Voltammetric e-Tongue Based on a Single Sensor and Variable Selection for the Classification of Teas

A Two-Step Simulated Annealing Algorithm for Spectral Data Feature Extraction

To address the shortcomings in many traditional spectral feature extraction algorithms in practical application of low modeling accuracy and poor stability, this paper introduces the "Boruta algorithm-based local optimization process" based on the traditional simulated annealing algorithm and proposes the "two-step simulated annealing algorithm (TSSA)". This algorithm combines global optimization and local optimization. The Boruta algorithm ensures that the feature extraction results are all strongly correlated with the dependent variable, reducing data redundancy. The accuracy and stability of the algorithm model are significantly improved. The experimental results show that compared with the traditional feature extraction method, the accuracy indexes of the inversion model established by using the TSSA algorithm for feature extraction were significantly improved, with the determination coefficient R² of 0.9654, the root mean square error (RMSE) of 3.6723 μg/L, and the mean absolute error (MAE) of 3.1461 μg/L.

Electronic Tongue Recognition with Feature Specificity Enhancement

As a kind of intelligent instrument, an electronic tongue (E-tongue) realizes liquid analysis with an electrode-sensor array and certain machine learning methods. The large amplitude pulse voltammetry (LAPV) is a regular E-tongue type that prefers to collect a large amount of response data at a high sampling frequency within a short time. Therefore, a fast and effective feature extraction method is necessary for machine learning methods. Considering the fact that massive common-mode components (high correlated signals) in the sensor-array responses would depress the recognition performance of the machine learning models, we have proposed an alternative feature extraction method named feature specificity enhancement (FSE) for feature specificity enhancement and feature dimension reduction. The proposed FSE method highlights the specificity signals by eliminating the common mode signals on paired sensor responses. Meanwhile, the radial basis function is utilized to project the original features into a nonlinear space. Furthermore, we selected the kernel extreme learning machine (KELM) as the recognition part owing to its fast speed and excellent flexibility. Two datasets from LAPV E-tongues have been adopted for the evaluation of the machine-learning models. One is collected by a designed E-tongue for beverage identification and the other one is a public benchmark. For performance comparison, we introduced several machine-learning models consisting of different combinations of feature extraction and recognition methods. The experimental results show that the proposed FSE coupled with KELM demonstrates obvious superiority to other models in accuracy, time consumption and memory cost. Additionally, low parameter sensitivity of the proposed model has been demonstrated as well.

Identification of Similar Chinese Congou Black Teas Using an Electronic Tongue Combined with Pattern Recognition

It is very difficult for humans to distinguish between two kinds of black tea obtained with similar processing technology. In this paper, an electronic tongue was used to discriminate samples of seven different grades of two types of Chinese Congou black tea. The type of black tea was identified by principal component analysis and discriminant analysis. The latter showed better results. The samples of the two types of black tea distributed on the two sides of the region graph were obtained from discriminant analysis, according to tea type. For grade discrimination, we determined grade prediction models for each tea type by partial least-squares analysis; the coefficients of determination of the prediction models were both above 0.95. Discriminant analysis separated each sample in region graph depending on its grade and displayed a classification accuracy of 98.20% by cross-validation. The back-propagation neural network showed that the grade prediction accuracy for all samples was 95.00%. Discriminant analysis could successfully distinguish tea types and grades. As a complement, the models of the biochemical components of tea and electronic tongue by support vector machine showed good prediction results. Therefore, the electronic tongue is a useful tool for Congou black tea classification.