交互式VOD视频源非线性智能分段算法

A Non-Linear Algorithm for Video Segmentation on Interactive Video-On-Demand System

为了实现视频点播系统的交互性功能 ,经典的方法是将视频源均匀分割存储 ,以实现其有限的交互性 .从VOD系统的实际出发 ,该文提出了非线性智能分段方法NLSM .它的核心思想是在一段时间内对一定步长范围内VCR交互性操作的到达情况进行统计求和 ,计算系统总体时频总面积 ,并按信道数目将各时间段与VCR发生频率的乘积把总面积均匀等分 ,从而实现视频源在时间轴上的非线性分割 .针对均匀的周期性广播分段方法ESM方案进行比较仿真研究 ,获得了用户VCR请求的响应等待时间的理想结果和最佳的用户等待时间标准偏差、系统拒绝概率和信道吞吐量等方面的优良性能 .NLSM算法对历史信息进行数据挖掘 ,揭示用户对视频节目内容感兴趣程度的潜在分布规律 ,具有良好的自适应能力和学习智能 .算法性能分析和仿真实验的结果说明该方法在实际应用中是可行且高效的 .

To realize interactive performance of a Video-On-Demand system, a traditional method to segment a video with equal fragments for storage is adopted to implement limit interactive performance of a VODs. According to the reality of VODs, A non-linear segmentation method(NLSM) presented by us was adopted in the paper. The key of NLSM is to realize non-linear segmentation of a video in time axis. After computing the total area of time and frequency, NLSM will equally segment the product of interval and frequency of VCR-Requests arrival according to statistic sum of VCR operation in certain step width in a period of time. Comparing with periodical broadcasting by equal segmentation method(ESM) in our simulations, we obtained good outcomes in the mean latency of VCR requests and the best performance of standard deviation of user latency, system defection probability and throughput of a channel. It is found that NLSM can make a data mining in historical information and reveal potential distribution law of users interest in video content. This scheme has good adaptive ability and learning intelligence. By the performance analysis and empirical results, it could be verified that the scheme is feasible and efficient.