In uplink large-scale multiple-input and multiple-output (MIMO) systems, the data detection at the receiver is the major challenge due to the considerable increase in dimensions of MIMO systems. In this paper, a Bayesian strategy is investigated for MIMO systems. The main idea of this algorithm is to improve the performance of a detector by finding the incorrectly detected symbols relying on the sparse property of the estimation error. In the proposed method, the conventional massive MIMO model is converted into a sparse model by inducing a sparse constrain on the symbol error vector obtained from a linear detector (minimum mean square error or zero forcing detectors). Then, by recovering the non-zero entries of error (i.e. the incorrectly detected symbols) via the sparse Bayesian approach, the primary estimate of the information vector is corrected. Numerical results reveal that exploiting the sparse characteristic of the estimation error leads to improvement in detection performance and the proposed Bayesian approach achieves better results than conventional and previous state of the art error recovery methods. Meanwhile, computational complexity of the sparse Bayesian error recovery (SBER) algorithm is the same as that of the linear detectors.

在上行链路大规模多输入多输出(MIMO)系统中，由于MIMO系统的维度显着增加，接收端的数据检测是主要挑战。在本文中，研究了用于 MIMO 系统的贝叶斯策略。该算法的主要思想是通过依赖估计误差的稀疏特性找到错误检测的符号来提高检测器的性能。在所提出的方法中，通过对从线性检测器（最小均方误差或迫零检测器）获得的符号误差向量引入稀疏约束，将传统的大规模 MIMO 模型转换为稀疏模型。然后，通过稀疏贝叶斯方法恢复非零错误条目（即错误检测到的符号），信息向量的主要估计被校正。数值结果表明，利用估计误差的稀疏特性可以提高检测性能，并且所提出的贝叶斯方法比传统的和以前最先进的误差恢复方法取得了更好的结果。同时，稀疏贝叶斯错误恢复（SBER）算法的计算复杂度与线性检测器的计算复杂度相同。