This thesis relates to knowledge capitalisation in informed virtual environment (IVE) for industrial maintenance training. An IVE is a virtual environment including knowledge based models in which it is possible to both interact and allow behaviours by interpretation of dynamic or static representations. Today there is no system proposing a gesture capitalisation in virtual environments. Following the creation and study tire production training by virtual reality, we have initiated a work on knowledge capitalisation in virtuo. Our approach is to consider gesture : technical, choreographic or communication gesture as knowledge. This empirical knowledge is usually associated to body movements : this is the concept of embodied cognition. We propose a model of gestural annotation allowing a capitalisation of the gesture as an extension of 3D annotations. A prototype based on this model has been developed (Ptol em ee), and allows knowledge capitalisation and sharing of for industrial maintenance. The main perspectives of our work are natural interaction, knowledge capitalisation and nally dynamic assistance for training in IVE.
