Research On Rate Adaptation Strategies For Video Streaming Clients

An important feature of mobile streaming media services is the requirement to transmit large amounts of high-definition video data in real time under an unstable network status. In order to solve the contradiction between the fluent watching of high-definition videos and bandwidth fluctuations, DASH(Dynamic Adaptive Streaming over HTTP) comes into being. DASH servers provide multiple bitrate video streamings, and clients adapt to the changes of bandwidth by switching bitrate based on rate adaptation strategies to provide a high Qo E(Quality of Experience). Therefore, it is significant for the application of DASH to design an effective rate adaptation strategy.In order to improve the quality of experience, specifically for scenarios like a single mobile client watching videos based on a DASH streaming service in the wireless network environment, this paper comes up with a rate adaptation strategy based on buffer partition. First, through the analysis of DASH systems and the streaming media Qo E model for DASH clients, this paper quantizes four assessment metrics of Qo E, analyzes the importance of each index qualitatively, and further reveals the restrictive relationship of them by experiment data. Secondly, this paper establishes a buffer state Markov model by analyzing the dynamic of the client buffer. According to the characteristics of transition between buffer states, the client buffer is divided into three regions logically, and a buffer partition model is established. Then the Qo E model is combined with the buffer regional characteristics, by adopting different bandwidth-based strategies depending on buffer regions, and emphasizing different metrics in corresponding regions, to get a general balanced performance. Finally, the sizes of buffer regions are approximately estimated by theoretical analysis.By building a DASH experimental platform in a simulation environment of wireless network,this paper makes a comparison between the rate adaptation strategy came up above which based on buffer regions partition and other adaptive models. The experiment results show that the bitrate adaptation strategy came up in this paper can avoid unnecessary rebuffering effectively, as well as improve average bitrate by 10%~45% and reduce the frequency of bitrate switching by 20%~50%.