基于稀疏分解的频谱感知方法研究

发布时间：2018-01-08 08:03

  本文关键词：基于稀疏分解的频谱感知方法研究　出处：《哈尔滨工业大学》2015年硕士论文　论文类型：学位论文

  更多相关文章： 稀疏分解 频谱感知 分布式压缩感知 过完备字典 字典训练

【摘要】：近年来,随着各种无线电新技术和新业务的广泛应用以及通信技术的迅速发展,对频谱资源的需求程度和数量日益增长,如何在频谱资源有限的情况下对其更为有效的利用成为一个急需解决的问题。认知无线电技术的提出为我们解决这个问题提供了方法。频谱感知技术是认知无线电的基础环节,它的好坏关系着系统性能的优劣。稀疏分解可以抓住信号内部的主要特征,提取信号成分,具有去噪的功能;近年来兴起的压缩感知技术也是以信号能够在变换域上稀疏分解为前提的,它能够突破奈奎斯特采样定理的限制,仅通过少量的压缩测量值便能够实现信号的重构。本文对基于稀疏分解的频谱感知方法进行了研究。在频谱能量检测方式中,门限的设置与接收信号的信噪比直接相关,因此,接收信号的信噪比是影响频谱感知性能的一个决定性因素。在一定范围内,随着信噪比的降低,恒虚警条件下的检测概率快速下降。在本文中,考虑到稀疏分解有一定程度的去噪作用,因此将稀疏分解引入到接收机前端,在接收信号后,首先进行去噪处理,再进入后端的频谱感知部分。在经过稀疏分解去噪的过程后,随着信噪比的改善,在目前的信噪比下设定新的门限,检测性能必然有很大程度的提高。在认知MIMO系统中,多天线提供了更高的检测可靠性,但也带来了采样数据量的急剧提升,而分布式压缩感知技术正是基于多信号的联合采样技术,降低采样的数据量。它要求在多信号的基础上满足联合稀疏模型,而MIMO技术由于天线之间的相关性,刚好满足JSM-2模型,因此基于认知MIMO的分布式压缩感知技术迫切需要一个针对多天线环境的联合稀疏字典。字典训练算法作为一种较为新颖的字典获取方法,只给出了单信源训练信号的情况下如何获得训练字典。本文结合字典训练算法与多天线下的联合稀疏模型,将普通的字典训练算法拓展为三种不同合并方式下的多天线联合训练字典算法。相比于一般的字典训练算法,联合字典训练算法能够在同样的训练次数的情况下,获得更好的稀疏表示效果。因此,在分布式压缩感知重构之后,重构概率显著提高,之后的频谱检测性能也有了进一步的提升。基于联合训练字典的频谱感知技术在提高系统检测可靠性的基础上,也降低了采样数据量。
[Abstract]:In recent years, with the wide application of various new radio technologies and new services and the rapid development of communication technology, the demand for spectrum resources is increasing day by day. How to make more effective use of spectrum resources under the condition of limited spectrum resources is an urgent problem to be solved. The development of cognitive radio technology provides a method for us to solve this problem. Spectrum sensing technology is cognitive wireless. The foundation of electricity. The sparse decomposition can grasp the main characteristics of the signal, extract the signal components, and have the function of denoising. The compression sensing technology developed in recent years is also based on the sparse decomposition of signals in the transform domain, which can break through the limitation of Nyquist sampling theorem. Only a small amount of compressed measurements can be used to reconstruct the signal. In this paper, the spectral sensing method based on sparse decomposition is studied. The threshold setting is directly related to the signal-to-noise ratio of the received signal. Therefore, the SNR of the received signal is a decisive factor affecting the spectrum sensing performance. In a certain range, the SNR decreases with the decrease of SNR. In this paper, considering the sparse decomposition has a certain degree of de-noising, the sparse decomposition is introduced to the front end of the receiver, after receiving the signal. After the sparse decomposition and denoising process, with the improvement of SNR, a new threshold is set under the current SNR. In cognitive MIMO systems, multiple antennas provide higher detection reliability, but also bring a sharp increase in the sample data. Distributed compression sensing technology is based on multi-signal joint sampling technology to reduce the amount of data sampled, it requires that the multi-signal on the basis of the United sparse model. Because of the correlation between antennas, MIMO technology just meets the JSM-2 model. Therefore, distributed compression sensing technology based on cognitive MIMO urgently needs a joint sparse dictionary for multi-antenna environment. Dictionary training algorithm is a novel dictionary acquisition method. Only how to obtain the training dictionary under the condition of single source training signal is given. This paper combines the dictionary training algorithm with the joint sparse model under multiple antennas. The common dictionary training algorithm is extended to the multi-antenna joint training dictionary algorithm under three different combinations, compared with the general dictionary training algorithm. The joint dictionary training algorithm can obtain better sparse representation under the same training times. Therefore, after distributed compression awareness reconstruction, the reconstruction probability is improved significantly. The spectrum sensing technology based on the joint training dictionary can improve the reliability of the system and reduce the sample data.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN925

【参考文献】

中国期刊全文数据库 前3条

1 蔡泽民;赖剑煌;;一种基于超完备字典学习的图像去噪方法[J];电子学报;2009年02期

2 杨俊;谢勤岚;;基于DCT过完备字典和MOD算法的图像去噪方法[J];计算机与数字工程;2012年05期

3 胡海峰;杨震;;无线传感器网络中基于空间相关性的分布式压缩感知[J];南京邮电大学学报(自然科学版);2009年06期

中国博士学位论文全文数据库 前1条

1 王春光;基于稀疏分解的心电信号特征波检测及心电数据压缩[D];国防科学技术大学;2010年

中国硕士学位论文全文数据库 前2条

1 邵君;基于MP的信号稀疏分解算法研究[D];西南交通大学;2006年

2 徐勇俊;基于信号稀疏表示的字典设计[D];南京理工大学;2013年

，

本文编号：1396262