Joint decoding of multiple-description network-coded data

This paper considers a transmission scheme combining multiple description coding (MDC) and network coding (NC). Our aim is to benefit from the property of MDC to provide progressive quality improvement with the number of received packets and from the efficient use of network resources provided by NC. MDC is performed either via frame expansion (NC-MDC-F) or via correlating transform (NC-MDC-T). With NC-MDC-F, when not enough NC have been received, reconstruction may be done via mixed integer quadratic programming (MIQP). With NC-MDC-T, the reconstruction involves simple Gaussian elimination. In both cases, a good robustness to missing NC packets is observed. When the number of missing packets is small, NC-MDC-F provides better SNR thanks to a reduction of a part of the quantization noise. The price to be paid is a decreased robustness to missing packets. When the number of lost packets increases, a reconstruction is still possible with NC-MDC-F for some packets, even if the number of missing packets is larger than n-k, the number of introduced redundancy packets. A typical application context is multicasting of multimedia contents with delay constraints. Users with very good channel conditions will benefit from the reception of redundant packets with NC-MDC-F, whereas these packets are useless with NC-MDC-T or with error-correcting NC.