[摘要]：The desorption part of MOX's response curves to gases represents the dynamic information of gas desorption and oxide recovery on the sensitive layer. It contains abundance information and lacks to be well investigated. This paper proposes a feature extraction method from the desorption part of response curves in phase-space. The method extracts only one feature, min (dS(t)/dt)/ max(S-t), which is found to be able to represent the recovery profile. This feature could be extracted in the front part of the recovery process. Besides, it is also found that this feature does not change with the reacting time (of the reaction between a sensor and a sample) changing. A dataset was use in the paper, which was consisted of 189 measurements of 9 VOCs by 5 TGS gas sensors at three concentrations. A common feature extraction method that extracted integral in phase-space was used as a comparison in sort classification with the proposed method. The correct recognition rates of Fisher Discriminant Analysis (FDA) with the proposed method and the phase-space integral extracting method were 91.5% and 79.9%, respectively. The results show that the proposed method is an efficient feature extraction method of the recovery profile of response curves.
