A QoS-Aware 3D Point Cloud Streaming from Real Space for Interaction in Metaverse

In this paper, we present a point cloud streaming method of real-space objects such as humans and animals for real-time 3D reconstruction in VR space. The system uses a depth camera to scan a human or animal, divides the point cloud into parts of the body, and then controls the quality of the point cloud (i.e., resolution and frame rate) for each part in real-time according to the object's motion and context. This enables point cloud streaming with limited resources (computational and network resources) and maximizes the user's quality of experience (QoE). We have implemented and evaluated a series of systems incorporating the proposed method to enhance the user experience in realistic environments and scenarios while maintaining interactivity in VR-based online communication. The results show that the proposed system is feasible under resource-constrained environments without significantly affecting the user's QoE.