Recently there has been an increasing interest on tangible games in which human motions are recognized rather than the handling of keyboards and mouses. Such games require a motion controller for recognizing the motions of users. In this paper, we analyze the characteristics of values of accelerator sensors which are generated by a user who perform a tennis swing while holding a smartphone with his/her hand, and propose a method for motion recognition based on DWT(Discrete Wavelet Transform). The proposed method enables a smartphone to serve as a motion controller, so that a user can enjoy a tangible tennis game without eliminates a need for buying the device. We developed a tennis game prototype using the proposed method. To our experiment, our method showed a high recognition rate and the usefulness in the game.