摘要
目前的频谱感知算法以模型驱动为主,其感知性能过于依赖预定的统计模型,这使得其在信道环境复杂的卫星通信场景中的部署变得困难。对LEO卫星过境期间的信噪比波动情况进行分析,结果显示信噪比的波动达到14 dB。针对该复杂场景提出了一种基于双向长短期记忆网络及贝叶斯似然比检验联合的频谱感知算法。该算法不需要任何主信号的先验知识,可自动从主信号中学习隐藏特征并做出决策。基于Neyman-Pearson准则,在神经网络输出端设计了一种基于阈值的检测方案,可方便地控制恒定的虚警概率。仿真结果表明,所提算法在信噪比为-14 dB的情况下,仍能达到83%的检测性能,且始终优于卷积神经网络、多层感知机和基于模型驱动的能量检测算法。
With LEO mega satellite constellations coming into operation,the available spectrum resources are more crowded.To improve spectrum utilization,the cognitive satellite communication network composed of GEO relay satellites and LEO satellites has become one of the important candidate technologies to solve the above problem.In this scenario,LEO satellites are permitted to access the authorized spectrum of the GEO satellites through spectrum sensing technology.To avoid interferences from secondary users,spectrum sensing,which is used to quickly determine the presence or absence of primary users,is the most critical step in the scenario of cognitive satellite communication.Since most current spectrum sensing algorithms are model-driven,they rely heavily on the predetermined statistical model for their detection performance,which makes it more difficult to be modeled and deployed in satellite communication scenarios with complex channel environments.In this paper,we first analyze the fluctuation of the Signal-to-Noise Ratio(SNR)at the LEO satellite’s receiving end with the satellite-toground link loss model.The results show that the SNR’s fluctuation reaches 14 dB during satellite transit.Secondly,in this complex channel environment,a spectrum sensing algorithm combining a Bidirectional Long Short-Term Memory(Bi-LSTM)network and a Bayesian likelihood ratio test is proposed.The algorithm can automatically learn hidden features from the primary user signals and make final decisions without requiring any prior knowledge of the primary user signals.Additionally,according to the Neyman-Pearson criterion,we design a threshold-based detection mechanism at the output of the Bi-LSTM network,which can conveniently control the false alarm probability.Finally,the simulation results show that even with an SNR of–14 dB,the proposed algorithm achieves an excellent detection performance of 83%and always outperforms convolutional neural networks,multilayer perceptrons,and model-driven energy detection algorithms.
作者
杨凯
胡圣波
张欣
YANG Kai;HU Shengbo;ZHANG Xin(School of Big Data and Information Engineering,Guizhou University,Guiyang 550025;Intelligent Information Processing Research Institute,Guizhou Normal University,Guiyang 550001)
出处
《空间科学学报》
CAS
CSCD
北大核心
2023年第3期567-575,共9页
Chinese Journal of Space Science
基金
国家自然科学基金项目(61561009)
贵州省教育厅项目(KY[2017]031,KY[2020]007)共同资助。
关键词
认知卫星通信
频谱感知
深度学习
LSTM
贝叶斯似然比检验
Cognitive satellite communications
Spectrum sensing
Deep learning
LSTM
Bayesian likelihood ratio test
