Calibration of a mini-spectrophotometer using a neural-network-based interpolator

The computer-based interpretation of spectrometric data, aimed at identification of the chemical composition of an analyzed substance, requires some information on the spectrophotometer, acquired during its calibration. This information is of particular importance if a low-cost, low-resolution mini-spectrophotometer is used for analysis since only a few data per spectral peak are then available for processing. Artificial neural networks seem to be a sufficiently flexible and numerically powerful tools for solving the related problem of data interpolation. In this paper, the applicability of two classes of neural networks for solving this problem, viz. RBF networks and Elman networks, is studied using synthetic spectrometric-type data.