Adaptive estimation of Raman chemical mixture spectra

We introduce an adaptive mixing algorithm for estimating the relative ratios of chemicals in a mixture spectrum. This procedure is particularly well suited to mixtures with a large dynamic range of mixture weights. It has the advantage of being able to be used in conjunction with a band-pass (difference-to-Gaussian or DOG) filter, and a correction of baseline off-set and tilting of the spectrum. Output of these filtering techniques is a cleaner signal retaining most of the relevant Raman spectral signature while minimizing artifacts due primarily to Rayleigh, dust, and atmospheric aerosols. We will describe the results of applying these algorithm to mixture spectra with both real and simulated additive noise.