您的位置:首页  > 论文页面

一种面向DASH的QoE客观评价模型

发表时间:2022-06-27  浏览量:265  下载量:17
全部作者: 张令雨,石东新
作者单位: 中国传媒大学信息与通信工程学院
摘 要: 近年来,流媒体视频服务正以前所未有之势影响着人们的生活。用户体验质量(quality of experience,QoE)作为衡量流媒体视频服务质量的重要指标,一直受到工业界和学术界的关注。基于HTTP的动态自适应流(dynamic adaptive streaming over HTTP,DASH)国际标准对QoE有了新要求,然而目前面向DASH的QoE客观评价模型较为简单,在高质量视频传输上仍有较大的提升空间。因此,面向DASH,提出了一种基于时序卷积网络(temporal convolutional network,TCN)和注意力机制的QoE客观评价模型,使用LIVE-NFLX-II数据集作为实验数据,通过特征分析从众多候选特征中选取最佳特征作为TCN的输入,同时使用注意力机制模块SE-Net对提取的时序特征进行加权,通过训练得到最终的QoE客观评价模型。实验结果表明,本文提出的QoE模型可以很好地拟合QoE特征与主观体验之间的关系,较当前面向DASH的最优的QoE模型在各项指标上平均提升8%左右。
关 键 词: 通信技术;QoE客观评价模型;时序卷积网络;注意力机制
Title: A QoE objective evaluation model for DASH
Author: ZHANG Lingyu, SHI Dongxin
Organization: School of Information and Communication Engineering, Communication University of China
Abstract: In recent years, streaming video services are affecting people’s lives like never before. As an important indicator to measure the quality of streaming video services, quality of experience (QoE) has always been the attention of industry and academia. Dynamic adaptive streaming over HTTP (DASH) international standard has new requirements for QoE. However, the current DASH-oriented QoE objective evaluation model is relatively simple, and there is indeed a lot of room for improvement in high-quality video transmission. Therefore, an objective evaluation model for user’s QoE based on temporal convolutional network (TCN) and attention mechanism is proposed for DASH. Using the LIVE-NFLX-II dataset as the experimental data, the best features are selected as the input of TCN from many candidate features through feature analysis, the attention mechanism module SE-Net is simultaneously used to weight the extracted time series features, and the final QoE objective evaluation model is obtained through training. The experimental results show that the QoE model proposed in this paper can well fit the relationship between QoE characteristics and subjective experience, which is about 8% higher on average than the current optimal QoE model for DASH.
Key words: communications technology; QoE objective evaluation model; temporal convolutional network; attention mechanism
发表期数: 2022年6月第2期
引用格式: 张令雨,石东新. 一种面向DASH的QoE客观评价模型[J]. 中国科技论文在线精品论文,2022,15(2):168-180.
 
3 评论数 0
暂无评论
友情链接