Research On Streaming Media Adaptive Strategy Based On The Receiver Buffer Dynamics

Abstract:Driven by an insatiate appetite for bandwidth in the Internet, advances in media compression technologies and accelerating user demand, video steaming over the Internet has quickly risen to become a mainstream "killer" application over the past two decades. In order to meet the requirements of high quality streaming media transmission and user’s higher quality of experience (QoE), Dynamic Adaptive Streaming over HTTP (DASH) technology arises at the historic moment.Through the study of existing DASH adaptive strategy, the thesis find that most of the adaptive strategy is based on the bandwidth estimation. However, bandwidth estimation will be affected by many factors such as video chunk size, dynamic TCP congestion control algorithm and competition of flows. Once bandwidth estimation became unstable and less accurate, it can lead to that the QoE of many existing adaptive strategy is not high enough with the problems of unnecessary re-buffering, bad average video quality and so on. To solve this problem, a DASH adaptive strategy is proposed in this thesis, which will not estimate the bandwidth and will make adaptive decision based on the receiver buffer dynamics. The thesis first establishes the mathematical model of receiver buffer in HTTP streaming, and then puts forward a video rate switch strategy based on buffer dynamics, which uses a scheme combines reference function decision-making and second-time decision-making. Then the thesis designs the second-time decision-making algorithm, and designs the reference function for rate decision based on related QoE index. In the design process, through theoretical analysis, the thesis proves that the scheme can avoid unnecessary re-buffering, has a relatively small startup delay and also good average video quality. The thesis also analyzes and explains why convergence speed and video rate switch frequency has the game relation. Then, the thesis designs a video segment request strategy. Finally, the thesis realizes a prototype system based on the proposed adaptive strategy of DASH. Through experiment based on the system, the thesis proves that the proposed adaptive strategy has relatively good performance of QoE...