基于YOLOv6框架实现典型通信信号调制快速识别方法

随着战场通信侦察对抗系统的快速发展，通信信号体制变得非常复杂，给非合作接收条件下的通信信号检测、调制识别及信号辐射源个体识别带来困难。为了全面掌握信号先验信息，对复杂多样的通信信号体制进行盲检与识别，本文提出基于时频图分析和深度神经网络的多种通信信号自动调制识别方法。首先，利用时频分析将不同典型通信信号转换为时频图像，再将标注后的时频图输入基于深度学习的YOLOv6（目标检测模型）网络中进行特征学习；然后，通过设计YOLOv6更高效的网络结构，使其能够对信号的时频图进行快速识别；最后，将训练后的网络权重对典型通信交叠信号进行测试，对提取的特征向量进行分类识别，完成6种调制方式识别与位置的快速确定，实现在非合作接收条件下的多个典型通信信号调制方式的检测和识别。

A fast recognition method of typical communication signal modulation based on YOLOv6 framework

With the rapid development of battlefield communication reconnaissance countermeasure system, communication signal system becomes more complex, which brings difficulties to communication signal detection, modulation recognition and signal emitter individual recognition under non cooperative reception conditions. In order to fully grasp the prior information of the signal, blindly detect and identify the complex communication signal system, this paper proposes an automatic modulation recognition method for multiple communication signals based on time-frequency diagram analysis and deep neural network. Through time-frequency analysis, different typical communication signals are firstly converted into time-frequency images. Then the marked time-frequency diagram is input into YOLOv6 network based on deep learning for feature learning. By designing a more efficient network structure, YOLOv6 can quickly identify the time-frequency diagram of signals. Based on the generated network weight, the typical communication overlapping signal is tested, the extracted feature vectors are classified and recognized, the identification of six modulation modes and the rapid determination of the position is completed. Finally, the detection and recognition of multiple typi-cal communication signal modulation modes under the condition of non-cooperative reception are realized.