【作者】 马志龙；

【导师】 周敬利；

【作者基本信息】 华中科技大学 ， 计算机系统结构， 2009， 博士

【摘要】 随着计算机技术、压缩技术以及通信技术的发展,流媒体业务得到了飞速的发展和应用。而Internet的无连接包转发机制主要是为突发性的数据传输而设计的,并不适用于对连续媒体流的传输。为了在Internet上有效的、高质量的传输流媒体,还需要多种技术的支持,例如流媒体编解码技术、传输控制技术、容错技术、同步技术和相关协议等。传统的网络应用通常以网络QoS(Quality of Service)参数来描述相应的服务质量,并不能完全适用于流媒体应用系统。从流媒体服务系统的角度来说,流媒体服务的最终目的是提供给用户满意的感知质量(QoE,Quality of Exeperience),QoE能够比较准确的反映终端用户端对流媒体服务的满意程度,例如图像是否清晰流畅、声音是否连续等。因此针对用户感知质量的优化来进行流媒体传输服务控制是研究的方向。研究内容包括以下几个方面:介绍了流媒体服务质量控制的相关理论和技术,提出了一个基于用户感知质量优化的流媒体服务系统框架。该服务框架中最重要的部分是QoE服务质量控制模块,包括QoE决策、QoE质量映射和服务区分三个子模块。QoE决策是针对流媒体应用定义其QoE指标参数,针对视频流应用,通常使用图像失真作为其QoE指标参数。QoE质量映射是指将当网络服务参数映射为QoE指标参数,将用户的QoE需求转化成对传输QoS服务的需求,可以协助流服务系统对QoE质量进定量的分析。区分服务则是根据媒体流数据包对QoE影响的不同对数据进行了分层处理,在传输系统中提供不同质量的服务。结合网络反馈参数和QoE服务质量模块的分析结果,框架中的自适应拥塞控制模块和自适应错误控制模块采取相应的优化控制策略,对传输服务进行有针对性的调整。该服务系统框架的特点是结合QoE质量分析进行传输策略优化控制,实现了高质量、高效率的流媒体传输服务。研究并实现了一个通用的视频QoE质量映射模型。研究了视频数据包丢失导致的图像失真和由于帧间解码依赖关系导致的错误扩散失真,结合视频错误隐藏技术进行建模分析,实现了一个通用的视频流QoE映射模型,也叫做传输失真模型。该模型在编解码算法和错误隐蔽算法确定的情况下,在发送端通过网络Qos参数和视频编码参数估算视频流媒体的QoE质量损失,能够为同步控制、传输控制等算法提供定量的参考指标。QoE质量映射模型实现了QoS参数到QoE的模糊映射,能够为视频流应用提供比较准确的定量分析依据。研究了视频流媒体拥塞控制机制,提出并实现了基于QoE优化的自适应拥塞控制算法及改进算法。自适应拥塞控制算法根据编码宏块的重要性测定参数将数据包进行分层处理,在进行拥塞控制的时候,主动丢弃重要性较低的数据包,尽量保证高重要性的数据包优先传输,使得QoE质量最优。改进的拥塞控制算法则是使用前向扫描和后向扫描的预取数据传输机制,保证了QoE质量的平稳性。实验证明基于QoE优化的自适应拥塞控制算法及改进算法相比其它同类算法,恢复图像的平均PSNR有所提高。研究了视频流媒体传输错误控制机制,提出并实现了基于QoE优化的自适应FEC算法。算法根据传输层反馈参数计算冗余数据可用的最大带宽,根据编码宏块的重要性测定参数对打包数据提供了不同的服务策略。包括对可自然恢复的数据包不增加冗余数据,对不同重要性的编码宏块分配相应比例的冗余数据,根据编码宏块重要性进行打包策略优化。实验证明了基于QoE优化的自适应FEC算法在保证终端QoE质量的同时,能够减少网络开销。在同等网络条件下,相比其它同类算法能够提高视频QoE质量,恢复图像的平均PSNR值提高了0.1-2dB。更多还原

【Abstract】 Along with deeply development of computer technology, compression technology and communications technology, the Streaming Media Business became fast developed and widespread used。But currently there are a lot of difficulties to transfer Streaming Media by Internet; the main reason is that the Internet was not designed for continuously connectionism Streaming Media transfer, but only for paroxysmal data transfer. For the purpose to efficiently transfer Streaming Media with high quality, there are many kinds of technology supports are needed, such as Streaming Media coding technique, QOS technique, fault-tolerant technique, simultaneous techniques and relevant agreements etc.Raditional network applications usually use net QOS parameter to describe the corresponding service quality, but cannot absolutely apply to Streaming Media Application System. In terms of Streaming Media service system, the final purpose of it is to provide sensible quality QOE for customers’ satisfaction, sensible quality QOE can be much more precisely mirror the end-user satisfaction about Streaming Media service, such as if the image clear and fluently transfer, or the voice continuously transfer etc. Therefore direction of research is to achieve the Streaming Media service system control with end-user sensible quality QOE.The contents of research as follows:This article particularly presents the theory and technology of Streaming Media service quality control system, advances a theory of Streaming Media service system frame based on QOE sensible quality optimization. Which is the most important improvement from that frame is to join QOS network service parameter into QOE sensible service quality mapping module and Streaming Media distinguish module, service quality mapping module will make the user’s QOE needs into the service needs of network’s QOS, provide assistance to system on the quantitative analysis of QOE. Differentiated Services Module can offer the differentiated service on Media internal data according to the importance of Streaming Media contents, the differentiated service can also assist system to have a pointed revision on service part by cache memory scheduling, adaptive transmission, error control those streaming media control mechanism. As a primary objective the frame improves the quality of user’s QOE perceived, and a wide range of practical value.After research, found the realization of a model which points against of QOS parameter to QOE sensible quality mapping model by the video Streaming Media transmission, also known as the transmission distortion model. Studied the common technology based on macro video streaming media codec technology, analysis of the video frame codec between the dependence caused by the proliferation of video streaming error distortion, the final combination of video streaming technology for a unified error analysis of hiding, to achieve the predictive coding based on video streaming media transmission distortion model. The model is able to codec algorithms and error concealment algorithm to determine the circumstances in the sending end through the network QOS parameters and parameter estimation video coding video streaming media QOE mass loss, can simultaneously control and transmission control algorithms and other quantitative indicators. QOS parameters of network services to QOE mapping perceived quality streaming media service module is the most important service module, as well as the video streaming services is the most complex and commonly used services. This research based on predictive coding of video streaming media transmission distortion model parameters to achieve the QOS service to the QOE perceived quality of the fuzzy mapping, to provide similar quantitative analysis for the current vast majority network of video streaming media applications.Research on the video streaming media congestion control mechanism, designed and implemented to optimize the QOE-based adaptive congestion control algorithm and improved algorithm. Algorithm based on quantitative analysis of video streaming media transmission from the QOS parameters to QOE perceived quality mapping , differentiated services based on macro block coding QOE perceived quality of the contribution. Algorithm based on the importance of hierarchical data processing parameters, in a time when deal with congestion control, as far as possible to ensure the best quality of QOE end. Experiments show that the control algorithm can effectively improve the performance of streaming media transmission and the image quality of the client, and it can also be effectively enhanced the QOE quality of video streaming.Research on the video streaming media transmission error control mechanism, a new algorithm was designed and implemented to optimize the QOE-based adaptive FEC algorithm. Algorithm based on quantitative analysis of video streaming media transmission from the QOS parameters to QOE perceived quality mapping , differentiated services based on macro block coding QOE perceived quality of the contribution. Mainly including the corresponding proportion of redundant data by the importance of coding for different macro block allocation, in accordance with the importance of encoding macro blocks to pack up the optimization strategy. Experiments show that the control algorithm can effectively improve the performance of streaming media transmission and the image quality of the client, and it can also be effectively enhanced the QOE quality of video streaming.更多还原