Qualitative and Quantitative Identification of Components in Mixture by Terahertz Spectroscopy

• Published in: IEEE transactions on terahertz science and technology Vol. 8; no. 6; pp. 696 - 701
• Main Authors: Peng, Yan, Shi, Chenjun, Xu, Mingqian, Kou, Tianyi, Wu, Xu, Song, Bin, Ma, Hongyun, Guo, Shiwei, Liu, Lizhuang, Zhu, Yiming
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.11.2018; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Absorption; Artificial neural networks; Brain; Correlation coefficient; Correlation coefficients; Data models; Identification; Identification methods; Mathematical model; Mixture identification; Neural networks; Regression analysis; Spectroscopy; Spectrum analysis; Support vector machines; terahertz (THz) spectroscopy; Wavelet transforms
• ISSN: 2156-342X; 2156-3446
• DOI: 10.1109/TTHZ.2018.2867816

Typical methods for the analysis of mixture components include multiple linear regression, partial linear squares, and artificial neural network. However, these methods need large amount of samples and time to improve recognition accuracy. In this paper, based on the data obtained from terahertz spectroscopy, an identification method with less sample requirements and lower calculation time but higher accuracy is proposed. Based on the wavelet transform, baseline elimination, support vector regression, and loop iteration of samples, the specific substance in the mixture can be identified effectively. For example, seven substances that exist in brain glioma are chosen as the components of a mixture, where the key substances used for glioma diagnosis are set as the target substances and the spectra of mixtures with different mix proportions serve as training data. The average correlation coefficient of identification achieves 99.135% and the root-mean-square error is 0.40%. These results have profound implications for the eventual practical application of exact qualitative and quantitative identification of components in mixtures.