Communicating Correlated Sources Over MAC and Interference Channels II: Joint Source-Channel Coding

We present the second part of our work on communicating correlated sources over multiple access (MAC) and interference channels (IC). Specifically, we undertake a Shannon-theoretic study of the above scenarios and focus on characterizing sufficient conditions for lossless recoverability of the sources at the decoder(s). We enhance the fixed block-length (B-L) coding technique by incorporating the technique of inducing source correlation onto channel inputs, originally discovered by Cover, El Gamal and Salehi. In contrast to the first part, performance analysis of a joint source-channel decoder poses new challenges. We enhance the earlier developed suite of coding and analytical tools to overcome these challenges and derive (simplified) single-letter characterizations for a new set of sufficient conditions for both scenarios. For both the MAC and IC problems, the derived sufficient conditions are (i) subsumed in the current known tightest, and (ii) strictly weaker for identified examples. Lastly, we propose simple 'plug-in' approaches that can further weaken the derived sufficient conditions. Our findings enable us to subsume Dueck's findings (1981) and go even further for the example considered therein.

Automated summary

In this work, a Shannon-theoretic study of communicating correlated sources over multiple access and interference channels is presented. The fixed block-length coding technique is enhanced by incorporating the technique of inducing source correlation onto channel inputs. New single-letter characterizations for sufficient conditions are derived for both scenarios, and simple 'plug-in' approaches are proposed to further weaken the derived conditions. Subsuming Dueck's findings and going beyond in the example considered therein are made possible by the findings.