基于射频指纹与行为序列的射频机器学习系统研究  

Research on Radio Frequency Machine Learning System Based on Radio Frequency Fingerprint and Behavioral Sequence

作　　者：王宁[1] WANG Ning(China Airborne Missile Academy,Luoyang 471009,China)

机构地区：[1]中国空空导弹研究院,河南洛阳471009

出　　处：《通信电源技术》2023年第11期65-67,共3页Telecom Power Technology

摘　　要：主要提出了一种基于射频指纹特征和行为序列的射频信号机器学习系统,本研究采用理论分析为主,结合建模仿真的方法。首先分析了目前射频信号处理系统面临的问题,然后提出了射频信号机器学习系统,接着详细分析了组成系统的射频指纹特征提取、个体与行为序列识别、信号性质分类3个部分的功能与实现,最后指出需要解决的关键科学问题,包括时频分析对射频指纹特征提取及个体识别的影响问题、神经网络的局部过拟合问题、行为序列中可识别模式组合的向量化问题等。创建了一种既能够对射频辐射源个体识别又能够在重要性、危害性、合作性等行为等级方面进行分类的机器学习系统框架,大大提高了频谱空间中的态势感知能力。The article mainly proposes an Radio Frequency(RF)signal machine learning system based on RF fingerprint features and behavioral sequences,and this study uses theoretical analysis as the main method combined with simulation of reality.The research process first analyzes the problems faced by the current RF signal processing system,then proposes this RF signal machine learning system,followed by a detailed analysis of the functions and implementations of the three parts of the system:radio frequency fingerprint feature extraction,individual and behavioral sequence recognition,and signal property classification,and finally points out the key scientific problems that need to be solved,induding the impact of time-frequency analysis on RF fingerprint feature extraction and individual identification,local overfitting problem of neural networks,vectorization of recognizable pattern combinations in behavior sequences.This study creates a machine learning system framework that is capable of both individual identification and classification of radio frequency radiation sources in terms of importance,hazard,cooperation and other behavioral levels,which greatly improves the situational awareness in the spectrum space.

关 键 词：射频指纹 行为序列 机器学习 射频机器学习 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]