Rehabilitation and training is generally performed under control of a specialist. However the possibility to self-train motion as effectively as it would be done with an expert has clearly many advantages. One can indeed choose time, duration and place of his training. This paper presents the development of a software based on real-time data collecting from IMUs (Inertial Measurement Units) for human motion self-training. By choosing adequate location for the IMUs, and parameters (anatomic angles) of the movement to control, the system will be able to monitor every single motion of the user. The software allows to choose any body segment and to compute angles between any IMUs in 3 dimensions or in projection in anatomical planes. After defining the movement to control, the user needs to record several optimal movements so as to specify the reference and the boundaries of what will be set as the acceptable motion. Then, the software allows the user to repeat the movement as many times as he wants and displays a visual feedback showing the compliance between the reference and the actual motion. Recording the reference motion with the instructions of an expert will allow using it as an efficient tool for self-rehabilitation or self-training.