太赫兹科学与电子信息学报2025，Vol.23Issue(10)：1045-1050,1059,7.DOI:10.11805/TKYDA2024198
基于GCN与LSTM的多频段复杂频谱动态预测
Dynamic prediction of multi-band complex spectrum based on GCN and LSTM
汪生 1张树森 1杨健 1邢伟宁 1许鲁彦1
作者信息
- 1. 军事科学院 系统工程研究院,北京 100191
- 折叠
摘要
Abstract
As the scarcity of spectrum resources becomes increasingly acute,spectrum-prediction-enabled dynamic spectrum access has attracted widespread attention.Owing to the highly bursty nature and intricate latent dependencies of spectrum-sensing data,achieving high-accuracy prediction for multi-channel spectra remains challenging.This paper proposes a multi-band,complex-spectrum dynamic-prediction method that integrates Graph Convolutional Networks(GCN)with Long Short-Term Memory(LSTM):firstly,a GCN extracts features and establishes spatial relationships,capturing and representing the complex structure and inter-dependencies within the spectrum data;secondly,an LSTM effectively learns and forecasts the temporal dynamics of the spectrum,mitigating the inability of conventional predictors to handle long sequences;finally,the approach is evaluated on real-world spectrum measurements.Experimental results demonstrate significant improvements in prediction accuracy and stability over existing methods.关键词
频谱预测/图神经网络(GCN)/长短期记忆网络(LSTM)/深度学习Key words
spectrum prediction/Graph Neural Network(GNN)/Long Short Term Memory(LSTM)/deep learning分类
电子信息工程引用本文复制引用
汪生,张树森,杨健,邢伟宁,许鲁彦..基于GCN与LSTM的多频段复杂频谱动态预测[J].太赫兹科学与电子信息学报,2025,23(10):1045-1050,1059,7.
