Improved Iterative Decoder for LDPC Codes with Performance Approximating to a Maximum Likelihood Decoder |
An ideal maximum likelihood decoder is approximated when iteratively decoding codewords by using stopping criteria appropriate for the "waterfall region" of significant noise interference for short block lengths (<10 kbits). In the case of nonconvergence according to a given stopping criterion, the received vector is systematically modified and re-input to the iterative decoder, and this process repeated until the stopping criteria are met. The weakest symbols in the vector (having lowest likelihood) are modified first. The stopping criterion may be based on a calculated Euclidian distance, and an example of performance is given for a b/b + 2 type Turbo code of rate 2/3 for a block length of 1024 bits.
Papagiannis E, Ambroze MA, Tomlinson M