On-body radio propagation in the 2.45-GHz industrial-scientific-medical frequency band (2.40–2.48 GHz) was investigated during three different activities: jogging, rowing and cycling. Four different links were examined, from the waist to the wrist, ankle, chest and back; it was observed that the channel behavior could be related to the repetitive nature of the activities. Furthermore, combinations of the wrist, ankle, and chest channels could potentially be used to identify the activity, while the waist-back channel shows little variation between activities (roughly 2 dB, compared to 6–15 dB for the other links). The results also show that dynamic on-body radio channels contain rich biomechanical information, such as motion pattern, heartbeat, and breathing process. It is demonstrated that physiological and kinetic features can, therefore, be extracted through some known signal-processing techniques. For example, the repetitive nature of the activities introduces harmonics into the signal via fading, which correspond to the speed of motion. In addition, the mechanical motion of the torso during respiration and cardiac activity also introduce harmonics due to changes in the path loss, albeit with low magnitudes. The detection of such signals is discussed.