Efficient Decoding Algorithms For LDPC Codes

From the raising of Turbo codes, LDPC codes has attracted more and more attention for its remarkable performance. The decoder algorithm of LDPC codes is already being a mature technology, but for the practice implement, especially small terminals such as mobile handsets, reducing the complexity of decoding algorithm is still desirable, since low complexity implies low cost of the hardware implementation and also low consumption of battery power. Thus, decoding LDPC codes with low complexity still attracts coding field study.In this thesis, LDPC coding methods and decoding algorithms are firstly reviewed in detail, we also study classic simplified algorithms after that, like MSA. Due to deep theoretical analysis and simulative research of inner character of LDPC iterative decoder, we developed a new reduced-complexity decoding algorithm, which depends on detecting correct bits after few iterations and deleting these correct bits from Tanner graph, also means check matrix. Through deleting, we can achieve smaller and smaller matrix and reduce the calculation amount of follow iterations. Even more, this deleting also can pass more efficient correct information to other bits and accelerate decoding convergence. Besides, we also design efficient and reasonable scheme to avoid false detecting and diffusion.Even more, when implement our new algorithm with high-level modulation, this paper proposed a new low complexity coded modulation idea and present two new algorithm. Through implement detecting and deleting into coded modulation, decoder can take more efficient advantage of correct information and avoid wrong information spreading, further more, deleting bits also can simplify the demodulation complexity through descending constellation level and calculate amount.Simulation results show that, when implemented with binary modulation, our new algorithm can keep decoder performance of SPA and achieve performance gain of MSA, while get 25-50% reduction of complexity. When implemented with high-level modulation, we proposed two schemes, both of them can achieve at least 0.3dB performance gain while get 30% complexity reduction.