Considering the natural frequency estimation and spurious frequency elimination are common in time-varying modal parameter identification, a subspace tracking algorithm based on the novel information criterion is introduced. A post-process procedure for spurious modes elimination is proposed. Firstly, the pseudo modal parameters are obtained through the proposed subspace tracking identification method, then the fuzzy C-means clustering algorithm is implemented to acquire the initial pseudo natural frequencies, finally a moving data window is applied to select the pseudo natural frequencies by data comparison. The proposed validation method only needs an estimated active modes number. Numerical simulations with different frequency changing types in dicats that the proposed method can reach higher calculation accuracy. The identification and validation method are verified through a modal parameters identification of a time-varying structure in the high temperature environment. The experiment shows that the proposed method has high application value.