One of the major problems experienced with LAN-based videoconferencing systems is the degradation of conference latency and fidelity due to network congestion. In this paper, we present a comprehensive transmission control framework describing the relationship of latency and fidelity to properties of human perception and network congestion. We discuss how network factors such as physical link capacity and size of maximum transmission units may exacerbate the effects of congestion on the conference and how the impact of congestion can be ameliorated through judicious adaptation of the media streams' bit and message rates. Finally, we demonstrate the ability of an adaptive transmission control algorithm based upon the framework to produce low-latency, high-fidelity conferences over congested internetworks.

One of the major obstacles facing designers of video conferencing systems is the problem of ameliorating the effects of congestion on interconnected packet-switched networks that do not support real-time communication. We present a framework for transmission control that describes the current network environment as a set of sustainable bit and packet transmission-rate combinations and show that adaptively scaling both the bit and packet-rate of the audio and video streams can reduce the impact of congestion. We empirically demonstrate the validity of adapting both packet and bit-rate using a simple feedback mechanism and simple adaptation heuristics to deliver audio and video streams suitable for low-latency, high-fidelity playout.