In this paper, we propose a belief-propagation (BP)-based algorithm which calculates with only k minimal value of LLRs from check nodes to bit nodes in the Horizontal step to reduce the complexity and the memory size of general BP algorithm for LDPC codes decoding. In order to find fc minimal value of LLRs in the horizontal step, a cyclic calculation scheme is provided. We describe this scheme as cyclic approximated min algorithm. Furthermore, an overlapped cyclic approximated rain algorithm is proposed to reduce the number of iteration, too. Some simulation results are given, which show this new decoding approach with δ-min algorithm can achieve an error performance very close to that of BP algorithm on the additive white Gaussian noise channel.
展开▼