基于观看经历的视频QoE评价研究

QoE evaluation of video services has significant implications for improving the service performance and guaranteeing the user experience. To accurately evaluate the user’s QoE for various video services, the user’s viewing experience is considered for video quality assessment from the technical viewpoint of QoE, and for user’s QoE evaluation in the true sense, respectively, from the perspective of real perception analysis. Targeting various video services and types of distortion, the distortion is attributed to different types of user’s viewing event according to the characteristics of user’s perception, and a unified event-based evaluation framework is established to study the instant perceptual video quality, which is affected by various types of distortion. Then, a user viewing experience-based, scalable, instant video quality assessment model is proposed. With respect to user’s QoE evaluation, a novel method is proposed by using the representation parameters of user’s viewing behaviors and operation behaviors are employed to evaluate the QoE, combining the user’s watching history. Then, a behavior-represented QoE evaluation model based on user watching history is proposed. This project will provide both the important reference and experience for analyzing the user’s experience of video service in real life, and effective theoretical basis and technical support for quantitative QoE evaluation.

视频业务QoE评估对改善视频服务性能和保障用户体验具有重要的指导意义。为了能够准确的评价广泛视频业务中用户的QoE，本项目拟从用户真实感知过程角度出发，结合用户的观看经历分别研究QoE技术域的视频质量评估以及真正意义上的用户QoE评估。针对纷繁复杂的视频业务类型和失真类型，结合用户感知特性将不同类型的失真效应划分成各类用户观看事件，从而构建基于事件的统一评估框架来研究各种失真效应对用户即时感知视频质量的影响，建立一种基于用户观看经历的、易扩展、即时网络视频质量评估模型。针对用户的QoE评估，提出结合历史观看经历，利用用户眼动行为和操作行为表征参数预测用户QoE的方法，建立了基于用户历史观看经历的行为参数表征QoE评估模型。为分析现实生活中用户对网络视频业务的体验感受提供重要的参考和经验，为定量预测用户体验质量提供了有效的理论依据和技术支持。

本项目针对纷繁复杂的视频业务类型和失真类型，深入探寻用户的实际感知过程，构建了基于事件的评估框架来研究不同失真效应对用户感知视频质量的影响。通过大量主观实验测试，着重研究了编码失真、网络传输引起的失真以及终端设备显示引起的失真等等对用户感知质量的影响，建立了基于用户观看经历的、易扩展、即时网络视频质量评估模型。另外，针对真正意义上的用户的体验质量QoE，本项目根据行为心理学相关理论研究用户内在状态的外显形式，首次提出结合历史观看经历，利用用户眼动行为和操作行为表征参数预测用户QoE的思路和方法，建立了基于用户历史观看经历的行为参数表征QoE评估模型。本研究成果为分析现实生活中用户对网络视频业务的体验感受提供重要的参考和经验，为定量预测用户体验质量提供了有效的理论依据和技术支持。整个项目的执行情况很好地完成了预期的研究计划。