Massive spatial modulation (SM) multi-input multi-output (MIMO) system is a promising technique in uplink communications for future mobile communication, due to its power and spectral efficiencies. However, these systems, like other MIMO communications, face the challenge of channel estimation. Pilot-based channel estimation methods result in data rate reduction as well as imposing additional complexity at the receiver side, which is intensified in time-varying channels. Therefore, blind channel estimation is an alternative way to avoid pilot transmission. Considering a time-varying channel and taking the advantages of machine learning techniques, blind channel estimation and data detection for SM uplink multi-user massive MIMO communications is presented in this article. In this regard, a blind multi-user detection based on the expectation-maximization (EM) algorithm, called BMU-EM, is presented first; however, this detector suffers from high computational complexity. In order to mitigate the complexity problem, a blind multi-user detection based on sparse Bayesian learning and expectation-maximization, called BMU-SBEM, is proposed. Simulation results show that the BMU-SBEM detector performs almost close to the optimum detector where the perfect channel information is available. Furthermore, the computational complexity of the BMU-SBEM detector increases linearly with the number of users, making it suitable for massive communications in time-varying channels.

中文翻译：

---

空间调制上行链路多用户大规模 MIMO 通信的盲检测：稀疏贝叶斯学习期望最大化方法

大规模空间调制 (SM) 多输入多输出 (MIMO) 系统由于其功率和频谱效率而在未来移动通信的上行链路通信中是一种很有前途的技术。然而，这些系统与其他 MIMO 通信一样，面临着信道估计的挑战。基于导频的信道估计方法会降低数据速率，并在接收端增加额外的复杂性，这在时变信道中会加剧。因此，盲信道估计是避免导频传输的一种替代方式。考虑时变信道，利用机器学习技术，提出了单模上行多用户大规模MIMO通信的盲信道估计和数据检测。在这方面，首先提出了一种基于期望最大化（EM）算法的盲多用户检测，称为BMU-EM；然而，这种检测器的计算复杂度很高。为了缓解复杂性问题，提出了一种基于稀疏贝叶斯学习和期望最大化的盲多用户检测，称为BMU-SBEM。仿真结果表明，BMU-SBEM 检测器的性能几乎接近于可获得完美通道信息的最佳检测器。此外，BMU-SBEM 检测器的计算复杂度随用户数量线性增加，使其适用于时变信道中的海量通信。提出了一种基于稀疏贝叶斯学习和期望最大化的盲多用户检测，称为BMU-SBEM。仿真结果表明，BMU-SBEM 检测器的性能几乎接近于可获得完美通道信息的最佳检测器。此外，BMU-SBEM 检测器的计算复杂度随用户数量线性增加，使其适用于时变信道中的海量通信。提出了一种基于稀疏贝叶斯学习和期望最大化的盲多用户检测，称为BMU-SBEM。仿真结果表明，BMU-SBEM 检测器的性能几乎接近于可获得完美通道信息的最佳检测器。此外，BMU-SBEM 检测器的计算复杂度随用户数量线性增加，使其适用于时变信道中的海量通信。