Sensors and features selection for robust gas concentration evaluation

International audienceThis paper seeks to highlight the importance of the knowledge of metal oxide gas sensor behaviour before conceiving an electronic nose for a dedicated application. Therefore, a depth study of sensor response properties is needed for the selection of the more appropriate sensors via optimized measurement conditions and extracted features. Especially for continuous gas evaluation, the most important aspects to consider are the measurement time and the drift of the gas sensors. In this work, for fast recognition of pine oil vapour dilutions, the performance of two features are shown: the maximum of the derivative curve (Peak), an unusual feature which needs a very short gas exposure time, and the sensor amplitude voltage (Vs-V0) obtained at the end of the gas exposition phase. The performance of the new feature Peak, validated by Principal Component Analysis results, leads us to work with the shortest gas exposition and sensor regeneration times, and allows us to choose the best sensors according to our application