Abstract

PARDO, M  and  SBERVEGLIERI, G. Sensors and data analysis for electronic nose. An. Asoc. Quím. Argent. [online]. 2005, vol.93, n.1-3, pp.89-98. ISSN 0365-0375.

The Pico Electronic Nose (EN), based on thin film semiconductor sensors, developed at the University of Brescia, is described. In particular we stress the advantages given by selecting the best features extracted from the response curves. We perform feature selection (FS) on an EN dataset composed of 30 features, obtained by extracting five diverse features from the response curves of six metal oxide sensors. We show that the performance (both classification error and PCA appearance) is always significantly better for the best features than for all thirty features. Results are not univocal regarding the best feature type. Yet, for three out of four datasets, in which the complete dataset can be decomposed, the features extracted over the sensor desorption lead to higher performances. The standard R/R₀ situates in the lower part of the ranking.