大规模MIMO系统中软输出信号检测方法  

Soft-output signal detection method for uplink multiuser large-scale MIMO systems

作　　者：景小荣[1] 李翱[1] 

机构地区：[1]重庆邮电大学通信与信息工程学院,重庆400065

出　　处：《华中科技大学学报（自然科学版）》2017年第4期116-121,共6页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家高技术研究发展计划资助项目(2014AA01A705);国家科技重大专项资助项目(2015ZX03001033-002);重庆市基础与前沿研究计划资助项目(cstc2015jcyjA40040)

摘　　要：为解决在上行多用户大规模多输入多输出(MIMO)系统中,迫零(ZF)检测算法可取得近似最优性能,但ZF检测算法涉及复杂度高达O(K3)的矩阵求逆运算(其中K为用户数)的问题,基于Lanczos算法提出一种软输出信号检测方法,避免了高阶矩阵求逆运算,使复杂度由O(K3)降为O(K2).该方法为了计算软输出信息——对数似然比(LLR),通过对基于Lanczos算法的迭代计算解向量的过程进行分析,给出了一种低复杂度的LLR近似计算方法.仿真结果表明:提出的软输出信号检测方法的误比特率(BER)性能与计算复杂度均优于基于Neumann级数近似的信号检测算法,同时,最多仅需5次迭代就可取得逼近于ZF检测算法的性能.Zero-forcing（ZF）detection algorithm can achieve near-optimal performance in large-scale MIMO（multiple input multiple output）systems,but it involves computation of the matrix inversion with complexity O（K^3）,where K denotes the number of users.To solve this problem,a soft-output signal detection method based on Lanczos algorithm was proposed,which avoids the direct computation of matrix inversion and reduces the computational complexity from O（K^3）to O（K^2）.To further compute the soft-output information—log likelihood ratios（LLR）,by analyzing the iterative procedure of Lanczos algorithm,an approximate low-complexity method was proposed to calculate LLRs.Simulation results show that the proposed method outperforms the Neumann series approximation algorithm on both the bit error ratio（BER）and the computational complexity,and achieves almost the same performance as the standard ZF algorithm with only 5 iterations.

关 键 词：大规模MIMO 多用户检测 迫零 软输出 LANCZOS算法 

分 类 号：TN92[电子电信—通信与信息系统]