Substance identification by error deformed spectra

Substance identification by their luminescence spectra is a highly sensitive and non distraction method. If a signal level is low then recognition errors may occur. The aim of this work was to define the identification algorithm with error probability control. For this purpose, the value of dissimilarity measure in the form of Spectral Angle Mapper (SAM) was analyzed. The relation between errors in measured spectra and the dissimilarity measure distribution was defined. The accuracy of the statistical hypothesis was used in spectral library search. The resulting algorithm was tested on more than 4000 sample spectra. The case when the measured spectra contained a signal of unknown source was analyzed, as well as the case when the measured spectra might contain either a signal or be equal to noise.