A software package for the evaluation of peak parameters in an analytical signal based on a non-linear regression method

Computer-based data handling of analytical signals to extract commonly used analytical parameters often produces poor results if the signals are affected by noise and a drifting baseline. Only a minor part of the information present in the signal is used for correction. A computer package is presented, in which the total information of the signal is used to give accurate evaluation of analytical parameters. The program uses a non-linear regression method to deconvolute analytical signals into a number of peaks and a baseline. To describe the peak shape any mathematical model can be used. In the computer package, a Gaussian curve-related model is used, with variable asymmetry. The baseline is described with a polynomial of variable order. The method is simultaneously a filter procedure Deconvolution of poorly separated peaks is possible. The software is developed on a minicomputer; however, test results of this study indicate the feasibility of implementation on a microcomputer without extremely time-consuming runs of the program.