Simple encoding of infrared spectra for pattern recognition. 1: Statistical examination of the effectiveness and information content by factor analysis

Abstract

SUMMARYThe spectral region from 700 to 3600 cm" is subdivided into several wave number intervals. The peaks in each interval are summarized by means of three encoding algorithms. Using a factor model of kcommon factors, the total extractable variance (corn) of a given set of intervals is calculated and correlated with the redundancy of information in all these intervals. The value of corn is verified by analysis of the factor loadings a, (factor pattern). Finally, the information content of some chosen sets of intervals coded by the three selected feature algorithms will be correlated to the probability of information flow through a serial-parallel network. The encoding using only wave numbers was found to be the most effective.