基于支持向量机的羰基化合物红外光谱研究  

Characterization of Carbonyl Compounds from Their Infrared Spectra Based on Support Vector Machines

作　　者：谢微[1] 李光明[1] 陆敏春[1] 聂伏生[1] 李梦龙[1] 

机构地区：[1]四川大学化学学院,成都610064

出　　处：《分析化学》2006年第U09期113-117,共5页Chinese Journal of Analytical Chemistry

基　　金：国家自然科学基金资助项目(No.29877016)

摘　　要：设计了一个4层分等级分类系统,通过支持向量机技术对羰基化合物的红外光谱成功地进行了信息分类和提取,并与人工神经网络进行比较,结果表明:支持向量机对羰基类化合物红外光谱分类效果优于人工神经网络;另外详细探讨了表征酰胺类化合物的特征光谱峰片段对识别伯仲叔酰胺的影响,证明了酰胺的N-H伸缩振动峰对识别伯仲叔酰胺贡献最大,为构建红外光谱智能解析系统提供了定量依据。A recently actively used pattern recognition method, support vector machine （SVM） was introduced to build classifiers for a 4-level hierarchical classification structure of carbonyl compounds from their infrared spectra successfully. Results were compared favorably with those obtained by using artificial neural networks （ANNs） methods. The effects of the amides＇ segmental spectra of characteristic frequency upon the identification of primary, secondary or tertiary substitute are discussed. The results testify that the N-H stretching absorption of amides contributes most to identification. Also, the quantitative foundation was provided for the establishment of infrared intelligent interpretation system.

关 键 词：红外光谱 支持向量机 人工神经网络 信息提取 羰基 酰胺 

分 类 号：O657.33[理学—分析化学]