To check the created sensor signal patterns from the point of view of recognition ability of the electronic nose, we have
investigated the distribution of the measurement points of all sensors in the multidimensional space. To inspect it in a graphical way, we have mapped the results of the measurements originally made in the 12-coordinate system (12 sensors) into 3-D space, formed by the three most important principal components in principal component analysis (PCA) [18], [19].
Fig. 6 presents visually the distribution of all 100 measured signals at the succeeding time instants (the dynamic mode) for
pure robusta, pure arabica, the 50% mixture of robusta and arabica, the 90% mixture of robusta and 10% of arabica, and the 10% mixture of robusta and 90% of arabica. The first principal component (PC1) represents 94.2% of the variance of the data.
The next components are responsible for 3.7% (PC2) and 1.3% (PC3) of the total variance of the data.