基于S变换的通信电台个体特征提取与识别新方法研究

The problem of individual recognition for communication station is a very important research subject in communication countermeasure, and the extraction of recognition feature and the design of individual recognizer are two key technologies for it. In one hand, the technology of ordinary Time-Frequency analysis has some defects, such as the limitation of Time-Frequency resolution, the weakness of Time-Frequency focusing, the interference of cross-term, the complexity of kernel function, and so on. In order to overcome those defects, a new technology of Time-Frequency analysis, which is usually called as S-Transform, is introduced in this subject. Through deducing the statistical characteristic of noise and analyzing the features of signal and noise in S-Transform domain, a new model of Adaptive Constant False-Alarm Detection and three new models of Adaptive Time-Frequency Filtering are presented in this subject, those new models will have better detection probability and signal-to-noise ratio in the case of Constant False-Alarm Rate than other Time-Frequency analysis technologies. Based on the research result mentioned above, some new methods of how to extract the modulation feature and fingerprint feature of communication station in S-Transform domain are presented, those new methods will have better testability, stability and reliability than other Time-Frequency analysis technologies. In the other hand, the recognition rate of communication station individual with single recognizer is usually limited in complex situations. Therefore, from the analysis of recognition feature, a new recognizer which is named as interval evidence recognizer is presented in this subject. In the precondition of stabilization and reliability, this new recognizer will have better recognition rate for the individual recognition of communication station. Finally, the research result of this paper can be used to improve the individual recognition rate of communication station in the complex situation, such as the signal-to-noise ratio is very low and changes in a greater range. Meantime, the research result of this paper will also make a positive effect on the development of relevant subject.

通信电台个体识别是通信对抗中非常重要的研究课题，而特征提取和识别器设计是其两个最关键的技术。一方面，针对一般时频特征提取技术存在的时频分辨率受限、时频聚焦性差、交叉项干扰、核函数复杂等缺点，引入一种全新的时频分析技术-S变换，推导噪声在S变换域的统计特征，分析噪声和信号在S变换域的特性，提出一种自适应恒虚警检测模型和三个自适应时频滤波模型，该模型具有更高的检测概率和信噪比。在此基础上，提出从信号S域时频谱中提取通信电台调制特征和指纹特征的方法，该方法提取的特征具有更好的可检测性、稳定性和可靠性；另一方面，针对单一分类器识别效果受限的缺点，从分析识别特征出发，结合灰色关联分析和证据融合的基本思想，提出一种区间证据识别器的设计方法，该识别器在保证稳定和可靠的前提下，具有更高的识别率。本课题的研究成果旨在低信噪比和信噪比大范围变化情况下提高通信电台的个体识别率，并对相关学科的发展产生积极的影响

通信电台个体的细微特征通常会体现在其所发射的通信信号上，被称为通信电台的“指纹特征”，可以用于标识通信电台的属性，该技术在电子战、无线电管理、无线通信电台个体的细微特征通常会体现在其所发射的通信信号上，被称为通信电台的“指纹特征”，可以用于标识通信电台的属性，该技术在电子战、无线电管理、无线网络安全及设备诊断等方面具有重大的应用价值。本课题针对低信噪比下通信电台的个体识别问题开展研究，主要成果如下：第一，引入全新的时频分析技术—S变换方法，提出了S变换的离散化快速算法，解决了计算复杂度问题。第二，推导了高斯白噪声在S变换域的统计特性，提出了基于S变换的自适应恒虚警时频检测模型，该模型相比其它时频变换，在满足虚警概率前提下，在-6dB时即可获得90%的检测概率。第三，提出了S变换时频滤波模型，在信号的S变换域上，通过图像处理的方法，分析信号的噪声分布情况，在此基础上设计时频滤波器组，达到抑制噪声的目的。第四，在S变换域上提取了信号的瞬时频率特征，构建了用于信号调制识别的决策树分类器，在信噪比大于7dB的情况下，平均识别率大于90%。第五，利用S变换提取信号的瞬时频率，进而提取通信电台码元持续时间内的细微特征，通过采集不同FSK电台的实测数据验证表明，通信电台的个体识别率达到96%以上。第六，研究了基于振荡器非线性特征的通信电台指纹特征，提出了基于局部散布差异特征提取的设备非线性分析方法。第七，提取了通信信号的调制特征集合，采用SFS特征选择算法从该集合中选择了最优特征集合。第八，提出了改进的自适应权重灰色关联算法和改进自适应区间的灰色关联算法，仿真结果表明，在信噪比0dB时，算法的识别率仍然高于92%。第九，引入证据融合模型来描述通信电台识别问题，解决了融合架构、BPA获得和冲突证据融合等关键问题，构建新的识别模型，在信噪比0dB时，信号识别率仍接近94%。第十，利用USRP、通信电台搭建了演示验证系统，可以用于算法的验证。综上所述，本课题达到了预期目的，取得了一批理论研究成果，部分成果已经在课题组承担的国防基础重点项目中得到应用，取得了非常好的应用效果。