摘要
针对腾讯QQ语音应用的识别问题,建立了客户源端模型,并对流量进行了离线分析。在此基础上,提出了基于流量统计的两种识别方法:Bayesian识别法和心跳识别法。一方面,利用QQ语音应用的源端编码特点,得出语音数据流量的本质特征,并使用Bayesian理论进行检测。另一方面,针对QQ语音应用中一些非语音数据包的周期性特点,利用卡方检验,提出了心跳识别的方法。实验结果表明,这两种方法综合使用,可以很有效地实时检测QQ语音流量。
This paper establishes the source model of QQ voice and analyzes its traffic offline in order to identify the QQ voice traffic. Based on that, it proposes two identification methods： the Bayesian method and the heartbeat method. On the one hand, the theory of Bayesian is employed to identify the essential character of voice traffic, which is extracted according to the encoding at source. On the other hand, the heartbeat method is presented based on the Chi-square test by identifi- cation of the periodic feature of non-voice traffic. The experiment results show that the integrated deployment of the two methods is very effective for real-time detection of QQ voice traffic flow.
出处
《高技术通讯》
EI
CAS
CSCD
北大核心
2010年第1期38-44,共7页
Chinese High Technology Letters
基金
863计划(2007AA01Z220)
天津市自然科学基金(08JCYBJC14200)资助项目
