Multiple description coding over multiple correlated erasure channels

This paper examines the use of Multiple Description Coding (MDC) for information transmission over multiple correlated erasure channels. It addresses the problem of optimal rate allocation between descriptions to attain the minimum possible average distortion at the receiver. The rate allocation problem is formulated on the basis of rate-distortion theory for the general case of unbalanced MDC and under the constraint that the total coding rate is fixed. The channel parameters, that is, erasure rates, for which MDC outperforms Single Description Coding are characterised. It is shown that MDC is optimal when the channel erasure rates are comparable and the total coding rate is large. A heuristic approximation for optimal rate allocation is also presented, which can be readily used in practice to determine the relative rates between descriptions needed to achieve the minimal average distortion. The relationship between the optimal average distortion and the erasure correlation is explored. It is shown that for some cases, such as for equal average erasure probabilities, the average distortion decreases as the channel correlation increases. Copyright © 2012 John Wiley & Sons, Ltd.