Destination buffering for low-bandwidth audio transmissions using redundancy-based error control

Digital audio is becoming increasingly prevalent with the advent of software for Internet telephony and real-time audio playback for World Wide Web browsers. Since audio quality depends on timely delivery of the packet stream, protocol mechanisms must address the control of delays and packet losses in an integrated fashion. Two critical mechanisms for high-quality audio delivery are the receiver buffering strategy and error control. Buffering at the audio receiver is required for continuous playback in the presence of network delay variations (jitter), and a number of algorithms have been proposed. Timely recovery of packet loss protects audio quality against network errors. Recent work has suggested the effectiveness of an approach that sends multiple copies of the audio data in consecutive packets so that small burst losses in the network are overcome. Packet-level traces of Internet audioconferencing software were collected over a network path including LANs and a 28.8 kbits/s dial-up connection. Using these traces in simulations, receiver buffering strategies for controlling packet jitter and for supporting redundancy-based error control are examined. The study determines that jitter control algorithms will not generally provide adequate buffering when the requirements of error control are included. This observation leads to a proposed modification for one popular jitter control algorithm, and the performance trade-offs are explored.