Low-latency live video streaming for mass-interactive experiences

The demand for ever more performing live video streaming only increases with every new release of technology in this fast-evolving industry. In this thesis, we explore the multiple ways of deploying a live video streaming system that would allow several cinema rooms to play a multiplayer crowd interactive game. The goal is to design a model that would allow the exchange of multiple streams, including video, between the participating theaters, with very low latency requirements and real-time monitoring and adaptation to the network conditions. In order to comply with the low latency demands, the new standard JPEG XS is given for the compression of the video. The literature detailing the different possible topologies and components of such a system made us choose a Selective Forwarding Unit (SFU) architecture. Based on the low complexity and low latency promises of JPEG XS, we placed encoders in the SFU to adapt to the network quality. We then proposed an implementation of the selected model and analyzed whether our model is the most adapted to answer the initial problem. We come to the conclusion that the principle of the SFU is indeed the best suited for the redirection of live streams between clients, but that the results are insufficient to validate the use of transcoding in the central server.