HRRP Data Augmentation Using Generative Adversarial Networks

Abstract

In radar automatic target recognition, high resolution range profile (HRRP) can promise satisfactory performance by deep learning when the training samples are affluent. Actually, it is difficult to acquire HRRP samples in battlefield environment. An approach using generative adversarial network (GAN) to augment HRRP data is proposed to deal with the lack of data. There are four GAN models adopted to explore the effect of data augmentation: deep convolutional GAN (DCGAN), Auxiliary Classifier GAN (ACGAN), Least Squares Conditional GAN (LSCGAN) and Wasserstein Conditional GAN (WCGAN). The experimental results show that ACGAN is more suitable than other models in HRRP data augmentation.