Abstract: The multi-layer fluid-solid interface model containing a soft medium layer is established. The dispersion equation of the fluid-solid interface wave is deduced by means of the transfer matrix method and the dispersion curve is calculated according to the dispersion equation. It is observed that the dispersion curves change with the depth, thickness and shear wave velocity of the soft medium layer. The finite element software is used to establish the simulation model, and the genetic BP neural network is introduced to inverse parameters of the model. The results show that the parameter inversion for depth, thickness and transverse wave velocity of soft medium layer can be effectively performed by BP neural network.