Identification of CO, H_2, C_2H_2 and their mixtures by wavelet transformation and Support Vector Machine

摘要

Gas identification is widely applied in agriculture and industrial fields. Temperature Modulated Dynamic Test (TMDT) is often used to improve the selectivity and stability of semiconductor gas sensors in gas identification field. But the keys of obtaining accurate and stable results in mixture gases identification are effective feature extraction from the sensor's response. In this paper, we present a new strategy for identification of hydrogen (H_2), carbon monoxide (CO), ethyne (C_2H_2) and their mixtures in air using a single gas sensor. Extracting features using wavelet decomposition transformation (DWT) from dynamic response combined with Support Vector Machine (SVM) pattern recognition method are introduced here. Experiment result shows that the proposed strategy can perform well in discrimination of CO, H_2, C_2H_2 and their mixtures. In addition, the structure of distributed SVM network introduced here outperforms a single SVM.