This work proposes a new method for the determination of the mass diffusion coefficient in hygroscopic materials. The experiment consisted of submitting one face of the sample to a variation in time of the relative humidity (RH) and measuring the RH on its back face. The imposed RH and temperature were measured during the test and served as boundary conditions in a comprehensive computational code to solve heat and mass transfer in porous media. This model uses a physical engine embedded in the inverse procedure to determine the mass diffusion coefficient. Compared with classical methods, this new method has several advantages: It allows several samples to be measured simultaneously, simply by multiplexing the RH sensors. Accurate values can be obtained even when starting and ending out of equilibrium, which allows the characterization time to be drastically reduced. The external mass transfer coefficient has a negligible effect on the identified value. Nonstandard Fickian behaviors can be detected by the disagreement between the measured and simulated curves. The results show that the diffusivity obtained for spruce wood is in good agreement with those found with classical methods. In contrast, the fiber board results differed between the experiment and model, or yielded unrealistic values, which confirms the dual-scale nature of mass transfer that occurs in this kind of material.