伪标签半监督通信辐射源个体识别方法

Research on pseudo label semi-supervised emitter identification technology

摘要:

针对通信辐射源个体识别技术中有标签信号样本不足导致个体识别准确率较低的问题，提出了基于伪标签半监督深度学习的辐射源个体识别方法，该方法利用加权平均思想改进了伪标签的赋值方式，有效增强了伪标签的质量，提升了网络模型的鲁棒性。介绍了如何基于伪标签思想设计半监督深度学习方法，并运用熵正则化算法的概念从理论方面解释了伪标签的有效性。实验设计了适合于信号样本的卷积神经网络，采取不同数目的有标签样本与无标签样本组建的训练集方案，得到了改进的伪标签半监督方法在测试集的识别准确率，结果表明，该方法较全监督方法和改进前的伪标签半监督方法有着更好的识别效果和更强的优越性。

Abstract:

Aiming at the problem of low recognition accuracy caused by insufficient labeled signal samples in communication emitter identification technology, a semi-supervised deep learning method for emitter identification based on pseudo label is proposed. This method uses the idea of weighted average to improve the assignment method of pseudo label, effectively enhances the quality of pseudo label, and improves the robustness of network model. This paper introduces how to design a semi-supervised deep learning method based on the idea of pseudo label, and explains the effectiveness of pseudo label theoretically by using the concept of entropy regularization algorithm. Convolutional neural network is designed for signal samples, Taking different number of labeled samples and unlabeled samples as training set, the recognition accuracy of the improved pseudo label semi-supervised method in the test set is obtained. The results show that the method has better recognition effect and stronger superiority than the full supervised method and the original pseudo label semi-supervised method.