This paper presented the problem of information redundance factor in spectral data analysis. Three-component mixtures fluorescence spectra data of anthracene, pyrene and phenanthrene, which were typical polycyclic aromatic hydrocarbons (PAHs), were investigated through principal component regression anaylsis and move windows wavelength selection in chemometrics. Theoretical analysis and experimental result demonstrated that there was information redundancy in fluorescence signal; the data which have been optimized by redundancy techniques could show more realistic information of the test samples quantitatively, and for the multi-component mixed system with spectral overlapped seriously, analysis with the data which have been optimized through redundancy techniques could improve the sensitivity and stability of the models.