Abstract: We analyze features of ground shaking signals and radiated noise of moving vehicles on highway using AR model, use hypothesis test to observe difference, discuss the mechanism of the signal, and carry out classification with BP neural networks. Difference between the AR model parameters of the ground shaking signals and radiated noise is observed. Because the difference of parameters of radiated noise is greater than that of ground vibrating signals, parameters of radiated noise can be used to identify the type of vehicles, with accuracy better than 80%.