Unequal channel protection of multiple description codes for wireless broadcast applications

详细信息	查看全文 | 推荐本文 |

• 作者：Abdul Bais ; ^{abdul.bais@uregina.ca} ; ^{baisabdul@yahoo.com} ; Tanay Dey ^{dey204@uregina.ca} ; Nima Sarshar ^{nima.sarshar@uregina.ca}
• 关键词：Wireline and wireless hybrid networks ; Multimedia broadcasting and streaming ; Multiple-description coding ; Error correction ; Quality of service ; Convex optimization
• 刊名：Signal Processing ; Image Communication
• 出版年：2012
• 期刊代码：110_09235965
• 类别：cp
• 出版时间：July, 2012
• 卷：27
• 期：6
• 页码：625-636
• 文件大小：731 K

摘要

We investigate the problem of optimal channel protection of multiple-description coded (MDC) multimedia contents at a wireless access point (WAP). For each MDC packet, the WAP has the option to protect and broadcast the packet using one of the available channel coders, or to drop the packet altogether. The goal is to optimize the channel protection strategy at packet level to maximize the expected reconstruction quality at wireless clients (in a rate-distortion sense), subject to the total bandwidth budget available to the WAP. For a fixed FEC-based MDC, we show how this optimization can be approximated by a convex optimization problem with linear constraints, and thus, can be solved efficiently. We verify the validity of our results through extensive simulations of a wireless image streaming system that employs multiple turbo channel coders to wirelessly broadcast images, where we report gains of more than 2.50 dB PSNR. Interestingly, while the MDC is balanced, and therefore all packets have equal importance in the reconstruction, the optimal channel protection may require unequal error protection of packets. In other 0words, some packets may have to be protected with stronger channel coders. Next, we consider the case where the design of the MDC itself can also be optimized. We investigate the joint optimization of the channel protection strategy and the design of the MDC through an iterative approach. Our simulations show further gains of up to 3.32 dB PSNR using this joint optimization.