Feature selection for identification of substituted benzenes from infrared spectra

Any coupling of a spectroscopic detection technique to chromatography for the analysis of complex mixtures inevitably leads to a tremendous bulk of collected data and here gas chromatography/infrared spectroscopy is no exception. To deal with the large amount of data produced, we already demonstrated computer methods that select certain spectra form the data set as being more interesting for the analysis at hand, and focussed attention on the interpretation of those spectra. In this paper, we report an extension of this work that includes automatic determination of proper ranges of frequencies within which characteristic patterns of absorbances for a given functional group will be found. Patterns of absorbances in those correlated regions are used to describe a functionality for selective reduction of infrared data. The method was successfully applied to substituted benzene compounds.