Pairwise joint probability propagation in BICM-ID

The iterative decoding of BICM is analyzed from an optimization point of view. The objective function underlying the standard iterative decoding is a coarse approximation of the optimal Maximum Likelihood Decoding. This approximation is suitable when large block of data are processed but may be critical for short blocks. In this contribution, it is proved that refined approximations can be obtained by propagating extrinsics on groups of bits. The particular case of pairwise joint probability propagation is investigated. Its computational complexity is similar to standard iterative decoding. Simulations show the favorable behavior of the proposed method.