QoE-Based Multi-Stream Scalable Video Adaptation over Wireless Networks with Proxy Hao Hu † , Xiaoqing Zhu * , Yao Wang † , Rong Pan * , Jiang Zhu * and Flavio Bonomi * † Electrical & Computer Engineering, Polytechnic Institute of NYU, Brooklyn, NY 11201 * Advanced Architecture & Research, Cisco Systems, San Jose, CA 95134 Abstract—In this paper, we present a proxy-based solution for adapting the scalable video streams at the edge of a wireless network, which can respond quickly to highly dynamic wire- less links. Our design adopts the scalable video coding (SVC) technique for lightweight rate adaptation at the edge. We derive a QoE model, i.e., rate-quality tradeoff model, that relates the maximum subjective quality under a given rate by choosing the optimal frame rate and quantization stepsize. The proxy iteratively allocates rates of different video streams to maximize a weighted sum of video qualities associated with different streams, based on the periodically observed link throughputs and the sending buffer status. Simulation studies show that our scheme consistently outperforms TFRC in terms of agility to track link qualities and overall quality of all streams. In addition, the proposed scheme supports differential services for different streams, and competes fairly with TCP flows. I. I NTRODUCTION Recent years have seen a proliferation of smart phones and constant bandwidth upgrades in broadband mobile networks. These two factors combined have fueled the rapid growth of mobile media traffic. The study in [1] predicts that by 2015, two-thirds of world’s mobile data will be video. On the other hand, mobile media streaming remains a daunting task, espe- cially for users in a highly dynamic environment. The chal- lenges are multifold. First, rate adaptation for streaming video needs to closely track fluctuations in the available wireless link bandwidth. Conventional techniques such as TCP-friendly rate control (TFRC) [2], however, typically rely on end-to- end packet statistics and fall behind abrupt changes in the underlying network conditions. Second, existing approaches achieve fairness by allocating equal rates to all competing flows, whereas video streams naturally differ in their utilities of rate depending on their contents. For instance, it would be desirable for an action movie sequence to be streamed at a higher rate than a head-and-shoulder news clip competing over the same bottleneck wireless link. Such content-aware alloca- tion is missing in today’s systems. Thirdly, clients connecting to the same access node may experience different throughputs over their respective wireless links, due to factors such as distance and channel fading characteristics. Without proper in-network information, rate adaptation decisions made at the senders can easily lead to inefficient resource sharing, e.g. the occurrence of head-of-line blocking. In this paper, we address the above issues in a novel rate adaptation scheme for streaming video over a highly dynamic environment. As shown in Fig. 1, our design introduces a Base Station Video Servers Proxy Mobile Clients Video Stream Video Substream Fig. 1. Architecture overview of the adaptive video streaming system. proxy at the edge of the network, right where congestion over the wireless links occurs. This allows the rate adap- tation module to constantly monitor the bottleneck buffer level, which, in turn, reflects variations in the throughput and delay of wireless links for all receivers. To strike a balance between computational complexity and efficiency, we adopt the latest H.264/SVC standard [3] for lightweight in-network rate adaptation. It can also improve the content caching performance [4]. By combined usage of temporal scalability and amplitude scalability, a wide bitrate range (with a factor of more than 10) is allowed. The resulting scalable video stream can be decoded at different frame rates (FR) and quantization stepsizes (QS). Based on the parametric models from our prior work [5], we derive a QoE model that relates the video subjective quality under a given rate by choosing the optimal FR and QS. This model (also called Rate-Quality model) works as a utility function for the rate adaptation module to maximize the overall viewing experience of all streams. We solved two specific problems: i) to allocate the video rate for each stream based on their respective rate-quality relations and wireless link throughputs and the common bottleneck buffer level; and ii) to extract video packets belonging to the appropriate temporal and amplitude layers from each SVC stream based on the allocated rate. The first problem is solved by an iterative solution, whereby the per-stream rate is calcu- lated based on periodic updates of bottleneck buffer level and relative link throughputs. The second problem can be solved offline, by pre-ordering the video temporal and amplitude lay- ers in a rate-quality optimized way so that each additional layer offers maximum quality improvement for the rate increment. Extensive simulation studies confirm that our scheme adapts more swiftly to the dynamic wireless environment than TFRC. The proposed scheme provides content-aware and channel condition-aware network resource allocation. Furthermore, it is flexible enough to support differentiated service for video streams with different user-specified importance levels.