Silhouette-Based Selectivity Analysis of a Staircase-Modulated Electronic Nose

In this paper, we show that we are able to enhance for selectivity of an electronic nose, through a staircase temperature modulation approach. This improvement was evaluated using a novel method based on Silhouette analysis. This fully reflects both the intra and inter cluster differences of samples compared with traditional methods. Additionally, the Silhouette method capable of assessing the selectivity of both the entire system and individual samples. An array of 30 MOX sensors modulated from 200 - 500°C was tested with 11 chemicals at three different concentrations. Experimental results show that the staircase temperature modulation significantly improved the selectivity of the electronic nose under both unsupervised and supervised conditions. The average Silhouettes of the original data and the one after LDA (linear discriminant analysis) increased from 0.13 and 0.94 to 0.32 and 0.99, respectively after the modulation.