| 摘要 |
Generating parity-check exponent matrixes H exp for both coding sequences of information bits and decoding parity-check codewords, by using LDPC codes supporting different block sizes for each code-rate R j , is optimized starting from a multi-rate random exponent table, the main table, from which sub-tables can be extracted for every requested code-rate. The generated exponent matrixes are built with random techniques, by which a first r b × r b deterministic sub-matrix accounting for a block of r redundancy bits is contiguous to a random r b × k b sub-matrix that accounts for a block of k information bits. Some constraints on the elements of the first column vector Cv of deterministic matrix, and some other distance rules on the random elements must be respected in order to simplify the encoder and prevent unwanted cycles in Tanner's graph at the decoder, as done in IEEE Std 802.16e (WiMAX). The main table is devoted to the only random matrixes plus the first column vector of deterministic matrixes for many code-rates. The main table is derived from an original table that only includes random exponent values which respect the distance rules. The respect of constraints on Cv vectors is achieved by forcing the value of some specific elements of each vector to the requested ones. So doing, the distance rules are lost, but when a sub-table is extracted for a specific code-rate, the remedy consists of forcing to the conventional element "-1" indicating the null permutation matrix those previously forced elements of the other code-rates to carry out. In this way possible cycles at the decoder are interrupted.
|
