University of California, Santa Cruz
Mismatched decoding theory is applied to study the error exponents (both random-coding and expurgated) and achievable rates for Bit-Interleaved Coded Modulation (BICM). The gains achieved by constant-composition codes with respect to the usual random codes are highlighted. Random-coding techniques are typically used to prove channel coding theorems for achievable rates and exponents under Maximum-Likelihood (ML) and maximum-metric mismatched decoding. For these problems, different random-coding ensembles, where code-words are generated in an i.i.d. manner, have been considered in the literature.