Robust source-channel coding for real-time multimedia

Abstract

Multimedia applications operating in today's Internet have to employ some form of error resilience to cope with losses. For interactive applications with strict delay constraints the latency introduced by these schemes has to be low as well. Furthermore the parameters of the applied scheme have to be set based on measurements in a possibly rapidly changing environment. In this paper we propose a robust method, called min---max-?, for optimal source-channel code rate allocation to deal with time-varying packet channels and channel state estimation errors. We evaluate its performance when used with forward error correction and multiple description coding in both stationary and non-stationary environments. We show that on a stationary channel robust rate allocation is suboptimal in terms of mean distortion, but it achieves a lower variance, while on a non-stationary channel it prevents severe degradation of the quality. We apply the proposed rate allocation method to motion compensated video and show that it performs better on a non-stationary channel than minimization of the mean distortion proposed earlier.