On the Utility of FEC Mechanisms for Audio Applications

摘要

FEC mechanisms have been proposed to recover from packet losses, and hence to improve the perceived quality in audio applications. Recently, it has been shown in [1] that the redundancy added by a FEC scheme increases the congestion of the network and deteriorates the audio quality instead of improving it. In this work we show via a simple queuing analysis that the impact of FEC on the audio quality is not always negative and that we can get better quality in some scenarios. In particular, we show that FEC is beneficial when a small number of flows implement it or when the audio applications have some particular utility functions. We derive conditions on when to get a gain in quality as well as bounds on the maximum gain that we can obtain.