Scalable Audio Streaming Over the Internet with Network-Aware Rate-Distortion Optimization

Reliable audio streaming over the unreliable Internet is a challenging task. Because during a connected session, the bandwidth can vary greatly and the data packets can be lost in the transmission, it is difficult to guarantee a smooth quality of the delivered audio over the Internet. In this work, a system for streaming of scalable coded audio over the Internet is proposed. Scalable audio coding generates a bitstream consisting a number of data units (DUs), where a subset of the DUs can be extracted to reconstruct audio at a lower quality level. A network aware rate-distortion optimization model is proposed so that the server can easily adapt to the network bandwidth fluctuation by sending less packets when the network is congested. The server can also deal with severe packet loss by only retransmitting the more important data packets. By using the scalable audio coding and network aware delivery scheme, the quality of the audio streaming over severe network condition is improved dramatically.