Waveform Reconstruction of DSSS Signal Based on VAE-GAN

Abstract

The complex electromagnetic environment will limit the efficacy of communication equipment. It is critical to construct a complex electromagnetic environment to test communication equipment in order to maximize its capability. One of the most important methods for constructing a complex electromagnetic environment is signal reconstruction. This paper proposes a VAE-GAN-based method for reconstructing direct sequence spread spectrum (DSSS) signals. In this method, the deep residual shrinkage network (DRSN) and self-attention mechanism are added to the encoder and discriminator of VAE-GAN. In feature learning, the DRSNs can reduce the redundant information caused by noise in the collected signal. The self-attention mechanism can establish the long-distance dependence between the input sequences, making it easier for the network to learn the samples’ pseudonoise (PN) sequence features. In addition, feature loss is applied to the encoder and generator to improve network stability during training. The results of the experiments indicate that this method can reconstruct DSSS signals with the characteristics of the target signal.