This paper describes the optimization of an
electronic nose (E-nose) equipped with metal-oxide gas sensors
and dedicated to continuous concentration monitoring of volatile
molecules. The optimization concerns particularly the selection of
more appropriate characteristic features coupled with
measurement conditions in order to minimize both the
measurement time and the gas sensor drifts. First, a promising
and fast feature corresponding to the maximum of the derivative
curve of the sensor time-response is explored. The performance
of this feature is demonstrated by comparison with a
conventional steady-state feature, especially regarding its
occurrence time, stability and sensitivity. Then the optimization
of the measurement time (delay between two successive
detections) has been illustrated and discussed. Optimized
operating conditions and feature were finally validated by using
non-supervised and supervised data mining analyses which show
robust concentration discrimination. This optimization work
constitutes an important step in real time applications for E-nose
users.