Tasks like Human Activity Detection require expensive and complicated devices. e.g., cameras, or wearable devices. However, the limits of wearable devices exist because many people don’t know how to use them, particularly the elderly. Also Cameras have the problem of privacy invasion. The radar sensor which is widely used commercially and militarily these days doesn’t need complicated operations and doesn’t intrude on people’s private lives. Among them, mmWave radar systems have especially a simple structure, with good resolution. Through this mmWave radar sensor, three-dimensional position which includes changes in height, direction, and movement can be estimated. In this paper, we propose the method to handle this position data with a graph representation. Then we applied st-gcn [1] to extract feature in spatial temporal domain. On the open datasets, RadHAR [2], we evaluated our approach to distinguish between 5 different human activities.