Artificial bionic taste sensors coupled with chemometrics for rapid detection of beef adulteration.

The purpose of this study was to investigate the potential of taste sensors coupled with chemometrics for rapid determination of beef adulteration. A total of 228 minced meat samples were prepared and analyzed via raw ground beef mixed separately with chicken, duck, and pork in the range of 0 ~ 50% by weight at 10% intervals. Total sugars, protein, fat, and ash contents were also measured to validate the differences between raw meats. For sensing the water-soluble chemicals in the meats, an electronic tongue based on multifrequency large-amplitude pulses and six metal electrodes (platinum, gold, palladium, tungsten, titanium, and silver) was employed. Fisher linear discriminant analysis (Fisher LDA) and extreme learning machine (ELM) were used to model the identification of raw and the adulterated meats. While an adulterant was detected, the level of adulteration was predicted using partial least squares (PLS) and ELM and the results compared. The results showed that superior recognition models derived from ELM were obtained, as the recognition rates for the independent samples in different meat groups were all over 90%; ELM models were more precisely than PLS models for prediction of the adulteration levels of beef mixed with chicken, duck, and pork, with root mean squares error for the independent samples of 0.33, 0.18, and 0.38% and coefficients of variance of 0.914, 0.956, and 0.928, respectively. The results suggested that taste sensors combined with ELM could be useful in the rapid detection of beef adulterated with other meats.