Generalizing the Transfer in Iterative Error Correction: Dissection Decoding

Iterative decoding with message-passing is consid-
ered. The message format is generalized from the classical, single
probability value for each code symbol to a probability distribu-
tion by introducing an additional logarithmic probability mea-
sure. Thereby, the representation of the probability distributions
underlying the constituent code constraints by the messages is
improved in terms of the Kullback-Leibler divergence. Simulation
shows that this improvement can transfer to the error correcting
performance.