Neural network for fitting vibrational-rotational line contours in high-resolution spectra
A.P. Shcherbakov
V.E. Zuev Institute of Atmospheric Optics of Siberian Branch of the Russian Academy of Science, Tomsk, Russia
Keywords: absorption line parameters, Fourier transform spectrometer, Voigt profile, automatic processing, neural network
Abstract
The article is devoted to the problem of automation of fitting lines parameters in high-resolution spectra recorded at modern spectrometers. When fitting a model contour, due to the presence of many local minima in the minimized standard deviation, a sufficiently accurate initial approximation of line profile parameters is required. The article proposes a method for finding a sufficiently accurate initial approximation of line profile parameters for the convergence of the fitting process. The method is based on the Kohonen neural network. Tests and comparison of other algorithms and networks for solving this problem are carried out. The suggested method can be used to process vibrational-rotational spectra and obtain databases for solving problems of atmospheric optics, molecular physics, and engineering problems.
|
