MESSAGE PASSING AND LOOPS SOLUTION ALGORITHM BASED ON CUT-NODE TREE

Abstract

Low-density parity-check (LDPC) codes are forward error-correction and linear
block codes. An LDPC code can be described by a bipartite graph called Tanner graph[1]. Loops,
especially short loops in tanner graph, degrade the performance of LDPC decoder, because they affect
the independence of the extrinsic information exchanged in the iterative decoding. In this paper, based
on graph theory and Tanner graph, the loop structure in LDPC codes are studied carefully, a new notion,
cut-node tree, is proposed to describe LDPC codes. Cut-node tree has full information of Tanner graph.
So all loop features in LDPCs can be calculated relatively easy by a computer. Traditional message
passing in graph is improved to avoid repeated iteration of information, a new decoding schemes for
LDPC codes is proposed and can suppress repeated iteration of information in SPA. The results help to
further research on related field.