基于PACGAN与差分星座轨迹图的辐射源个体识别

深度学习解决个体识别的一个突出问题是难以获得足够样本对网络进行训练，针对该问题，提出了一种基于PACGAN(Pooling Auxiliary Classifier Generative Adversarial Network)的辐射源个体识别算法。该算法针对输入信号的差分星座轨迹图进行处理，并对辅助分类生成式对抗网(ACGAN)进行了适应性改进。在判别器网络中引入池化操作，增强其在多分类任务中的特征提取能力；针对样本图像特征大量边缘分布的情况，添加零填充层并以增强其边缘特征提取能力，增大卷积层感受野以提取全局性特征。通过对五种ZigBee设备的实验，结果表明本文提出算法在小样本条件下相较于其他方法具有更高的准确性。 

Individual Identification Method based on PACGAN and Differential Constellation Trace Figure

One of the outstanding problems of applying deep learning to solve Individual Identification is that it is difficult to collect enough samples to train the network. In order to solve this problem, an individual identification algorithm based on PACGAN is proposed. The algorithm processes the Differential Constellation Trace Figure of input signals, and improves the adaptability of ACGAN. This paper improves the adaptability of ACGAN, introduces pooling layer in the discriminator network to enhance its feature extraction ability in multi classification task; for the situation of a large number of edge distribution of sample image features, adds zero filling layer and increases convolution kernel receptive field to enhance its edge feature extraction ability. The results of five kinds of ZigBee devices show that the proposed algorithm has higher accuracy than other methods in the case of small sample set.