Improved Decodings for LDPC Coded Modulation Systems

In this paper, Two improved iterative decoding algorithms are proposed for medium-short length LDPC codes coded BICM-ID systems. The first improved algorithm is based on averaged Log-Likelihood Ratio (LLR). The extrinsic information delivered by every BP iteration is averaged and the averaged LLR values are used as a prior LLR fed back to the input of demodulator for the next iteration. This method can alleviate the LLR oscillation of certain variable nodes during BP iterations and thus make the extrinsic output of decoder is more credible. The second improved algorithm is based on the Gaussian assumption of the extrinsic LLRs delivered by BP decoder. It multiplies the extrinsic output of the decoder by a scaling factor and then pass it as a priori information to the demodulator. The algorithm can decrease the correlation of coded bits within the symbol which is caused by the existing cycles in the bipartite graph of short length LDPC codes and thus reduce the error propagation. Simulation results show that both improved algorithms achieve noticeable performance gains with only modest increase in computation complexity.