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

Radar signal deinterleaving is an important content of electronic reconnaissance. In this paper, a new radar signal deinterleaving method based on semantic segmentation thought is proposed. We select representative sequence modeling neural network architectures, and input the difference of time of arrival (DTOA) of pulse stream to them. According to the semantics contained in different categories of radar signals, each pulse in the pulse stream is marked according to the category of semantics contained, and radar signals are deinterleaved. Compared with the traditional einterleaving method, this method can adapt to any pulse repetition interval (PRI) modulation mode and does not require PRI periodicity. Compared with other deinterleaving methods using neural network, this method does not need to digitize the data and train a network for each type of target. This method also eliminates the need to iterate the input and output of data. The proposed method has high robustness under the condition of pulse loss and noise pulses. The research also shows that recurrent neural network (RNN) still has more advantages than convolutional neural network (CNN) in this sequence modeling problem.