【作者】 杨远；

【导师】 张海林；

【作者基本信息】 西安电子科技大学 ， 通信与信息系统， 2009， 博士

【摘要】 多输入多输出(MIMO: Multiple Input Multiple Output)技术给我们指出了一条利用空间维度来提高系统容量和(或)可靠性的有效途径,而MIMO系统接收端的设计最终会极大地影响系统的传输速率、误码率性能和系统复杂度。本文从性能和复杂度两方面的考量中,针对适用于各种信道条件(缺秩和满秩信道;相关性和非相关性信道)和天线配置下的MIMO接收信号的检测问题进行了研究。其主要工作概括如下:1.针对迭代检测解码MIMO系统中软输入软输出MMSE检测算法进行研究。提出了一种简化算法,它利用矩阵求逆引理和信道矩阵的奇异值分解,避免了大量的求逆运算,适用于各种调制方式。2.针对高维调制下的MIMO检测,提出了比特级的树搜索算法。该算法使用MMSE准则的信道预处理方法,同时考虑格雷和自然映射的QAM信号的比特映射结构可以得到一个比特级上三角分层结构。使用这样的上三角结构,并结合树搜索算法,可以得到比特级的树搜索算法。这样的树搜索算法适用于缺秩MIMO系统,相对于原始的树搜索算法,在高维调制时其度量更新和排序算法的计算量获得了明显降低,而且该算法还可以进一步利用反馈的先验信息来降低度量更新的计算复杂度。3.针对树搜索算法中的检测顺序进行研究,给出了在树搜索算法下利用对数似然比信息进行排序的算法。该算法充分利用接收信号、信道信息和先验信息来决定信号的检测顺序,能有效提高树搜索算法性能。4.利用分组检测结构和迭代反馈思想,提出了迭代分组最大后验概率检测算法。信号分组将高维检测问题分割为多个低维检测问题,未检测分组可以利用已检测分组得到的后验信息,在所有分组信号检测完成之后可以将其它分组的后验信息传递给第一个分组,重新进行一次检测过程,从而完成一次迭代。迭代分组最大后验概率算法可以充分利用检测器得到的后验信息,性能相对于分组最大后验概率算法有较大提高,而其复杂度可以由分组大小和迭代次数控制。5.提出了一种基于降格算法(LR: Lattice Reduction)的软输出栈检测算法。LR算法将等效信道矩阵转换为一个良态矩阵,在转换域内LR能够提供一个很好的发射信号估计。降格之后的每层信号的上下界可以计算出来。在得到初始序列估计和信号的边界之后就可以使用最佳优先的栈算法来进行转换域内的序列搜索,得到最有可能的发射序列集合,由此可以计算得到信号的软信息。仿真结果表明基于LR的软输出栈算法与其它算法相比能以较低的复杂度获得相同的误码率性能。更多还原

【Abstract】 Multiple-Input Multiple-Output (MIMO) technique is an effective way to achieve high capacity and (or) to improve the reliability. The design of MIMO detector can greatly affect the data rate, bit error ratio and the computational complexity. Based on the tradeoff between the performance and the complexity, this dissertation is intended to design effective and efficient MIMO detectors that are applicable for different channel conditions (full-rank or rank-deficient channel; correlated or uncorrelated channel) with any antenna configurations. The main results are summarized as followed.1. Investigate the efficient implementation of the soft-input soft-output MMSE detector for Turbo-MIMO systems and a simplified MMSE detector is proposed. By employing the matrix inversion lemma and the singular value decomposition of the channel matrix, the proposed algorithm decreases complexity greatly and can be applicable for any constellation.2. Based on the successive interference cancellation algorithm using QR decomposition, the MMSE preprocessing is introduced to form an equivalent upper triangular channel. By employing the MMSE preprocessing and exposing the underlying bit mapping structure of QAM signals, a bit-level tree structure can be obtained. When tree search algorithm is applied to this bit-level tree structure, a bit-level tree search algorithm is obtained. The proposed bit-level tree search algorithm can be employed in the rank-deficient MIMO systems. Compared with the original tree-search algorithm, the proposed algorithm can greatly reduce the computational complexity when high order constellation is employed. In addition, the proposed algorithm can utilize the a priori information from the decoder to further reduce the computational complexity.3. The detection ordering algorithm in the tree search detection is investigated and the ordering algorithm employing log-likelihood ratio information in the tree search detection is proposed. The novel algorithm can fully utilize the received signal, channel state information and a prior information to determine the detection order. The performance of the tree search detection can be improved by the ordering algorithm. 4. The group detection strategy operates by dividing the set of transmitted symbols into small groups. In the proposed iterative group MAP detection algorithm, information obtained by the detected groups is shared by other groups. After all the groups are detected, the information about all the transmitted signals can be feed back to the first detected group to improve the performance; all the other groups are detected again. The algorithm works in an iterative way. The iterative group MAP detection can fully utilize the information obtained by the detectors and it achieves a better performance compared with the group MAP detection. Its complexity can be controlled by the group number and iteration number.5. A soft-output detector with lattice-reduction (LR) is proposed. LR can transform the system model into an equivalent one with better condition. LR algorithm provides a good initial estimation of the transmitted signals. The boundary of signals in the reduced lattice for each layer can be obtained. After the boundary and the initial estimation are obtained, the stack algorithm can be employed to search the candidate list. When the list is generated, the soft bit of the transmitted signals can be calculated. Simulation results show that compared with other detectors, the proposed algorithm can provide the same performance with significantly reduced computational complexity.更多还原