Based on the multivariate function Taylor formula and optimal control principle, first, using the 10.65 GHz vertical polarization data of observed brightness temperature from FY-3C microwave radiation imager and simulated brightness temperature from community radiative transfer model (CRTM), the linear retrievals model between the microwave land surface emissivity and two (four) influencing factors are constructed in the central part of the Taklimakan Desert (retrieval area) in January. Second, considering the observed brightness temperature as a reference, the land surface emissivity obtained by two kinds of linear retrieval models are provided to CRTM model to simulate brightness temperature, it is found that the average deviations of the simulated brightness temperatures are reduced to 56.36% and 49.37% of the original simulated brightness temperature, respectively. Finally, in the whole Taklimakan Desert (test area), the time and space independence tests of two kinds of linear retrieval models are carried out by selecting Jan 18, which is used for retrieval date, and Jan 29, which is not used for retrieval date. The results show that the simulated brightness temperatures of the land surface emissivity obtained by two kinds of linear retrieval models are still closer to the observation than the original simulated brightness temperature, and the average deviations are also significantly reduced. The two linear retrieval models are reasonable and universal for the retrieval of land surface emissivity in the desert area in January, and the linear retrieval model with 4 factors is better than 2 factors.