Robust analysis of spectra with strong background signals by First-Derivative Indirect Hard Modeling (FD-IHM)

The presented FD-IHM is validated using Raman spectra of ethanol/acetone mixtures. To introduce a variety of background signals, we used fluorescence dye, scattering bodies (yeast) and various background light sources. Classical IHM allows us to predict the test sets with a root-mean-square error of prediction (RMSEP) ranging from 0.60 wt% to 2.06 wt%, but careful manual background treatment had to be applied. With FD-IHM, we reduce the RMSEP error by 21%-73% without any background treatment. Thus, FD-IHM allows for both, efficient and accurate analysis of spectra with large background signals.