拉曼光谱结合稀疏非负最小二乘算法用于混合物组分识别

引用本文: 颜凡, 朱启兵, 黄敏, 刘财政, 张丽文, 张恒. 拉曼光谱结合稀疏非负最小二乘算法用于混合物组分识别[J]. 分析化学, 2020, 48(2): 298-305. doi: 10.19756/j.issn.0253-3820.191256 shu

Citation: YAN Fan, ZHU Qi-Bing, HUANG Min, LIU Cai-Zheng, ZHANG Li-Wen, ZHANG Heng. Identification of Mixture Components Using Sparse Non-Negative Least Squares Algorithm Base on Raman Spectroscopy[J]. Chinese Journal of Analytical Chemistry, 2020, 48(2): 298-305. doi: 10.19756/j.issn.0253-3820.191256 shu

拉曼光谱结合稀疏非负最小二乘算法用于混合物组分识别

颜凡 ^1, ,
朱启兵 ^1, ,
黄敏 ^1, ,
刘财政 ^1, ,
张丽文 ^2, ,
张恒 ^2,

1.
江南大学轻工过程先进控制教育部重点实验室, 无锡 214122;
2.
北京卓立汉光仪器有限公司, 北京 101102
基金项目:
本文系国家自然科学基金项目（No.61775086）资助

摘要: 拉曼光谱数据含有与被测物质组分相对应的指纹谱信息，是混合物组分识别的有效方法。传统的拉曼光谱法用于混合物组分检测时，存在光谱特征提取困难、搜索比对算法性能容易受数据库大小影响、识别精度难以保证等问题。针对此问题，本研究提出了一种基于稀疏非负最小二乘算法的混合物组分拉曼光谱识别方法。本方法将待识别的混合物光谱数据看作是各种纯净物光谱数据的线性表示；考虑到混合物组分数量相对于数据库中纯净物数量具有稀疏特性，利用稀疏最小二乘算法获得混合物光谱在纯净物光谱数据中的线性表示系数；并根据统计学中的2δ准则确定疑似组分；在此基础上，利用迭代最小二乘算法并结合T检验方法，实现混合物组分的最终识别。本研究基于自建的500种纯净物拉曼光谱数据库，对组分等体积比混合的19个混合物样本和不同体积比的81个样本进行了组分识别。结果表明，在等体积比情况下，本算法的查准率为90.24%，查全率为88.10%；对于不同体积比的混合物样本，整体查准率为93.22%，查全率为83.65%，表明此算法具有良好的稳定性和准确度。

关键词:

English

Identification of Mixture Components Using Sparse Non-Negative Least Squares Algorithm Base on Raman Spectroscopy

1.
Key Laboratory of Advanced Process Control for Light Industy, Ministry of Education, Jiangnan University, Wuxi 214122, China;
2.
Beijing Zhuoli Hanguang Instrument Co. Ltd., Beijing 101102, China
Received Date: 06 May 2019
Revised Date: 15 November 2019
Fund Project:
This work was supported by the National Natural Science Foundation of China (No. 61775086).

Abstract: Raman spectral data contain fingerprint spectral information corresponding to the components of the measured substances, which is an effective method to identify the components of mixtures. The traditional mixture component detection methods based on Raman spectrum have some issues such as the difficulty in extracting spectral features, the performance of search and peak matching algorithms is easily affected by the database, and the recognition accuracy is difficult to guarantee. To overcome these problems, a method of mixture components recognition using Raman spectrum based on sparse non-negative least squares algorithm is proposed. In this method, the spectral data of the mixture to be recognized is regarded as the linear representation of the spectral data of all kinds of pure substances. Considering the sparse characteristic of the mixture components quantity relative to the pure substance quantity in the database, the linear representation coefficient of the mixture spectrum in the pure substance spectrum data is obtained by the sparse non-negative least squares algorithm. And the suspected components are determined according to the statistical 2δ principle. On this basis, the iterative least squares algorithm combined with the T-distribution test method is used to realize the final identification of the mixture components. The Raman spectral data of 500 pure substances are used to build a standard database for identification of the experimental sample of 19 mixtures with equal volume ratio of components and 81 mixtures with different volume ratio of components. The results show that the precision is 90.24% and the recall is 88.10% under the condition of equal volume ratios, and the precision is 93.22% and the recall is 83.65% under the condition of different volume ratios, which proves the good stability and accuracy of the proposed algorithm.

Key words:

Identification of mixture components
/ Raman spectroscopy
/ Sparse non-negative least squares algorithm

1. [1]
  LIU Bin, LIU Yun-Hu. Food Science and Technology,2018,43(6):317-321 刘彬, 刘云虎.食品科技,2018,43(6):317-321
2. [2]
  WU Li-Sa, ZHAO Ming-Yue, GE Chang, CAI He-Qing, JI Ling-Bo, HU You-Chi, HU Jun. Tobacco Science & Technology,2018,51(4):46-52 吴丽洒, 赵明月, 葛畅, 蔡何青, 姬凌波, 胡有持, 胡军.烟草科技,2018,51(4):46-52
3. [3]
  Stöckel S, Kirchhoff J, Neugebauer U, Rösch P, Popp J. J. Raman Spectrosc.,2016,47(1):89-109
4. [4]
  Penido C A F D, Pacheco M T T, Lednev I K, Silveira L. J. Raman Spectrosc.,2016,47(1):28-38
5. [5]
  Li Y S, Church J S. J. Food Drug Anal.,2014,22(1):29-48
6. [6]
  Shashilov V A, Lednev I K. Chem. Rev.,2010,110(10):5692-5713
7. [7]
  Vignesh T, Shanmukh S, Yarra M, Botonjic-Sehic E, Grassi J, Boudries H, Dasaratha S. Appl. Spectrosc.,2012,66(3):334-340
8. [8]
  MA Jing. Spectroscopy and Spectral Analysis,2014,34(7):1865-1868 马靖.光谱学与光谱分析,2014,34(7):1865-1868
9. [9]
  Zhang Z M, Chen X Q, Lu H M, Liang Y Z, Fan W, Xu D, Zhou J, Ye F, Yang Z Y. Chemometr. Intell. Lab. Syst., 2014,137:10-20
10. [10]
  HUANG Pei-Xian, YAO Zhi-Xiang, SU Hui, SUN Kuo. Journal of Instrumental Analysis,2013,32(3):281-286 黄培贤, 姚志湘, 粟晖, 孙阔.分析测试学报,2013,32(3):281-286
11. [11]
  Fan X, Ming W, Zeng H, Zhang Z M, Lu H M. Analyst,2019,144(5):1789-1798
12. [12]
  HU Zhi-Yu, WANG Qiang. Journal of Test and Measurement Technology,2016,30(5):400-405 胡志裕, 王强.测试技术学报,2016,30(5):400-405
13. [13]
  Koh K. Stanford University,2009:59-66
14. [14]
  Zhang Z M, Chen S, Liang Y Z, Liu Z X, Zhang Q M, Ding L X, Ye F, Zhou H. J. Raman Spectrosc.,2010,41(6):659-669

扫一扫看文章

计量

PDF下载量: 0
文章访问数: 101
HTML全文浏览量: 3

通讯作者: 陈斌, bchen63@163.com

1.
沈阳化工大学材料科学与工程学院沈阳 110142

拉曼光谱结合稀疏非负最小二乘算法用于混合物组分识别

English

Identification of Mixture Components Using Sparse Non-Negative Least Squares Algorithm Base on Raman Spectroscopy

计量

文章相关

通讯作者: 陈斌, bchen63@163.com

中国化学会秘书处