Résumé : Infrared spectroscopy is a powerful tool in protein science due to its sensitivity to changes in secondary structure or conformation. In order to take advantage of the full power of infrared spectroscopy in structural studies of proteins, complex band contours, such as the amide I band, have to be decomposed into their main component bands, a process referred to as curve fitting. In this paper, we report on an improved curve fitting approach in which absorption spectra and second derivative spectra are fitted simultaneously. Our approach, which we name co-fitting, leads to a more reliable modelling of the experimental data because it uses more spectral information than the standard approach of fitting only the absorption spectrum. It also avoids that the fitting routine becomes trapped in local minima. We have tested the proposed approach using infrared absorption spectra of three mixed α/β proteins with different degrees of spectral overlap in the amide I region: ribonuclease A, pyruvate kinase, and aconitase.