Geographical Origin Identification of Goji Berries Using a Voltammetric Electronic Tongue and MVMD–MGRU/I-TCN Hybrid Model

Goji berries are a well-known medicinal and edible plant that has been consumed as both a health food and a traditional herbal medicine since ancient times. Rapid and accurate identification of their geographical origin is crucial for quality control and consumer protection. In this study, a fast and effective traceability method for Goji berry origin is proposed, based on a voltammetric electronic tongue(E-tongue) combined with a hybrid MVMD–MGRU/I-TCN (Multivariate Variational Mode Decomposition–Multi-layer Gated Recurrent Unit/Improved Temporal Convolutional Network) Hybrid model. Taste information from Goji berries collected across nine different production regions was obtained using the E-tongue, and a time-series modeling framework was established through the proposed hybrid architecture. The MVMD–MGRU branch captures dynamic temporal dependencies and time-domain feature representations, while the I-TCN branch enhances global temporal perception and local feature extraction through a pyramid convolution mechanism that expands the receptive field of dilated convolutions. The features extracted by both branches are adaptively fused to leverage their complementary strengths, and the fused representation is fed into a decision network for recognition and classification. Experimental results demonstrate that the proposed multi-branch feature fusion strategy outperforms single-modal feature extraction methods, achieving test accuracy, precision, recall, and F1-score of 99.26%, 99.31%, 99.26%, and 0.9926, respectively. This study provides a novel and efficient approach for the rapid identification of Goji berry origin and offers valuable insights into the traceability of other traditional Chinese medicinal materials.