A hybrid least squares and principal component analysis algorithm for Raman spectroscopy

The least squares fitting algorithm is the most commonly used algorithm in Raman spectroscopy. In this paper, however, we show that it is sensitive to variations in the background signal when the signal of interest is weak. To address this problem, we propose a novel algorithm to analyze measured spectra in Raman spectroscopy. The method is a hybrid least squares and principal component analysis algorithm. It explicitly accounts for any variations expected in the reference spectra used in the signal decomposition. We compare the novel algorithm to the least squares method with a low-order polynomial residual model, and demonstrate the novel algorithm's superior performance by comparing quantitative error metrics. Our experiments use both simulated data and data acquired from an in vitro solution of Raman-enhanced gold nanoparticles.