A Novel Bionic Olfactory Network Combined with an Electronic Nose for Identification of Industrial Exhaust

Jia Yan,Huisheng Zhang,Xinran Ge,Wenzheng Yang,Xiaoyan Peng,Tao Liu
DOI: https://doi.org/10.1016/j.microc.2024.110287
IF: 5.304
2024-01-01
Microchemical Journal
Abstract:Traditional electronic nose (E-nose) signal processing methods and classification techniques are often cumbersome, heavily dependent on the experience of the implementers and poor in biological interpretability. In view of the above problems, a novel bionic olfactory network is proposed for processing E-nose data for industrial exhaust identification. First, the bionic olfactory network (BON) consists of a bionic olfactory bulb system (BOBS) and a bionic olfactory cortex (BOC). The BOBS highly simulates the odor information conduction function of the mammalian olfactory bulb and is composed of a novel bionic olfactory bulb model and recurrence quantification analysis (RQA), which can transform the original sensor responses into neuron spike sequences and extract features from the sequences, simplifying traditional data processing steps. The BOC is a convolutional spiking neural network that primarily imitates the odor information recognition function of the olfactory cortex in the mammalian olfactory system. The whole BON exhibits excellent biological interpretability due to its unique mode of information transmission. Finally, the experimental data of ten types of industrial exhaust were collected by using a self-built E-nose system. The experimental results indicate that the classification accuracy of the BON is 94.4%, which is much better than that of the combination of traditional signal processing methods and classification techniques. An even better recognition accuracy of 79.2% can be obtained in small sample training, which is still the best performance.
What problem does this paper attempt to address?