Abstract
Non-invasive Raman spectroscopy has been used in an increasing number of applications in recent years. However, in situations where surface signal is excessive, the acquired spectrum of probed sample suffers from surface interference in either conventional backscattering Raman or specially designed Raman methods. A computational method for Raman spectral recovery is required. Strong overlapping of Raman bands and intense fluorescence are the main obstacles that hinder the spectral recovery. In this paper, we present a modified version of an indirect hard modeling algorithm to extract the true Raman spectrum of the probed sample in a two-layer system. The proposed algorithm requires two spectra. By an iterative stepwise optimization, it models one spectrum as a combination of a scaling of the other spectrum, a polynomial baseline and the Raman peaks of the probed sample. It addresses the issue of Raman bands overlapping as well as intense fluorescence interference. The performance of the algorithm is evaluated on experimental Raman spectra. Comparative studies show that the proposed algorithm provides better results for Raman spectral recovery.
