Cloud-assisted real-time transrating for http live streaming
Cloud-assisted real-time transrating for http live streaming
复制标题
DOI：
10.1109/mwc.2013.6549284
复制
发表时间：
2013-07
影响因子：
12.9
通讯作者：
Chin-Feng Lai;H. Chao;Ying-Xun Lai;J. Wan
中科院分区：
文献类型：
--
作者：
Chin-Feng Lai;H. Chao;Ying-Xun Lai;J. Wan
随着智能手机、平板电脑等手持设备的普及，用户对网络媒体流媒体服务的需求越来越大。然而，在流媒体播放过程中，网络带宽不稳定往往会导致播放卡顿或延迟，降低用户体验。本文提出了一种云辅助实时转码机制，该机制包含HTTP直播流协议、编码模式转换状态机和三种错误模式的带宽评估。在所提出的机制中，可以根据带宽评估结果来观察当前网络的方差；如果在不同的模式下采用不同的转码策略，就能够计算出合适的媒体分段码率，并转码成满足当前带宽条件的分段，使得用户可以通过HTTP重定向技术自动获得合适的媒体质量。利用云计算，用户可以根据自己的需要，将不同的媒体片段转码到不同的转码器中，利用该技术提高转码效率，实现实时转码。最后，分析峰值信噪比（PSNR）和带宽利用率，作为媒体质量改善的参考指标。实验结果显示，在一般网络行为下PSNR可提升约2.8dB，流媒体时带宽利用率可维持在80%以上。
With the increasing trend for hand-held devices, such as the intelligent mobile phone and tablet PC, users have an increasing demand for network media streaming service. However, in the course of streaming media, the unstable network bandwidth often causes frozen play or delay to degrade user experience. A cloudassisted real-time transcoding mechanism is proposed in this paper, which contains HTTP live streaming protocol, a coding mode transition state machine, and three bandwidth evaluations of error patterns. In the proposed mechanism, the variance in current network is able to be observed according to the bandwidth evaluation results; if different transcoding strategies are used in different modes, the appropriate media segment bit rate is able to be calculated, which is transcoded into the segment to meet the current bandwidth conditions, it makes the users can obtain appropriate media quality automatically through the HTTP redirection technique. Using cloud computing the user will take different transrating media clips into different transrators according to the users' needs, using this technology to increase the transrating efficiency to achieve real-time transrating. Finally, the peak signal-to-noise ratio (PSNR) and bandwidth utilization rate are analyzed as the reference index of improvement in media quality. According to the experimental results, the PSNR can be increased by about 2.8 dB in general network behavior, and the bandwidth utilization rate can be maintained above 80 percent during streaming.