HUMANN-based computational neural system for the determination of pollutants using fluorescence measurements

Abstract

Polychlorinated Biphenyls (PCBs) and Polychlorinated Dibenzofurans (PCDFs) are chlorinated aromatic compounds and are emitted into the environment. It has been shown lo have toxicity and carcinogenic potential characteristics. Because of this their identification and quantification is a matter of great concern. However. the similar structure of PCBs and PCDFs can produce overlapping in fluorescence spectra, which add difficult to their determination. We present in this paper. an HUMANN-based computational neural systeni [1][2] for the identification of these compounds. HUMANN is a multilayer neural net with high biologicnl plausibility. Its adaptive character is essentially embodiment in the labelling module, because of its dynamic dimension The determination of the different analytes will be indicated by the firing neurons in the labelling layer and by the activation level of these neurons. In this work its have also been developed a model for spectral data. íiuorescence spectrurn of single compounds and complex niixture. via Gaussian distribution. Our final proposal consists in putting to work together fluorescence spectrometry and neural computation approach, and to analyse the good results and the troubles found in this new method using three type of spectra: excitation, emission and synchronous.