基于SVM回归模型的混合气体组分种类光谱识别方法  

A Method of Mixed Gas Component Infrared Spectrum Recognition Based on SVM Regression Model

作　　者：白鹏[1] 王建华[2] 王宏柯[3] 张发启[3] 刘君华[1] 

机构地区：[1]西安交通大学电气工程学院,西安710049 [2]空军工程大学理学院,西安710051 [3]空军工程大学工程学院,西安710038

出　　处：《光子学报》2008年第4期754-757,共4页Acta Photonica Sinica

基　　金：国家自然科学基金(60276037,60772016)资助

摘　　要：针对混合气体红外光谱分析中无法采用同一模型同时进行混合气体组分浓度的定量分析和组分种类的定性分析的问题,本文提出了基于SVM回归模型的混合气体组分种类光谱识别方法.通过详细推导,证明混合气体组分种类识别完全可以通过组分浓度分析的SVM回归模型来求解,混合气体组分种类识别是一种特殊的回归.实验结果显示,该方法的混合气体组分种类的正确识别率不小于92.5%.In view of that spectrum quantitative analysis of component concentration, and qualitative analysis of component recognition can not be carried out using the same model simultaneously during infrared spectrum analysis of the mixed gas,a method of mixed gas component infrared spectrum recognition based on SVM regression model is proposed. The detailed derivation proves that the component recognition of mixed gas can be solved by the SVM regression model of component concentration analysis, and the component recognition of mixed gas is one special regression. The experimental result shows that, the correct rate of component recognition is not less than 92.5%.

关 键 词：支持向量机 回归模型 识别 红外光谱 

分 类 号：TH744.4[机械工程—光学工程]