As the number of base station (BS) antennas in massive multiple-input multiple-output (MIMO) systems is dramatically increased, linear detection methods are able to achieve the near optimal performance at the expense of performing the complicated matrix inversion of a high dimensional matrix, which makes hardware implementation infeasible. Meanwhile, the straightforward application of the hard symbol-to-bit decision algorithm yields a very poor performance for the higher order quadrature amplitude modulation (QAM) that is widely adopted as an efficient tool to accommodate high demand of data throughput in modern wireless systems. In this work, we propose a low-complexity soft-output signal detector for multiuser massive MIMO communication systems employing Gray-coded higher order QAM with square constellations. In the proposed detector, the two-dimensional double successive projection (2D-DSP) algorithm is utilized to iteratively realize zero-forcing (ZF) algorithm for multiuser signal recovery, which circumvents the matrix inverse operation. Moreover, a simplified implementation of Max-Log-MAP algorithm is developed to calculate log likelihood ratios (LLRs) for soft-input channel decoding by fully exploiting the bit-flipping property of Gray-coded modulation scheme along with the utilization of the channel hardening property of the massive MIMO systems. Numerical results show that the proposed soft-output detector provides a relatively good tradeoff between the complexity and performance compared with several existing detectors, and approaches the performance of the ZF algorithm with only 3 iterations.

随着大规模多输入多输出（MIMO）系统中基站（BS）天线的数量急剧增加，线性检测方法能够以实现复杂的高维矩阵求逆为代价实现近乎最佳的性能。矩阵，这使得硬件实现不可行。同时，硬符号到位决策算法的直接应用对于高阶正交幅度调制（QAM）产生了非常差的性能，而高阶正交幅度调制（QAM）被广泛用作满足现代无线系统中对数据吞吐量的高要求的有效工具。在这项工作中，我们提出了一种低复杂度的软输出信号检测器，该检测器用于采用带有正方形星座的格雷编码高阶QAM的多用户大规模MIMO通信系统。在建议的检测器中，利用二维双连续投影（2D-DSP）算法迭代实现了多用户信号恢复的迫零（ZF）算法，从而避免了矩阵逆运算。此外，开发了Max-Log-MAP算法的简化实现，以通过充分利用格雷码调制方案的比特翻转特性以及信道硬化的利用来计算软输入信道解码的对数似然比（LLR）。大规模MIMO系统的特性。数值结果表明，与几种现有的检测器相比，所提出的软输出检测器在复杂度和性能之间提供了一个相对好的折衷，并且仅需3次迭代即可达到ZF算法的性能。