| This work opens up a new line of enquiry into networked signal communications, which we call joint network-source coding (JNSC). JNSC puts the research on network coding in a rate-distortion perspective and can be con sidered as the extension of network coding to communication of degradable signals. We show that higher overall bandwidth economy can be achieved by allowing network clients to reconstruct a source signal at different fidelities, rather than insisting on an identical reconstruction at all sinks as in lossless network coding. We then introduce a constructive algorithmic approach to solving the JNSC problem via collaborative multiple description diversity coding and diversity routing. We also study information-theoretical limiting behaviors of the proposed design. Unlike existing works on network coding, our JNSC approach respects the existing network hardware design and protocols. Considering that the bulk of network traffic is for signal communications with fidelity (e.g., multimedia streaming), efficient solutions to JNSC problem, such as those proposed by this thesis, have the potential to improve the quality of network services. |
