基于支持张量机的多维光谱定量分析方法  

Multi-way Spectral Analysis Methods Based on Support Tensor Machines

作　　者：杜树新[1] 李林军[1] 

机构地区：[1]浙江大学宁波理工学院信息科学与工程分院,浙江宁波315100

出　　处：《发光学报》2013年第4期523-528,共6页Chinese Journal of Luminescence

基　　金：国家自然科学基金(60974111);国家"863"计划(2009AA04Z123)资助项目

摘　　要：现代光谱仪器强大的多维光谱数据产生能力使得多维光谱数据定量分析方法成为迫切需要研究的课题。针对多维光谱定量分析中的多维光谱数据以张量模式表达的特点,研究了基于支持张量机的多维光谱定量分析方法。该方法保留了多维光谱数据所固有的结构信息及数据的内在相关性,减少了模型中的待定模型参数,也克服了平行因子法、多维偏最小二乘等方法中需要预估组分数的缺点。对水体中化学耗氧量和总有机碳的检测进行了实验检验。实验结果表明:与现有的多维光谱定量分析方法比较,本方法提高了校正模型性能,并且模型对需预先确定的参数C和ε的变化不敏感。The powerful ability of generating multi-way spectral data in modern spectrum instruments makes multi-way spectral quantitative analysis method become one of the important topics. Since the multi-way spectrometry is represented as a tensor, a new multi-way spectral quantitative analysis method based on support tensor machines （STM） is presented. The presented methods preserve the intrinsic structure of the multi-way spectrometry. Due to the reduction of model parameter number, the methods represented by tensor reduce the over-fit problem in the case of small samples. Unlike the conventional methods such as parallel factor analysis （PARAFAC） and multi-way partial least squares （N-PLS）, the presented methods do not need to estimate the component number. The ex- periments for detecting chemical oxygen demand （COD） and total organic carbon （TOC） in water are carried out. The experimental results show that the presented methods are improved with compar- ison to the existing multi-way spectral quantitative analysis methods, and the models are not sensitive to the pre-determined parameters C and ε.

关 键 词：多维光谱 定量分析 支持张量机 校正模型 

分 类 号：O657.3[理学—分析化学]