Performance of fluorescence spectroscopy for beef meat authentication: Effect of excitation mode and discriminant algorithms

This study evaluated the performance of classical front face (FFFS) and synchronous (SFS) fluorescence spectroscopy combined with Partial Least Square Discriminant Analysis (PLSDA), Support Vector Machine associated with PLS (PLS-SVM) and Principal Components Analysis (PCA-SVM) to discriminate three beef muscles (Longissimus thoracis, Rectus abdominis and Semitendinosus). For the FFFS, 5 excitation wavelengths were investigated, while 6 offsets were studied for SFS. Globally, the results showed a good discrimination between muscles with Recall and Precision between 47.82 and 94.34% and Error ranging from 6.03 to 32.39%. For the FFFS, the PLS-SVM with the 382nm excitation wavelength gave the best discrimination results (Recall, Precision and Error of 94.34%, 89.53% and 6.03% respectively). For SFS, when performing discrimination of the three muscles, the 120nm offset gave the highest Recall and Precision (from 57.66% to 94.99%) and the lowest Error values (from 6.78 to 8.66%) whatever the algorithm (PLSDA, PLS-SVM and PCA-SVM). •Fluorescence spectroscopy is able to discriminate three beef muscles (LT, RA and SM) with a high accuracy.•For FFFS, the 335nm excitation wavelength was identified as the best one for muscle discrimination.•For SFS, the 120nm excitation offset was identified as the best one for muscle discrimination.•FFFS worked better than SFS for discrimination of the three beef muscles.