Abstract
Analytical methods for meat-species identification and detection of adulteration are always needed for quality control and the safety of consumers. A novel application of PLS-Kernel with MIR for quick detection of pork in beef mixes down to 1.4 wt% is outlined. PLS-Kernel algorithm showed an excellent performance for handling many variables spectral data, which makes it suitable for food analysis by IR. Raw spectral data indicated a nonlinear relationship between pork level and IR band intensities of proteins and fats in the mixes. The ratios A(1654cm-1) (amid-I) / A(1745cm-1) (C=O band of ester), A(1540cm-1) (amid-II) / A(1745cm-1), and (A(1395cm-1) + A(1450cm-1)) / A(1175cm-1)were correlated with pork level to establish a useful analytical signal for detection purposes. Chemometric analysis was carried out on 900-1900 cm(-1) as an informative spectral range for protein. PLS-Kernel model is recommended over PCR and PLS for handling many variables in IR data due to the lower computation cycles and less-memory storage. The proposed method would replace advanced techniques for quick pork detection in minced meat mixes.