Abstract: Nearinfrared spectroscopy combined with partial least squares (PLS) method was applied to establishingan optimal mathematic calibration model, by which the concentrations of amino nitrogen, total acid and salt content in the soy sauce were predicted, and the color ratio of the soy sauce was determined. Furthermore, the influences of spectral preprocessing methods and the numbers of principal components on the prediction ability of the PLS model were discussed. The results show that the calibration model established with firstderivative preprocessing spectra was the best, 18 samples of tested set were predicted with this model, the relative standard errors for the prediction of ami no nitrogen, total acid, salt content and color ratio were 1.516%, 1.811%, 1.798% and 1.893%, respectively. Experimental results demonstrate that this method is fast and convenient, and it has a high ability of prediction, thus it is promising for the nondestructive quality control of soy sauce.

Key words: NIR spectroscopy, partial least squares (PLS), soy sauce, nondestructive analysis