摘要
自动调制样式识别分类是解调前的重要步骤,在频谱管理、认知无线电、智能调制解调器、监视和干扰识别等许多应用中发挥着重要作用。深度学习具有强大的分类能力,基于深度学习中的卷积神经网络,将映射成星座图的具有不同调制样式的通信信号馈送进神经网络,从而达到通信信号调试样式识别分类的目的。基于实验目的,提出一种改进的卷积神经网络结构可实现对七种不同的调制样式的分类,在信噪比≥5dB时,识别率可达97.99%,信噪比≥9dB时,识别率可达100%。
Automatic modulation pattern recognition classification is an important step before demodulation and plays an important role in many applications such as spectrum management,cognitive radio,smart modem,surveillance and interference recognition.Deep learning has powerful classification ability.Based on the convolutional neural network in deep learning,the communication signals with different modulation patterns mapped into constellations are fed into the neural network,so as to achieve the purpose of communication signal debugging pattern recognition and classification.Based on the experimental purpose,an improved convolutional neural network structure can realize the classification of seven different modulation patterns.The signal-to-noise ratio is≥5db,the recognition rate can reach 97.99%,the signal-to-noise ratio is≥9db,the recognition rate can reach 100%.
作者
陈昌美
李艳斌
CHEN Chang-mei;LI Yan-bin(The 54th Research Institute of CETC,Shijiazhuang 050081,China)
出处
《信息技术》
2020年第1期101-106,共6页
Information Technology
关键词
自动调制分类
星座图
深度学习
卷积神经网络
automatic modulation classification
constellation
deep learning
conv-olutional neural network(CNN)
