【作者】 王赟；

【导师】 汪晋宽；

【作者基本信息】 东北大学 ， 导航、制导与控制， 2008， 博士

【摘要】 无线多输入多输出(MIMO)技术可以显著提高系统容量,是下一代移动通信系统的关键传输技术。多输入多输出技术充分开发空间资源,利用多个天线实现多发多收,在不需要增加频谱资源和天线发送功率的情况下,可以成倍地提高信道容量。典型的空间复用技术是贝尔实验室的空时分层结构。本文对独立平坦衰落与频率选择性衰落信道环境中的MIMO信号检测技术进行了研究,在不降低频谱效率的前提下,力图通过提高检测算法性能改善系统的误比特性能,并降低信号检测的计算复杂度。研究着眼于采用空时分层结构的MIMO系统检测技术中若干关键技术,包括GOLDEN检测技术、均方根迭代检测技术、按序QR分解检测技术、自适应最小二乘按序判决反馈检测技术做了详细的研究,主要研究内容如下：针对常规V-BLAST检测算法在MIMO系统接收端进行检测时,需要进行大量伪逆运算导致检测复杂度增加的问题,提出了一种基于迭代QR分解的MMSE V-BLAST算法,把对信道矩阵求伪逆的过程转化为利用迭代QR分解近似逼近的过程,从而避免了伪逆运算,有效的降低了检测算法的复杂度。并且由于避免了伪逆运算而无需满足信道矩阵的行数必须大于或等于列数的要求,因而对于发射和接收端无需满足发射天线数必须小于或等于接收天线数的要求,扩展了系统的一般适应性。考虑到传统的均方根检测算法计算复杂度高。提出快速均方根V-BLAST检测算法,所提算法利用矩阵酉变换的性质,仅对信道矩阵进行一次排序,并且无需迫零向量,该算法在系统检测性能总体保持一致的情况下,计算复杂度下降,对于响应要求及时的系统有很好的适应性。用于多输入多输出通信系统检测的按序QR分解算法在多径瑞利慢衰落信道中系统复杂度低,但误码率较高。提出一种基于列正交变换的串行干扰消除算法,该算法对信道矩阵按列正交变换,避免了求上三角矩阵的运算,并且在判决信号过程中,将先判决出的信号通过信道后的输出向量作为干扰进行消除,从而避免了已判决信号对未判决信号的影响。在基于列正交变换的串行干扰消除算法基础上,借鉴并行处理的思想,提出并行QR分解检测算法,并对其检测性能进行了分析。所提算法首先对信道矩阵按列正交变换,并且利用矩阵酉变换的性质仅对信道矩阵进行一次排序,在判决信号过程中,采用部分判决信号反馈和接收信号干扰消除并行处理的检测算法。在低信噪比的情况下,由于避免了功率强度大的信号对其余信号产生干扰,所以此算法适用于信道条件恶劣的情况下。理论证明,利用修正的Gram Schmidt正交化方法对矩阵进行正交化的计算量大,而基于Householder变换的正交化方法的计算量大约为其2/3。因此,在MIMO系统检测时,考虑采用基于Householder变换的串行干扰消除算法。所提算法对信道矩阵进行Householder变换并且利用矩阵酉变换的性质仅对信道矩阵进行一次排序,在判决信号过程中,将先判决出的信号通过信道后的输出向量作为干扰进行消除,因而计算量降低。提出一种在MIMO频率选择性无线环境中得到高数据传输率的接收机结构。基于RLS按序串行干扰消除的MIMO判决反馈算法在应用迭代最小二乘算法得到的前馈数据向量中按序将已检测并判决出的信号进行消除,并将前馈数据向量合并到反馈数据向量中,避免了已判决出的信号对前馈数据向量的干扰,利用矩阵求逆公式,重新定义迭代过程中权重向量。使迭代过程均方根误差降低,检测性能提升。在基于RLS按序串行干扰消除的MIMO判决反馈算法的基础上,借鉴并行处理的思想提出了基于RLS并行干扰消除的MIMO判决反馈算法。所提算法在应用迭代最小二乘算法得到的前馈数据向量中运用并行干扰消除的方法将已检测并判决出的信号进行消除,并将前馈数据向量合并到反馈数据向量中。所提算法无需对信道矩进行排序,因而计算量降低；在低信噪比的情况下,由于避免了功率强度大的信号对其余信号产生干扰,所以此算法适用与信道条件恶劣的情况下。传统的最小二乘恒模算法(LSCMA)误差曲线不具对称性,是导致LSCMA算法收敛速度慢、收敛后均方误差大的主要原因。为此将LSCMA算法进行了改进,将其误差曲线定义为对数正态误差曲线,并在次基础上加入了判决条件。将运用改进的对数正态误差恒模算法得到的前馈数据向量合并到反馈数据向量中,并按序将已检测并判决出的信号进行消除,避免了已判决出的信号对接收数据向量的干扰。由于LSCMA算法利用了信号的恒模性质,因而比传统的RLS算法复杂度低,适用于快衰落信道。更多还原

【Abstract】 Wireless multiple-input multiple-output (MIMO) antenna systems offer significant improvements in performance and capacity when used in wireless communications and have received much attention recently. The MIMO exploits the space resource to improve the channel capacity effectively without additional frequency spectrum and transmission power. A typical MIMO multiplexing technique is Bell laboratories layered space-time (BLAST) architecture. This dissertation makes researches on signal detection techniques for MIMO systems over independent flat fading channels and frequency-selective fading channels aiming at improving the performance of detecting algorithm, ameliorating the Bit Error Rate (BER) performance, and decreasing the computational complexity, without loss in spectrum efficiency. The study focused on some key signal detection technologies of MIMO system with space-time layered structure, such as GOLDEN detection technique, sorted QR decomposition detection technique, adaptive RLS ordered decision feedback equalization detection technique. The thesis work can be summarized as follows.An improved MMSE V-BLAST algorithm based on iterative QR decomposition is proposed to overcome the shortcoming of increasing system detection complexity caused by a lot of pseudo-inverses operations when detecting using MMSE detection algorithm in MIMO system receiver. The complexity is reduced and the performance is improved. The MMSE V-BLAST algorithm extended the system’s adaptability, because of avoiding the pseudo-inverse operations and the system needn’t satisfy the requirements that the number of transmitted antennas must less then that of received antennas.A fast square-root detection algorithm for V-BLAST was proposed. The algorithm aimed at the question of high complexity when traditional square-root detection algorithm detecting in MIMO receivers. It simplified the computing process of traditional square-root detection algorithm and proposed a parallel processing algorithm of judged signal’s feedback and received signal’s interference cancellation. The detection complexity of the algorithm was reduced, while the detection performances retain. It has good adaptability for the system with response in time.In view of the fact that the sorted QR decomposition (SQRD) algorithm for multiple input multiple output (MIMO) communication detection has a higher bit error rate when working in multi-path Rayleigh slow fading channels, the paper proposes a serial interference cancellation (SIC) algorithm for MIMO detection based on column orthogonal (CO) transform named COSIC. The COSIC algorithm transforms the channel matrix column orthogonally to avoid solving the upper triangular matrix. In the processing of judging signals, it takes the output of the judged signals which are transmitted by the channel matrix as interference to cancel and the detection performance is improved distinctly on the basis of increasing system time complexity a little.Theory proving, the complexity of the modified Gram-Schmidt method to decompose the channel matrix is high. The complexity of Householder transformation is as 2/3 times as it compared to the modified Gram-Schmidt method. An improved parallel detection algorithm based on Householder transform (HIP) is proposed. The HIP algorithm transforms the channel matrix column orthogonally based on Householder transform to avoid the operation of upper triangular matrix and sorted the channel matrix only once. In the processing of judging signals, it proposes a parallel processing algorithm of judged signals’ feedback and received signals’ interference cancellation and the complexity is reduced distinctly.We present a new receiver structure able to deliver high data rates in a multi-input-multi-output (MIMO) frequency selective wireless environment. The ordered successive interference cancellation MIMO decision feedback equalization based on recursive least square algorithm (RLS-OSIC-DFE) is obtained by canceling decided symbols from the received symbols successively with the decision feed-forward equalizer solution as a well known expression encountered in fast recursive least squares adaptive algorithms, and the decision feedback equalizer as a convolution of the decision feed-forward equalizer with the channel. The RLS-OSIC-DFE algorithm avoids the the interference of the decided symbols and improves the detection performance and the detection performance of proposed algorithm is improved dramatically.On the basis of the RLS-OSIC-DFE, We present a parallel interference cancellation MIMO decision feedback equalization based on recursive least square algorithm (RLS-PIC-DFE) by using the parallel processing. It is obtained by canceling decided symbols from the received symbols parallelly with the decision feed-forward equalizer solution as a well known expression encountered in fast recursive least squares adaptive algorithms, and the decision feedback equalizer as a convolution of the decision feed-forward equalizer with the channel. The RLS-PIC-DFE algorithm avoided sorting the channel matrix and reduced the complexity. While with low SNR, it avoided the interference caused by the signal of high power. It is suitable for the bad channel condition.The error curves of LSCMA have no symmetry and it is the main cause of slow convergence and large mean square error. Then we proposed the LSCMA and defined the error curves as a novel lognormal error function and added a decision condition. It is obtained by canceling decided symbols from the received symbols successively with the decision feed-forward equalizer solution as a well known expression encountered in improved log-normal error function based on CMA (ILNCMA-OSIC-DFE) adaptive algorithms. The decision feedback equalizer replaces a convolution of the decision feed-forward equalizer with the channel. The ILNCMA-OSIC-DFE algorithm avoids the interference of the decided symbols and improves the detection performance. Because the LSCMA used the constant modulus property, it is suitable for fast fading channel.更多还原