【作者】 李元杰；

【导师】 杨绿溪；

【作者基本信息】 东南大学 ， 信号与信息处理， 2006， 博士

【摘要】 在现代通信中,数据高速传输的要求与有限带宽之间的矛盾促进了人们对无线通信系统空间资源的利用。近年来的研究表明,MIMO系统由于采用了空间分集和复用技术,大大提高了传输速率,因而获得更大的系统容量。本论文研究了在MIMO无线通信系统中的信道估计与跟踪问题,主要的工作如下:1.提出了一种基于误差修正的单载波多天线系统信道半盲估计的方法,该方法利用了已知训练序列和全部的数据帧信息。估计算法分为两步进行,在第一步,信道参数的估计初值可以利用由训练序列构成的循环前缀得到,因而有效地节省了带宽。在第二步,通过比较接收信号和估计信号之间的误差,使用误差信息对信道的估计初值进行修正,从而得到更为精确的信道估计值,且有比较低的计算复杂度,更易于实际实现。2.提出了一种快速高效的MIMO CDMA系统信道盲估计方法,在接收机的前端,我们设计了一种特殊结构的解相关匹配滤波器。该解相关器参数不依赖于未知的信道,而且具有很好的块对角阵结构,易于参数的实时更新。然后用一种简单的基于一阶统计量的方法从解相关匹配滤波器的输出信号估计出多径信道。由于只使用了一阶统计量,该算法的计算量远小于其他有效的盲估计算法,而且具有很高的估计精度。且易于对缓慢时变信道进行跟踪。3.提出了基于聚类分析的MIMO CDMA信道盲估计算法,对于解相关器的输出信号,使用K-均值聚类算法提取其中的信道信息,从而避免了矩阵的特征分解运算。而且均值运算可以很好地消除加性噪声的影响,对于有色高斯噪声无需进行预白化,因而降低了计算量。另外还根据发射信号的特点,对K-均值聚类算法的初始值设定进行了改进,以提高收敛速度。而且该算法对于加性噪声及信道阶数过估计不敏感,有很好的鲁棒性。4.针对通信系统发射信号是取自有限符号集的特点,提出了基于EM算法的MIMO CDMA系统极大似然盲信道估计算法,在接收端将解相关器的输出信号矢量建模为混合高斯分布的随机变量,每个高斯分布的峰值即对应于信道矢量,从而可以用EM算法通过迭代的方法得到信道参数的极大似然估计,在较少观测样本情况下就能得到很好的估计性能。5.提出了一种基于隐训练序列的信道估计方法,无需专门为训练序列分配时隙,在没有带宽损失的情况下估计出信道参数。不同于以往的利用接收信号循环周期性的信道估计方法,该算法主要是利用了训练序列与信息序列的不相关特性,因而具更多还原

【Abstract】 The increasing demand for high data rate and the limited available bandwidth motivates the investigation of wireless systems that efficiently exploit the spatial domain. It has been recently shown that the use of spatial diversity can improve throughput and coverage in addition to allowing a high degree of spatial reuse and thereby increase the system capacity. In this dissertation, we mainly focus on the topics of channel estimation and tracking in MIMO communication systems. The main contents are as follows:1. Propose an error-adjustment scheme for semiblind channel estimation. The algorithm consists of a two-step iteration to achieve even better performance. The initial channel estimate is computed using a variant of cyclic prefix. Then the estimate error of the received symbol is used to adjust the initial channel estimation. Computer simulations show that the proposed algorithm offers good estimation behavior for MIMO systems.2. A low-complexity blind channel estimation technique is proposed for long-code MIMO CDMA systems. Simply based on the first-order statistic of the decorrelating matched filter output, the channel parameters can be estimated effectively without training sequences. Computer simulations show that the proposed algorithm offers good estimation behavior for MIMO CDMA systems.3. A blind channel estimation method is presented for the multiple-input-multiple-output (MIMO) CDMA system employing long spreading codes. To estimate channel parameters, we apply a decorrelating matched filter as front-end at the receiver. The matched filter outputs consist of the signal space spanned by the users channel vector and we can find the centers of the sets by clustering procedure. Then the channel parameters are estimated via the first-order statistics of the filter outputs. We also improve the K-mean clustering algorithm in class centers initialization for fast convergence. The proposed method has low computational complexity and estimates the channel in an effective way.4. A blind channel estimation method is presented for the multiple-input-multiple-output (MIMO) CDMA system employing long spreading codes. By exploiting the finite alphabet p.d.f. of the transmit symbols, the matched filter outputs can be modeled as a Gaussian Mixture Model. Then the channel parameters are estimated using expectation-maximum (EM) algorithm that fully exploits the statistical features of the transmitted signal.5. A new estimation method based on the superimposed (implicit) training is proposed for更多还原