【作者】 郑勇；

【导师】 冯大政；

【作者基本信息】 西安电子科技大学 ， 信号与信息处理， 2010， 博士

【摘要】 本论文主要研究无线多天线系统的信道估计与信号检测问题。多天线的应用模式主要包括了空间分集(如空时码)、空间复用(如V-BLAST系统)等。文章对这两类问题都进行了研究,提出了新的改进并与原算法进行了比较。迭代方法可以有效提高算法性能,因此在本文集中研究了迭代算法及联合信道估计和信号检测算法在不同多天线应用模式下的实现及算法改进。文章最后还研究了在没有训练信号情况下的盲估计和均衡问题。研究内容主要包括:1、研究了一些最基本的V-BLAST系统的连续干扰消除(SIC)检测算法,包括了基于迫零和最小均方误差准则的连续干扰消除技术,基于QR分解的算法等。文中首先给出对系统模型和现有算法的描述和分析,针对连续干扰消除算法中存在的误差传递问题,然后提出一种在QR分解算法基础上,基于混合SIC和并行干扰消除(PIC)的改进算法。通过仿真比较和复杂度分析,对各种算法的性能和实现复杂度做了对比。2、研究高性能的基于后验概率的MIMO检测算法,其主要包含两个问题,一是如何利用软信息,二是如何实现最大后验所需的有效搜索算法。文中首先给出了软输入软输出接收机系统中如何在解码器和检测器件通过软信息的交换来实现Turbo迭代。另外,基于后验概率的算法主要是要能够实现高效的符号路径搜索来最大化后验概率。因此文中给出了几个主要的搜索算法,包括列表球形译码算法、堆栈算法以及M算法。文中提出了对其中一些算法的改进或新算法。针对堆栈算法中原有算法在搜索过程中需要通过额外的辅助堆栈计算来得到搜索路径长度偏移量的问题,本文提出一种更为有效的在算法中考虑长度偏移量的方法,在性能和算法复杂度上都有所改善。另外对于在搜索类算法中复杂度最小的M算法,其性能有时在MIMO检测中不能令人满意的情况,本文讨论了将结合Soft-Viterbi算法的M算法应用于MIMO检测的情况。在最后通过仿真验证了所提算法的有效性。3、研究了基于最小均方误差(MMSE)的MIMO检测算法。首先,我们依然先介绍了如何在MMSE MIMO检测中利用来自解码器的软信息,这与后验概率检测算法中的情况有所不同。基于最小均方误差的检测算法包含了线性最小均方误差检测(LMMSE)、非线性最小均方误差检测(NMMSE)以及各种干扰消除等技术,文中都对此进行了分析研究,并在此基础上提出了新的改进。首先,对基于NMMSE检测的GPDA算法进行了改进。引入了分组迭代GPDA算法,即将待检测信号先进行分组,组内用MAP检测,而其它分组信号作为干扰消去后残留干扰的叠加被近似为高斯噪声,并计算相应的匹配均值和匹配方差。这样一方面减少了干扰信号,一方面对组内的少量信号使用MAP算法提高了检测性能并控制了实现复杂度。通过调整分组大小,算法可以取得性能和复杂度的折中。在章节最后的实验部分对比了几种算法,验证了本文算法的较好性能。4、研究了空时块码正交频分复用(STBC-OFDM)系统的联合迭代信道估计与信号检测问题。STBC的解码性能很大程度上受到信道估计精度的影响,而实际系统中由于训练信号有限或者时变信道环境导致信道估计通常有较大偏差。文章研究通过联合迭代信道估计与信号检测的方法来改善信道估计和STBC解码的性能。在现有文献中包含了基于EM算法和循环最小算法的迭代实现。而这些算法在解码环节通常是通过搜索来实现解码的,而本文中提出了一种解析求解的方法大大降低解码复杂度。在信道估计问题上,本文在时域信道估计中通过有效利用FFT和IFFT在时域和频域信道间转换避免了原有算法中为了降低复杂度而限制信号为恒模的问题。仿真中在等间隔插入导频的OFDM系统和前置训练信号的系统信道跟踪问题中验证了算法有效性。另外在利用信道解码输出的迭代算法中,本文提出利用解码输出的硬判决反馈实现迭代,仿真中性能接近已知信道参数的理想情况性能曲线。5、研究了多天线系统的盲信道估计与均衡问题。本章给出了一种将子空间方法从单输入多输出扩展到多输入多输出系统的简洁的形式,并研究了有色信号输入下信道的辨识与均衡。多输入多输出系统盲均衡后会存在一个模糊矩阵,即变成一个信号瞬时混合的问题,文中通过联合对角化方法求解了该问题,得到了较为满意的结果。更多还原

【Abstract】 In this dissertation, we research the problem of channel estimation and signal detection in wireless multiple-input-multiple-output (MIMO) system. The applications of MIMO technique include space diversity (such as space-time coding) or space multiplex (such as V-BLAST). We have studied two systems in this dissertation, and improved the existing algorithms and compared with them. Iterative processing can efficiently improve the performance of algorithm, and in this dissertation, we concern with the iterative algorithm and the joint estimation of channel and signal under different MIMO systems. Further more, we also studied the problem of blind estimation and equalization of MIMO systems without training symbols. The main contents in this dissertation include:1、Sequential interference cancellation (SIC) algorithm for V-BLAST, including zero-forcing or MMSE based SIC and QR decomposition (QRD) based method, are studied. We have given the system model and description and analysis of these algorithms. To tackle the problem of error propagation, we propose an improved QRD based method which combines SIC and parallel interference cancellation (PIC). Advantages of the new algorithm have been verified by the simulations and complexity analysis.2、A prior probability (APP) MIMO detection algorithms with error correct coding (ECC) are studied. The two problems of such detection are, how to explore the soft information from the decoder, and how to find the optimal sequence to achieve APP detection by efficient searching. In the dissertation, we show how to implement the Turbo iteration by exchanging soft information between detector and decoder in the soft-in-soft-out (SISO) receiver. To achieve high efficient searching algorithm for APP MIMO detector, some of the prevalent algorithm are given, which include the list sphere decoding (LSD), the list sequential algorithm (LISS) and the M algorithm. In addition, we have proposed an improved LISS algorithm and an method applying soft-Viterbi M algorithm (SOMA) to MIMO detection. The improved LISS algorithm performs stack searching with the length bias, but avoids the time-consuming auxiliary stack operation in the original LISS algorithm. Both of the improved LISS algorithm and the SOMA for MIMO detection are verified via the simulations.3、The minimum mean square error (MMSE) based MIMO detector is also studied in this dissertation. At first, we still present how to explore soft information from decoder in MMSE based MIMO detector, which is different from the cases of the APP detector. In the relevant chapter, we have given and analyzed some MMSE based MIMO detecting techniques, such as linear MMSE (LMMSE) detector, nonlinear MMSE (NMMSE) detector, PIC/SIC methods and soft IC and hard IC algorithms. A new group-GPDA algorithm based on the NMMSE detection is proposed. Different from the original GPDA algorithm, new algorithm groups the signal to be detected and alternately perform MAP detection in each group, at the meantime, approximate the other groups as Gaussian noise with matched mean and matched variance. Such processing reduces the interference components and improves the performance by applying MAP detector to the group under detecting. By adjust the size of group, tradeoff between performance and complexity can be achieved. The performance improvement has been verified via the simulations.4、Another important MIMO technique is space-time coding (STC). The performance of STC decoding is greatly affected by channel estimate error, which is hard to avoid in practice due to limit numbers of pilots in OFDM symbol or time varying channel. In the chapter, an iterative joint channel estimate and space-time block coding (STBC) algorithm is proposed, based on the analysis of existing iterative algorithm, such as the EM algorithm and the cyclic minimization algorithm. The new algorithm greatly improves the performance of STBC decoding and channel estimate but avoid the time-consuming searching operations in those existing STBC decoding methods. In addition, the time-domain channel estimate avoids the limitation on constant-modular signals of existing method, which is to reduce the complexity of channel estimate, by efficiently using the FFT and IFFT to convert the channel parameters between time domain and frequency domain. In the simulations, our method has been verified to be approaching the curve with ideal channel state information (CSI), in the case of OFDM system with evenly inserted pilots, and the case of time-varying channel.5、Blind channel estimate and equalization for MIMO system is a problem to be studied in this dissertation. In the relevant chapter, we proposed a neat derivation for subspace based method for MIMO system. With this method, the problems of channel estimate and equalization with colored signal input have been studied. There exists an ambiguous matrix after the step of MIMO blind equalization. That is a problem of instantaneous mixture to be resolved. In our method, we explore the joint diagonalization to resolve it and the satisfied results have been gained.更多还原