Optimization design of polar-LDPC concatenated scheme based on deep learning

Abstract In 5G communication systems, because of incomplete polarization of polar codes, the performance of polar codes is poor at low and medium SNR. The concatenated scheme for polar codes and low-density parity-check codes (Polar-LDPC) was proposed to improve decoding performance. Although the performance optimization of Polar-LDPC has been improved, the decoding delay will be increased compared with the non-concatenated schemes. In order to solve this problem, we propose a deep learning-based concatenated scheme for Polar-LDPC codes by paralleling multiple neural networks. The neural network in the proposed Polar-LDPC concatenated scheme, which assists the decoding of polar codes, can reduce the decoding delay. The experiment simulation results show that the proposed concatenated scheme can reduce time complexity of the polar codes under the same code length.