摘要
在现今的通信中,接收信号样本不完全会造成数据的缺失,给数字信号的识别带来困难。因此小样本条件下的数字通信信号调制识别研究具有重大意义。生成式对抗网络(GAN)作为一种拟合生成数据的热门方法备受关注。在原始GAN的基础上将深度卷积对抗网络用于条件生成式对抗网络,实现小样本数据的扩充和识别。实验仿真结果表明,所提出的方法可以有效地生成数据并进行分类识别。此外与相关算法的比较,验证了算法的可行性。
In today’s communication,incomplete received signal samples will result in missing data and bring difficulties to the identification of digital signals.Therefore,the research on the modulation recognition of digital communication signals under the condition of small samples is of great significance.As a popular method of fitting generated data,GAN(generative adversarial network)has attracted much attention.On the basis of the original GAN,the deep convolutional GAN(DCGAN)is used in the conditional GAN(cGAN)to realize the expansion and recognition of small sample data.The simulation results indicate that the proposed method could effectively generate data for classification and recognition.In addition,the comparison with related algorithms verifies the feasibility of the proposed algorithm.
作者
马小博
张邦宁
郭道省
曹林
MA Xiao-bo;ZHANG Bang-ning;GUO Dao-xing;CAO Lin(Army Engineering University of PLA,Nanjing Jiangsu 210007,China)
出处
《通信技术》
2020年第11期2641-2646,共6页
Communications Technology
基金
江苏省自然科学基金(No.BK20191328)。
