With the recent development of wireless and sensing technology, wireless sensor devices have invaded the medical area with a wide range of capability. These devices allow medical practitioners to monitor the patients remotely and timely. However, it is known that the body movement can result in significant fluctuations of the radio signal, affecting the data communication. Real-time activity recognition is necessary to ensure that data transmission is avoided when the signal quality is bad. In this paper, we propose an Activity-based Transmission Protocol (ATP) that can recognize different user's activities and adapt the transmission protocol accordingly. By generating the acceleration waveform using the accelerometer reading, we classify the human activities into simple, complex and static movements and map it according to the transmission protocol. The classification will be used to select the transmission protocol. Extensive experiments performed on real hardware show that the ATP reduce the energy consumption by identifying and adapting the transmission protocol according the user's activities without affecting the packet reliability.