多媒体业务OoE建模及会话流分类

发布时间：2018-11-22 07:51

【摘要】：随着互联网技术的飞速发展、多媒体业务也迎来了发展的高潮,逐步多样化。直播视频、2D游戏、3D游戏等大量高速率、高吞吐量、多用户并发的业务在网络中产生了大量的混合业务流。从网络服务提供商的角度来说,怎样在有限的网络资源中,高效地对混合流进行识别分类,以及提供较好的多媒体业务感受质量是需要认真思考的问题。多媒体业务中采用服务质量(QoS,Quality of Service)和体验质量(QoE,Quality of Experience)来表征网络中相关参数和用户的感知状况。多媒体流根据业务形式分成四种流:交互类、背景类、流类、会话类。不同的类型其QoS参数的影响程度不同。本文从会话类流入手,考虑其低时延、时延抖动低、吞吐量大的特点,进而分析建立其QoS-QoE的模型,得出不同QoS特征下与QoE的MOS(Mean Opinion Score)值概率分布关系。并把其中最明显的特征概率分布以网络流特征的形式加入到会话流原有的特征集当中,对QQ视频、2D游戏、3D游戏、Skype语音四种会话流进行分类。本文在会话类业务流的基础上进行考虑,结合该类业务流的特点,综合选取了时延抖动、吞吐量、平均包间隔三个具有代表性的网络QoS参数,同时对已有的模型进行改进,将模型中的抽象QoS具体为三个参数,并设置权重系数和参数影响因子,综合得出在不同情况下,特征与MOS值概率分布的关系。随后从统计角度进行分析,进而将MOS值概率分布以众数的形式加入到多媒体混合流分类特征集中,以研究对会话类混合流的分类效果。通过实验数据表明:改进的QoE模型能够提高多媒体混合业务流分类的准确率。同时,能得出具体的QoE模型函数表达式。从而验证了思路的正确性。
[Abstract]:With the rapid development of Internet technology, multimedia business also ushered in the climax of development, gradually diversified. Live video, 2D games, 3D games and other high speed, high throughput, multi-user concurrent services in the network has produced a large number of mixed traffic. From the point of view of network service provider, how to identify and classify the mixed flow efficiently in the limited network resources, and how to provide better multimedia service feeling quality are the problems that need to be seriously considered. Quality of Service (QoS,Quality of Service) and experience quality (QoE,Quality of Experience) are used to represent the related parameters and user perception in multimedia services. Multimedia streams are divided into four streams according to business form: interactive class, background class, stream class, session class. The influence of QoS parameters varies with different types. In this paper, we consider the characteristics of low delay, low delay jitter and large throughput of session class inflow hand, and then analyze and establish its QoS-QoE model, and obtain the probability distribution relationship between MOS (Mean Opinion Score) value and QoE under different QoS characteristics. The most obvious feature probability distribution is added to the original feature set of session flow in the form of network flow features. The four session streams of QQ video, 2D game, 3D game and Skype speech are classified. Considering the characteristics of session traffic, this paper selects three typical network QoS parameters, such as delay jitter, throughput and average packet spacing, and improves the existing model. The abstract QoS in the model is divided into three parameters, and the weight coefficient and the parameter influence factor are set up, and the relationship between the characteristics and the probability distribution of MOS value is obtained in different cases. Then from the statistical point of view, the probability distribution of MOS value is added to the multimedia hybrid stream classification feature set in the form of mode to study the classification effect of session class mixed flow. Experimental data show that the improved QoE model can improve the classification accuracy of multimedia mixed traffic. At the same time, the concrete expression of QoE model function can be obtained. Thus, the correctness of the train of thought is verified.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.0

【相似文献】