A Radar Signal Deinterleaving Method Based on Semantic Segmentation with Neural Network

Abstract

Radar signal deinterleaving is an important part of electronic reconnaissance. This study proposes a new radar signal deinterleaving method based on semantic segmentation, which we call "semantic segmentation deinterleaving" (SSD). We select representative sequence modeling neural network (NN) architectures and input the difference of time of arrival of the pulse stream into them. According to semantics contained in different radar signal types, each pulse in the pulse stream is marked according to the category of semantics contained, and radar signals are deinterleaved. Compared to the traditional deinterleaving method, the SSD method can adapt to complex pulse repetition interval (PRI) modulation environments without searching the PRI or PRI period. Multiple rounds of search and merging operation are not required for radar signals with multiple pulses in a period. Compared to other deinterleaving methods based on NNs, the SSD method does not need to digitize the data and train a network for each target type. The SSD method also does not need to iterate input and output data. The proposed method has high robustness to pulse loss and noise pulses. This research also shows that recurrent NNs still have more advantages than convolutional NNs in this sequence modeling task.

arXiv comment10 pages, 18 figures

References in corpus (8)

• Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling
• Rethinking Atrous Convolution for Semantic Image Segmentation
• Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation
• An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling
• Semantic Image Segmentation with Deep Convolutional Nets and Fully Connected CRFs
• Recurrent Neural Network Regularization
• Attention-Based Models for Speech Recognition
• Grid Long Short-Term Memory