新一代短距离无线通信系统信号检测与接收技术研究

【作者】 李斌；

【导师】 周正；

【作者基本信息】 北京邮电大学 ， 通信与信息系统， 2013， 博士

【摘要】 以高速宽带、低功耗、低成本及个性化应用为特点的短距离无线通信,已成为未来泛在无线网络的重要组成部分,因而受到人们的普遍关注。本文选题来源于国家自然科学基金及国家科技重大专项等项目,具有重要的理论意义及广阔的应用前景。本文针对未来短距离无线通信系统信号检测与接收技术进行深入研究,重点分析了密集多径传输下的信号检测与接收机制。主要完成了以下具有创新性的研究成果：1.针对密集多径分簇信道下的参数提取、理论建模及模型评估问题,i)提出了一类基于突变点检测与仿生聚类机制的自动分簇识别方案,两种新算法均可显著提升密集多径信道下分簇识别与参数提取的准确度；ii)提出了一种基于Fresh反射理论的新簇内多径功率时延剖面(PDP)模型,能更为合理地匹配密集多径测量数据,并可揭示其中的大尺度级频率选择性衰落特性；iii)提出了一种基于分簇PDP包络加权的低复杂度非相干检测方案,能有效提升检测性能,也可验证密集多径分簇提取算法与新PDP模型的准确性。2.针对密集多径信道下低复杂度短距离无线传输场景,提出了一种统一的非相干信号检测新框架,其中包括：i)一种基于协方差矩阵及特征谱量化特征模式分类的仿生信号处理机制；ii)四种不同应用需求下的密集多径信号非相干检测方案,即基于模糊C-均值聚类的非相干盲检测方案、基于改进蚁群聚类分析的非相干盲检测方案、基于Parzen窗估计概率神经网络的在线式非相干检测方案、以及基于新数值优化算法的离线式非相干检测方案；iii)一种基于文化基因与量子计算的量子文化进化数值算法,可实现高性能模式分类与信号检测。3.针对密集多径传输与射频功放器件非线性失真问题,提出了一种非线性动态状态空间模型及一种基于贝叶斯统计推理框架的联合信道估计与盲信号检测方案,可直接在接收端对非线性失真与多径串扰后的接收星座图进行行盲校准。针对蒙特卡罗序贯重要性采样理论难以直接处理非线性盲均衡的难题,提出了一种通用非线性估计粒子滤波算法,并设计了一种能同时应对线性串扰与非线性失真的序贯联合盲估计方案,可有效避免传统的复杂预失真机制。4.针对新一代高速短距离无线局域网／个域网中波束赋形机制复杂度过高的难题,提出了两种基于无约束数值优化的新波束赋形训练算法,包括：i)一种基于Rosenbrock数值算法与新型预搜索机制的高维空间波束寻优方案,能克服Rosenbrock算法易陷入局部解的固有局限,通过逐步迭代细化方式缩小目标搜索区域,可获得100%搜索成功率,并显著降低搜索复杂度、协议包头与设备功耗；ii)一种融合了概率扰动与新型两级参数调控、并适宜于离散空间寻优的全局性数值搜索算法,能以100%概率收敛至最优解,在非视距密集多径传输下搜索到最优波束。5.针对时间选择性衰落信道,提出了一类基于新颖动态状态空间模型的联合信道估计与盲信号检测机制,包括：i)一种基于贝叶斯统计推理框架的时变慢衰落信道下分布式认知频谱检测算法,通过迭代估计方案可联合估计出授权用户状态与时变信道增益,在无需多节点协作检测的情况下,能显著提升时变衰落信道下的分布式频谱检测性能；ii)一种基于序贯检测原理的时变密集多径信号盲检测算法,通过高效迭代机制可联合估计出时变密集多径响应与未知数据序列,能在时变性密集多径衰落下获得良好的盲检测性能。最后,对全文研究内容进行了总结,并对下一代短距离无线接入与互联中的信号检测与接收技术发展方向及今后工作进行了展望。更多还原

【Abstract】 The short-range wireless communications, characterized by high-data-rate, wide bandwidth, low-power consumption, low cost and personalized ap-plications, have emerged as an important part of future ubiquitous wireless ac-cessing networks, which hence nowadays have drawn general attention. Sup-ported by the Natural Science Foundation of China as well as the National Sci-ence and Technology Major Projects, this thesis is mainly devoted to promot-ing signal processing in the future short-range wireless communications, which may have the great theoretic significance and the wide application prospect.This thesis mainly investigated the signal receiving techniques in next-generation short-range wireless communications, with the emphasis especially put on signal detection and reception in intensive multipath propagations. Ac-cordingly, the main innovative contributions of the investigation can be sum-marized as follows.1. For the issues of parameters extraction, parametric modeling and model evaluation in the intensive multipath channels,1) two kinds of efficient auto-matic cluster identification algorithms, inspired by two promising perspectives of discontinuity/singularity detection and biological pattern clustering, have been proposed, which can significantly improve the accuracy of cluster identi-fications and parameters extractions;2) a novel parametric intra-cluster power delay profile (PDP) model derived from the Fresh reflection theory is present-ed, which, for the first time, reveals a new kind large-scale frequency selectivity and can match the multipath measurement more effectively;3) a clustered PDP weighted low-complexity signal detector is proposed, which can enhance the receiving performance and is hence of significance to verify the effectiveness and accuracy of the proposed cluster identifications and PDP modeling.2. For the low-complexity short-range wireless communications in the p- resence of intensive multipath propagations, a unified new non-coherent signal detection framework is designed.1) A pattern classification-based noncoher-ent signal detection algorithm, based on the covariance matrix of received ran-dom signals and a developed novel characteristic spectrum, is proposed, and a promising biologically inspired signal processing (Bio-SP) framework is then suggested by combining nature intelligence inspired algorithms;2) Four non-coherent detection schemes are designed for several scenarios with different realistic requirements, i.e., the fuzzy c-means (FCM) clustering based blind noncoherent detector, the modified ant colony clustering based blind nonco-herent detector, the Parzen probabilistic neural network (PPNN) based non-coherent detector for the online data-aided (DA) case, and a new biological algorithm-based off-line supervised noncoherent detector;3) A novel quantum memetic algorithm (QMA) inspired by the memetic gene and quantum comput-ing mechanics is developed, which can realize the high-performance numerical optimization and signal classification/detection.3. For the complex scenarios with both intensive multipath propagations and nonlinear power amplifier distortions, a novel nonlinear dynamic state-space model (DSM) is properly formulated. On this basis, relying on the Bayesian statistical inference, a joint channel estimation and blind signal detec-tion algorithm is developed. By excluding the pre-distortion in transceiver-end with high complexity, the distorted signal constellations, after propagated from both nonlinear PA and linear multipath, are directly calibrated in receiver-end. Since the existing Monte-Carlo sequential importance sampling can hardly deal with the nonlinear equalization, a local linearization technique is introduced and a generalized particle filtering is presented. Thus, a sequentially joint and blind estimation technique is proposed which can simultaneously address both multipath interference and nonlinear distortion.4. Considering the high complexity of the existing beam-forming train-ing in the next-generation ultra-high speed WLANs/WPANs, from a promis-ing perspective of non-constrained numerical optimization, two efficient beam- forming training algorithms based on numerical optimization are developed.1) The first algorithm, premised on a Rosenbrock numerical search, is suggested to identify optimal beam-patterns. To overcome the shortcoming of classical Rosenbrock numerical search which may easily fall into local optimums, a pre-search algorithm inspired by the small region dividing-and-conquering is presented. The new algorithm can find the optimal beams with a success prob-ability of100%, which thereby significantly reduces the search complexity, the protocol overhead and power consumption.2) A new numerical search algo-rithm is developed by integrating the probabilistic disturbances mechanic and a novel two-level disturbances control scheme. The developed numerical algo-rithm is of special promise to the discrete-space optimization, which basical-ly avoids local optimums and the computation demanding local exploitations. This new algorithm can identify optimal beams with an probability of100%in intensive multipath propagations.5. In order to address the difficulty of signal detections in the presence of dynamic time-varying fading propagations, a novel dynamic state-space model is developed and then a unified joint channel estimation and blind signal de-tection is proposed.1) A distributed spectrum sensing algorithm for cognitive radios is proposed which relies on the Bayesian statistical inference and Monte-Carlo random sampling theorem. The time-varying fading gain, accompanying the primary user working state, is iteratively estimated in time. To the best of our knowledge, for the first time, this new spectrum sensing algorithm real-izes the joint estimation of unknown primary state and time-varying channel gain, which can significantly improve the sensing performance in time-varying fading channel.2) A blind signal detection algorithm in the presence of time-varying intensive multipath propagations is proposed. An efficient iterative detection algorithm is designed which can jointly estimate the time-varying multipath channel and the unknown transmitted symbols. The new algorith-m can significantly enhance the blind detection performance in time-varying intensive multipath propagations. Finally, the investigations of the whole dissertation are summarized. On this basis, several valuable research directions on signal reception techniques in future short-range wireless connecting are discussed.更多还原