Low-frequency vibrations are generated by heavy-weight impacts in multi-story residential buildings, such as running or heavy footsteps. These structural vibrations are transmitted into the receiving room as acoustic noise through the dynamic interaction between the oscillating slab and the enclosed air. Conventional prediction methods for heavy-weight floor impact sound often employ simplified acoustic assumptions. The complex characteristic of fluid-structure interaction (FSI) is frequently neglected by these approaches. To address this issue, the interaction mechanisms are investigated using experiments and numerical simulations. Experimental data are acquired by dropping a rubber impact ball on a concrete slab. Sound pressure levels in the receiving room are captured by microphones. The physical conversion of structural vibration into acoustic pressure waves is analyzed using a three-dimensional explicit finite element model. Lagrangian solid elements model the concrete slab. The underlying air layer is modeled using the arbitrary Lagrangian-Eulerian (ALE) formulation to accommodate the dynamic boundary deformation of the slab without the severe mesh distortion. A penalty-based FSI coupling algorithm is utilized to simulate energy transfer by calculating coupling forces from the penetration of the Lagrangian structure into the ALE fluid mesh. Comparative analysis is performed in both time and frequency domains, focusing on one-third octave bands up to 630 Hz. Peak sound pressure levels are successfully reproduced by the numerical model. The correlation between the structural eigenmodes of the slab and the acoustic cavity modes of the receiving room is analyzed to determine the mechanisms of peak noise generation. Standing waves and local acoustic hotspots are identified by analyzing the spatial pressure distribution within the receiving room. A robust methodology for predicting low-frequency floor impact noise is provided by the ALE-based FSI framework, effectively overcoming the limitations of simplified acoustic assumptions.