Communicating by Cosets and Applications to Broadcast

We describe a constructive coset-based coding framework for the general broadcast problem. Our construction is inspired by the information-theoretic framework of Marton [16], who used the concept of random binning to provide the best-known achievable broadcast region to date. Our approach leverages our recent coset-based construction for the Slepian-Wolf coding problem. The key enabling idea is to use cosets of codes to communicate. For the special case of degraded broadcast, our approach has the practical advantage of not compromising the privacy of the message sets of the individual users, with no user being able to decode any other user’s message. A recent work [6] demonstrated that it is possible to apply the framework of Costa [7] into more general cases beyond the strictly Gaussian case. We describe a Trellis-based coset coding framework based on our previous work on the dual problem [15] which illustrates signal design for broadcast inspired by our previous work in source coding with side information, and provide encouraging preliminary simulation results that validate the efficacy of our approach.