【作者】 高洪元；

【导师】 刁鸣；

【作者基本信息】 哈尔滨工程大学 ， 通信与信息系统， 2010， 博士

【摘要】 随着无线移动通信技术的快速发展,人们的工作和日常生活变得更加便捷、丰富。无论是第三代还是第四代移动通信系统都需要更大的系统容量,才能为用户提供更丰富的多媒体业务及高速数据传输业务。众所周知,CDMA通信系统是一种严重受干扰限制的系统,多址干扰和远近效应是这种通信系统很难避免的主要干扰。如何有效地抑制多址干扰和远近效应,提高通信系统性能和容量具有很重要的理论价值和现实意义。解决这些问题的一种有效方法就是在接收端使用多用户检测技术,多用户检测不是把多址干扰和远近效应简单地看作干扰噪声来处理,而是把他们作为一种有用的信息,充分地利用各用户间的关联进行联合检测,提高系统的检测性能和系统容量,因此多用户检测成为CDMA通信系统的一个关键技术。最优多用户检测使用的穷尽搜索方法具有指数级的计算复杂度,这在当前的硬件水平下是不可能实现的。无论DS-CDMA还是MC-CDMA系统的最优或准最优多用户检测都可以看作一个组合优化问题,可以用智能计算方法解决。因此,深入研究智能优化理论,将智能优化的优化机理和多用户检测技术相结合,研究能够抑制多址干扰(MAI)和远近效应并具有低误码率(BER)和低计算复杂度的智能检测方法具有深远的意义,也是本文要解决的主要难题。本文研究了在高斯和冲击噪声环境下DS-CDMA系统和MC-CDMA系统的多用户检测模型,同时深入研究了智能信息处理理论,提出了一系列智能计算新方法,结合工程难题设计了多种基于智能计算的多用户检测方法。在多用户检测和智能信息处理理论研究方向,本论文的主要内容和创新如下：1.为了有效控制各个用户的功率,研究了三个测向难题,提出了解决这些技术难题的测向目标函数,并且设计了三种智能计算方法去分别求解目标函数：文化量子算法、差分粒子群算法和文化蜂群算法。所设计的三种测向方法不仅可有效用于多用户检测技术的功率控制,而且还可推广到其它应用测向技术的领域。提出的基于文化量子算法的广义加权子空间拟合测向突破了基于四阶累积量测向的一些局限。提出的基于差分粒子群算法的分数低阶协方差子空间拟合测向方法更适于冲击噪声环境下测向。所提基于文化蜂群算法的非圆极大似然测向方法更有效地利用了信号的非圆信息。2.针对DS-CDMA系统最优多用户检测器计算量大的缺点,提出了使用智能计算方法解决这个矛盾的三种框架。在每一种框架下,设计了一种新的智能计算方法完成最优多用户检测器的设计。仿真结果表明基于神经网络粒子群、免疫克隆量子算法和克隆量子算法的三种智能多用户检测器都具有结构简单和检测性能优的特点,适用于不同环境下的应用要求。3.结合神经网络和量子计算的特点,提出了新型的量子神经网络和量子混沌神经网络。所提的量子神经网络和量子混沌神经网络把量子演进机制和神经元的特点较完美的结合起来,具有更好的检测性能。使用所提的量子神经网络和量子混沌神经网络不仅可设计出有效的多用户检测方法,而且还可推广到一些可用Hopfield神经网络解决的组合优化问题。然后,基于量子机制和蛙跳算法的原理设计了量子蛙跳算法,与量子神经网络结合得到一种快速收敛的智能多用户检测方法。4.在给出了MC-CDMA系统的数学模型基础上,基于随机Hopfield神经网络、量子神经网路和两种群集智能,提出了神经网络鱼群算法多用户检测器和免疫蚁群算法多用户检测器,在多径衰落信道环境下验证了所设计的两种MC-CDMA检测器具有接近最优检测器的优良性能。5.在讨论了非高斯噪声DS-CDMA和MC-CDMA系统的多用户检测数学模型基础上,给出两种鲁棒多用户检测模型。结合DNA计算、群集智能和免疫系统的相关理论,提出了DNA克隆选择算法和DNA鱼群算法,设计了三种鲁棒多用户检测器以适合不同的冲击噪声背景的检测需要。更多还原

【Abstract】 With the rapid development of wireless mobile communication techniques, it is evident that the wireless mobile communication will greatly facilitate and enrich our work and daily life. In order to provide colorful multimedia service and high rate data service, the 3rd Generation (3G) and beyond 4th Generation (4G) mobile communication systems need higher wireless capacity and systems performance. It is known that Code-Division Multiple-Access (CDMA) mobile communication systems are severe interference-limited systems. Multiple access interference (MAI) and near far problem (NFP) are the main interference in the communications systems. It is important to suppress MAI and NFP of suppressing so that the system performance and capacity are increased. An efficient method suppressed MAI and NFP is multiuser detection (MUD) in which the MAI and NFP are viewed as useful information resource and the relationship between users is sufficiently used to improve the detection performance. So the multiuser detection is one of key techniques in CDMA communication systems.Optimal multiuser detection can not be implemented for its computation complexity of exponent. Since optimal multiuser detection problem of DS-CDMA and MC-CDMA can be viewed as a combinational optimization problem, intelligence computation can be used to resolve multiuser detection. This thesis is dedicated to the application of intelligence computational methods based on bionics to solve the difficult issue of MUD design capable of canceling the so-called multiple access interference and near far problem to reach low bit error rate (BER) with acceptable computation complexity. Our aim is focusing on the novel intelligence MUD algorithm development of DS-CDMA and MC-CDMA systems in Gassian noise and impulse noise environment. So we proposed a series of novel intelligence computation algorithms, and designed some multiuser detectors based on the proposed intelligence computation and classical problem.The main contribution of this thesis for multiuser detection and intelligence information processing can be summarized as follows:1. In order to control user powers, three direction finding problem are researched and corresponding objection functions are proposed. Three intelligence computation methods are designed to resolve corresponding objection functions:differential particle swarm optimization, cultural quantum algorithm and cultural bee colony algorithm. The three direction finding methods are not only effective for power control of multi-user detection technology, but also can be extended to other applications of finding technologies. The proposed cultural quantum algorithm based generalized weighted signal subspace fitting overcome some limitations the fourth-order cumulant-based methods. The proposed fractional lower order covariance subspace fitting method based on PSO algorithm based on differential particle swarm optimization is more suitable for impulsive noise environment. The proposed noncircular signal maximum likelihood method based cultural bee colony effectively used information of noncircular signals information.2. To resolve computation complexity of optimal multiuser detection in DS-CDMA systems, three intelligence computation frameworks are proposed. Based on every intelligence computation framework, a novel optimal multiuser detection based on intelligence computation is designed. Simulation results show that the three kinds of quasi-optimal multiuser detectors based on neural network particle swarm optimization, immune clonal quantum algorithm and clonal quantum algorithm have simple structure and good detection performance for different applications.3. Combined the advantages of artificial neural networks and quantum computation, we proposed a novel quantum neural networks and quantum chaotic neural network. The proposed neural network is the perfect combination of evolution mechanisms of quantum characteristics and neurons which lead to better detection performance. Based on the propose quantum neural network can not only design effective multiuser detection method, but also can be extend to some combinatorial optimization problem which can be solved by Hopfield neural network. Then, based on theory of quantum optimization and shuffled frog leaping, a quantum shuffled frog leaping is proposed and a multiuser detector based on the quantum shuffled frog leaping is designed.4. Based on the stochastic Hopfield neural networks, quantum neural network and two swarm intelligences, neural network fish school and immune ant colony are proposed in MC-CDMA systems model. The performance of the proposed detectors using neural network fish school and immune ant colony close to the optimal detector in multipath fading channel of MC-CDMA systems.5. Discussed multiuser detection model of the non-Gaussian noise in DS-CDMA and MC-CDMA systems, robust multiuser detections for different communications systems are presented. Combined DNA computing theory, swarm intelligence and the theory of the immune system, DNA clonal selection algorithm and DNA fish school algorithm are proposed. Then, three robust multiuser detectors are proposed for different noise environment.更多还原