From the inversion of the speech data of native and non-native speakers, we can find the difference in the vocal tract shape. Schroeder's theory assumes that the vocal tract is a uniform cylinder. The axial distribution of deformation of the cross sectional area is obtained as a perturbation from the uniform cylinder by applying the Boltzmann-Ehrenfest theorem. We extend Schroeder's theory to a case of the non-uniform cross sectional area of the vocal tract for an application to the foreign language learning process. Furthermore, we discussed constraint conditions to obtain a special solution to Schroeder's inverse problem and verifies it by numerical examples.