Data Augmentation for Signal Modulation Classification using Generative Adverse Network

Deep learning has been widely investigated for radio applications. The classification performance of the deep learning greatly depends on the quality of dataset. However, the deficiency of the training data is a critical issue limiting the classification accuracy in practical scenarios. In this paper, we proposed to use the generative adversarial network (GAN) as a data augmentation tool to solve the problem of inadequate training issue under the lack of sufficient data samples. The data augmentation process could be realized by Nash equilibrium of generator and discriminator. The result shows that the accuracy of the classifier is increased by nearly 4 percentage in the signal to noise ratio range of 0 to 20 dB after data augmentation.