Polar code is a channel coding method that has been proved to be able to reach Shannon capacity in the binary discrete memoryless channel. Because of the superior performance and low encoding and decoding complexity, polar code has attracted extensive attention in the industry and been chosen as the channel coding scheme for the control channel in the scenario of EMBB in 5G mobile communication. In this work, we propose an intelligent BP decoding algorithm of polar code based on smart post-processing. We employ the neural network to classify the output data of regular BP decoding into “good-bit” and “bad-bit” categories. We also design a strategy to search the bits, which are most probably incorrect from the “bad-bit” group for post-processing. Then, we can invert the “bad-bit” to correct the residual error in the Belief Propagation (BP) iterative process. Simulation results prove that the proposed algorithm can achieve at least 0.5dB error correction performance enhancement compared with the regular BP decoding with slight computation complexity and energy consumption increase.

极地码是一种信道编码方法，已被证明能够在二进制离散无记忆信道中达到香农容量。由于极好的性能和低的编解码复杂度，极地码在业界引起了广泛的关注，并被选为5G移动通信中EMBB场景中控制信道的信道编码方案。在这项工作中，我们提出了一种基于智能后处理的极地码智能BP解码算法。我们使用神经网络将常规BP解码的输出数据分类为“好位”和“坏位”类别。我们还设计了一种策略，用于搜索“坏位”组中最有可能不正确的位以进行后处理。然后，我们可以反转“不良位”以纠正“信念传播（BP）”迭代过程中的残留错误。仿真结果表明，与常规的BP解码相比，该算法在纠错性能上至少提高了0.5dB，且计算复杂度小，能耗增加。