| 摘要: |
| 针对短波频段内海量电台的快速自动识别问题,本文提出一种基于稀疏分量分析的快速识别方法,该方法基于各电台发射信号时域所具有的独特周期性及稀疏性,基于高速频谱扫描数据对各信道的多个电台进行自动分离和识别。首先,针对单信道多电台混合问题,将信道的功率时间序列建模为包含多个电台成分的混合信号,并基于混合信号所具有稀疏的性质,使用稀疏向量分析方法(Sparse Component Analysis,SCA)对混合信号进行自动分离,实现对同信道上不同电台的识别。在此基础上,针对短波时变信道衰落下的电台信号进行分离问题,提出一种基于时间特征聚类的稀疏分量分析的算法,该算法将时间特征与幅值特征相融合进行聚类,实现对混合矩阵的估计。最后,针对混合噪声对分类结果的影响,在聚类结果基础上,将信号向类心向量进行投影,去除时变信道衰落引入的噪声。在仿真实验阶段,发射源设置不同播放时间,不同占空比和不同周期,在8个地点布置信号采集系统,使用短波预测软件(Voice of Amercian Coverage Analysis Program,VOACAP)对接收功率进行仿真,该算法的识别正确率为98.1%,相比聚类稀疏分量分析和快速独立成分分析分别提升了7.3%和16.8%,能够很好解决短波电台分离和识别的问题。 |
| 关键词: 短波电台 信号识别 信号分离 稀疏分量分析 无线电监测 |
| DOI：10.19363/J.cnki.cn10-1380/tn.2023.08.13 |
| 投稿时间：2021-02-04修订日期：2021-03-25 |
| 基金项目:本课题得到中国科学院青年创新促进会(No. Y9YY015104)项目资助。 |
|
| A Fast Automatic Identification Method of Massive Shortwave Radio Stations Based on Sparse Component Analysis |
| HUANG Weiqing,WANG Yuankun,ZHANG Qiaoyu,LI Jing,WEI Dong |
| Institute of Information Engineering, Chinese Academy of Sciences, Beijing 100093, China;School of Cyber Security, University of Chinese Academy of Sciences Beijing 100049, China |
| Abstract: |
| Aiming at the problem of rapid automatic identification of a large number of radio stations in the high frequency band, a fast identification method based on sparse component analysis is proposed. Based on the unique periodicity and sparseness of each station’s transmitted signal in the time domain, the high-speed spectrum scanning data are used to separate and identify multiple radio stations on each channel automatically. First, to solve the single-channel multi-station mixing problem, the power time series of the channel is modeled as a mixed signal containing multiple radio components, and based on the sparse nature of the mixed signal, Sparse Component Analysis (SCA) is used to automatically separate the mixed signal to achieve the same Identification of different stations on a channel. Then, in order to separate the radio signals under shortwave time-varying channel fading, a sparse component analysis algorithm based on time feature clustering is proposed, in which clustering is performed with both the time features and the amplitude features to realize the estimation of the mixing matrix. In addition, in order to solve the problem of mixed noise on the classification results, based on the clustering results, the algorithm projects the signal to the centroid vector to remove the noise introduced by time-varying channel fading.. In simulation experiments, the transmitting source is set with different playing times, different duty cycles and different periods, the signal receive system is arranged in 8 locations, and the received power is simulated by using the shortwave prediction software Voice of American Coverage Analysis Program (VOACAP) our algorithm’s accuracy rate of station identification is 98.1%, which is 7.3% and 16.8% higher than clustering based sparse component analysis and fast independent component analysis, respectively, providing a good solution to the problem of separation and identification of shortwave radio stations. |
| Key words: shortwave radio stations signal identification signal separation sparse component analysis radio monitoring |
