In this paper, we present 'Virtual Observation' (VO) a software tool for contextual observation and assessment of user's directly from within the virtual reality (VR) simulation framework. Unlike other recording systems, the VO system described in this paper focuses on recording and reconstructing VR user's positional, rotational and input data to recreate the same experience the user had with a VR simulation. Different from animation-based approaches, VO records user inputs and reconstructs the simulation from them and the user positional data. Moreover, the system allows the broadcast of this information to a remote machine enabling remote live observation of the simulation. Datasets recorded by the system can be shared by exporting them as XML files or, optionally, into a standalone online application, such as browser WebGL, allowing researchers, developers and educators to share and review a VR user simulation through a free-moving camera using a web browser. In this paper, the consistency of the data generated from the software by the client, server and reconstructed datasets acquired during real-time live observations was evaluated. We conclude that this Virtual Observation software offers detailed reconstruction of low-level information and visual information of user actions during simulations for both live and offline observations. We envision that our system will be of benefit for researchers, developers and educators that work with VR applications.

Unlike conventional desktop simulations which have constrained interaction, immersive Virtual Reality (VR) allows users to freely move and interact with objects. In this paper we discuss a work-inprogress system that 'virtually' records participants movement and actions within a simulation. This system recovers and rebuilds recorded data on request, accurately replaying individual participants motions and actions in the simulation. Observers can review this reconstruction using an unrestricted virtual camera and if necessary, observe changes from recorded input devices. Reconstruction of each participants' skeleton structure was created using tracked input devices. We conclude that our system offers detailed recreation of high-level knowledge and visual information of participant actions during simulations.