摘要
由于超宽带信号频谱不规则,且信号强度低,容易受到其它信号干扰,对超宽带频谱信号的识别带来了严重的干扰。为了精准识别超宽带频谱信号,提出一种基于深度卷积网络的超宽带频谱信号识别方法。通过形态学滤波对原始超宽带频谱信号滤波处理,获取噪声频谱;采用经典阈值识别干扰频率阈值,删除噪声频谱,重构超宽带频谱信号。对超宽带频谱数据展开多帧叠加,提取其中的弱信号特征,将叠加处理的频谱图像输入到深度卷积网络中展开超宽带频谱信号识别。实验结果表明,所提方法的归一化均方差小,且频谱信号识别率在90%以上。
Currently,the frequency spectrum of ultra-wideband signals is irregular and the signal strength is low,thus making it easy to be interfered with by other signals,which brings serious interference to the recognition of ultra-wideband spectrum signals.In order to accurately recognize the ultra-wideband spectrum signal,this paper put forward a method of identifying ultra-wideband spectrum signals based on deep convolutional neural network.Firstly,we filtered the original ultra-wideband spectrum signal through morphological filtering,thus obtaining the noise spectrum.Then,we adopted the classical threshold to identify the interference frequency threshold and delete the noise spectrum,thus reconstructing the ultra-wideband spectrum signal.Next,we superimposed multiple frames of ultra-wideband spectrum data and extracted the weak signal features.Finally,we inputted the superimposed image into a deep convolutional neural network for signal recognition.Experimental results prove that the proposed method has a small normalized mean square error.And the recognition rate of frequency spectrum signal is more than 90%.
作者
陆海锋
梁卓明
LU Hai-feng;LIANG Zhuo-ming(Zhaoqing University,Zhaoqing Guangdong 526061,China;South China Normal University,Guangzhou Guangdong 520631,China)
出处
《计算机仿真》
2024年第11期220-224,共5页
Computer Simulation
基金
中央财政支持地方高校改革发展资金项目(2023-156028)。
关键词
深度卷积网络
超宽带
频谱信号
识别
Deep convolutional network
Ultra-wideband UWB
Spectrum signal
Distinguish
