Video adaptation proxy for wireless Dynamic Adaptive Streaming over HTTP

This research studies Dynamic Adaptive Streaming over HTTP (DASH) in cellular broadband access networks (e.g. HSDPA, HSPA+, LTE). We propose a wireless DASH (WiDASH) proxy to enhance the Quality-of-Experience (QoE) of wireless DASH. WiDASH proxy locates at the edge between Internet and wireless core networks. Different from conventional DASH, whose rate adaptation logics are either implemented locally in the user equipment or in the DASH server, WiDASH proxy is in charge of video adaptation, which makes it feasible to perform global optimization over multiple concurrent DASH flows. WiDASH proxy improves DASH QoE by three novel technologies. The first one is architecture level innovation. WiDASH simultaneously explores the potential of both split TCP and parallel TCP. It splits the original TCP connection from DASH server to a wireless user into one wired TCP and multiple wireless TCPs. This approach is able to increase video streams´ average throughput without sacrificing inter user fairness. Second, WiDASH proxy employs video quality aware dynamic prioritization. Low bit rate video streams are entitled with high priority to guarantee minimum QoE for users with bad wireless channel quality. A video stream´s priority is gradually dropped to normal state with the growth of its bit rate so that inter flow fairness between video streams and background best effort flows are still maintained. Third, we design a multiple-input multiple-output adaptive optimal controller, based on adaptive control theory. This rate controller minimizes a quadratic cost function which is defined as the weighted sum of multiple concurrent DASH videos´ distortion, bit rate variation, and playback jitter. The solution of the optimization problem implicitly coordinates multiple DASH streams to improve overall streaming QoE. Simulation results verify that WiDASH proxy is able to improve QoE by providing smoother video streams with higher average visual quality for wireless HTTP video streamin- .