宽带频谱压缩感知算法研究

发布时间：2018-04-05 13:00

本文选题：宽带频谱感知　切入点：压缩感知　出处：《国防科学技术大学》2014年博士论文

【摘要】：频谱感知(Spectrum Sensing)技术旨在寻找授权频带内未被使用的空闲频谱资源,是利用动态频谱接入机制解决可分配频谱资源日益匮乏与授权频带被大量闲置之间矛盾的前提条件。随着无线电技术与业务的迅猛发展,宽带频谱感知技术因其能提供更灵活的接入选择等诸多优点受到业界的广泛关注,已成为频谱感知研究中一个非常重要的发展方向。然而经典的奈奎斯特采样定理要求以两倍于信号最高频率的采样率获取采样数据,因此宽带频谱感知需要极高的采样速率,从而对模数转换器的实现以及采样数据的存储提出了严峻挑战。通过发掘频谱资源利用率低而具备的稀疏性,宽带频谱压缩感知技术将压缩感知理论与宽带频谱感知研究相结合,有效地降低了宽带频谱感知对采样速率的要求,从而解决了过高采样速率带来的一系列问题。目前,宽带频谱压缩感知的研究虽然取得了一定的进展,但仍然存在着许多问题有待解决。论文主要围绕着宽带频谱压缩感知研究中存在的若干问题进行讨论、展开研究,旨在对不同的感知场景设计有效、可靠的宽带频谱压缩感知算法。本文的主要研究成果如下:(1)在单节点宽带频谱压缩感知算法研究方面,针对无线衰落信道以及噪声不确定性等诸多不利因素带来的感知性能不理想等问题,通过发掘实际应用场景中除了宽带频谱稀疏性之外的一些较易获得的先验信息,如部分信道占用状态、信道占用概率以及信道划分信息,分别提出了基于部分占用状态的宽带频谱压缩感知算法、基于占用概率的增强型宽带频谱压缩感知算法以及基于非零子块迭代检测的宽带频谱压缩感知算法,并仿真了所提算法在无噪声和有噪声时的性能。仿真结果表明所提算法均能有效地利用相应的先验信息,显著地提高单节点宽带频谱压缩感知算法的频谱重构和频谱感知性能。此外,仿真实验还讨论了非理想先验信息以及算法参数对所提算法的影响。(2)在多节点合作式宽带频谱压缩感知算法研究方面,针对现有算法大多存在的合作开销过大问题,利用多节点合作式感知场景中单个感知节点接收信号的大动态范围特征及不同节点接收信号的联合稀疏特性,提出了基于支集融合的分布式宽带频谱压缩感知算法,算法采用迭代机制,利用邻近节点之间的局部通信自适应地获得可靠的融合支集信息,并将其作为先验信息参与下一次迭代时的本地频谱重构。仿真实验以现有的分布式算法作为参照,比较了所提算法的精确重构概率、检测概率、计算复杂度和通信负担。仿真结果表明所提算法能够以较低的合作开销,获得较好的频谱重构和频谱感知性能。此外,仿真实验还分析了所提算法对算法参数选取的稳定性。(3)在多节点合作式宽带频谱压缩感知算法研究方面,分析了现有算法由于均涉及频谱重构过程,因而普遍计算量过大且需要一些诸如环境噪声能量或者接收信号稀疏度等难以获得的额外信息,严重限制了其应用范围。鉴于宽带频谱感知通常仅关注于获得整个宽带频谱的占用状态,提出了基于Karcher均值的分布式宽带频谱压缩感知算法,算法中使用联合稀疏信号的Karcher均值作为表征频谱占用状态的统计量,实现了从压缩采样数据到频谱占用状态的直接估计,从而省去了需耗费大量运算资源的频谱重构过程。为节省单个节点有限的通信资源,设计了一种分布式交替乘子法,仅利用邻近节点之间的局部通信,以分布式计算的方式实现了整个感知过程。仿真实验以现有的分布式算法作为参照,比较了所提算法的检测概率、计算复杂度和通信负担,仿真结果表明无需额外信息的所提算法能够以较低的通信资源开销和极低的计算资源开销获得较好的频谱感知性能。此外,仿真实验还测试了算法参数和感知网络参数对所提算法性能的影响。(4)针对复杂电磁环境中宽带非平稳信号增多导致的频域不再具备稀疏性的情况,分析了现有宽带频谱压缩感知算法将由于频域稀疏性不复存在而无法应用。通过将感知域从频域扩展到时频域,发掘感知对象在时频域上的稀疏性,提出了一种基于短时傅里叶变换的时频域信息压缩感知算法,从以远低于奈奎斯特采样率获取的压缩采样数据重构出短时傅里叶变换时频域信息,并以典型的宽带非平稳信号作为感知对象,仿真了所提算法的时频域信息重构性能。仿真结果表明,所提算法能够以较低的采样开销获得性能较好的短时傅里叶变换时频域信息。
[Abstract]:Spectrum sensing technology (Spectrum Sensing) in order to find the authorization of the idle spectrum resources within the band is not in use, is the use of dynamic spectrum access mechanism to solve the distribution of spectrum resources shortage and contradiction is the premise condition of authorized frequency band between a large number of idle. With the rapid development of radio technology and business, the wideband spectrum sensing technology because it can provide access selection more flexible advantages such as attention by the industry, has become a very important research direction of spectrum sensing. However the classical Nyquist sampling theorem to two times the highest frequency of the signal sampling rate to obtain sample data, so the wideband spectrum sensing requires a high sampling rate of ADC and the realization of the sampling data storage challenges. By exploiting the low utilization of spectrum sparsity and ability, wide band The spectrum of compressed sensing technology will be compressed sensing theory and Research on wideband spectrum sensing combination, effectively reducing the sampling rate for wideband spectrum sensing requirements, in order to solve a series of problems brought by the high sampling rate. At present, the research of broadband spectrum compressed sensing has made some progress, but there are still many problems to be solved. This thesis mainly focuses on the broadband spectrum compression problems in the study of perception are discussed, studied, aimed at the effective design of the perception of different scenes, the broadband spectrum reliable compressed sensing algorithm. The main research results are as follows: (1) in the compressed sensing algorithm of single node spectrum sensing performance, the problem for wireless fading channel and noise uncertainty and many other adverse factors is not ideal, by exploring the practical application in the broadband frequency Some of the spectrum is easy to obtain the prior information of sparse outside, as part of the channel occupation, occupation probability and channel division of information channel are proposed for wideband spectrum partial occupancy state of compressed sensing algorithm based on the probability of occupation enhanced broadband spectrum compressed sensing algorithm and broadband spectrum non zero block iterative detection of compressed sensing algorithm based on the simulation, and the performance of the algorithm in the absence of noise and noise. The simulation results show that the proposed algorithm can effectively use the prior information, spectrum reconstruction and spectrum sensing performance of compressed sensing algorithm of single node broadband frequency significantly improved. In addition, the simulation experiment the effect of non ideal prior information and the algorithm parameters of the proposed method is also discussed. (2) in the compressed sensing algorithm of multi node cooperative wideband spectrum, most of the existing algorithms The cooperation overhead problem, using joint sparse characteristics of the large dynamic range of the received signal characteristics of single node and multi node cooperative sensing scenarios and different nodes of the received signal, we propose a distributed broadband spectrum support fusion compressed sensing algorithm based on iterative algorithm using local adaptive mechanism, communication between neighboring nodes to obtain reliable the support of information fusion, and as a priori information in local spectrum reconstruction of the next iteration. Simulation experiments with existing distributed algorithms as a reference, accurate reconstruction probability, the proposed algorithm of detection probability, computational complexity and communication burden. The simulation results show that the proposed algorithm can lower the cooperation overhead, to obtain the spectral reconstruction and spectrum sensing performance. In addition, the simulation experiment is also analyzed the stability of the algorithm of selecting the parameters of the algorithm (3). In the compressed sensing algorithm of multi node cooperative wideband spectrum, analysis of the existing algorithms are involved in the reconstruction process because of the spectrum, so it is generally too large amount of calculation and some environmental noise such as energy or receiving the signal sparsity is difficult to obtain additional information, severely limits the scope of its application. In view of the wideband spectrum sensing usually only focus on the the broadband spectrum occupancy, we propose a distributed broadband spectrum mean Karcher compressed sensing algorithm based on the combination of sparse signal algorithm Karcher mean as characterization of spectrum occupancy statistics, from the implementation of the direct compression sampling data to estimate spectrum occupancy, spectrum reconstruction process which eliminates the need to spend a lot of computing resources in order to save communication resources. A single node is limited, the design of a distributed alternating multiplier method, using only the adjacent Local communication between the nodes in distributed computing mode to realize the whole process of perception. The simulation experiment with the existing distributed algorithms as a reference, the proposed detection probability algorithm, computational complexity and communication burden, the simulation results show that without additional information of the proposed algorithm is able to communication and very low resource cost low computational resource overhead spectrum obtained good performance. In addition, the simulation experiment also tested the effect of algorithm parameters and sensing network parameters on the performance of the proposed algorithm. (4) based on frequency domain broadband non-stationary signal in complicated electromagnetic environment to increase the no longer have the sparsity, the analysis of the existing wideband spectrum compressive sensing algorithm due to the frequency domain sparsity does not exist and cannot be used. The sensing domain from frequency domain extension in time-frequency domain, sparse frequency domain to explore perceived objects when presented A time domain and frequency domain information of short-time Fourier transform algorithm based on compressed sensing, lower than the Nyquist sampling rate gets compressed frequency sampling data to reconstruct the Fourier transform from beyond, with a typical broadband nonstationary signal as the sense objects, the proposed algorithm of time domain information reconstruction performance simulation. The simulation results show that the the proposed algorithm can lower the sampling cost performance of short time Fourier transform good time domain information.

【学位授予单位】：国防科学技术大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN925

【参考文献】