Ad Hoc Networks. 10th EAI International Conference, ADHOCNETS 2018, Cairns, Australia, September 20-23, 2018, Proceedings

Research Article

Predicting Freezing of WebRTC Videos in WiFi Networks

Download
260 downloads
  • @INPROCEEDINGS{10.1007/978-3-030-05888-3_27,
        author={Suying Yan and Yuchun Guo and Yishuai Chen and Feng Xie},
        title={Predicting Freezing of WebRTC Videos in WiFi Networks},
        proceedings={Ad Hoc Networks. 10th EAI International Conference, ADHOCNETS 2018, Cairns, Australia, September 20-23, 2018, Proceedings},
        proceedings_a={ADHOCNETS},
        year={2018},
        month={12},
        keywords={WiFi WebRTC QoS Freezing Machine learning},
        doi={10.1007/978-3-030-05888-3_27}
    }
    
  • Suying Yan
    Yuchun Guo
    Yishuai Chen
    Feng Xie
    Year: 2018
    Predicting Freezing of WebRTC Videos in WiFi Networks
    ADHOCNETS
    Springer
    DOI: 10.1007/978-3-030-05888-3_27
Suying Yan1,*, Yuchun Guo1,*, Yishuai Chen1,*, Feng Xie2,*
  • 1: Beijing Jiaotong University
  • 2: ZTE Inc.
*Contact email: 13120210@bjtu.edu.cn, ychguo@bjtu.edu.cn, yschen@bjtu.edu.cn, xie.feng@zte.com.cn

Abstract

WebRTC is an open source project which enables real-time communication within web browsers. It facilitates web-based multimedia applications, e.g. video conferencing and receives great interest from the academia. Nevertheless understanding of quality of experience (QoE) for the WebRTC video applications in wireless environment is still desired. For the QoE metric, we focus on the widely accepted video freezing event. We propose to identify a freezing event by comparing the interval of receiving time between two successive video frames, named , with a threshold. To enable automatically tracking of video freezing, we modify the original WebRtc protocol to punch receiving timestamp on the frame overhead. Furthermore, we evaluate the correlation between video freezing and quality of service (QoS) in WiFi network based on experiments in typical indoor environment. We build a machine learning model to infer whether QoE is unacceptable or not in the next time window based on current QoS metrics. Experiments verify that the model has good accuracy and the QoE state is mainly relevant to quality metrics of , and . This model is helpful to highlight the providers in system design and improve user experience via avoiding bad QoE in advance.