Low-Density Parity-Check Lattices: Construction and Decoding Analysis (original) (raw)

2000, IEEE Transactions on Information Theory

Low-density parity-check codes (LDPC) can have an impressive performance under iterative decoding algorithms. In this paper we introduce a method to construct high coding gain lattices with low decoding complexity based on LDPC codes. To construct such lattices we apply Construction D 0 , due to Bos, Conway, and Sloane, to a set of parity checks defining a family of nested LDPC codes. For the decoding algorithm, we generalize the application of max-sum algorithm to the Tanner graph of lattices. Bounds on the decoding complexity are derived and our analysis shows that using LDPC codes results in low decoding complexity for the proposed lattices. The progressive edge growth (PEG) algorithm is then extended to construct a class of nested regular LDPC codes which are in turn used to generate low density parity check lattices. Using this approach, a class of two-level lattices is constructed. The performance of this class improves when the dimension increases and is within 3 dB of the Shannon limit for error probabilities of about 10 06. This is while the decoding complexity is still quite manageable even for dimensions of a few thousands. Index Terms-Additive white Gaussian noise (AWGN) channel, iterative decoding, lattice codes, lattices, low-density parity-check (LDPC) codes, low-density parity-check (LDPC) lattices, min-sum algorithm, progressive-edge growth (PEG) construction, Tanner graphs.