Research On Low Complexity Decoding Algorithms For LDPC Codes

Low-density parity-check (LDPC) codes have attracted much attention due totheir excellent performance. With the increase of communication traffics, highdecoding throughput and ultra low bit error rate (BER) are required in the nextgeneration high speed communications. Thus, the implementation complexity ofdecoding algorithm for LDPC codes needs to be reduced. Meanwhile, someapproaches should be adopted to lower the error floors of LDPC codes.In order to reduce the complexity of decoding algorithm for LDPC codes, a lowcomplexity quantized decoding algorithm is proposed in this thesis. Based onadaptive-offset min-sum (AOMS) algorithm, the low complexity quantized decodingalgorithm is obtained by introducing a pre-determined iteration number as thecondition of adaptive selecting the offset factors. Furthermore, a4-bit nonuniformquantization scheme is designed, which could guarantee the dynamic range of theextrinsic information and easily utilize the optimized quantized offset factors.Simulation results show that, compared with floating-point decoding algorithms, theperformance degradation of the low complexity quantized decoding algorithmproposed in this thesis can be neglected.For the purpose of lowering the error floors of LDPC codes, a kind of lowcomplexity post-processing method for residual errors of LDPC codes is proposed.According to this method, a simple product code is designed using LDPC codes withvery high rate algebraic block codes by exploiting the sparsity of LDPC decodingfailure and sparse error distribution in an LDPC frame with errors. Specifically,different with traditional product codes, the information bits are first encoded withvery high rate algebraic block codes of low encoding and decoding complexity. Then,LDPC codes are used as the inner code to obtain a low decoding threshold usingiterative decoding. Simulation results show that this design scheme can achieve ultralow bit error rate at relatively low signal-to-noise ratios.