Correcting Fluctuating Baselines and Spectral Overlap with Genetic Regression

The application of a new algorithm, known as genetic regression (GR), to calibration problems with spectra containing complex fluctuating baselines is illustrated with the use of synthetic data. The ability of the algorithm to automatically compensate for the presence of linear and polynomial (quadratic and cubic) baselines in the presence of complex spectral overlap is investigated along with the effect of noise. GR is unique in that it provides an effective wavelength optimization technique by sorting through the spectr al data and selecting and appropriately combining wavelengths that compensate for structured baseline and spectral overlap. The results obtained with GR are compared with those obtained with background-corrected linear regression. GR is shown to give much better results and, in constrast to traditional background correction, is much faster and can compensate for the presence of both structured baseline and complex spectral overlap simultaneously. The results of a noise study show that the method works at low signal-to-noise ratio (SNR) and that the error in the final result is a function of the noise.