Study On Serial Decoding Algorithms Of LDPC Codes

With the growing demand for efficient and reliable digital transmission, Low-density parity-check(LDPC) codes have attracted wide attentions and obtained intensive research for their near-capacity performance, and low hardware implement complexity. Recently LDPC codes have been widely used in optical fiber communications, fixed wireless communication, magnetic/optical storage and other fields. In this thesis, the theory of LDPC codes and their serial decoding algorithms are deeply investigated. The main works are given as follows:Firstly, we introduce the principles of channel coding and make a detailed analysis on the BP decoding algorithm and several factors which affect the decoder.Secondly, two kinds of serial decoding algorithms are studied: dynamic serial decoding algorithms and static serial decoding algorithms. Dynamic serial decoding algorithms,which are also known as informed dynamic schedule algorithms, include RBP, NW-RBP,IVC-RBP. And static serial decoding algorithms mainly include LBP, shuffled iterative algorithm and M2I2-based algorithm. Meanwhile, hybrid scheduling algorithms which mixed different schedules in the decoding process are also discussed. The comparisons of the convergence speed, complexity and performance of deferent serial decoding algorithms are given.Finally, two static serial decoding algorithms are proposed and realized. One algorithm is Error Pattern(EP)-based algorithm. This algorithm uses probabilistic statistical methods to generate schedule sequence. Compared with the LBP algorithm, EP-based algorithm behaves better in convergence speed and decoding performance. The other algorithm we proposed is node-wised maximum mutual information increase based(NW M2I2-based)algorithm. This algorithm arranges decoding schedules with the guidance of maximum mutual information of check nodes and then the decoder decodes according to the fixed node sequence. The simulation results show that this algorithm can effectively improve the decoding performance.