Real-time 3D content distribution over a network requires facing several challenges, most notably the handling of the large amount of data usually associated with 3D meshes. The scope of the present paper falls within the well-established context of real-time capture and streaming of OpenGL command sequences, focusing in particular on data compression schemes. However, we advance beyond the state-of-the-art improving over previous attempts of ¿in-frame¿ geometric compression on 3D structures inferred from generic OpenGL command sequences and adding ¿inter-frame¿ redundancy exploitation of the traffic generated by the typical architecture of interactive applications.

The work addresses the topic of software integration in the context of complex virtual reality applications. We propose a novel method to render in a single graphical context objects handled by separate programs based on different VR frameworks, even when the various applications are running on different machines. Our technique is based on a network rendering architecture, where both origin and destination applications communicate through a set of protocols that deliver compressed graphical instructions and keep the machines synchronized. Existing applications require minimal changes in order to be compatible with our system; in particular we tested the validity of our approach on a number of rendering frameworks (Ogre3D, OpenSceneGraph, Torque, XVR). We believe that our technique can solve many integration burdens typically encountered in practical situations.

In this paper, we propose a real time 3D gesture recognition system that relies on the state based approach. The novelty of this work is the introduction of probabilistic neural networks (PNNs) to characterize the uncertain boundaries of each state. The 3D gestures are modeled as a sequence of states in a configuration space; the number of states and their spatial parameters are calculated by dynamic k-means clustering on the training data of the gesture without temporal information. Gesture recognition is performed using a simple Finite State Machine (FSM), where, each state transition depends only on the output of its corresponding PNN and optionally on its time restrictions (minimum and maximum time permitted in the state). If a recognizer reaches its final state, then it could be said that a gesture is recognized. The approach is illustrated with the implementation of a real time system that recognizes the semantic meaning of seven basic gestures of the Indian Dance, the description of the system and the technologies used, it will be described in detail in the paper.