Noninvasive monitoring of blood glucose is the current international academic research focus. Near-infrared (NIR) spectroscopy is the most prospective method of the present study, however, with the flaw of insufficient specificity to glucose. Tissue polarimetry has recently received considerable attention due to its specificity to glucose. Thus the glucose predicting accuracy would be improved by combining spectral intensity and polarization characteristics. However the backscattering spectral polarization characteristics of turbid media have not been reported within the wavelength range from visible to near-infrared light. In this paper, we simulated the backscattering spectral Mueller matrix of turbid medium by vector Monte Carlo. And the polarization characteristics, which are linear/circular degree of polarization (DOP) and linear/circular diattenuation, can be extracted from the simulated Mueller matrix by polar decomposition. Circular diattenuation is not discussed because it remains almost zero on the backscattering plane. While reduced scattering coefficient increases linearly with increasing wavelength, the spectral curves show distinct wavelength dependencies. Interestingly, the wavelength dependencies at center position are different from those at off-center position for linear/circular DOP and linear diattenuation. As expected, it is shown that both linear DOP and linear diattenuation increase with the increasing wavelength. However it is not the case for linear DOP in the central area around the incident point. In this area linear DOP decays approximately exponentially with increasing wavelength. As for circular DOP, it varies with wavelength non-monotonically. These results should be meaningful when spectral polarization characteristics are used to combine with spectral intensity to extract glucose concentration by chemometrics.