Abstract:

Soymilk protein content is a major index for quality evaluation. In this study, near infrared spectra were obtained
for 83 adulterated soymilk samples and then the spectra and optical density of these adulterated samples were analyzed
by statistical methods. The feasibility for qualitative discrimination of soymilk quality was studied by using protein as the
major qualitative index. At last, qualitative discrimination standard was established. The experimental results indicated
that the spectral peak changed obviously in the wavelength range from 742.59 to 810.96 nm with an increase in soymilk
protein content. The optical density OD810.96 nm and OD742.59 nm were used to plot distribution diagram. Based on optical
density distribution diagram of 83 calibration samples, the soymilk classification model of OD difference was confirmed as
following: when its value of ΔOD742.59-810.96 nm was more than 0.062 9, the soymilk sample was classified into adulterated
sample, otherwise it was normal soymilk sample. Totally 37 prediction set samples were classified according to the model;
100% of adulterated soymilk in the prediction set were classified as adulterated samples, and 2 of 20 normal soymilk
samples were classified wrongly into adulterated samples. The preferable prediction results indicated that the accuracy of the
developed method was superior. The feasibility of soymilk quality evaluation based on optical density combined with near
infrared spectral data was confirmed. The method was simple and reliable, and could provide certain references for the rapid
detection of soymilk quality.

Key words: soymilk, optical density, near infrared spectroscopy, qualitative discrimination