Abstract: The random seismic acceleration signals of four floors of a reinforced concrete frame structure model are sampled, whose random noise of high frequencies and trend noise of low frequencies are eliminated efficiently by the de-noising method based on virtual response information. Using integral operator transform, the displacement, velocity and acceleration signals in transform space of the four floors are obtained. With these signals, the stiffness and damping parameters of upper three floors of frame structure models are identified by the inversion arithmetic in the time domain though no information of the seismic excitation is known. Utilizing the information of the first order frequency, the ground floor stiffness is identified by solving an inversion eigenvalue problem. The identification results of stiffness parameters tally with the numerical results by D-value method on the whole.