Are Multiple Descriptions Better Than One?

In this paper we compare three schemes proposed for error recovery for real-time multimedia applications: media-dependent forward error correction (MD-FEC) proposed for real-time audio, media-independent forward error correction (MI-FEC) proposed for real-time video and the recently re-discovered multiple description coding (MDC). We provide a detailed queueing analysis for these schemes considering bursty traffic sources, and combine results from information theory and queueing theory to analyze their performance bounds. We conclude that MDC always performs better than MD-FEC, and that the average loss probability plays a key role in the choice of the optimal parameters. We also show that MDC outperforms MI-FEC if packet losses are highly correlated, like in the current Internet and the available delay for error control is low.