Assessing and Improving User QoE of Video Telephony

Video telephony is an important application. It is extremely challenging to deliver good user Quality of Experience (QoE) over the best-effort Internet. In this paper, we summarize our previous work on measurement study of commercial video telephony solutions and outline our current and future work on assessing and improving user QoE of video telephony. 1 Background and Motivation The Internet has fundamentally changed the way people communicate, from emails, text-messages, blogs, tweets, to Voice-over-IP calls, etc. We are now experiencing the next big change: Video Telephony. Although video telephony was originally conceived in 1920s, largely due to its stringent bandwidth and delay requirements, it has little success, especially in the end-consumer market, until very recently. The proliferation of video-capable consumer electronic devices and the penetration of increasingly faster residential network accesses paved the way for the wide adoption of video telephony. Video telephony is more than two-party video chat. It also supports Multi-party Video Conferencing (MPVC) between multiple geographically distributed participants by transmitting audio and video signals among them in realtime. Compared with voice, video is much more bandwidth-demanding. While Skype encodes high-quality voice at data rate of 40kbps, a Skype video call with good quality can easily use up bandwidth of 1Mbps [4]. Compared with video streaming, video telephony has much tighter delay constraints. While seconds of buffering delay is often tolerable in video streaming, in video telephony, user Quality-of-Experience (QoE) degrades significantly if the one-way end-to-end video delay goes over 350 milli-seconds [1]. To deliver good video telephony QoE over the best-effort Internet, video telephony solutions have to cope with user device and network access heterogeneities, dynamic bandwidth variations, and random network impairments, such as packet losses and delays. All these have to be done through video generation and distribution in realtime, which makes the design space extremely tight. 2 Measurement of Commercial Solutions We recently conducted a comprehensive measurement study on three most popular video conferencing systems on the Internet: Facetime/iChat, Google+ Hangout, and Skype. Our study was focused on their key design choices and their delivered user QoE. The ultimate goal is to understand how different design choices impact video telephony QoE perceived by end consumers. All the systems use proprietary protocols and encrypt data and signaling. There is very limited public information about their architecture, video encoding and distribution algorithms. Through a series of carefully designed black-box measurements, we unveiled important information about their key design choices [4, 2, 3].