Synergistic Integration of Laser Oxidation and Long Short-Term Memory for Advanced Odor Classification in Next-Generation Artificial Olfactory Systems

Abstract

Emulating and enhancing human olfactory capabilities, artificial olfactory technology provides adept detection of subtle odors, gases, and various chemical substances. Metal oxide semiconductors (MOSs) are ideal materials for next-generation artificial olfactory devices due to their outstanding gas sensing performance, characterized high sensitivity, high response speed, and robust stability, as well as their compatibility with microfabrication. For broader applications, developing a comprehensive database of diverse odorants is crucial, which necessitates expanding the types of MOS channels in artificial olfactory devices. This paper reports a laser-induced oxidation-based artificial olfactory device using a 7 × 3 sensor array composed of three metal oxides (SnO2-x, ZnOx, and WO3-x). By analyzing the response pattern of various odorants using a deep neural network, the device achieved 95.2% accuracy in classifying eight single odor molecules. Additionally, it successfully deconvoluted the types and concentrations of two odor mixtures and classified ten types of wine with accuracies of 91.3% and 92.5%, respectively. Furthermore, this study identified the proper number and arrangement of sensors for next-generation e-nose development. Our innovative artificial olfactory system can be integrated into various fields, such as the aromatic industry and virtual reality, making it a beneficial technology for future artificial olfaction applications. © 2025 The Authors. Published by American Chemical Society.