| 摘要: |
| 针对非合作通信场景下的跳频信号自动化检测识别问题，本文提出了一种基于方向梯度直方图与支持向量机的跳频信号检测识别算法。该算法将无线通信信号转化为包含时间、频率和幅度的时频瀑布图，采用方向梯度直方图特征提取算法将不同跳频序列在瀑布图上产生的独特结构特征提取出来。然后利用支持向量机将特征序列映射到高维空间，通过寻找最大间隔分离超平面，实现跳频信号的检测与多种跳频序列的识别，并依此建立跳频信号检测识别原型系统。最后在室内多径信道环境下进行了测试验证，该算法能够完全自动化的精确检测到开放电磁环境下的跳频信号并且能够实现对多种跳频序列的识别。在信干噪比不超过20dB时，针对不同跳频序列的平均识别正确率能够达到98.01%。 |
| 关键词: 跳频信号 检测识别 方向梯度直方图 支持向量机 |
| DOI：10.19363/J.cnki.cn10-1380/tn.2020.05.06 |
| Received:December 28, 2018Revised:April 01, 2019 |
| 基金项目:本课题得到中国科学院战略性先导C类（No.XDC02000000）专项资助。 |
|
| Detection and Recognition Algorithm for Frequency Hopping Signals Based on HOG-SVM |
| ZHANG Meng,WANG Wen,REN Junxing,WEI Dong,HUANG Weiqing,YANG Zhaoyang,LV Zhiqiang |
| 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: |
| In this paper, an algorithm based on Histogram of Oriented Gradient (HOG) and Support Vector Machine (SVM) is proposed to detect and recognize the frequency hopping signals in non-cooperative communication environment. The algorithm transforms wireless communication signals into time-frequency waterfall graph which include time, frequency and amplitude information. The algorithm of HOG is used to extract the special structure features on the waterfall graphs of different frequency-hopping sequences. Then the feature sequences are mapped to high dimensional space based on SVM to find the maximum margin hyperplane, thus the detection of frequency hopping signal and the recognition of different frequency-hopping sequences are realized, and the prototype system of frequency hopping signal detection and recognition is established. In the indoor multipath channel environment, the algorithm can automatically detect the frequency hopping signal accurately in open electromagnetic environment and can recognize many kinds of frequency-hopping sequences. Under the circumstance that the signal interference noise ratio is not above 20dB, the accuracy rate of recognizing different frequency hopping sequences can reach 98.01%. |
| Key words: frequency hopping signal detection and recognition histogram of oriented gradient support vector machine |
