中文摘要：

针对平滑流（SF）算法在带宽预测时存在的“毛刺”现象以及仅依靠带宽预测而没有缓存区控制所导致的频繁播放停滞的问题,提出一种动态自适应混合控制码率算法。首先,通过使用标准差来代替原SF算法中波动参数的计算,消除了预测带宽存在的“毛刺”现象;其次,针对原SF算法存在没有考虑缓存区状态所带来的频繁播放停滞问题以及传统缓存区控制算法存在分级困难的问题,引入一种基于缓存区容量的新型缓存区控制策略;最后,将改进SF算法与新型缓存区控制策略相结合形成混合算法来选择视频码率。实验结果表明,混合算法不但消除了原SF算法在带宽预测中存在的毛刺现象,而且弥补了仅依靠单一算法选择码率的不足,使得所选码率视频既降低了播放停滞频率（恶劣网络环境下效果明显,降低43%左右）又符合实际网络情况,提高了用户的观看体验。

英文摘要：

Concerning the problem that Smooth Flow （SF） algorithm in bandwidth prediction has flash phenomenon and frequent play stagnation caused by bandwidth prediction without cache control, a dynamic adaptive hybrid rate control algorithm was proposed. First of all, calculation of fluctuation parameter in original SF algorithm was replaced by standard deviation, which Can eliminate the flash phenomenon. Secondly, for the original SF algorithm, there is frequent play stagnation problem because the bandwidth prediction does not consider the cache state, at the same time, the traditional cache control method has problem with hierarchical difficulties, a new cache control strategy was intodueed to solve these problems. Finally, the improved SF algorithm was combined with the new cache control strategy to form a hybrid algorithm to select the video bitrate. The experimental results show that the hybrid algorithm not only eliminates the flash phenomenon of SF algorithm in bandwidth prediction, but also overcomes the shortcoming of selecting bit rate by only single algorithm; the selected video not only reduces the frequency of play stagnation （the frequency was significantly decreased by about 43% under bad network environment）, but also obey the actual network situation, improving the users＇ viewing experience.