Signal Deinterleaving Based on U-Net Networks

Abstract

Signal deinterleaving is a key part of electronic warfare radar signal processing. As signal environments become more complex, classical sorting methods are being put to the test. This paper draws on the idea of using fully convolutional network to achieve image segmentation, proposes a new signal deinterleaving method based on U-Net network, and introduces the algorithm principle and network training method in detail. Simulation results show that if the PDW sequence of a single target can be fully intercepted, signal sorting based on U-Net segmentation network can be well realized in a complex signal environment as long as the target signal parameters are separable in any dimension.

Topics & Concepts