Feasibility study on prediction of the grain mixtures for black sesame paste recipe with different chemometric methods

An attractive strategy for quantitative analysis of the grain mixtures for black sesame paste was proposed in this research by combined application of FT-IR spectroscopy and chemometrics. The FT-IR spectra of three grain (black sesame, black rice, and black bean) mixtures, which are used for producing black sesame paste were collected. Based on the FT-IR spectra, different regression methods were utilized and optimized to estimate each mixture component content. The PCA-Class results revealed that each mixture class was highly distinct and allowed discrimination from others. PLSR, siPLS, and PCR were comparatively performed to calibrate regression models. The optimal model was achieved with the R2 higher than 0.98 and the RMSE less than 4. The identification results of hold-out validation samples proved the accuracy of the model and indicated that PCR was the most suitable regression model. The results indicate that FT-IR spectroscopy combined with chemometrics could be potentially applied in the industry for quantitative analysis of mixed foods in the future.