企业级视频会议的服务质量保证研究

发布时间：2018-03-11 20:03

本文选题：企业级　切入点：视频会议　出处：《华南理工大学》2014年博士论文　论文类型：学位论文

【摘要】：视频会议系统为与会者提供一个多媒体的实时交互环境,使得人们能够远程面对面交流,方便人们交互信息,提高工作效率,被广泛应用于政府、商务、教育等行业。新一代视频会议使用互联网作为传输平台,能提供高清晰度视频、音频、文字等多种灵活的交互方式,这对传输过程中的带宽、延时、抖动等参数有更高的要求。但是现有的互联网尚不具备这样的服务质量保证能力。该矛盾成为当前新一代视频会议面临的服务质量挑战。本文针对互联网及其之上的视频会议系统,在企业级网络的通信环境下,展开了视频会议服务质量保证研究。1、提出一个企业级视频会议服务质量保证系统(EVCQoS)的框架,并建立它的形式化体系结构,从数学上分析EVCQoS系统模型的连通性和行为特性。为网络传输QoS和码率适配的解决方案提供理论依据。2、以EVCQoS体系结构为理论依据,构建一个基于网络的视频会议QoS机制。机制首先建立资源矩阵(资源模型),组织网络和会议的资源,以应对各种会议对单播、组播等资源的需求。监视和控制企业域内的网络资源,在网络的边界使用支持SAVI的交换机来控制合法流量的进入,在网络核心采用资源预约方式来管理带宽的使用。最终实现视频流的传输服务质量保证。3、对于传输网络不能提供足够的网络资源的情况,使用基于应用层面的服务质量保证机制,本文针对这一问题,提出了基于QoE的码率适配算法。采用H.264 SVC对视频编码,利用H.264 SVC编码算法能够在分辨率、帧率、视频质量三个维度伸缩的特点,来提供视频码率调整手段。对于给定带宽,使用伪主观的神经网络模型来评估视频的QoE质量,根据网络资源的情况,选择最优的编码参数对视频编码,以实现用户QoE的最大化。4、服务质量保证系统使用神经网络QoE模型来优化视频编码,也采用该模型来监视视频会议的运行情况。神经网络QoE模型的训练需要大量的主观测试数据,成本较高并且无法实时获得。本文针对这一问题,首先研究了环境不一致情况下QoE模型的适用性问题,实验结果显示,在环境不一致情况下QoE预测模型的正确率明显下降。然后提出一种QoE模型适配训练方法,在已经训练好的QoE模型基础上,使用一个线性转换将参数矢量进行适配后再输入ANN模型,使用较少的主观数据能够训练出应对不同评估环境的QoE模型。基于上述研究,本文实现了EVCQoS系统。实验结果表明,EVCQoS系统能够在CERNET2上保证企业级视频会议的传输服务质量和应用服务质量。基于网络的QoS机制和码率适配这两项关键技术具有互补性,当网络通信资源不足而导致传输网络QoS保证技术不能提供足够资源情况,基于QoE的码率适配技术仍能尽力保证视频会议的服务质量。
[Abstract]:The video conference system provides a multimedia real-time interactive environment for the participants. It enables people to communicate face to face remotely, facilitate people to exchange information, improve work efficiency, and is widely used in government and business. Education and other industries. The new generation of video conferencing uses the Internet as a transmission platform, can provide high-definition video, audio, text and other flexible interaction, this process of transmission bandwidth, delay, Jitter and other parameters have higher requirements, but the existing Internet does not yet have such a quality of service assurance capability. This contradiction has become a new generation of video conferencing facing the quality of service challenges. This paper aims at the Internet and its. Videoconferencing system based on. In the communication environment of enterprise network, the research on quality of service assurance of videoconferencing is carried out, and a framework of enterprise quality assurance system of videoconference service quality is proposed, and its formal architecture is established. The connectivity and behavior characteristics of the EVCQoS system model are analyzed mathematically. It provides a theoretical basis for the solution of network transmission QoS and bit-rate adaptation, and takes the EVCQoS architecture as the theoretical basis. A Web-based QoS mechanism for videoconferencing is constructed. Firstly, the resource matrix (resource model) is set up to organize the resources of the network and the conference, in order to deal with the unicast of all kinds of meetings. Monitor and control the network resources in the enterprise domain, use the switch supporting SAVI to control the access of the legitimate traffic at the boundary of the network. In the core of the network, resource reservation is adopted to manage the use of bandwidth. Finally, the quality of service (QoS) of video stream is guaranteed. 3, which can not provide enough network resources for the transmission network. In order to solve this problem, this paper proposes a bit-rate adaptation algorithm based on QoE, which uses H.264 SVC to encode video and H.264 SVC coding algorithm to solve this problem. For a given bandwidth, a pseudo-subjective neural network model is used to evaluate the QoE quality of video. The optimal coding parameters are selected to maximize the user's QoE. The QoS assurance system uses the neural network QoE model to optimize the video coding. The training of neural network QoE model requires a lot of subjective test data, which is expensive and can not be obtained in real time. In this paper, the applicability of QoE model under environmental inconsistency is studied. The experimental results show that the correct rate of QoE prediction model decreases obviously under the circumstance of environmental inconsistency. Then, a training method for QoE model adaptation is proposed. Based on the trained QoE model, a linear transformation is used to adapt the parameter vector and then input into the ANN model. Using less subjective data, the QoE model can be trained to deal with different evaluation environments. In this paper, the EVCQoS system is implemented. The experimental results show that the EVC QoS system can guarantee the quality of service (QoS) and application quality of service (QoS) of enterprise video conferencing on CERNET2. The two key technologies of network-based QoS mechanism and bit rate adaptation are complementary. When the network communication resources are insufficient, the transmission network QoS guarantee technology can not provide enough resources, the bit-rate adaptation technology based on QoE can still try its best to ensure the quality of service of video conference.
【学位授予单位】：华南理工大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN948.63;TN919.81

【参考文献】