The following is an abstract for the selected article. A PDF download of the full text of this article is available here. Members may download full texts at no charge. Non-members may be charged a small fee for certain articles.

A Mixed Linear/Nonlinear Least-Squares Method for Determining Line Parameters from Noisy Spectra

Volume 43, Number 3 (April 1989) Page 558-566

Premoli, Amedeo; Rastello, Maria Luisa

The determination of the parameters characterizing line profiles embedded in a noisy spectrum is an important task in IR and Raman spectroscopy. A direct analysis from an inspection of the spectrum is often difficult in the presence of overlapped profiles and/or considerable back-ground. This paper presents an algorithm based on alternate stages of linear and nonlinear optimization for extracting Lorentzians or Gaussians from the measured data, taking into account the presence of a frequency-dependent background superimposed onto white noise. The analysis of several experimental spectra shows the capability of the algorithm in discriminating noisy and partially overlapped profiles.