Abstract

Networked Music Performances (NMPs) involve geographically-displaced musicians performing together in real-time. To date, scarce research has been conducted on how to integrate NMP systems with immersive audio rendering techniques able to enrich the musicians' perception of sharing the same acoustic environment. In addition, the use of wireless technologies for NMPs has been largely overlooked. In this paper, we propose two architectures for Immersive Networked Music Performances (INMPs), which differ for the physical positions of the computing blocks constituting the 3D audio toolchain. These architectures leverage a backend specifically conceived to support remote musical practices via Software Defined Networking methods, and takes advantage of the orchestration, slicing, and Multi-access Edge Computing (MEC) capabilities of 5G. Moreover, we illustrate how to integrate in the architectures machine learning algorithms for network traffic prediction and audio packet loss concealment. Traffic predictions at multiple time scales are utilized to achieve an optimized placement of Virtual Network Functions hosting audio mixing and processing functionalities within the available MEC sites, depending on the users' geographical locations and current network load conditions. An analysis of the technical requirements for INMPs using the two architectures is provided, along with their performance assessment conducted via simulators.