菱形中继网络中基于叠加训练的信道估计技术研究

针对放大转发(Amplify-and-Forward, AF)模式下的菱形中继网络，为了高效获取级联和单跳链路信道状态信息(Channel State Information, CSI)，本文提出基于叠加训练的信道估计方案，以消除多址接入干扰和训练间互干扰为目标，进行最优的多训练序列设计。新方案将中继训练叠加到源训练序列上，通过对中继识别符号以及中继训练组进行联合优化设计，设计了一种基于频域循环移位的正交扩展序列组生成算法。为了消除非高斯复合噪声对单跳信道估计造成的严重干扰，提出了一种中继噪声消除算法。通过两路中继链路获取的信息副本，能够在端节点实现分集合并，有效提高符号检测性能。仿真实验对比了同类型的信道估计方案，分析验证了方案的有效性。 

Research on Channel Estimation Technology Based on Superimposed Training in Diamond Relay Networks

In order to obtain the channel state information (CSI) of cascaded and single-hop links in amplify-and-forward (AF) diamond relay networks efficiently, this paper proposes a channel estimation scheme based on superimposed training to eliminate multi-access interference and mutual interference between trainings. Then the optimal multi-training sequence design is performed. The new scheme superimposes relay training on the source training sequence. Through joint optimization design of relay identification symbols and relay training groups, a generation algorithm of orthogonal spreading sequence group based on frequency domain cyclic shift is designed. In order to eliminate the serious interference caused by non-Gaussian compound noise on single-hop channel estimation, a relay noise cancellation algorithm is proposed. Diversity combining can be implemented at destination through the copies of information obtained from the two relay links. This can improve symbol detection performance effectively. The simulation experiment compares the same type of channel estimation scheme, and the analysis verifies the effectiveness of this scheme.