A non-iterative method for fitting of overlapping Gaussian peaks

ArticleinNuclear Instruments and Methods 180(2-3):553-556 · April 1981with42 Reads
DOI: 10.1016/0029-554X(81)90099-9
A method for fitting two overlapping Gaussian peaks without an iterative procedure is presented. The method utilizes a linearization technique of the sum of the Gaussian functions by the use of the difference equation. The validity of the method has been tested against pseudo-experimental spectra and comparison with the conventional non-linear least-squares method has been made.
  • [Show abstract] [Hide abstract] ABSTRACT: The methods used for mapping extended objects can also be applied to star counting, provided that the star density is not too high. The method considered for star counting in the present investigation is more reliable than other methods using similar principles, because the point source response is actually measured, and any beam asymmetries or noninteger chop/sampling-interval ratios are unimportant. The method has been tested by analyzing some data in parallel with the 'manual' deconvolution performed by Eaton et al. (1982). Only a modest improvement is obtained for uncrowded fields, but the method shows a clear advantage in crowded fields such as those located near the Galactic Center. A major advantage of the considered method of infrared star counting is that a large amount of data can be easily processed in a completely consistent way.
    Article · Apr 1982
  • [Show abstract] [Hide abstract] ABSTRACT: Extracting parameters iteratively from two component exponential decays requires initial guesses. These are avoided if the treatment is made linear. In practice it is found that the method is badly biased or fails where it is most necessary, i.e., where the decay rates (lifetimes) are close. The origins of the difficulties are detailed and explained with the aid of analyses of simulated decays. Ways of broadening the regions of applicability of the non-iterative method are given.
    Article · Oct 1982
  • [Show abstract] [Hide abstract] ABSTRACT: A method has been developed of distinguishing the spectral peaks detected as local maxima on a measured curve from the noise-induced artefacts when peak positions are not known beforehand. The method is based on a derivation of the distribution function for local maxima of the random signal component in the dependence on their relative heights. The resulting probability of a 'noise peak' being formed within a given number of experimental points determines the mean statistical authenticity of any detected maximum. Examples are given for white Gaussian noise, for this noise filtered by the Savitzky-Golay algorithm and also an application to Auger electron spectra.
    Article · Apr 1987
Show more