Chromatographic signal processing for PAH in methanol solution

In this paper we describe two methods to estimate the concentration of polycyclic aromatic hydrocarbons (PAHs) in a methanol solution, from a gas chromatography analysis. We present an innovative stochastic forward model based on a molecular random walk. To infer on PAHs concentration profiles, we use two inversion methods. The first one is a Bayesian estimator using a MCMC algorithm and Gibbs sampling. The second one is a sparse representation method with non-negativity constraint on the mixture vector based on the decomposition of the signal on a dictionary of chromatographic impulse response functions as defined by the forward model. Some results provided by those two methods are finally shown with a comparison of the computational and the quantification performances.