Resumen:
HTTP Adaptive Streaming (HAS) offers media players the possibility to dynamically select the most appropriate bitrate according to the connectivity performance. A best-effort strategy to take instant decisions could dramatically damage the overall Quality of Experience (QoE) with re-buffering times, and potential image freezes along with quality fluctuations. This is more critical in environments where multiple clients share the available bandwidth. Here, clients compete for the best connectivity. To address this issue, we propose LAMB-DASH, an online algorithm that, based on the historical probability of the playout session, improves the Quality Level (QL) chunk Mean Opinion Score (c-MOS). LAMB-DASH is designed for heterogeneous contents and changeable connectivity performance. It removes the need to access a probability distribution to specific parameters and conditions in advance. This way, LAMB-DASH focuses on the fast response and on the reduced computing overhead to provide a universal bitrate selection criterion. This paper validates the proposed solution in a real environment which considers live and on-demand Dynamic Adaptive Streaming over HTTP (DASH) and High-Efficiency Video Coding (HEVC) services implemented on top of GStreamer clients.
