基于软件无线电与神经网络的频谱监测识别系统

为了解决目前城市频谱监测工作依赖于固定的监测站、持续性监测能力较差和频谱异常判定人工依赖性高等问题,提出了一种利用软件无线电搭配人工智能的新型频谱监测识别系统。首先利用GNURadio软件无线电平台,实现对某一频段的实时监测,获得所需要的频域数据;再利用一系列预处理手段,优化数据样本;最后,在前馈(BP)神经网络中,对频域状态波形进行识别,确定其信号数量、类型及信号所处信道,可以实现持续性频谱监测和智能频谱状态识别判定,其神经网络识别准确率高达96.1%。该系统可以嵌入手持频谱监测设备,并结合云端服务器持续智能地监测区域频谱环境。

Joint Software Radio and Neural Network Signal Monitoring and Identification System

Current urban spectrum monitoring relies on fixed monitoring stations,and there are problems of poor continuous monitoring ability and high dependence of manual identification of spectrum anomaly.A new spectrum monitoring and identification system using software radio and artificial intelligence is proposed.Firstly,real-time monitoring of a certain frequency band is realized by using GNURadio software radio platform to obtain the frequency domain data required by researchers.Then a series of pre-processing methods are used to optimize the data samples.Finally,in the BP neural network,The frequency domain state waveform is identified to determine the number,type and channel of the signal.Continuous spectrum monitoring and intelligent spectrum state recognition can be realized.Its neural network recognition accuracy rate is as high as 96.1%.The system can be embedded in handheld spectrum monitoring equipment,and combined with the cloud server to monitor the regional spectrum environment continuously and intelligently.