The mineral composition of rock is one of the main factors affecting the spectral reflectance characteristics, and it’s an important reason for generating various rock characteristic spectra. This study choose the rock samples provided by Jet Propulsion Laboratory (JPL) (including all kinds of mineral percentage of rocks, and spectral reflectances range from 0.35 to 2.50 μm wavelength measured by ASD spectrometer), and the various types of mineral spectral reflectances contained within the rocks are the essential data. Using the spectral linear mixture model of rocks and their minerals, firstly, a simulation study on the mixture of rock and mineral composition is achieved, the experimental results indicate that rock spectral curves using the model which based on the theory of the linear mixture are able to simulate better and preserve the absorption characteristics of various mineral components well. Then, 8 samples which contain biotite mineral are picked from the rock spectra of igneous, biotite contents and the absorption depth characteristics of spectral reflection at 2.332 μm, furthermore, a variety of linear and nonlinear normal statistical models are used to fit the relationship between the depth of absorption spectra and the content of the mineral composition of biotite, finally, a new simulation model is build up with the Growth and the Exponential curve model, and a statistical response relationship between the spectral absorption depth and the rock mineral contents is simulated by using the new model, the fitting results show that the correlation coefficient reaches 0.998 4 and the standard deviation is 0.572, although the standard deviation using Growth and Exponential model is less than the two model combined with the new model fitting the standard deviation, the correlation coefficient of the new model had significantly increased, which suggesting that the new model fitting effect is closer to the measured values of samples, it proves that the simulation results of new model is closer to the measured value.