A New Approach to Near-Infrared Spectral Data Analysis Using Independent Component Analysis

Abstract

This paper presents a new approach to near-infrared spectral (NIR) data analysis that is based on independent component analysis (ICA). The main advantage of the new method is that it is able to separate the spectra of the constituent components from the spectra of their mixtures. The separation is a blind operation, since the constituent components of mixtures can be unknown. The ICA based method is therefore particularly useful in identifying the unknown components in a mixture as well as in estimating their concentrations. The approach is introduced by reference to case studies and compared to other techniques for NIR analysis including principal component regression (PCR), multiple linear regression (MLR), and partial least squares (PLS) as well as Fourier and wavelet transforms.

Extraction of multiple pure component 1H and 13C NMR spectra from two mixtures: Novel solution obtained by sparse component analysis-based blind decomposition

“…Quantification and identification of the components present in the mixture is a traditional problem not only in NMR spectroscopy [3][4][5] but also in infrared (IR) spectroscopy [6,12], EPR spectroscopy [7,8], mass spectrometry [9,10,12], Raman spectroscopy [11], etc. Identification of the spectra of mixtures proceeds in majority of the cases by matching the mixture's spectra with a library reference compounds, [3,[6][7][8]. This approach is ineffective with the accuracy strongly dependent on the library's content of the pure component spectra.…”

“…ICA assumes that pure components are statistically independent and that at most one is normally distributed. The two requirements: to have more linearly independent mixtures than pure components and to have statistically independent pure components seem to be most critical for the success of the BSS approach to blind extraction of the pure components, [6,8,9,12].…”

“…Significant amount of efforts has been devoted to relax statistical independence assumption: (i) raw data preprocessing technique by first or second order derivative has been proposed in [6,12] to reduce level of statistical dependence among pure components. This technique belongs to the generalization of the ICA known as dependent component analysis (DCA), [16,[21][22][23]; (ii) an algorithm for blind decomposition of EPR spectra has been derived in [8] minimizing contrast function that exploits sparseness rather than statistical independence among the pure components; (iii) the so called mean field ICA has been proposed in [9] to cope with statistically dependent components.…”

Correction of Spectral Interference of Ethanol and Glucose for Rice Wine Evaluation

“…ICA has been applied to electroencephalograms (EEGs) for removing blink and muscle artifacts [11] and facial images for face recognition [12]. Recently, ICA has been applied to NIR and MIR spectra to decompose their spectra into statistically independent spectra [13], [14]. ICA is introduced to an NIR spectrum to decompose pure spectra from the original spectrum of the mixture [13].…”

“…Recently, ICA has been applied to NIR and MIR spectra to decompose their spectra into statistically independent spectra [13], [14]. ICA is introduced to an NIR spectrum to decompose pure spectra from the original spectrum of the mixture [13]. Independent component regression (ICR), a combination of ICA and MLR, has been applied to NIR spectra to quantitatively evaluate glucose concentrations from spectra of mixtures [14].…”