Error Probability of Distributed Arithmetic Coding

Distributed Arithmetic Coding (DAC) is an interesting realization of Slepian-Wolf coding that performs nonlinear coset-like partition for equiprobable binary sources. Just as in other coding problems, it is a very important issue to theoretically predict the decoding error probability of DAC. This letter proposes a plausible solution to this difficult issue. Our analyses are based on the concept of coexisting interval introduced in our previous work. Specifically, we use it to deduce the probability that two given codewords coexist in the same coset. Experimental results confirm the correctness of our theoretical analyses.