| 摘要: |
| 视频用户体验的质量要求给网络传输带来了巨大挑战，未来的虚拟现实/增强现实(Virtual Reality/Augmented Reality，VR/AR) 视频业务则对交互时延提出了更高要求。视频业务识别是网络运营商及视频业务提供商对视频进行优先网络资源分配和自适应速率控制的基本前提。综述了业务识别领域的一般研究方法及其对视频业务的具体应用。首先，总结了每种业务识别方法的工作原理，对比了这些方法应用于视频业务时的优缺点及在准确率和召回率指标上的识别性能;然后，特别介绍了可以提升识别准确率和计算性能的特征选择算法;最后，阐述并分析了高吞吐量网络分类速度不够、无成熟视频业务识别平台和样本分布不均衡等视频业务识别面临的挑战，并针对现有业务识别方法存在的问题，给出了引入移动边缘计算、基于端口的方法串联其他识别方法等可能的解决方案。 |
| 关键词: 视频业务 业务识别 交互时延 资源分配 特征选择 |
| DOI： |
|
| 基金项目: |
|
| A survey on traffic identification methods for video services |
| LIU Xiaoyu,WANG Yumei,LIU Yu |
| (School of Information and Communication Engineering,Beijing University of Posts and Telecommunications,Beijing 100876,China) |
| Abstract: |
| User′s requirements for high quality video have brought great challenges for network transmission.Future virtual reality/augmented reality(VR/AR) video services will impose higher requirements on interaction delay.For network operators and video service providers,video traffic identification is a precondition for network resource allocation and adaptive rate control of video services.This paper reviews the basic methods in the field of traffic identification and applies them into video services.Firstly,the different methods for traffic identification are summarized,and the advantages,disadvantages,and the identification performances are compared in terms of accuracy and recall of these methods for video services.Then,a feature selection algorithm that can improve identification accuracy and computational performance is introduced particularly.Finally,the challenges of video traffic identification are described and discussed,e.g.,insufficient classification speed in high throughput networks,absence of mature video traffic identification platform,unbalanced sample distribution,and so on.According to the advantages and disadvantages of the existing traffic identification methods,possible solutions such as the introduction of mobile edge computing,and the combination of the port-based method and other traffic identification methods are proposed . |
| Key words: video service traffic identification interaction delay resource allocation feature selection |
