Wavelet based Raman spectra comparison

In this paper a novel blind peak detector for Raman spectra based on the continuous wavelet transform is proposed. The peak detector is shown to correctly identify peaks of different widths and intensities, and is well suited for subsequent peak-to-peak matching for the process of identifying unknown Raman spectra with the help of a library of reference spectra. No background reduction or initial noise filtering is necessary. The described methods were verified by successfully identifying 73 paint pigments by a spectral search through a Raman library for 638 pigments. In all except 7 cases, the Schmincke–Kremer pairings from the truth reference were found to be amongst the two top pairings with the largest matching cost function. These 7 cases were again inspected by the paintings conservator and doubts were raised regarding the truthfulness of the initial pairing. The proposed methods are equally applicable to other spectral methods.