摘要：

根据终端拉服务模式的码流自适应匹配技术，提出一种实时自适应HTTP流化码流切换算法。利用滑动窗口分析最近视频分片的下载时间，采用假设检验方法确定在滑动窗口内是否发生速率跳变，以此估计平均下载速率和方差。应用Ek/D/1/N排队模型计算缓存区下溢的概率，对下溢风险和码流质量分别赋以损益值和收益值，计算出综合收益最高的码流作为目标码流。仿真结果证明，该算法能根据网络可用带宽实时调整传输码流，具有良好的码率跟踪性能。

关键词: HTTP流化, 码流切换, 滑动窗口, 假设检验, Ek/D/1/N排队模型

Abstract:

According to the adaptive bitstream matching technology of terminal services pull mode, this paper proposes a real-time adaptive bitstream switching algorithm for HTTP streaming. It uses the sliding window to analyze the time for downloading each video segmen. Hypothesis testing is used to judge whether there is a rate jump, which degrades the estimation significantly, in present observing window. The Ek/D/1/N queue is employed to model the number of segments in the receiver buffer and to figure out the underflow probability. Customized values of loss and benefit are set to underflow risk and stream quality respectively. A stream with optimal integrated benefit is considered to be the target stream. Experimental results show that the proposed algorithm can determine the appropriate downloading stream based on the current available bandwidth, with good performance of rate adaption.

Key words: HTTP streaming, bitstream switching, sliding window, hypothesis test, Ek/D/1/N queue model

中图分类号: 

• TP391