Authentication of duck blood tofu binary and ternary adulterated with cow and pig blood-based gel using Fourier transform near-infrared coupled with fast chemometrics.

This work aims to investigate a feasible and practical technique for the authentication of edible animal blood food (EABF) using Fourier transform near-infrared (FT-NIR) coupled with fast chemometrics. A total of 540 samples were used, including raw duck blood tofu (DBT), cow blood-based gel (CBG), pig blood-based gel (PBG), and DBT binary and ternary adulterated with CBG and PBG. The protein, fat, total sugar, and 16 kinds of amino acids were measured to validate the difference in basic organic matters among EABFs according to species. Fisher linear discriminate analysis (Fisher LDA) and extreme learning machine (ELM) were implemented comparatively to identify the adulterated EABF. To predict adulteration levels, four extreme learning machine regression (ELMR) models were constructed and optimized. Results showed that, by analyzing 27 crucial spectral variables, the ELM model provides higher accuracy of 93.89% than Fisher LDA for the independent samples. All the correlation coefficients of the optimized ELMR models' training and prediction sets were better than 0.94, the root mean square errors were all less than 3.5%, and the residual prediction deviation and the range error ratios were all higher than 4.0 and 12.0, respectively. In conclusion, the FT-NIR paired with ELM have great potential in authenticating the EABF. This work presents amino acids content in EABFs for the first time and built tracing models for rapid authentication of DBT, which can be used to manage the EABF market, thereby preventing illegal adulteration and unfair competition.