无线电通信技术 Radio Communications Technology ISSN 1003-3114,CN 13-1099/TN 了充分发挥 CNN 的能力，对于经过高阶调制的复数形式信号数据，通 过预处理转化为实数 形式。仿真结果表明，相较于已有研究中提出的联合迭代检测译码结构，GF(64)域上的（42,21）多元 LDPC 码经 64-QAM 调制传输，CNN 方案能够获得最高 1 dB 的性能增益，验证了其在抵抗相关信道噪声的有效 性。 关键词：多元 LDPC 码；卷积神经网络；迭代检测译码算法；相关噪声 中图分类号：TN919 文献标志码： A
CNN-aided Iterative Detection and Decoding for Nonbinary LDPC Coded Modulation Systems WAN Fei, BAI Baoming, ZHU Min (State Key Laboratory of ISN, Xidian University, Xi’an 710071, China)
Abstract: In this paper, we propose a deep learning assisted iterative detection and decoding method for nonbinary LDPC (NB-LDPC) coded modulation systems under correlated channel noise, in which the convolution neural network (CNN) is incorporated into iterative hard-reliability based majority logic algorithms. The CNN-based architectur
