Abstract
A self-developed voltammetric electronic tongue (VE-tongue) was used as rapid techniques to classify the wolfberry samples from different geographical origins. The VE-tongue is comprised of eight working electrodes, a reference electrode and an auxiliary electrode. The signals collected by the sensor array are extracted using discrete wavelet transform (DWT). Multivariate statistical data analysis techniques such as principal component analysis (PCA), linear discriminant analysis (LDA) and support vector machine (SVM) are used to classify the wolfberry samples. The results show that neither PCA nor LDA can classify the samples of different geographical origins correctly, and some samples are scattered together. The parameters of SVM are optimized by two optimization methods: leave-one-out cross validation, and particle swarm optimization (PSO). The classification accuracy is 95% and 100%, respectively. The SVM model based on PSO is superior to other methods in distinguishing the geographical origin of wolfberry. This study provides a new method for the quick classification of Chinese wolfberry in the market.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have