Transmission of Correlated Messages over the Broadcast Channel
Suhan Choi, S. Sandeep Pradhan · 2005
Abstract — New architectures for transmission of correlated messages over the broadcast channel is presented. We propose to use graphs as digital interface between source coding and channel coding since the correlation structure of sources can be maintained. An achievable rate region for broadcast channels with correlated messages is presented. We consider graphs to represent correlated messages, and to translate the correlation in the given messages into channel encoder. It is shown that such correlated messages can be sent with arbitrarily small error probability over the broadcast channel, given by p(y1, y2|x), by using a special channel code which exploits the existing correlation in the messages, if the sizes of messages and the correlation structure of the messages satisfy certain conditions. We prove this by using the random coding argument, random binning, and jointly typical sequence property. If the messages are independent, this rate region is exactly same as Marton region, which is the best known inner bound of the capacity region of the broadcast channel. However, the achievable rate region can be larger than Marton region if the messages are correlated. I.