中文摘要：

针对混合气体种类光谱识别中组分气体特征谱线重叠严重的问题，将支持向量机用于混合气体种类光谱识别中，提出了一种基于支持向量机二值分类识别模型的逐一混合气体种类识别方法。利用支持向量机的核函数变换，将特征谱线重叠严重的光谱在高维空间变换为线性可分后再逐一进行混合气体种类识别。在天然气气体种类识别实验中，比较了不同核函数、数据预处理、特征提取、训练样本数等条件与识别结果的关系，结果表明，方法对1％浓度以上的天然气组成气体的正确识别率大于97％，在理论和实际应用中具有重要的推广价值。

英文摘要：

In view of the the difficulty of mixture gas＇s characteristic absorption spectrum line overlapped, a novel seriatim mixture gas recognition method based on Support Vector Machine （SVM） two classes classification and recognition model was presented. Through transformation of the kernel function, multi-dimensional and overlapped spectrum data were mapped into the high dimension space, so that classification and recognition of mixture gas were carried out in the high dimension space of Support Vector. The method was applied in the experiment of natural gas components recognition, and the relation between recognition result and different kernel function, characteristic selection and number of training samples were respectively compared. The experiment shows that the recognition rate of component gas is more than 97% when component concentration of natural gas is not less than 1%.