Identifying the structural modal parameters based on the measured responses of structure is a major application for structural health monitoring. This study presents a procedure based on Gram-Schmidt process to improve the accuracy of the calculated stiffness matrix of a structure. The continuous wavelet transform is applied to the measured responses of a structure and the time series model could be reconstructed to denoise for the measured noisy responses. The identified modal parameters would be obtained from the coefficient in time series model. Next, correcting the identified mode shapes via Gram-Schmidt process. Finally, the structural stiffness matrix can be created via the identified natural frequencies and the corrected mode shapes. The accuracy of this procedure is numerically confirmed; the effects of the wavelet parameters and noise on the ability to accurately estimate the stiffness matrix are also investigated.