Low-delay distributed source coding for time-varying sources with unknown statistics
We consider the multiple access source coding (MASC) problem (also known as the SlepianWolf problem) for situations where the joint source statistics are unknown a priori. Since neither encoder receives information about the joint source statistics, we allow an asymptotically negligible amount of communication between the encoders. We prove the existence of universal 2-encoder Linked MASCs (LMASCs) with rates approaching the Slepian-Wolf bound, demonstrate the tightness of this bound, and calculate the rate of convergence of the proposed universal LMASC. The result generalizes to M > 2 encoders. We also consider scenarios where the number of bits passed between the system encoders is allowed to grow linearly in the code dimension; in these scenarios one encoder can act as a conduit for the flow of another encoder's information.