基于深度学习的极化码串行抵消译码优化算法

针对5G场景下极化码串行抵消译码算法低输出高延迟的问题,提出加快串行抵消译码过程中深度学习译码器整体译码速度的方案。该方案根据信道极化理论计算不同子信道的可靠性,通过调整参数的不同取值,剪掉译码树上均为固定位的叶子节点所在的子二叉树,从而减少深度学习译码器的数量,加快了整体的译码速度。仿真结果表明,所提出的方案不仅具有和原串行抵消算法相同的译码性能,而且降低了极化码串行抵消深度学习译码的时间复杂度。

On Successive Cancellation Decoding Based-Deep Learning Methods of Polar Codes

To solve the low output and high latency of successive cancellation decoding for polar codes in 5G scene, accelerate scheme of successive cancellation decoding based on deep learning for polar codes is proposed. According to the reliability of the polarized different channels, the scheme will prune the children binary tree which leaf nodes are all frozen bits under the chosen different parameters. Then, it can improves the whole decoding speed since the amount of the deep learning decoder is fewer. Simulation shows that the proposed scheme can obtain the same decoding performance with successive cancellation decoding, and also can reduce the time complexity of the successive cancellation decoding based on deep learning.