基于变分推理的Turbo频域均衡与信道估计

Variational inference-based Turbo frequency domain equalization and channel estimation

针对时变频率选择性信道下的单载波编码传输系统,提出了基于变分推理的低复杂度Turbo频域均衡与信道估计联合迭代算法。所设计的软输入软输出检测器和子块间干扰抵消器能统一处理信道估计误差的影响。通过分析接收信号因子图,提出了算法实现的双迭代结构,内迭代完成联合频域均衡与软信道估计,外迭代进行软干扰抵消和解码。仿真结果表明,所提出算法的误码率性能和对叠加训练信号强度的稳健性均优于具有相似复杂度的期望最大化方案。

A new low complexity Turbo frequency domain equalization （FDE） and channel estimation algo- rithm based on variational inference framework is proposed for the coded single-carrier transmission in time-var- ying frequency selective channels. The algorithm enables an uniform handling of channel estimation error in the deign of soft-input soft-output detector and the inter block interference canceler. Furthermore, by analyzing the receive signal＇s factor graph, a dual iterative implementation scheme of the algorithm is considered, in which the inner iterative serves for joint FDE and soft channel estimation and the outer one performs interference can- cellation and decoding. Simulation results show that the proposed receiver outperforms, both in bit-error-rate and in robustness to the strength of superimposed training signal, the expectation maximum based scheme with a similar computational complexity.