Abstract: Compared wifh one-dimensional（1D）spectroscopy, two-dimensional（2D）correlation spectroscopy technology is of higher spectral resolution, which is suitable for the discriminant analysis of the similar samples. Since 2D correlation spectroscopy is calculated on the basis of the one-dimensional spectra, noise and false information on the 1D spectrum can all result in the change of the 2D correlation spectrum, which consequently generates some false correlation peaks. Due to these false information, especially for the analysis of complex biological system, it is difficult to extract useful information using 2D correlation spectroscopy. According to the Beer-Lambert law, two components spectral models are built without and with overlapped peaks using simulated spectra. The influences of different normalization methods on 2D correlation spectroscopy are studied. The results show that whether the characteristic peaks in analysis system are overlapped, the appropriate normalization method not only can effectively eliminate the false information, but also can be sufficient to extract the feature information of the material under test.