A generative adversarial network for jamming waveform generation in the non-cooperative scenario

The generation of interference waveforms is a crucial issue in the field of communication engineering. One of the most challenging problems faced with interference waveform generation is constrained by the lack of annotated data and prior information. This work proposes an effective approach for generating interference waveforms using the generative adversarial network in the scenario without the need for prior information. An additional network as a frequency-domain discriminator is incorporated to enhance the interference effect of the generated waveforms. The experimental results on simulated communication datasets show that the interference effect of the generated waveforms is far superior to the most common noise interference and simultaneously achieves results close to or even superior to matched interference. The interference symbol error rate of the generated waveforms experiences a further improvement after incorporating frequency-domain information for auxiliary training.