Biomimetic Enhancement of Robustness in E-Nose Systems: Addressing Noise and Sensor Damage

This study aims to enhance the robustness of electronic nose (e-nose) against noise interference and sensor damage through a biomimetic approach. Inspired by the structure and biological mechanisms of the mammalian olfactory system, we have proposed a novel biomimetic model for processing the signals from the gas-sensitive sensor arrays of the e-nose. This model simulates the functions of the olfactory epithelium and olfactory bulb, encoding sensor signals into spike sequences, and incorporates the concept of wavelet filter banks from wavelet scattering networks, effectively capturing local variations in the signals. Utilizing a laboratory-developed e-nose system, we collected two datasets for validation. The first dataset comprised volatile samples from five traditional Chinese medicines, testing the capability to identify complex odors. The second dataset included 14 common volatile organic compounds (VOCs), such as ketones and aldehydes, to assess the recognition performance for individual volatiles. Experimental results demonstrate that the proposed biomimetic approach improves the accuracy and reliability of the e-nose in complex and unstable environments.