COMPARISON OF FOURIER SELF-DECONVOLUTION AND MAXIMUM-LIKELIHOOD RESTORATION FOR CURVE-FITTING

The advantages of combining deconvolution and curve-fitting for the analysis of spectra with heavily overlapped bands in the presence of noise and baseline errors are examined. Two methods of deconvolution are used, namely Fourier self-deconvolution (FSD) and maximum likelihood restoration (MLR). It is shown that for spectra with heavily overlapped bands and baseline errors there is an improvement in the conditioning if spectra are deconvolved prior to curve-fitting, so that more accurate band parameters are derived. This result is true for both FSD and MLR, but it is more to determine the optimum degree of deconvolution when MLR is used. The combination of deconvolution and curve-fitting also allows the objective optimization of the parameters used in either FSD or MLR.