The force between wheels and rails of the high-speed train was firstly extracted and applied into the computational model of radiation noises of wheels and rail respectively. As a result, the radiation noise of wheels and rails was obtained. As can be seen from the result, radiation noises of wheels had an obvious directivity on the body surface, while radiation noises of rails had an obvious periodicity on the body surface. With the increase of the analyzed frequency, both directivity and periodicity were shown more obviously. Then the aerodynamic model of the high-speed train was established, and the pressure and velocity distributions on the train surface were computed. The maximum pressure was at the tip of the nose of the high-speed train, the maximum velocity was at the transition of the cabin, and more serious eddy was in the rear of the high speed train. Based on the computed pressure distribution, the aerodynamic noise was distributed evenly on the entire body surface, which was gradually increased with the increasing analyzed frequency. Finally, the wheel radiation noise, rail radiation noise and aerodynamic noise were extracted as excitations and applied into the SEA (Statistical Energy Analysis) model of the high-speed train, in order to compute its full-spectrum noise under multi-physics coupling excitations. The computational result was compared with the experimental result. It was presented that the difference of average sound pressure level (SPL) was 2.8 dB between the experimental and numerical simulations within the entire analytical frequency band. The SEA model with considering the multi-physics coupling was effective.