Tissue phantoms with different optical parameters are designed to study the influence of absorption and scattering on tissue fluorescence and diffuse reflection spectra. An empirical recovery algorithm is optimized to correct the influence of absorption and scattering and obtain the intrinsic fluorescence spectrum of tissues. The results reveal that the empirical recovery algorithm (experience parameters kx and km are 0.9 and -0.8, respectively) can effectively reduce the influence of absorption and scattering on the fluorescence intensity, and the fluorescence intensity is linearly correlated with the concentration of fluorescence components. When we apply spectral recovery algorithm to the screening of diabetes based on skin tissue fluorescence spectrum, the results reveal that, comparing to the fluorescence spectrum before recovery, the area under receiver operating characteristic (ROC) curve increases from 0.54 to 0.81; besides, the sensitivity also increases from 38.6% to 77.6% when the specificity is 70.6%. Therefore, this study makes a major contribution to research on clinical application by optimizing fluorescence spectrum empirical recovery algorithm with tissue phantoms.