互联网电视(over the top,OTT)视频业务逐渐成为最流行的在线业务之一,然而网络视频往往由于网络质量差、服务平台过载等原因,出现播放失败、卡顿次数增加、缓冲时间过长等质量问题,导致用户感知质量(quality of experience,QoE)下降.因...互联网电视(over the top,OTT)视频业务逐渐成为最流行的在线业务之一,然而网络视频往往由于网络质量差、服务平台过载等原因,出现播放失败、卡顿次数增加、缓冲时间过长等质量问题,导致用户感知质量(quality of experience,QoE)下降.因此,运营商需要精确评估和掌握用户在使用网络视频业务过程中的质量体验,以便提前发现质量问题,进一步开展网络和业务优化工作.为了解决该问题,提出一种基于用户呼叫/事务/会话记录数据(extend data record,XDR)的无参考网络视频质量评估方法.该方法从大量XDR数据中提取出与视频质量相关性高的少量信息,将大规模、低价值的XDR话单数据转化为高价值、小规模的视频质量特征信息,有利于后续人工智能算法的应用和视频业务质量评价,降低进一步数据挖掘的资源成本,提升机器学习的输入样本质量和QoE评价结果的准确性.实验表明:使用该方法提取后的数据进行QoE预测,得到的预测结果在准确性方面明显优于目前基于原始XDR数据的QoE机器学习评估方法.展开更多
本文主要介绍了美国研究图书馆协会(ARL)提供的图书馆评价工具StatsQUAL系列中的LibQUAL+、MINES for library方法与应用进展,针对海南省高校图书馆综合评估和电子资源评估现状,探讨了StatsQUAL评价工具对海南省高校图书馆评估的意义和...本文主要介绍了美国研究图书馆协会(ARL)提供的图书馆评价工具StatsQUAL系列中的LibQUAL+、MINES for library方法与应用进展,针对海南省高校图书馆综合评估和电子资源评估现状,探讨了StatsQUAL评价工具对海南省高校图书馆评估的意义和借鉴价值,并提出个人建议。展开更多
The effective radio resource allocation al-gorithms, which satisfy diversiform requirements of mobile naltimedia services in wireless cellular net-works, have recently attracted more and more at-tention. This paper pr...The effective radio resource allocation al-gorithms, which satisfy diversiform requirements of mobile naltimedia services in wireless cellular net-works, have recently attracted more and more at-tention. This paper proposes a service-aware scheduling algorithm, in which the Mean Opinion Score （MOS） is chosen as the unified metric of the Quality of Experience （QoE）. As the network needs to provide satisfactory services to all the users, the fairness of QoE should be considered. The Propor- tional Fair （PF） principle is adopted to achieve the trade-off between the network perfonmnce and us- er fairness. Then, an integer progranming problem is formed and the QoE-aware PF scheduling princi-ple is derived by solving the relaxed problem. Simu-lation results show that the proposed scheduling principle can perform better in terms of user fair-ness than the previous principle maximizing the sum of MOS. It also outperfoms the max-rain scheduling principle in terms of average MOS.展开更多
文摘互联网电视(over the top,OTT)视频业务逐渐成为最流行的在线业务之一,然而网络视频往往由于网络质量差、服务平台过载等原因,出现播放失败、卡顿次数增加、缓冲时间过长等质量问题,导致用户感知质量(quality of experience,QoE)下降.因此,运营商需要精确评估和掌握用户在使用网络视频业务过程中的质量体验,以便提前发现质量问题,进一步开展网络和业务优化工作.为了解决该问题,提出一种基于用户呼叫/事务/会话记录数据(extend data record,XDR)的无参考网络视频质量评估方法.该方法从大量XDR数据中提取出与视频质量相关性高的少量信息,将大规模、低价值的XDR话单数据转化为高价值、小规模的视频质量特征信息,有利于后续人工智能算法的应用和视频业务质量评价,降低进一步数据挖掘的资源成本,提升机器学习的输入样本质量和QoE评价结果的准确性.实验表明:使用该方法提取后的数据进行QoE预测,得到的预测结果在准确性方面明显优于目前基于原始XDR数据的QoE机器学习评估方法.
基金This paper was supported partially by the Program for New Century Excellent Talents in University under Crant No. NCET-11-0600 the National Natural Science Foundation of China under Crant NN76022 and the France Telecom R ＆ D Beijing Co. Ltd.
文摘The effective radio resource allocation al-gorithms, which satisfy diversiform requirements of mobile naltimedia services in wireless cellular net-works, have recently attracted more and more at-tention. This paper proposes a service-aware scheduling algorithm, in which the Mean Opinion Score （MOS） is chosen as the unified metric of the Quality of Experience （QoE）. As the network needs to provide satisfactory services to all the users, the fairness of QoE should be considered. The Propor- tional Fair （PF） principle is adopted to achieve the trade-off between the network perfonmnce and us- er fairness. Then, an integer progranming problem is formed and the QoE-aware PF scheduling princi-ple is derived by solving the relaxed problem. Simu-lation results show that the proposed scheduling principle can perform better in terms of user fair-ness than the previous principle maximizing the sum of MOS. It also outperfoms the max-rain scheduling principle in terms of average MOS.
