Approximating Sensors' Responses of Odor Mixture on Machine Olfaction

Abstract

An increasing interest in current research on machine olfaction is to try to approximate or predict the sensor response to odor mixtures. Previously, the aid of special active odor sensing system was proposed. This system is able to produce the target odor recipe based on iteratively adjusting the ingredient from odor palette. Here, a new algorithm solution is proposed by combining the signal decomposition and reconstruction techniques, and Support Vector Machine (SVM). The prediction results of the proposed method are investigated by comparing with the real sensor responses recorded from a commercial e-nose machine. The results demonstrate that the new proposed method provides good approximation when applied to different mixing ratios of some coffees and green tea.