自适应多分辨率宽带频谱压缩感知

Compressed sensing is an efficient way to reduce the sampling rate below Nyquist sampling rate while still containing the whole signal's information presuming that the signal is sparse under certain basises. In order to better utilize the radio spectrum, cognitive radio is proposed to dynamically sense the wideband spectrum and agily reuse the licensed spectrum, which will greatly increase the efficiency of the radio spectrum. However, the wideband spectrum consists of many wireless communications with different signal modulations and bandwidth. Moreover, the fading environment may deteriorate the sensing signal and lead to sensing mistakes, that could probably interfere the licensed users greatly. Thus, an efficient wideband spectrum sensing method is necessary to consider the different bandwidh of signals and fading environments.. In this proposal, we try to develop efficient multi-resolution based compressed sensing methods and algorithms for the wideband spectrum containing different bandwidth signals. Different resolution sensing can help the cognitive radio to allocate diffent bandwidth to different cognitive users according to their different QoS requirements. More practically, we consider the radio spectrum is dynamic itself that may caused by transmission environment, licensed user occupying and quiting from the spectrum according to their system scheduler. The sparsity of the wideband spectrum will change unregularly. How to adaptively tracking the spectrum supporting sets is another concern in our proposal. We try to analysis the relationships of the supporting sets under certain compressed sensing recovery algorithms and design efficient adaptive compressed sensing schemes for streaming signals that are commonly encountered in the practice. . To conclude, we try to develop efficient compressed sensing scheme and algorithms with different resolution ability of the spectrum sensing in this proposal.The solution may be used in the spectrum surveiling systems, cognitive radio systems and other similar applications.

宽带频谱压缩感知利用信号的稀疏特征，为低于Nyquist采样率恢复原始信号提供了一种可行的方法，这大大降低了宽带频谱的复杂度。认知无线电提出通过频谱感知和动态再用频谱的概念来提高无线频谱的效率，需要有效感知宽带无线频谱。然而实际宽带频谱通常由多个不同无线系统组成，其信号调制方式、带宽不同，要求频谱感知具有不同分辨率；同时认知用户的不同业务需求也要求频谱感知具有多分辨率。.本项目主要研究具有多分辨率能力的压缩感知架构与算法。考虑到实际应用中，无线信号的传输环境衰落、宽带频谱中授权无线系统用户动态占用与退出，造成感知的无线宽带频谱呈现时变、动态特征，其信号的支撑频率是随时间变化的，分析捕获信号前后稀疏支撑集变化特征，研究针对稀疏度变化的自适应压缩感知方法。研究成果可用于监测雷达系统、无线电监测、认知无线电等系统。

根据项目申请，本项目主要针对宽带无线信号的稀疏特征，从分辨率、压缩测量等方面进行压缩感知理论与算法方面的研究。项目取得的主要成果体现在如下几个方面：.1）.针对宽带无线信号流特征，研究分析了压缩采样前后的信号相关性，建立了压缩采样下的流信号高斯-马尔科夫模型，并基于该模型，提出了基于卡尔曼的压缩感知算法，分别研究了未知稀疏度下、未知信道噪声参数下的全盲卡尔曼压缩感知算法。研究结果表明，采用高斯-马尔科夫信号流模型的卡尔曼压缩感知算法相比其他流信号压缩感知算法具有更高的压缩效率。.2）.针对宽带无线信号流特征，提出一种基于连续观察窗缓存机制的多分辨率下的压缩感知框架，研究了多分辨率下的压缩重构性能，提出了一种高低分辨率重构的方法。针对宽带无线信号的动态特性，研究了自适应采样下的压缩感知算法。.3）.在成果转化方面，开发了分布式电磁监测系统软件，申请了软件著作权1项。搭建了相关仿真平台，并基于FPGA和AD9361，开发了相应的宽带频谱感知系统硬件平台，为验证算法的实际有效性提供了基础。.4）.人才培养方面，完整培养硕士生毕业6名，培养博士生2名。.5） 发表SCI/EI检索论文15篇，其中SCI检索2篇，EI检索12篇。