【作者】 肖曼琳；

【导师】 魏平；

【作者基本信息】 电子科技大学 ， 信号与信息处理， 2013， 博士

【摘要】 为了使未来的无线通信系统能满足更多的商业和公共服务的需求，扩大系统容量已经成为当今无线通信领域最重要的研究课题。认知无线电技术是通过机会主义的频谱资源共享来提高整个系统的容量，所以近年来该技术作为最有竞争力的备选方案已经被广泛研究。认知无线电应用中第一步也是最重要的技术是在感兴趣的频段内检测频谱的使用情况，即判断原始用户是否存在。为保障原始用户不受到干扰并且为认知无线电用户争取到更多的可用时间，原始用户的检测工作必须在低信噪比和短时间内完成。所以，认知无线电中信号检测问题的最大挑战是寻找在低信噪比和小样本数情况下性能优良的检测算法和检测策略。本论文提出用多部车辆做为检测设备的载体，对原始用户进行快速检测。根据不同的检测要求和环境情况，研究了不同的检测策略和检测技术，包括时频技术、合作协作、空域技术等。主要工作和贡献如下：1.以已开放频段的原始用户信号——ATSC数字电视信号为例，提出了不同的时频信号检测算法。在完成ATSC数字电视信号建模的基础上，通过分析其频谱特性，提出了导频信号检测算法。根据ATSC数字电视信号的特殊编码结构，提出了数据域同步信号检测算法和数据段同步信号检测算法。通过分析ATSC数字电视信号的统计特性，提出了周期平稳信号检测算法。通过仿真实验，给出了这几种检测算法的性能分析，为工程实现提供了可靠的理论依据。2.对车载合作感知系统进行讨论并完成系统仿真。在该系统中，多部车载设备通过中心协调式合作策略对ATSC数字电视信号进行合作检测。整个仿真系统可以分为信号产生，信号检测以及数据融合三个部分。信号产生部分主要生成需要检测的原始用户信号，即将通过车载信道后的ATSC数字电视信号为接收数据。车载信道包括不同环境下的衰落、阴影效应和多径等。信号检测部分可以采用简单的盲检测算法也可以采用特征检测算法。数据融合部分主要使用了软判决和硬判决的数据融合策略。在仿真实验中，采用经典的环境参数并且对检测算法和数据融合算法进行不同的组合。仿真实验结果为不同情况下的策略选择提供了理论依据。3.通过多信号特征分析，提出了基于样本特征值的检测算法。在该算法中，首先将多部车辆接收到的样本组成数据矩阵，然后通过特征分解得到样本特征值并且利用样本特征值构造出检测统计量，最后通过和检测门限做比较获得判决结果。提出了两种适用于样本数较少情况下的检测统计量，并且根据样本特征值的波动特性给出相应的检测门限。通过仿真实验可以证明这两种算法能够有效地在短时间内完成对原始用户的检测。4.讨论了原始用户个数估计算法。分析了传统个数估计算法在样本数较少时性能下降的原因，给出了利用特征值完成信号检测时所需的极限能量。提出了利用样本特征值构造检测统计量的两种方法，并结合序列假设检验方法给出了原始用户个数估计算法。通过理论分析和仿真实验，证明了这两种估计算法在样本数较少的环境下，比传统信号源个数估计算法和许多近年来提出的改进算法具有更优的估计性能。5.讨论了原始用户的来波方向估计算法。首先，分析了短数据和低信噪比情况下，基于子空间的测向算法性能衰减的原因，然后，提出了一种计算简单的参数化IAA-APES算法。该算法一定程度上克服了由于子空间泄露所造成的测向性能下降。仿真实验结果表明所提算法在较少样本下测向精度优于MUSIC算法且具有很高的分辨率。另外，针对分布式接收机本振信号的同步问题，采用了一种利用指示信号双向传输的方式，使得各接收机在各自的时钟下产生同步的本振信号。更多还原

【Abstract】 In order to enable future wireless communication systems for more commercial useor public services, the development of system capacity has been an important researcharea in wireless communications. In recent years, the cognitive radio technology hasbeen widely researched as the most competitive candidate since it utilizes theopportunistic spectrum sharing to increase system capacity. The first and foremost stepin cognitive radios is to detect the usage of spectra of interest, i.e., to decide whetherthere are primary users. In order to avoid interferences to primary users and to savemore available time for cognitive radio users, the primary signal detection should beconducted under the cases of low signal to noise radio (SNR) and/or small observationperiod. Therefore，the challenge of detection problem in cognitive radios is to design thenew algorithm and strategy which should keep good performances under the scenariosof low SNR and small samples. In this dissertation, multiple vehicles are used ascarriers of detection devices to detect primary users in a short time. To meet differentdetection requirements and various detection environments, some detection techniquesand strategies are explored, including time-frequency domain detection techniques,cooperative strategies and spatial processing technologies. The main work andcontributions are presented as follows.1.Taking the ATSC digital TV signal as an example which is the primary signal inan open frequency band for cognitive radio users, various time-frequency domaindetection algorithms are proposed. Based on the modeling of the ATSC digital TVsignal, the pilot tone detection algorithm is proposed by analyzing the characteristic ofspectrum. The data field synchronization signal detection algorithm and the datasegment synchronization signal detection algorithm are designed by utilizing thespecific coding scheme of the ATSC digital TV signal. The cyclostationary detectionalgorithm is explored based on statistical properties of the ATSC digital TV signal. Theperformances of these algorithms are analyzed by numerical simulations which providethe reliable theoretical basis for the engineering realization.2. The vehicular spectrum sensing system is discussed and modeled in which multiple vehicular sensors cooperatively detect the ATSC digital TV signal. This systemconsists of signal generation，signal detection and data fusion. The part of signalgeneration is to produce received ATSC digital signals which are transmitted viacomplicated vehicular channels. Path loss, shadowing and multipath in variousenvironments are considered. Simple blind detection algorithms and specific detectionalgorithms could be chosen in the part of signal detection. The options of data fusion aresoft decision and hard decision. The classical channel coefficients and signal parametersare used, and different combinations of detection algorithms and data fusion techniquesare adopted in numerical simulations. The best detection strategies in different scenarioscould be chosen on the basis of simulation results.3.The eigenvalue-based detection algorithms are proposed with the multi-signalcharacteristic analysis. In new detection algorithms, the data matrix consists ofobservation samples received by multiple vehicles. Then, sample eigenvalues obtainedby the eigen decomposition are used to construct new detection statistics. Comparingwith the detection threshold, the result could be decided lastly. Two new detectionstatistics are proposed which have good performances in the cases of low SNR andsmall samples. Taking into full account of the fluctuation of sample eigenvalues, thecorresponding detection thresholds are derived. Numerical simulations illustrate that thenew algorithms could figure out primary users in a short period.4.The estimation of the number of primary users is explored. The reason whyperformances of traditional estimation algorithms break down when observationsamples are in starvation is analyzed. The required energy of a signal detected fromnoise is derived under the limiting condition. Two new detection statistics are proposedby utilizing sample eignvalues. Combining the sequential hypothesis testing method,these two new statistics are used to estimate the number of primary users. According totheoretical analyses and simulation results, new estimation algorithms perform betterthan traditional algorithms and the improved algorithms proposed in recent years in thecase of small samples.5. The direction of arrival (DOA) estimation algorithm is discussed. The reasonwhy performances of the subspace-based DOA estimation algorithms deteriorate in thecases of low SNR and small samples is analyzed. Then, a parametric iterative adaptivealgorithm based on amplitude and phase estimation (PIAA-APES) with low computational complexity is proposed. The subspace swap phenomenon which is themain reason of the performance breakdown of the subspace-based algorithms is avoidedin the proposed algorithm. The simulation results illustrate that the accuracy andresolution of the orientation detection of the new algorithm are better than those ofMUSIC algorithm. Moreover, this dissertation proposes a new synchronization strategyby utilizing the two-way beacon signal in the distributed array system to generatesynchronous local oscillator signals under their local clocks.更多还原