Details of Simulated Experiments

Each sensor is characterized by different value of RBaseiine, AR and x, where AR and X change also in presence of the different odorants. 

To explore the potentialities of this approach, two different kind of digital glomerular blocks were considered. They differ in the number of contemporaneous input spikes necessary to produce the same event at the output, the cases of two (G2) and four (G4) spikes were investigated first separately and subsequently together (G2h-G4) to understand the information content brought by the two blocks. 

In order to extract the features from the glomerular spiking signal, the output is divided in seven equally distributed intervals over the entire simulation time. The number of spikes occurring in each interval represents the features pattern of the array to the odorant exposure (Rospars et al. 2000). To evaluate the odour identification properties of the proposed signal processing architecture, a discriminant analysis of the feature patterns were performed. 

To realize the dataset for the data analysis the following characteristics were taken into consideration  

The performances of the proposed method are compared with the shift between the two steady resistance values (called DR) corresponding to the equilibrium in the cleaning and measuring phase. Each of two equilibrium values has been calculated in presence of noise as the mean value of the ten consecutive samples. 

---