Abstract: The rapid detection of lipid oxidation in frozen pork was carried out by using Raman spectroscopy combined with chemometric methods. Frozen pork stored for 0–360 days was selected in this study, and the regression relationship between acid value or peroxide value and storage time was established by the least squares method. Raman spectra were collected with a portable Raman spectrometer and were preprocessed by Savitzky-Golay with five-point smoothing (SG-5), standard normal variate correction (SNV) or adaptive iteratively re-weighted penalized least squares (airPLS). The partial least square regression (PLSR) was used for modeling and analysis. Then competitive adaptive reweighted sampling (CARS) was used to select the Raman characteristic variables for the best preprocessed spectra to establish a CARS-PLSR model. It was found that the correlation between peroxide value and storage time was more significant during the storage period (P = 0.000 3 < 0.05). The prediction models with spectral preprocessing by using SNV and airPLS performed best. The CARS-PLSR model showed improved prediction performance. The CARS-PLSR models for prediction of acid and peroxide values had the following performance parameters: coefficient of determination for calibration (R2c), 0.88 and 0.84; root mean square error of cross-validation (RMSECV), 0.31 and 2.33; coefficient of determination for prediction (R2p), 0.76 and 0.75; root mean square error of prediction (RMSEP), 0.17 and 1.87, and ratio of performance to deviation (RPD), 2.59 and 1.90, respectively. In conclusion, our results show that the peroxide value of frozen pork is significantly correlated with storage time. The degree of lipid oxidation in frozen pork can be rapidly determined by using Raman spectroscopy.

Key words: Raman spectroscopy; frozen pork; storage time; acid value; peroxide value