针对舰船平台非合作信号的认知抗干扰技术研究

Cognitive anti-interference of non-cooperative signals is one of the key techniques for EMC between multiple equipment on a vessel platform. Due to the closely spaced antennas, the complex electromagnetic environment and the small characteristic differences between different types of interferences, traditional interference detection, classification and recognition methods are inaccurate. In addition, the traditional design methods of anti-interference waveforms have high computational complexity thus cannot meet real-time requirements. This project is founded on a number of subjects including cognition of non-cooperative interference signals, detector design based on adaptive background noise, extraction of interference features based on high order statistics, classification and recognition of interference based on deep neural network. Then, based on the interference cognition, real-time anti-interference decision is implemented through multi-objective function optimization. Finally, the design of anti-interference waveform is accomplished by spectrum fitting. The cognition-decision-behavior technology architecture is formed, which is mainly composed of cognition, real-time decision and anti-interference waveform optimization technologies. The successful implementation of this project will offer significant assistance for collaborative work of communication, detection and electronic warfare equipment of vessel platforms, and it is of great value in practical applications. Moreover, this research will enrich the knowledge and techniques of interference cognition and waveform design, and it is also valuable from the theoretical and academic perspective.

针对非合作信号的认知抗干扰是实现舰船平台多设备电磁兼容需要亟待解决的关键技术问题之一。由于舰船平台天线密集、电磁环境复杂以及干扰信号特征参数差异小，导致传统的干扰检测、分类识别方法结果出现偏差，同时传统的抗干扰波形设计方法也存在计算复杂度较高的缺陷，不能有效满足实时性需求。本项目首先从对非合作信号的干扰认知出发，设计自适应背景噪声的干扰检测器，利用高阶统计方法提取干扰特征，并基于深度神经网络实现干扰的分类识别，然后在干扰认知的基础上，通过多目标函数优化完成实时抗干扰决策过程，最后完成频谱拟合的抗干扰波形设计。形成由干扰认知技术、实时抗干扰决策技术、抗干扰波形优化设计技术为主要内容的“认知-决策-行为”技术体系结构。本项目的成功实施将对实现舰船平台内通信、探测及电子战等设备的协同工作提供帮助，具有很大的应用价值，相关研究也将丰富干扰认知、波形设计等领域的研究内容，具有重要的学术价值。

非合作信号干扰问题是复杂舰船电磁环境亟需解决的突出问题，认知抗干扰技术具有实时自适应识别和抑制非合作信号的能力，是实现舰船平台多用频设备智能抗干扰的有效技术途径。本项目从电磁环境分层认知的角度出发，围绕舰船平台干扰建模与检测、干扰特征提取与识别、干扰抑制合并策略、抗干扰波形设计等四个方面展开研究工作，构建了由干扰认知技术、实时抗干扰决策技术、抗干扰波形优化技术为主体的“认知-决策-行为”闭环技术体系结构。取得的主要研究成果有：突破了大动态低信噪比条件下的干扰检测技术，提出了融合自编码器和深度支持向量描述的干扰检测方法，典型干扰场景下该方法的ROC曲线面积接近于1；突破了多径衰落等非理想信道下的调制识别技术，提出了基于卷积神经网络和稀疏滤波的调制识别方法，针对开源数据集中11类数字调制的平均识别率超过90%，性能优于经典方法；攻克了分集合并抗干扰技术需精确估计信道参数的难题，设计了功率失效区最小准则的输出选择策略，提出了无需信道估计的干扰抑制合并及实时决策方法；突破了多约束多域联合波形优化技术，提出了基于时频分析的抗干扰波形快速迭代方法，定义了合成自相关函数，解决了稀疏频谱与特定区间低旁瓣联合约束的在线优化难题；研制成功了多通道智能感知原理样机，实现了针对舰船通信频段电磁干扰的实时检测、识别与评估。依托本项目已发表学术论文20篇（SCI检索8篇，EI检索11篇）。