The measurement of diffusion coefficients in multicomponent liquid mixtures is still an open question. A new technique aimed at overcoming this deficiency is presented. It is based on Raman spectroscopy, which has been extended to obtain liquid concentration data of all components in a mixture simultaneously with high temporal and spatial resolution. The following data analysis uses model-based methods. This provides a very flexible framework, allowing substantial reduction of experimental effort and time and the direct estimation of the diffusion coefficient as a function of concentration. Furthermore, the experimental procedure is improved using optimal experimental design techniques. This new methodology is validated for binary mixtures.