基于多步下溢概率估计的自适应媒体播放算法

Adaptive Media Playout Based on Multi-step Underflow Probability

自适应播放(AMP)是一种通过动态调整播放速率减少这种中断的技术。多数AMP算法都基于缓存的满溢度或者其变化调整播放速率,其难点在于如何选择合适的调整门限值。本文算法,利用滑动窗口估计到达帧率的经验分布,再利用统计计算方法估计多步缓存下溢概率,然后根据一步和多步下溢概率调整播放速率。仿真实验证明在一般的信道假设和多种片源的测试中,本算法较其他同类算法有更好的表现。

Adaptive Media Playout（AMP） is an effective method to reduce the playout interruptions of the video streaming applications. Most proposed AMP algorithms trigger the adjustment of playback speed based on the buffer＇ s fullness or its variations. Where the conundrum is hou to select the appropriate threshold. Our AMP algorithm firstly estimates the empirical distribution of arrival frame rate in sliding windows and then works out the one - step and multi - step underflow probabilities using statistical method. Finally, the play- back speed is altered according to the underflow probabilities. Experiments with various encoded video sources and general VBR channels show our AMP algorithm outperforms other congeneric algorithms.