Iterative decoding for multilevel codes using reliability information

Coded modulation schemes based on multilevel codes have been proposed previously. Here, a sequence of subsets of the signal set is coded using block and/or convolutional codes. For complexity reasons the decoding is done in several stages (multistage decoding). Punctured convolutional codes and parity check codes are applied to this scheme in channel with additive white Gaussian noise. The usual decoder structure does not exploit the asymptotic gain of a code combination at a bit error rate of 10/sup -5/. The decoder structure is expanded by three means: interleaving between the coded bits of every level, use of reliability information about decoding, and iterative decoding. It is shown that such a scheme can gain about 1 dB in signal-to-noise ratio compared to trellis coded modulation with a small increase of complexity. The price to be paid is additional decoder delay.<
>