Reactive Mechanisms for Recovering Audio Performance in Multimedia Conferencing Over Packet Switched Networks

The impact of multimedia traffic on the performance of networking applications is significant because the vast majority of currently available tools send data at a rate that does not depend on the state of the network. This is unlike rate-controlled applications, such as Transport Control Protocol-based (TCP) applications, which adjust their output rate and bandwidth requirements according to the state of the network. Audio is the most important component in a multimedia session. It is thus important to grant audio the maximum attention even if it necessitates sacrificing the throughput (but not the quality of service) of other less important applications. An example of such a sacrifice is to reduce the video bit-rate if the audio quality drops. Reducing the bit-rate may or may not be observed by the user. A video source may adapt to this change by changing the target frame rate if the drop in bit-rate is significant. In this paper we address feedback algorithms for dealing with audio loss problems. We propose a modification to a Real-time Control Protocol-based (RTCP) congestion control algorithm that is sensitive to audio performance. We analyze the RTCP-based congestion control technique and present its limitations. We then uncover a relationship between audio jitter and audio packet loss. Finally, we present a novel scheme for predicting loss using jitter information.