An Adaptive Drift Compensation Method Based on Integrated Dual-Channel Feature Fusion for Electronic Noses

Long-term accurate qualitative and quantitative analysis based on gas sensor arrays is a critical issue in the application of electronic nose systems, as the instability of gas sensors poses a great challenge. The instability is mainly manifested as long-term or short-term drift, and the existence of multidimensional correlation and noise interference among the raw data of the array increases the complexity. To address this issue, an adaptive gas sensor array drift compensation method based on integrated dual-channel feature fusion (IDCF) is proposed to improve the long-term accuracy of electronic nose detection in this article. It automatically learns important features through a neural network, splits the input feature channels according to their weights, and extracts important and supplementary information using different sizes of convolution kernels, respectively, to adequately capture the representative features of the gas sensor drift data. A soft threshold function is used as a nonlinear transformation layer embedded in the neural network to further eliminate or compensate the features associated with drift or interference for the purpose of improving detection stability and accuracy. The results show that the proposed method improves the prediction accuracy of gas classification by more than 10% compared with typical relative methods, effectively compensating for the effects of drift.