Robust Transmission of Lossless Audio with Low Delay over IP Networks

A system for streaming lossless audio with low delay over IP networks is presented. To achieve error robustness, the signal is divided into a base and an enhancement layer when the network is approaching congestion. The base layer is perceptually encoded using a time-varying pre- and postfilter, and this layer is transported using a high priority traffic class in a Differentiated Services (DiffServ) network. The enhancement layer is the difference between the original signal and base layer, and is transmitted using a regular Best Effort traffic class. In our experiments the system delay is just 256 samples, and it can be seen that the layering only introduces moderate amounts of redundancy, while improving the error resilience significantly.