Efficient Recognition of Volatile Organic Compounds Using Low-Cost e-Nose Device and Random Forest Classification

This paper introduces a novel method for the efficient recognition of volatile organic compounds (VOCs) using an electronic nose (e-nose) device coupled with advanced signal processing and classification techniques. The proposed e-nose is a cost-effective sensor array designed to detect a broad spectrum of VOCs with applications in several domains including industrial safety. Data collected from the sensors are processed using the Filter Diagonalization Method (FDM), a novel non-Fourier spectral estimation algorithm that provides reliable footprints from the VOCs. The classification stage employs a Random Forest (RF) model to analyze the FDM footprints and predict the VOC present in the e-nose’s surroundings. Experimental results show that the FDM-RF approach achieves a 96.4% classification accuracy.