Expert System For Characterization Of Fluorescence Spectra For Environmental Applications

Publisher Website
Google Scholar
Abstract
A potentially very powerful and feasible expert system for intelligent characterization of spectra is proposed. Unlike the conventional expert systems it does not solely rely on human expert knowledge; instead, it has the capability to generate the appropriate domain dependent problem solving knowledge itself. The structure of the system is based on supervised information processing techniques and machine-based pattern recognition methods. Feature extraction techniques are used to measure appropriate spectral characteristics and then a hierarchical pattern classifier is used to characterize the spectra. This system has the potential of using different classifiers as appropriate at each of the nodes. This system was used with reasonable results on a relatively simple example of 24 ultraviolet-visible synchronous fluorescence spectra of petroleum oils (heavy crudes and No. 6 fuel oils) with Euclidean distance as a measure of dissimilarity.

This publication has 0 references indexed in Scilit: