基于图域数字信号处理的通信信号调制识别方法研究

Automatic modulation classification (AMC) is an indispensable intermediate step between signal detection and demodulation, and plays a key role in various civilian and military applications, such as cognitive radio, spectrum monitoring, adaptive TTC and communication, surveillance and electronic warfare, etc. The development of communication technology and the revolutionary expansion of new applications of AMC, not only result in the implementation of AMC become complex and difficult, but also drastically increase the requirements for real-time and low-complexity AMC solutions. Thus, it is a great challenge to implement AMC for the broadband complex-modulated communication signals. However, the proposed digital signal processing on graphs (DSPG) theory provides a new idea to solve this knotty problem. Motivated by this point, we attempt to transform the broadband complex-modulated communication signals into graph domain, and study the theory and high-efficient algorithms of automatic modulation classification on graphs. Specifically, the objective of this project is to investigate the following problems:.1).the mapping mechanism and representation of the broadband complex-modulated communication signals in graph domain;.2).the feature model of communication signal in graph domain and its efficient extraction method；.3).the automatic modulation classification methods based on the graphical features identification.In this work, the whole theoretical system and implementation of the automatic modulation classification on graphs for communication signals will be developed. The outcome of this project will provide comprehensive technical support for the broadband complex-modulated communication signals cognition, and also lay the foundation for the application of DSPG in the communication signal processing.

通信信号识别作为信号检测与解调之间不可或缺的中间处理，被广泛应用于认知无线电、自适应测控通信、信号侦察和电子战等民用和军用领域。无线通信技术的快速发展以及通信信号识别应用领域的拓展，不仅使通信信号识别变得愈发复杂和困难，而且对信号识别方法的实时性和复杂度要求不断提高，造成复杂宽带通信信号识别面临巨大挑战。图域数字信号处理(DSPG)理论的出现为通信信号识别研究提供了一条新途径。本项目通过将通信信号转换到图域，结合DSPG理论，对复杂宽带通信信号识别理论和方法进行研究。主要研究内容包括：1）复杂宽带通信信号图域映射及表示；2）通信信号图域特征模型及其高效提取方法；3）基于信号图域特征的调制类型识别。从而构建一套完整的基于DSPG的通信信号识别理论体系和实现方法。本项目研究不仅可以为新体制复杂宽带通信信号识别提供全面的技术支撑，而且可以为DSPG理论的通信信号处理应用奠定基础。

自动调制识别作为信号检测与解调之间不可或缺的中间处理，被广泛应用于认知无线电、自适应测控通信、信号侦察和电子战等民用和军用领域。本项目基于图域数字信号处理(DSPG)理论，通过将通信信号转换到图域，对复杂宽带通信信号识别理论和方法进行研究，构建了一套完整的基于DSPG的通信信号调制识别理论和方法。具体成果如下：.1）在不同噪声背景下研究了各种通信调制信号图域映射表示方法。在高斯白噪声背景下，提出了一种基于二阶循环谱的通信信号图域映射及表示方法，并利用广义二阶谱和分数低阶循环谱实现了α稳定分布噪声环境下通信信号的图域映射及表示；针对高斯噪声背景下的M-QAM调制信号类内识别，提出了一种基于星座图统一网格模型的图域映射表示方法。.2）针对不同图域映射方法设计了对应的高效图域特征提取算法。面向基于循环谱分析的图域映射及表示方法，项目组提出了一种基于KL散度(Kullback-Leibler divergence)的通信信号图域特征自动提取方法，包括并行提取和迭代提取两种快速算法，建立起了基于邻接矩阵元素索引的图域特征模型；针对M-QAM信号图域表示，项目组提出了一种基于特征分解的M-QAM信号图域特征自动提取算法。.3）设计了高效分类器并从理论上对算法性能进行了分析。面向基于元素索引的图域特征模型，利用归一化的汉明距离来对信号进行分类；面向基于特征向量的M-QAM信号图域特征，利用最小特征向量夹角来实现M-QAM调制信号的类内识别；从理论上推导了α稳定分布噪声下图域调制识别方法的正确识别率估计公式。最后，搭建复杂宽带通信信号调制识别验证平台对研究成果进行了试验验证。. 本项目研究成果不仅可以为新体制复杂宽带通信信号调制识别提供全面的技术支撑，而且可以为DSPG理论的通信信号处理应用奠定基础。