Journal article
Automated selection of library subsets for infrared spectral searching
Analytical chemistry (Washington), Vol.62(3), pp.226-233
02/01/1990
DOI: 10.1021/ac00202a001
Abstract
A method for decreasing the time required to perform a standard library search is described based on principal components analysis of infrared spectra. Principal components analysis calculates a new set of axes and coordinates which reduce the dimensionality of the original data space. Spectra are projected onto a principal plane where they are represented by a single point in a two-dimensional space. The angle of the point In the plane representing the unknown spectrum Is determined, and only those library spectra with similar angles in the same plane are searched by use of the Euclidean distance metric. The principal components analysis was based on 2000 spectra in the EPA vapor phase infrared library. The plane angle procedure accurately selects a subset of spectra around the unknown which contains the 15–20 nearest matches as obtained by a previous Euclidean distance search of all 2000 spectra. The selection of the appropriate principal plane to use for a particular unknown spectrum is also discussed. The methodology is tested with Intralibrary searches and with searches of laboratory data. © 1990, American Chemical Society. All rights reserved.
Details
- Title: Subtitle
- Automated selection of library subsets for infrared spectral searching
- Creators
- Joanne M BjergaGary W Small
- Resource Type
- Journal article
- Publication Details
- Analytical chemistry (Washington), Vol.62(3), pp.226-233
- Publisher
- American Chemical Society
- DOI
- 10.1021/ac00202a001
- ISSN
- 0003-2700
- eISSN
- 1520-6882
- Language
- English
- Date published
- 02/01/1990
- Academic Unit
- Chemistry
- Record Identifier
- 9984216666902771
Metrics
2 Record Views