应用科学学报 ›› 2015, Vol. 33 ›› Issue (2): 117-128.doi: 10.3969/j.issn.0255-8297.2015.02.002

• 通信工程 • 上一篇    下一篇

无监督的视频业务特征分析与分类

姚利涛, 董育宁   

  1. 南京邮电大学通信与信息工程学院,南京210003
  • 收稿日期:2014-07-17 修回日期:2014-12-16 出版日期:2015-03-30 发布日期:2014-12-16
  • 作者简介:董育宁,教授,博导,研究方向:多媒体通信与无线网络,E-mail: dongyn@njupt.edu.cn
  • 基金资助:

    国家自然科学基金(No.61271233, No.60972038);教育部博士点基金(No.20103223110001)资助

Unsupervised Feature Analysis and Classification of Video Streams

YAO Li-tao, DONG Yu-ning   

  1. College of Telecommunications & Information Engineering, Nanjing University of Posts and
    Telecommunications, Nanjing 210003, China
  • Received:2014-07-17 Revised:2014-12-16 Online:2015-03-30 Published:2014-12-16

摘要: 基于机器学习的流统计特征识别的方法关键在于如何找到具有区分力度的业务流统
计特征. 为此,提出了一些能够较好地区分视频业务的QoS 相关的统计特征. 为了充分地发
挥多级聚类算法的优势,以灵活的特征选择策略标记不同层级的网络视频流,通过大量的真
实网络视频数据进行实验验证. 结果表明,该方法能比现有同类方法取得更高的分类准确率.

关键词: 视频流, 统计特征, QoS, 流分类, 多级聚类

Abstract:  For recognition of flow statistical features based on machine learning, the key is
to select distinguishable features of different traffic flows. This paper presents several QoSrelated
statistical features that can well discriminate video traffics. To make full use of the
advantages of hierarchical clustering algorithm, this paper uses a flexible feature selection
strategy to mark the network video streaming of different levels. Experiments are performed
on a large scale real network video data. The results show that the proposed method can
achieve significantly higher classification accuracy compared to existing methods.
Keywords: video streaming, statistical features, QoS, traffic classification, hierarchical
clustering

中图分类号: