Offline real wireless interference signal classification based on denoising diffusion probability model

The reliability and effectiveness of wireless communication data transmission depends largely on the detection and classification of interference signals. In recent years, deep learning (DL) algorithms have been widely used to detect and classify interference signals. Deep learning algorithms require high-quality training samples, however, in wireless communication systems, it is challenging to obtain a large number of high-quality interfered signal samples in real time. To address these challenges, this paper proposes a method for classification of offline real wireless interference signals based on a denoising diffusion probability model (DDPM), which utilizes a denoising diffusion probability model to process the collected signals offline before feature extraction, and then send the signals to a predefined classifier. The simulation results shows that the algorithm proposed in this paper can improve the accuracy of interference signal classification and recognition to 91% in the case of 4 sample signals, effectively solving the problem of low accuracy of interference recognition caused by small sample size and poor data quality in real wireless communication scenarios. In addition, this paper proves the feasibility of using generative models in signal processing and achieves high-precision identification performance in real communication scenarios.