Indirect identification of bridge frequencies, which refers to identifying bridge frequencies from the dynamic response of a vehicle moving on the bridge, has the potential to fast inspect bridges in large quantities. Accurate analysis on the dynamic response of the vehicle used for sensing is crucial to the indirect frequency identification. Most published research on this topic, however, is based on a simplified two-dimensional (2D) analysis of vehicle–bridge interaction (VBI), which cannot fully reproduce the mechanism of a real three-dimensional (3D) vehicle. As a complement, this study carries out the 3D simulation of VBI and accordingly proposes a novel frequency-domain method to identify bridge natural frequencies from the vertical acceleration of a full-car model’s wheels. The wheels’ equations of motion are first transferred into the frequency domain, then the frequency responses of the front and rear wheels are subtracted with a time lag to eliminate the adverse effect of road roughness. Also proposed by this study is a new method to identify the speed of the sensing vehicle with the correlation function of wheel acceleration so that the time lag for subtraction can be calculated. In order to investigate the performance of the proposed methods, a series of numerical simulations are conducted, including sensitivity analysis on the vehicle speed, the class of road roughness, the noise level, and the vehicle frequency. Finally, this paper remarks on the accuracy and robustness of the proposed method and challenges faced by the engineering practice of the indirect bridge structural identification methodology.