SVD based tea quality prediction using electronic tongue signal

The electronic tongue (ET) system is basically a multi-electrode system where the response of each electrode in presence of tea samples are multi dimensional combinations of different chemical compounds and represented by large number of measured points. It is expected that an ET system will examine and identify these signals precisely. Relevant feature extraction with appropriate signal processing of the responses generated by the electrode array may help to achieve this task of ET. In this work, a feature extraction method using singular value decomposition (SVD) method has been used to represent the ET signal before sending them to appropriate pattern classifier. The efficiency of the proposed method is verified on three types of ET data sets using support vector machine (SVM) classifiers. More than 98% of accuracy is obtained in all the three data sets which prove the efficacy of the proposed method.