Recognition of Radar Compound Jamming Based on Convolutional Neural Network

The modern electromagnetic environment is becoming more and more complicated, and during detection, radar may face not only single jamming but also compound jamming signals that belong to different varieties, which is more challenging to recognize. Traditional methods are difficult to extract effective features from a variety of jamming signals and their compound signals. Here, a fractional Fourier transform (FRFT)-based multifeature fusion network has been proposed, which combines the multibranch fractional features of the jamming signals and improves the recognition performance. By combining the local and global features of the fractional domain of the jamming signals and adding the attention mechanism, the attention ability of the network to the notable features of images can be further improved. Meanwhile, to make use of the correlation and complementarity between multiple types of information, the time-frequency images of jamming signals are fused based on this network model to realize a more effective and comprehensive expression of features. Simulation results show that, compared with the existing four classical network models, this algorithm has better recognition performance and generalization ability. When the jamming-to-noise ratio is −3 dB, the recognition accuracy of this algorithm can reach more than 99%.