Origin identification of Angelica dahurica using a bidirectional mixing network combined with an electronic nose system

Abstract

The medicinal value of Angelica dahurica is closely related to its origin. Variations in climate, soil, altitude, and other ecological factors across different origins can lead to significant differences in the quality of Angelica dahurica, and high-quality products are often subject to counterfeiting. To provide a rapid and effective method for quality identification, this paper proposes a Bidirectional Mixing Network (BM-Net) combined with an electronic nose (e-nose) system to distinguish Angelica dahurica from various origins. The e-nose system collects gas information from Angelica dahurica from four different origins with a wide range and four different origins with a small range. A Bidirectional Mixing Module (BMM) is proposed to adaptively calculation both local and global gas features from the time-series and sensor dimensions, with residual connection incorporated to enhance feature representation. Based on the BMM, the BM-Net is designed for effective identification of gas information from Angelica dahurica across different origins. The effectiveness of BM-Net is validated through ablation analysis and comparison with state-of-the-art gas information classification methods. For the gas information dataset of Angelica dahurica from a wide range of origins, BM-Net achieves an accuracy of 97.75 %, a precision of 97.64 %, and a recall of 97.94 %. For the dataset of Angelica dahurica from a small range of origins, BM-Net achieves an accuracy of 96.08 %, a precision of 96.60 %, and a recall of 96.05 %.