首页|LDACS系统基于循环谱和残差神经网络的频谱感知方法

LDACS系统基于循环谱和残差神经网络的频谱感知方法

扫码查看
原文链接
  • 万方数据
  • 维普
针对L波段数字航空通信系统(L-band digital aeronautic communication system，LDACS)可用频谱资源有限且易受大功率测距仪(distance measuring equipment，DME)信号干扰的问题，提出一种基于降维循环谱和残差神经网络的频谱感知方法。首先理论推导分析了 DME信号的循环谱特征;然后利用Fisher判别率(Fisher discriminant rate，FDR)提取循环频率能量最大的向量，通过主成分分析(principal component analysis，PCA)进行预处理特征增强;最后给出数据处理后的循环谱向量与卷积神经网络相结合的实现过程，实现了 DME信号的有效检测。仿真结果表明，该方法对噪声不敏感，当信噪比不低于-15 dB时，平均检测概率大于90％。当信噪比不低于-14 dB，检测概率接近100％。
Spectrum sensing method based on cyclic spectrum and residual neural network in LDACS system
To solve the problem that the available spectrum resources of L-band digital aeronautic communication system(LDACS)are limited and vulnerable to interference from high-power distance measuring equipment(DME)signals,a spectrum sensing method based on reduced dimension cyclic spectrum and residual neural network is proposed.Firstly,the cyclic spectrum characteristics of DME signal are analyzed theoretically.Then Fisher discriminant rate(FDR)is used to extract the vector with the highest cycle frequency energy,and the pre-processing features are enhanced by principal component analysis(PCA).Finally,the process of combining the cyclic spectral vector and convolutional neural network after data processing is given,and the effective detection of DME signal is achieved.Simulation results show that the method is not sensitive to noise,and the average detection probability is greater than 90％when the signal-to-noise ratio is no less than-15 dB.When the signal-to-noise ratio is not less than-14 dB,the detection probability is close to 100％.

L-band digital aeronautic communication system(LDACS)distance measuring equipment(DME)spectrum sensingcyclic spectrumresidual neural network

王磊、张劲、叶秋炫

展开 >

中国民航大学民航航班广域监视与安全管控技术重点实验室,天津 300300

L波段数字航空通信系统 测距仪 频谱感知 循环谱 残差神经网络

国家自然科学基金天津市多元投入基金中国民航大学民航航班广域监视与安全管控技术重点实验室开放基金

U223321621JCQNJC00770202101

2024

10.12305/j.issn.1001-506X.2024.09.34

系统工程与电子技术
中国航天科工防御技术研究院 中国宇航学会 中国系统工程学会

系统工程与电子技术

CSTPCD北大核心
影响因子：0.847
ISSN：1001-506X
年,卷(期)：2024.46(9)