E-nose-assisted classification of Traditional Chinese Medicine using Adaptive Kalman Filtering

This study proposes an Adaptive Kalman Filtering (AKF)-based sensor data preprocessing method to enhance the accuracy and reliability of electronic nose (E-nose) technology for odor-based quality monitoring in agricultural cultivation of Traditional Chinese Medicinal (TCM) herbs. E-nose devices, inspired by the human olfactory system, are increasingly applied in precision agriculture to assess the quality and authenticity of medicinal herbs throughout cultivation. However, conventional denoising techniques such as moving average and wavelet transform show limitations when processing complex, dynamic odor data from TCM herbs under real agricultural conditions. To address this challenge, the AKF method dynamically adjusts filtering parameters, effectively reducing sensor noise and drift, thus significantly improving odor data stability and precision. Experimental results confirmed that AKF achieves superior performance in agricultural odor classification tasks compared to conventional Kalman Filtering and classical denoising methods. Specifically, AKF-preprocessed data reached 95.65% classification accuracy using Support Vector Machines (SVM), exceeding moving average and wavelet transform methods by 4.35% and 13.04%, respectively. Integrating AKF preprocessing with feature normalization and Particle Swarm Optimization (PSO) further improved the classification results. These findings demonstrate AKF’s promising application potential in TCM agricultural odor sensing, offering critical theoretical and practical advancements in precision medicinal agriculture and herbal quality control.