【作者】 苏钢；

【导师】 朱光喜；

【作者基本信息】 华中科技大学 ， 信息与通信工程， 2007， 博士

【摘要】 针对未来移动通信系统架构中的关键技术问题,无线移动通信已成为通信领域的研究热点,包括能有效克服多径效应,提高系统数据传输速率的正交频分复用(OFDM, Orthogonal Frequency Division Multiplexing)技术;以及能显著提高系统容量或有效改善系统性能的多输入天线、多输出天线(MIMO, Multiple Input Multiple Output)技术等。本文在国家“十五”863计划重大课题“新一代蜂窝移动通信系统无线传输链路技术研究(2003AA12331005)”和国家自然科学基金重大项目“基于MIMO-OFDM系统的空中接口自适应技术研究(60496315)”的资助下,针对恶劣的无线移动通信中,多天线多载波环境下,发射机只有部分信道状态信息时的自适应传输方案进行了创新研究。本文首先研究多径衰落信道的信道脉冲响应模型,利用分形几何和分形测度等方法分析多径衰落信道的自相似特征,提出利用时间序列理论对信道观察值在欧式空间进行映射,实验表明信道估计值序列具有一定的分形相似性。其次针对当前高速TDD通信系统中,发射机通过链路反馈方式获取接收机的信道信息。当反馈时延和处理时延较大时反馈或估计获得的信道衰落将不能和未来的信道衰落有很好的匹配,导致系统性能恶化。本文针对这一特点,在发射机端只有不完全信道状态信息下,基于Jakes模型和自相关理论,利用MMSE准则,提出了对应的改进方法。该改进方法基于分形理论和扩展自相似过程,不仅能够有效的提高系统平均频谱效率,而且能有效的减少对"过期"的信道状态信息值的依赖,减少存储空间需求,明显减少系统复杂度;在单天线自适应OFDM系统中,当信道变化不剧烈时,基于扩展自相似过程的方法和经典的基于自相关方法性能区别不大,都能逼近理想信道容量极限;对于多普勒频移导致的时变快衰落信道,基于扩展自相似过程的方法能很好的拟合和跟踪信道变化,维持较高的系统平均频谱效率。而对于多天线自适应OFDM系统,处理方法相对要复杂得多,我们采用信道等效设计,定量地分析了多天线系统的平均频谱效率,并基于扩展自相似过程,提出一种适合多天线系统的改进方法。信道特性实际的非高斯特性,常常导致高斯假设下的系统性能严重降级。因此,在模型复杂性和算法结构的复杂度、精确度之间有一个选择的平衡,需要使用更符合信号噪声特性的实际的统计模型和处理系统。对于复杂的移动通信信道系统,分别用最大似然法、分位点法、抽样特征函数法对多径衰落的实测时间序列的进行参数分布估计,来考察信道的稳定性质,从而揭示出多径衰落信号和广义高斯分布的内在联系,因此,广义高斯分布可以更好的拟合信道响应序列。对慢衰落时变信道,信道频率响应序列可以视为弱偏离高斯过程,基于部分信道状态信息的信道估计修正性能与经典高斯理论下的性能相差不大。对于快衰落时变信道,其剧烈冲击过程,必然导致信道响应序列的方差趋向无限大,我们提出基于分数低阶矩的统计方法能提高系统的广义信噪比,保持一定的系统平均频谱效率,韧性也强于传统的高斯信号处理方法。终端移动速度越快,TDD系统需要增加更多的导频块,对时变信道的跟踪要求就越高,因此在实际系统中需要收敛更快的LMS算法。通过对信道预测值序列进行基于分数阶傅立叶变换以去其相关性或衰减其相关性,改善算法收敛性能,提高了信道跟踪性能。更多还原

【Abstract】 Recently, wireless mobile communication has been the hot topic in the domain of communication. Nowadays, many techniques have been comprehensively considered as the major candidates of the future radio communication architecture, including OFDM and MIMO. Supported by the National High Technology Research and Development Program of China under Grant No.2003AA12331005 and National Science Foundation of China under Grant No.60496315, our research work concentrates on the severe communication circumstances, for example, under multi-antenna and multi-carrier scenarios, we focus on the transmitter design of adaptive transform techniques with imperfect channel state information. First, we overview the current impulsive response model of multi-path fading channel.Then we analyze the self-similar characters of multi-path fading channel via fractional geometry and fractional metrics. As a time sequence, the multiple outdated estimates can be mapped into Euclid space to reduce uncertainty of the actual channel estimation. Experiments results show that the channel estimation sequence has some self-sililar character. Second, the transmitters acquire the channel state information (CSI) from the feedback of receivers in TDD systems. Unfortunately, system performance degrades due to imperfect CSI, caused by noisy channel estimation and delay. To improve the average spectral efficiency of adaptive OFDM system, we propose to use multiple outdated estimates and extended self-similar process to reduce the uncertainty of current channel estimate.Simulation results show that the novel method can track the actual channel state, so the average spectral efficiency of time-varying fast fading channels is significantly improved. The number of outdated estimates can be reduced significantly, which also means that the system can tolerate larger channel estimation error with reasonable complexity. For multi-antennae OFDM system, an identical scheme is applied to improve performance based on equated channel design, which provide theory guidance for their practical applications.The signal processing literature has traditionally been dominated by the Gaussian assumption. Unfortunately, Non-Gaussianity often results in significant performance degradation for systems optimized under the Gaussian assumption. Thus, more realistic statistical models must be used. There is a trade-off between model complexity and accuracy. The parameter estimation for multi-path fading channel estimations is calculated via method of maximum likelihood, method of sample quantiles and method of sample characteristic functoins, respectively. Analysed results show that Generalied Gaussian distribution assess stable fits to data sets. The data set of channel estimations is only slightly deviate from the Gaussian, so the performance changes little. But the outliers of fast fading cause infinite variance. So, the data set of multiple estimations is statistically processed via fractional low order moments. Simulation results show signal processing based on generalized Gaussian can improve generalized SNR and average spectrum efficiency with more robusticity than the Gaussian.In order to track the time-varying fast fading channel, LMS adaptive filtering is a very popularalternative technique to reduce quantity of pilots in real wireless systems. We introduce fractional Fourier transform (FRFT) and part of its properties, and then present its interpretation as a rotation of the timefrequency plane. The FRFT’s relationships with timefrequency representations have a very simple and natural form and supports LMS adaptive time-frequency filtering in the FRFT domain. The simulation results indicate that the novel method can speed convergence effectively with low complexity, which can benefit adaptive channel tracking.更多还原