Detecting Small Variations in Shape
Résumé
A method for the detection of very small differences between the shapes of two signals is presented. This method uses the normalized integrals of the signals to be compared; hence it filters the noise. After a general presentation giving an especially detailed account of the definition of equality of two shapes, the possible application fields with respect to the intrinsic properties of the method is analyzed. It makes it possible to test the fitting of a signal to a model, but chiefly, to detect small shape differences between two signals without needing any modeling of the signals or the systems from which they come. Finally it is shown how the method was applied to the particular problem of detecting twocomponent chromatographic peaks visually confounded and with low resolution.