Application of Dynamical Analysis Techniques in the Extraction of Compound-Specific Information from Fourier Transform Infrared Interferograms

Abstract

An automated compound-detection procedure is described that is based on the application of dynamical analysis techniques to digitally filtered Fourier transform infrared interferogram data. With the use of benzene as a test analyte, surface-of-section maps are computed from the filtered data. This calculation is resistant to phase variation in the interferograms and serves to remove the oscillating component of the filtered signal. Information regarding the overall shape or profile of the signal is enhanced, thereby resulting in an improved sensitivity for distinguishing analyte information from that due to interferences. Data vectors constructed from the surface-of-section maps are examined by use of principal component analysis. Linear discriminant analysis is then used to implement an automated algorithm for detecting benzene in the presence of nitrobenzene. When tested with data not included in the discriminant development computations, the computed linear discriminants are shown to perform with a detection accuracy of greater than 99%.