A FFT Preprocessing Algorithm to Improve the Reproducibility of Gas Sensors Signals

In a conducting polymers electronic nose, the conductivity of chemical sensors is modified by adsorption-desorption interactions between odorant molecules and sensing material. These phenomena are not always reversible and this fact could reduce the reproducibility of sensor response in multiple experiments. A pre-processing algorithm was developed able to apply a frequency analysis technique to sensor responses and its effectiveness was validated in the classification of simple substances such as acetone, n-butanol, and o-xylene with repeated tests performed on different days. In the classification phase a clear improvement was obtained, with an increase of the distance among the clusters in the principal components space and a negligible rise of computing time.