A New Quality Optimization Framework for DASH Streaming over Wireless Channels

Mobile devices are increasingly used as terminals for playback of multimedia content. However, maximizing the user's quality of experience is challenging due to the highly variable conditions of the wireless channels. A possibility to cope with such a variability is to dynamically adapt the source coding rate during the transmission, which is the underlying idea of the DASH standard. This work proposes a new framework to improve the quality of the DASH-based streaming experience by allowing to adjust the tradeoff between the quality of received content and the risk of playback freeze due to an empty buffer, which is a strong quality-disruptive event. The problem is analytically formulated and an efficient method to compute the playback freeze probability as a function of the representation choices over time is presented. Numerous simulation results using real download rate traces of 3G channels show the performance improvement compared to other bandwidth-adaptive algorithms as well as the robustness of the framework to variations of its most important parameters.