2.3. Data acquisition and multivariate data analysis
The instrument remote control and data acquisition was per-
formed by an external personal computer through the standard
communication port RS232. The software, named “Nose Pattern
Editor”, was used for data handling, pre-processing (classical fea-
ture) and explorative analysis (Principal components analysis,
PCA). This software was implemented by Sacmi.
The response of the sensors yields an exponential-like shape
but not all this information was used. In this case, only the classical
feature (R/R0) was extracted, where R0 was the initial resistance of
the sensor balanced in air and R was the resistance of the sensor in
the presence of the volatile compounds emitted from the perfumes
(which decreases with respect to R0).
In order to make a discrimination study, statistic multivariate
analysis was carried out after the classical feature selection by
applying PCA [16]. This is an unsupervised learning technique that
allows reduction of multidimensional data to a lower-dimensionalapproximation, thus simplifying the interpretation of the data by
displaying only the first and second principal components (PC1 and
PC2) in two dimensions while preserving most of the variance in
the data.