Abstract: Considering the problem of a false modal and the decision of a real IMF in an improved Empirical Mode Decomposition (EMD) by using frequency-band filters, a signal analysis method based on the combination of improved EMD and Independent Component Analysis (ICA) was proposed. This proposed method does not require setting a threshold in advance, and the real IMF component can be isolated automatically. This method also can eliminated the false modal components, which appeared in the process of an improved EMD, thusly it can make sure the validity of an EMD very well. Afterwards, the random decrement technique (RDT) is used to obtain free vibration modes of each IMF. Finally, Hilbert Transform (HT) and the least square fitting are employed to identify structural modal parameters from these free vibration modes. Finally, the presented method is used to identify modal parameters of two numerical examples and a seven-story steel frame. The results show that the proposed method can resolve the drawbacks of an improved EMD more effectively, and the modal parameters can be identified successfully.