Jamming Recognition Based on AC-VAEGAN
Yan Tang, Zhijin Zhao, Xueyi Ye, Shilian Zheng, Lijun Wang
Abstract
To solve the performance deterioration of jamming recognition method based on deep learning in the case of small sample set, a jamming recognition method based on AC-VAEGAN is presented. Interference signal time-frequency diagrams are used as the small sample training set. AC-VAEGAN network is designed by modifying the structure of ACGAN network, that is, adding the core idea of VAE in the ACGAN network, so that the latent space of a continuous and meaningful sample set can be obtained through the encoder. By combining of regularization loss, reconstruction loss, source loss and classification loss, loss functions of three parts of AC-VAEGAN are obtained, and the three parts are trained to make the AC-VAEGAN network more efficient. The simulation results show that when the jamming-to-noise ratio is -10dB~10dB, the AC-VAEGAN network has a higher correct recognition rate for the five kinds of jamming than ACGAN and CNN networks in the case of small sample datasets.