一种基于多层人工神经网络的感知频谱预测策略  

A Cognitive Radio Spectrum Prediction Scheme Based on Multilayer Neural Network

作　　者：庞博[1,2] 谢卓辰[1,2] 马陆[2] 刘会杰[2] 

机构地区：[1]中国科学院上海微系统与信息技术研究所,上海200050 [2]上海微小卫星工程中心,上海201203

出　　处：《遥测遥控》2014年第1期1-5,共5页Journal of Telemetry,Tracking and Command

摘　　要：认知无线电可以使低轨卫星系统在运行周期中利用频谱中的空闲信道完成通信。根据感知得到的频谱信息对信道的未来状态进行预测,可以维持通信过程的稳定性和效率。结合人工神经网络的理论,提出一种基于多层BP神经网络的频谱预测策略,根据信道的历史状态对其下一时刻的状态进行预测。利用星上实测数据进行仿真验证,结果表明此预测策略具有较好的预测性能。The LEO satellite communication system can use idle channels in spectrum with the cognitive radio. By predicting future states of channels according to the sensed spectrum information, the stability and efficiency of communication process are guaranteed. Based on the theory of multilayer BP neural network, a spectrum prediction scheme is proposed to predict the channels＇ states by history states. The measurement data on satellite is used for simulation and verification. The result shows the good performance of the proposed scheme.

关 键 词：干扰预测 频谱感知 低轨卫星 多层神经网络 

分 类 号：TN925[电子电信—通信与信息系统]