THz Spectroscopic Decomposition and Analysis in Mixture Inspection Using Soft Modeling Methods

Abstract

At present, the terahertz time-domain spectroscopic information of each component in multi-composition compounds detection is comprehensively combined. The lower resolution level of mixed spectra has posed many difficulties in signal analysis due to the overlapping characteristic information in the mixed ones. In this paper, to compare the performances of Nonnegative Matrix Factorization (NMF), Self-modeling Mixture Analysis (SMMA) and Multivariate Curve Resolution-Alternating Least Squares (MCR-ALS) on complex systems, a binary mixture and a ternary mixture are employed during THz-TDS testing. The position of the absorption peak (PK) and the correlation coefficient are used to evaluate the decomposition effects. The experimental results show that the component spectra resolved by MCR-ALS demonstrate good consistency with the sample components. Further, MCR-ALS presents excellent results in comparison with NMF and SMMA in terms of decomposing precision and computing speed. MCR-ALS thus appears a promising algorithm to resolve the THz multi-way data, and may be useful for many applications, such as medicine quality assurance and unknown component identification in the general area of terahertz science and technology.