基于卷积神经网络学习的盲突发通信信号检测方法

Burst Communication Blind Signals Detection Based on CNN

为了解决工程应用中传统数字信号处理方法对于低信噪比条件下盲突发通信信号检测存在检测正确率低和虚警率高的问题,提出了一种基于卷积神经网络(Convolutional Neural Network, CNN)的突发通信信号检测方法。该方法将传统数字信号处理提取检测特征与CNN深度学习结合,提高了低信噪比条件下突发通信信号检测的正确率,降低了虚警率。对比实验表明,对于QPSK突发通信信号,该方法比多分辨分析数字信号处理方法性能提高大约4 dB,比自适应门限能量数字信号检测处理方法性能提高大约8 dB。

In engineering application,traditional digital signal processing methods for low signal-to-noise-ratio burst communication signals detection encounter a low detection probability and a high false alarm probability.A new kind of burst communication signals detection method is proposed.It combines traditional digital signal processing methods for low signal-to-noise-ratio burst communication signals with the Convolutional Neural Network(CNN),and turns the signal detection to object detection using deep learning network.It increases the detection probability and decreases the false alarm probability for burst communication signals detection.Simulation experiments verified that the performance increases about 4 dB compared with the traditional multi-resolution signal analysis method,and increases about 8 dB compared with the traditional power method.