Classification of tea samples using SVM as machine learning component of E-tongue

This article introduces a new approach for identification of tea sample using pulse voltammetry method in an electronic tongue based instrumentation. The classifier system consists of a principle component (PCA) based feature extraction module followed by support vector machine based discrimination. The PCA score of unknown tea sample is undergone through different pair-wise (binary) classification using SVM for repeated times. For six different categories of tea samples in the present case, unknown sample is examined for fifteen times. The result of classification is six membership grades. Finally these membership grades are analyzed by decision directed acrylic graph method (DDAG) for decision making task about the exact authentication of unknown tea sample belonging to six categories. The proposed method could be equally followed for more than six categories.