大规模天线系统中基于数理统计方法的信道估计方法的研究

发布时间：2018-05-27 23:38

本文选题：大规模MIMO系统 + PCA　；参考：《杭州电子科技大学》2017年硕士论文

【摘要】：作为下一代通讯的关键技术之一,大规模天线(massive MIMO)技术如今越来越受到人们的重视。该系统中的每个基站配备着几十根到上百根天线,这些天线在同一频率上同时为多个用户提供服务。这样,在此系统中,空间分集能被充分地利用,从而频谱效率得到极大的提升。并且,即使利用简单的线性信号处理,例如迫零(ZF,Zero-Forcing)或者MMSE(Minimum Mean Square Error),系统的数据传送速率和能量效率都可以得到很大的提高。但是,上面提到的大规模MIMO系统所具有的各种优势是基于基站可以精确地估计出系统的信道状态信息(CSI)这一假设而得到的。一般来说CSI对于系统是未知的。传统的做法是利用基于训练序列的方法来获取CSI。具体来说,就是每个用户先发送预先设计好的且正交的导频信号,然后基站借助接收到的信号来估计CSI。但是针对大规模MIMO系统来说,其服务的用户数量非常多,为了保证正交性,需要很长的导频序列,因而耗费了大量的时频资源。而且,在此系统中,为了满足数据的高速率和大容量传输,会缩小小区的范围,使得小区的分布更加密集,相邻小区间的距离更加的接近。因此,当为了满足高频谱效率的要求使得相邻小区分享相同的导频序列时,会导致相邻小区用户的信号矢量干扰本小区用户的信道估计,从而大大降低估计的准确性。而且这种小区间的干扰会一直存在,与天线数无关。这种现象称为导频污染(pilot contamination)。该问题已经成为影响大规模MIMO系统性能的瓶颈。近些年,叠加导频(SP,Superimposed Pilot)技术已被用于大规模MIMO系统中。在SP技术中,由于没有时频资源用于训练,从而增加了系统的传送速率。在本文中,首先,我们对已有的SP技术作了改进。此项改进是基于严格MMSE的定义,从理论上作了严格的推导而得到的。因而它的估计性能要优于已有的方法。其次,本文还基于MMSE解码方法给出了新的解码方法。最后,仿真结果表明,新提出的方法的解码误码率要低于已有的SP方法。在本论文中,我们又提出了一种结合SP方法和二阶统计特性的新的信道估计方法。在该方法中,用户先传送数据符号和导频信号叠加在一起的信号,基站接收到信号后,利用主成分分析法(PCA,Principle Component Analysis)对接收信号矩阵进行降维处理,利用奇异值分解相应的性质得到信道估计剩下的步骤,再结合前面估计的信道及上述两种性质,得到一个新的信道估计。此外,基于MMSE解码方法设计出一种新的解码方法。仿真结果显示新提出的方法的系统性能要优于传统的基于训练的方法、现有的SP方法以及之前提出的方法。
[Abstract]:As one of the key technologies of the next generation communication, large scale antenna passive MIMOs have been paid more and more attention. Each base station in the system is equipped with dozens to hundreds of antennas that serve multiple users at the same frequency. In this way, spatial diversity can be fully utilized in this system, and the spectral efficiency can be greatly improved. Moreover, even using simple linear signal processing, such as zero forcing ZFG Zero-Forcing) or MMSE(Minimum Mean Square error, the data transfer rate and energy efficiency of the system can be greatly improved. However, the advantages of the large-scale MIMO system mentioned above are based on the assumption that the base station can accurately estimate the channel state information of the system. Generally speaking, CSI is unknown to the system. The traditional method is to obtain CSI based on training sequence. Specifically, each user sends a pre-designed and orthogonal pilot signal, and then the base station estimates the CSI with the received signal. However, for large scale MIMO systems, the number of users serving is very large. In order to ensure orthogonality, a long pilot sequence is required, which consumes a lot of time-frequency resources. In this system, in order to meet the high speed and large capacity of data transmission, the range of cells will be reduced, the distribution of cells will be more dense, and the distance between adjacent cells will be closer. Therefore, when the adjacent cells share the same pilot sequence in order to meet the requirements of high spectral efficiency, the signal vector of the adjacent cell users will interfere with the channel estimation of the local cell users, thus greatly reducing the accuracy of the estimation. Moreover, the interference in the interval will always exist, regardless of the number of antennas. This phenomenon is called pilot pollution pilot contamination. This problem has become a bottleneck affecting the performance of large-scale MIMO systems. In recent years, superimposed pilot technique has been used in large scale MIMO systems. In SP technology, there is no time-frequency resource for training, thus increasing the transmission rate of the system. In this paper, firstly, we improve the existing SP technology. This improvement is based on the strict definition of MMSE and is derived theoretically. Therefore, its estimation performance is better than the existing methods. Secondly, a new decoding method based on MMSE decoding method is presented. Finally, the simulation results show that the decoding error rate of the proposed method is lower than that of the existing SP method. In this paper, we propose a new channel estimation method combining SP method with second-order statistical properties. In this method, the user first transmits the signal superimposed by the data symbol and the pilot signal. After the base station receives the signal, the principal component analysis (PCA) is used to reduce the dimension of the received signal matrix. The remaining steps of channel estimation are obtained by using the corresponding properties of singular value decomposition (SVD), and a new channel estimation is obtained by combining the previously estimated channel and the above two properties. In addition, a new decoding method based on MMSE decoding method is designed. The simulation results show that the system performance of the new method is better than that of the traditional method based on training, the existing SP method and the previous method.
【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN919.3

【相似文献】