Chemical sensors clustering with the dynamic moments approach

One of the most important step in pattern recognition is the definition of features describing the entities that have to be assigned to classes. From the point of view of chemical sensors, the extraction of features consists in the selection of some characteristics of the temporal sequence of sensor signal taking place during the interaction between the sensors and the compound present in the environment. s paper, the morphological descriptors of the sensor trajectory in the phase space are introduced and investigated. The performance of these new features have been compared with the usual features in three typical electronic nose applications.