Feature Extraction and Data Modeling of Multi-Frequency Electronic Tongue Signals for Monitoring the Processing Stages of Ginger-Processed Pinellia ternata (Zhejiang)

The processing of ginger-processed Pinellia ternata (Zhejiang) has long relied on empirical judgment, lacking objective and real-time monitoring methods. This study introduces an intelligent framework that combines a multi-frequency electronic tongue with chemometric modeling—including principal component analysis–discrimination index (PCA–DI) and wrapper-based support vector machine (SVM) classification—for dynamic process monitoring. Taste-response signals were systematically collected from key processing, water-leaching, and pickling stages. PCA–DI analysis demonstrated clear separability among seven key processing nodes (DI = 93.77%). Notably, samples from days 2 and 3 of water-leaching showed high similarity, suggesting an optimal soaking duration, while a marked transition on pickling day 6 indicated a critical transformation point. The wrapper–SVM models achieved high classification accuracies of 95.51% for key nodes, 100% for water-leaching, and 89.32% for pickling. These findings demonstrate that integrating electronic tongue sensing with machine learning effectively captures dynamic quality variations, offering a robust and objective strategy for the standardization and optimization of traditional medicine processing.