A local pre-processing method for near-infrared spectra, combined with spectral segmentation and standard normal variate transformation

Pre-processing of near-infrared (NIR) spectral data has become a necessary part of chemometrics modeling and is widely used in many practical applications. The objective of the pre-processing is to remove physical phenomena in the spectra in order to improve subsequent qualitative or quantitative an...

Full description

Saved in:
Bibliographic Details
Published in:Analytica chimica acta 2016-02, Vol.909, p.30-40
Main Authors: Bi, Yiming, Yuan, Kailong, Xiao, Weiqiang, Wu, Jizhong, Shi, Chunyun, Xia, Jun, Chu, Guohai, Zhang, Guangxin, Zhou, Guojun
Format: Article
Language:English
Subjects:
Algorithms
Chemometrics
Databases, Factual
Linearity
Local method
Mathematical models
Meat - analysis
Near infrared radiation
Near-infrared spectroscopy
Pharmaceutical Preparations - analysis
Pre-processing
Quantitative analysis
Segmentation
Spectra
Spectroscopy, Near-Infrared - methods
Spectroscopy, Near-Infrared - standards
Standard normal variate
Tablets - analysis
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Pre-processing of near-infrared (NIR) spectral data has become a necessary part of chemometrics modeling and is widely used in many practical applications. The objective of the pre-processing is to remove physical phenomena in the spectra in order to improve subsequent qualitative or quantitative analysis. Herein, a localized version of standard normal variate (SNV) is proposed, in which the correction parameters are estimated from local spectral areas. The method of determining the optimal spectral segmentation is also presented. Compared with full range methods, the local method demonstrates advantages in spectral linearity correction, model interpretation and prediction accuracy. Several benchmark NIR data sets were studied in our experiments; the proposed method achieved comparable performance against proven full range methods, with the reduction of prediction errors being statistically significant in many cases. [Display omitted] •A local pre-processing algorithm is proposed for near-infrared spectra.•The optimal segmentation of local areas can be automatically determined by a cross validation scheme.•Experiments show that the proposed local method outperformed the full range pre-processing methods.•The proposed method has no manual parameter and can be easily used in many applications.
ISSN:0003-2670
1873-4324
DOI:10.1016/j.aca.2016.01.010