Trend: A dynamic bandwidth estimation and adaptation algorithm for real-time video calling

Abstract

Video calling over the internet is becoming more and more popular in recent years. The quality of video in such applications is a major contributor to the overall user experience during the call. As received video quality degrades significantly when the network path used is either under-utilized or over-utilized, a method for estimating network bandwidth availability for adapting video bit rate in video calls is highly desirable. This paper presents TREND, a novel technique for dynamic bandwidth estimation and adaptation designed for real-time, interactive video calling. By detecting the delay in the received video frames, TREND estimates available network bandwidth in the path and adapts the encoded bitrate sent to best fit it before packets are lost. TREND uses only standard protocols for delay detection and flow control, and can therefore be easily and effectively used with traditional visual communications systems. After testing TREND under various use cases we have found that it efficiently estimates available bandwidth while outperforming competing tools in terms of detection time and accuracy. By implementing TREND in its video conferencing solutions, RADVISION was able to show a superior quality of experience for its video calls over other solutions available today.