Univ. Grenoble Alpes, Grenoble, France, CEA, LETI, MINATEC Campus, 17 rue des Martyrs, Grenoble Cedex 9, France

We address the issue of remote processing of micro-chromatographic signals for point-of-care body fluid analysis in a home-care environment, or for on-site air and water quality assessment. We propose to use a random walk model considering the chromatographic signal as a distribution of molecule arrival time (MAT) at the chromatography column output and we associate to each analyte a cluster of MAT. The signal processing should realize automatically an unsupervised clustering of the MAT to identify the cluster list, and a classification to assign a cluster label to each MAT. On such complex fluids, the list is only partially known. Thus, we propose to introduce an infinite list of cluster candidates described as an Infinite Gaussian Mixture Model (IGMM) and to estimate a finite list using a variational non parametric Bayesian inference. Experimental results are given for the gas chromatographic analysis of polycyclic aromatic hydrocarbons (PAHs) pollutants in solution. © 2017 IEEE.

Bayesian estimation, Chromatography, Environmental monitoring, Medical diagnostics, Variational inference

Air, Aromatic hydrocarbons, Bayesian networks, Column chromatography, Diagnosis, Electronic nose, Gas chromatography, Gaussian distribution, Inference engines, Polycyclic aromatic hydrocarbons, Quality control, Signal processing, Water quality, Bayesian estimations, Environmental Monitoring, Infinite Gaussian mixture models, Medical diagnostics, Non-parametric bayesian inferences, Polycyclic aromatic hydrocarbons (PAHS), Variational inference, Water quality assessments, Chromatography