To understand the performance state of rubber bearings under seismic action, a new method is proposed to identify time-varying nonlinear properties of the rubber-bearing isolated system based on wavelet multi-resolution analysis technique. First, based on the discrete sequences of wavelet multi-resolution analysis theory, the approximate unknown time-varying nonlinear restoring force of rubber bearings is expressed by the Daubechies wavelet coefficient. Then, the wavelet coefficients are estimated by the least-squares technique, and the nonlinear restoring force of the rubber bearing is reconstructed. Finally, the identification method has been applied to numerical examples. The results show that the proposed approach can accurately identify the nonlinear characteristics of rubber bearings when the superstructure stiffness changes and can also track the stiffness degradation of the superstructure, thus demonstrating its validity and accuracy.