Energy-efficient video streaming over wireless networks
Chuah, Seong Ping
Date of Issue2013
School of Electrical and Electronic Engineering
Recent advances in several key areas such as video coding, wireless radio and silicon technologies have enabled pervasive and ubiquitous access to multimedia-rich mobile applications. As video standards evolve, more bits are being squeezed out while visual quality improves spectacularly. Meanwhile, datarate capacity of wireless links increases exponentially as wireless radio technology evolves from 2G to 4G. On the other hand, silicon technologies are empowering mobile devices with computing capability at exponential rate. However, large scale deployments often pose serious challenges to the resource allocation of the network and mobile devices. Insatiable demands for better visual (from HD to multiview/3D and to ultra HDTV) experience increases the size of video data. User-generated videos and their sharing put signi¯cant strains on the wireless networks. Coding and delivery of video often incur signi¯cant power cost to mobile devices and wireless networks. Energy-e±cient resource allocation and scheduling are thus essential for sustainability of large scale deployment of mobile multimedia applications. This thesis investigates the energy e±ciency perspective of resource allocation in video streaming over wireless networks. We consider three scenarios which are typical and common in the applications of wireless video streaming. We ¯rst investigate the best-e®ort video delivery to heterogeneous clients. For this scenario, we consider the wireless multicast of scalable video over wireless mesh networks. The video multicast is a best-e®ort service which is limited by channel access and energy constraints. We formulate the optimal multicast strategy to maximize the video quality of heterogeneous clients under limited resources constraints. We show that the optimal multicast strategy can be obtained e±ciently via a dynamic programming approach. To schedule the scalable video bitstreams for multicast, we propose a quality-impact index to rank the priorities of video frames. Next, we consider the reliable delivery of video frames over a wireless link. For this scenario, we consider energy minimal transmissions of a video frame with reliability and deadline constraints. In wireless video transmissions, encoded video frames are often large in data load and truncated into many transport packets for reliable transmissions. These transport packets need to be delivered before a deadline at a target reliability which depends on the importance of the video frame. We seek the optimal transmission strategy that minimizes the transmission energy under the reliability and deadline constraints of a group of transport packets. We investigate this problem for transmissions in the slow fading channel and the Rayleigh fading channel. In the slow fading channel, we propose a solution algorithm which can be solved e±ciently. In the Rayleigh fading channel, we derive probabilistic transmission policies which exploit the fading statistics to achieve the energy e±ciency gain. Finally, we further extend our focus to joint video coding and transmission. For this scenario, we consider the rate and power allocation framework of joint coding and transmission in wireless video chat applications. Unlike video streaming which is generally one way tra±c, video chat features distributed two way tra±c relayed via a base station. The base station imposes service charges on the clients for relaying their video bitstreams. For the wireless clients, we propose a complexity scalable video coding and its power-rate distortion (PRD) modeling. Based on the PRD model, we derive the optimal rate and power allocation scheme to minimize the video distortion and the network service charge under a power constraint. For the base station, we propose a power-aware hybrid pricing scheme that ensures balance between fairness and e±ciency in network utility, and that the base station operates a surplus budget. We show that the network dynamics can be analyzed in the Stackelberg game framework, and our proposed solutions converge to the Stackelberg equilibrium.
DRNTU::Engineering::Computer science and engineering::Information systems::Information systems applications