Rate-distortion comparisons between FEC and MDC based on Gilbert channel model

Abstract

Abstracl-To facilitate real-time audio-visual communications thmugh the Internet, low-delay packet-loss recovery techniques, such as forward e m r c o m t i o n (FEC) aud multiple description coding WDC), can be used. Based on a 2-state Gilbert channel model, we compare the SNR performance of FEC with XOR operation (XOR-FEC), FEC with Reed-Solomon code WFEC), side distortion optimized MDC (SD-MDC), and channel optimized MDC (optMDC). If we know the channel condition by feedback thmugh the Real-Time Contml Pmtocol (RTCP), optMDC can adjust the ratio between central distortion and side distortion so that it works at the best operating point. From a ratedistortion viewpoint, optMDC gives the best performance both in the single-channel and two-independent-chel transmission cases via packet networks with burst loss nature. Multiple-channel transmission is preferred if independent channels can be established between connections.