LAUSR

Abstract This paper demonstrates a novel approach for qualitative analysis of different lemon slices employing a large-amplitude pulse Voltammetric electronic tongue. Date preprocessing methods including Principal Component Analysis (PCA) and Discrete Wavelet Transform (DWT) were provided. Then Linear Discriminant Analysis (LDA) was used to compare the compression effect. According to the result of Linear Discriminant Analysis (LDA), the DWT was selected as the feature extraction method. Then Extreme Learning Machine (ELM) was used to qualitatively analyze different lemon slices and compare the result with the common classification model: Random Forest (RF) and Support Vector Machine (SVM). The models were compared according to the accuracy rate of training set, the accuracy rate of testing set and Kappa statistic (K). The results show that ELM has much better performance than other models in distinguishing the quality of lemon slices.