Abstract:
Commercial metal oxide semiconductor (MOS) sensors usually employed in electronic noses (e-noses) are well known for being low-cost, portability, and ease of use. However, these sensors can only identify a limited number of odors due to insufficient selectivity. Recent studies improved the selectivity by jointly integrating the MOS sensors with additional sensor sources. However, the published hybrid systems involved complex fabrication and measurement procedures. On the contrary, this work utilizes paper-based colorimetric sensors which are simpler and easier to use. This proposed hybrid system consists of 8 commercial metal oxide sensors and a paper-based colorimetric sensor coated with phenol red, methyl red, and methylene blue. Six volatile organic compounds (VOCs) are classified using this developed hybrid system. Each sensor system is compared with the hybrid system using principal component analysis (PCA) and hierarchical clustering analysis (HCA). It was found that the metal oxide sensors alone can identify 5 VOCs, while the colorimetric sensors can identify 2 VOCs at best. Finally, the hybrid system can discriminate all the 6 target VOCs based on 12 features selected by ANOVA (Analysis of Variance) feature selection coupled with support vector machine (SVM) classfier.
