基于非参数统计的盲频谱感知算法研究

发布时间：2018-08-23 19:52

【摘要】：认知无线电技术实现了频谱资源的动态分配,提高了频谱资源的利用率,而准确高效的频谱感知是认知无线电的核心环节。快速高性能的频谱感知算法研究已经成为了一个亟待解决的问题。传统的频谱感知算法通常需要授权用户信号以及噪声的分布信息,这些信息往往难以获得。非参数方法不涉及有关分布的参数,因而基于此可以设计盲频谱感知算法。本文对认知无线电中频谱感知算法进行了深入的研究,将频谱感知问题转化为非参数检验问题,主要工作和创新成果如下:1.研究了抗噪声不确定性的拟合优度类感知算法。基于拟合优度检验的频谱感知算法检测性能较好但易受到噪声不确定度的影响。利用仅对均值偏移敏感,对方差偏离不敏感的Cramer-vonMises(CM)统计量第一分量,设置新的检验统计量,克服了估计噪声方差过程中产生的误差对传统拟合优度算法性能的影响,并推导了频谱空闲时检验统计量的概率密度函数和判决门限,从而提出了利用CM分量的频谱感知算法。理论分析和仿真表明,所提算法在减小拟合优度检验中的CM算法复杂度的同时,大大减小了噪声不确定度对CM算法性能的影响。2.研究了多天线协作频谱感知场景下的快速频谱感知算法。在高斯信道场景下,协方差矩阵的非对角元素在主用户信号不存在时的概率分布呈现偶对称,而主用户信号存在时这一结论不再成立。利用二项检验和符号秩检验来衡量分布的对称性,设计了两种基于协方差矩阵的无噪声不确定度盲感知算法。理论分析和仿真表明,基于符号秩检验的感知算法性能优于基于二项检验的感知算法,但是复杂度更高。实际使用中,可以根据时效性和性能的需求选择合适的算法。3.研究了瑞利衰落信道下适用于低天线相关性场景的频谱感知算法。传统的特征值类算法及协方差类算法利用天线相关性可以实现无噪声不确定度的高性能频谱感知。但是该类算法假设天线之间信号具有相关性而噪声相互独立,在天线分集增益场景下,天线之间相关性变低时,这些算法性能急剧下降甚至失效。针对这一问题,提出一种基于卡方拟合优度检验的多天线协作频谱感知算法,从而将频谱感知问题转化为一个多项分布检验的问题,然后利用卡方拟合优度检验判决频谱是否空闲,从而实现频谱感知。理论分析和仿真表明所提算法可以在低天线相关性场景下工作并且性能不受噪声不确定度的影响。总之,所提算法和方案有效地解决了认知无线电系统中频谱感知技术存在的一些问题,并通过仿真验证了其有效性,所研究和提出的算法具有一定的理论和实际意义。
[Abstract]:Cognitive radio technology realizes the dynamic allocation of spectrum resources and improves the efficiency of spectrum resources. Accurate and efficient spectrum sensing is the core link of cognitive radio. Fast and high performance spectrum sensing algorithm research has become a problem to be solved. Traditional spectrum sensing algorithms usually require authorized user signals and noise distribution information, which are often difficult to obtain. The nonparametric method does not involve the parameters of the distribution, so blind spectrum sensing algorithm can be designed based on this. In this paper, the spectrum sensing algorithm in cognitive radio is deeply studied, and the spectrum sensing problem is transformed into a non-parametric test problem. The main work and innovative results are as follows: 1. A preference-class perception algorithm for anti-noise uncertainty is studied. The spectrum sensing algorithm based on goodness of fit test has better detection performance but is easily affected by noise uncertainty. Using the first component of the Cramer-vonMises (CM) statistic which is only sensitive to mean deviation and deviation from the other side, a new test statistic is set up to overcome the influence of the errors in the process of estimating the noise variance on the performance of the traditional goodness of fit algorithm. The probability density function and the decision threshold of the spectrum idle test statistic are derived, and a spectrum sensing algorithm using CM component is proposed. Theoretical analysis and simulation show that the proposed algorithm not only reduces the complexity of CM algorithm in the goodness of fit test, but also greatly reduces the influence of noise uncertainty on the performance of CM algorithm. A fast spectrum sensing algorithm for multi-antenna cooperative spectrum sensing scene is studied. In Gao Si channel scenario, the probability distribution of non-diagonal elements of the covariance matrix presents even symmetry when the primary user signal does not exist, but this conclusion is no longer true when the primary user signal exists. By using binomial test and symbolic rank test to measure the symmetry of distribution, two noiseless blind perception algorithms based on covariance matrix are designed. Theoretical analysis and simulation show that the performance of the perceptual algorithm based on symbolic rank test is better than that based on binomial test, but the complexity is higher. In practical use, the appropriate algorithm. 3. 3 can be selected according to the requirements of timeliness and performance. A spectrum sensing algorithm for low antenna correlation scenes in Rayleigh fading channels is studied. The traditional eigenvalue class algorithm and covariance class algorithm can realize high performance spectrum sensing without noise uncertainty by using antenna correlation. However, these algorithms assume that the signals are correlated and the noise is independent of each other. In the antenna diversity gain scenario, when the correlation between antennas becomes low, the performance of these algorithms drops sharply or even fails. To solve this problem, a multi-antenna cooperative spectrum sensing algorithm based on chi-square goodness of fit test is proposed, which can transform the spectrum sensing problem into a multi-item distribution test. Then, the chi-square goodness of fit is used to test whether the decision spectrum is idle or not, so as to realize spectrum sensing. Theoretical analysis and simulation show that the proposed algorithm can work in low antenna correlation scenarios and its performance is not affected by noise uncertainty. In a word, the proposed algorithm and scheme effectively solve some problems of spectrum sensing technology in cognitive radio system, and verify its effectiveness by simulation. The algorithm studied and proposed has certain theoretical and practical significance.
【学位授予单位】：西安邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN925

【相似文献】