Multispectra Analysis for Environmental Monitoring

The identification of chemical elements is an essential task in the characterization of environmental chemistry, food chemistry, biology, medicine, etc. Identification of unknown compound materials is usually performed by automatic searches for reference spectra that are most similar to the measured spectrum. The similarity of spectra is for instance defined by the Euclidean distance or correlation coefficient. Automatic searches for similar spectra is the most popular approach in computer-assisted evaluation of spectra. A number of approaches are possible to determine which features in the spectra of the mixtures correspond to the chemical molecules of interest. This process of assigning spectral features to distinct chemical molecules represents a type of pattern recognition: the spectrum of one chemical molecule represents a pattern which is searched for in several spectra. We are developing a SVRDM-based spectra classification product for environmental monitoring industry.