文献类型：期刊文献

英文题名：Robust Odor Detection in Electronic Nose Using Transfer-Learning Powered Scentformer Model

作者：Ni, Wangze[1,2];Wang, Tao[3];Wu, Yu[4];Chen, Lechen[1,2];Zeng, Min[1];Yang, Jianhua[1,2];Hu, Nantao[1,2];Zhang, Bowei[3];Xuan, Fuzhen[3];Yang, Zhi[1]

机构：[1]Shanghai Jiao Tong Univ, Natl Key Lab Adv Micro & Nano Manufacture Technol, Shanghai 200240, Peoples R China;[2]Shanghai Jiao Tong Univ, Sch Elect Informat & Elect Engn, Dept Micro Nano Elect, Shanghai 200240, Peoples R China;[3]East China Univ Sci & Technol, Sch Mech & Power Engn, Shanghai Key Lab Intelligent Sensing & Detect Tech, Shanghai 200237, Peoples R China;[4]Shanghai Marine Diesel Engine Res Inst, Natl Key Lab Marine Engine Sci & Technol, Shanghai 201108, Peoples R China

年份：2025

卷号：10

期号：5

起止页码：3704

外文期刊名：ACS SENSORS

收录：;EI(收录号：20252218492360);WOS:【SCI-EXPANDED(收录号:WOS:001489087700001)】；

基金：This work was supported by the National Key Research and Development Program of China (2022YFB3205500), the National Natural Science Foundation of China (62371299, 62301314, and 62471298), the Science Fund for Creative Research Groups of the National Natural Science Foundation of China (52321002), and the Natural Science Foundation of Shanghai (23ZR1430100). We also acknowledge analysis support from the Instrumental Analysis Center of Shanghai Jiao Tong University and the Center for Advanced Electronic Materials and Devices of Shanghai Jiao Tong University. The computations in this paper were run on the pi 2.0 cluster supported by the Center for High Performance Computing at Shanghai Jiao Tong University.

语种：英文

外文关键词：electronic nose; odor detection; convolutionalneural network; multihead attention; transfer learning

摘要：Mimicking the olfactory system of humans, the use of electronic noses (E-noses) for the detection of odors in nature has become a hot research topic. This study presents a novel E-nose based on deep learning architecture called Scentformer, which addresses the limitations of the current E-nose like a narrow detection range and limited generalizability across different scenarios. Armed with a self-adaptive data down-sampling method, the E-nose is capable of detecting 55 different natural odors with the classification accuracy of 99.94%, and the model embedded in the E-nose is analyzed using Shapley Additive exPlanations analysis, providing a quantitative interpretation of the E-nose performance. Furthermore, leveraging Scentformer's transfer learning ability, the E-nose efficiently adapts to new odors and gases. Rather than retraining all layers of the model on the new odor data set, only the fully connected layers need to be trained for the pretrained model. Using only 1 parts per thousand data of the retrained model, the pretrained model-based E-nose can also achieve classification accuracies of 99.14% across various odor and gas concentrations. This provides a robust approach to the detection of diverse direct current signals in real-world applications.