A Multimedia Signal Processing Cloud Concept for Low Delay Audio and Video Streaming via the Public Internet

摘要

A rehearsal environment for conducted orchestras via the public Internet requires a specialized server infrastructure, in order to provide minimal latencies between the musicians involved. In this document, we present a cloud computing concept for digital signal processing of audio and video data in realtime. Since 60 musicians and one conductor shall connect to the cloud, it is most important to distribute the signal processing and machine learning algorithms over multiple processing servers. The server infrastructure under investigation is built on top of an AVB network segment to be scalable and to maintain low latencies and jitter under heavy load. Latency and jitter are the most important properties of the realtime streams that are connected to the cloud, and are analyzed and discussed. The results have proven the proper design of the concept, but revealed the need for further optimization.