Studies On Decoding Performance Of LDPC Based On Parallel Computing

Low-density parity-check (LDPC) codes have become one of the most promisingtechnologies in the channel coding field due to their Shannon capacity-approachingperformance and low decoding complexity recently. With the development of intensiveresearch, LDPC codes have been widely used in many fields. How to improve thedecoding speed and decoding performance is always a key issue for medium andlonger LDPC codes. And yet, parallel computing technology is expected to be animportant manner of increasing decoding speed.In this paper, based on the research of decoding algorithm, an advanced algorithmhas been implemented in this thesis as well a parallel computing technology has beenused to improve the speed of decoding. The conclusions are as follow：1. The basic principles of LDPC codes are introduced, and the Tanner graphicalmodel for LDPC codes is provided. The regular and irregular LDPC codes are depicted.Then the principles of iteration decoding are illustrated simply.2. The basic theories of parallel computing platform are introduced, and the MPIparallel computing technology is provided. We also make some analysis on howparallel computing technology can be used to increase decoding speed.3. The decoding algorithms in BEC and AWGN channel are studied thoroughly.According to the speculation on some nodes after the failure of decoding, a newdecoding is processed again, the frame error rate has been reduced by this method, andthe decoding performance is improved.4. Based on the analysis of the LDPC decoding characteristic, LDPC parallelcomputing platform is designed. Meanwhile, LDPC parallel decoding algorithm isgiven and implemented according to MPI parallel computing technology. Experimentfor LDPC parallel decoding algorithm suggests that using LDPC computing platformand MPI parallel technology can largely improve LDPC decoding speed.