Maglev train is a new vehicle without support wheel and its movement speed is gained by means of a special measure equipment. By using BP neural network arithmetic , the maglev train speed can be estimated by means of combining characteristics of its traction linear induction motor. Based on a dynamic simulation experiment , real speed measure result is near to theory calculation. This shows the method is feasible. In view of the method , it can economize expensive cost for speed measurement of maglev train , reduce system equipment complexity, and also advance work reliability of entire train system.