unbalanced rotor angle, reduced unbalanced vibration of the rotor system, and shortened the start-times in the balancing process.