Neural-network-based calibration of a mini-spectrophotometer

A new method for calibration of mini-spectrophotometers is proposed. The method is designed to overcome two important drawbacks of existing methods, viz. their inability to deal with the problems implied by insufficiency of the number of output data and the effects of light polarization. It is based on the use of a tunable laser for acquisition of calibration data, and an RBF (radial basis function) neural network for modeling the polarization effects. The results of a preliminary study of this method, based on semi-synthetic data, are given.