Design of radar pulse deinterleaving scheme using deep learning

• Main Author: Zhang, Chenxin
• Other Authors: Teh Kah Chan
• Format: Thesis-Master by Coursework
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Engineering
• Online Access: https://hdl.handle.net/10356/173767
• Institution: Nanyang Technological University

As an important part of modern electronic countermeasure technology, radar reconnaissance is the basis of electronic jamming and electronic defence. Radar signal deinterleaving, as a key step of radar reconnaissance, is the premise and guarantee of information processing and analysis of reconnaissance system. With the progress of technology, the electromagnetic environment in which the reconnaissance system is located is becoming more and more complex, how to effectively and efficiently sort radar signals in the interference and noise has become a matter of great concern. This dissertation investigates the traditional signal deinterleaving method and deep learning algorithm, and draws on the processing methods of image and speech in deep learning, which reduces the number of parameters to be set in the traditional signal deinterleaving, improves the precision of signal deinterleaving, and makes the process of signal deinterleaving more automated and intelligent. In this dissertation, the Faster RCNN network is applied to signal deinterleaving, and the pulse width (PW) and radio frequency (RF) parameters of the radiation source contained in the received signals are estimated by simulating the pulse stream of the receiver for training, so as to achieve the deinterleaving of radar signals. The main work of this dissertation is as follows. This dissertation firstly studies the traditional radar signal deinterleaving algorithm, and analyses the histogram method in the main signal deinterleaving, discusses the deinterleaving of two signals intertwined, and compares and analyses the performance of the cumulative difference histogram (CDIF) algorithm and the sequence difference histogram (SDIF) algorithm on the deinterleaving of multiple signals intertwined. This dissertation introduces some ideas of clustering algorithms that have been applied to signal deinterleaving and then introduces some basic techniques of Convolutional Neural Networks, with a focus on CNN networks, and further introduces the basic structure of the Faster RCNN and its suitability in signal deinterleaving. Aiming at the shortcomings of the traditional signal deinterleaving and predeinterleaving clustering algorithm, this dissertation proposes a pre-deinterleaving algorithm based on the FasterRCNN network, which detects the PW and RF parameters of the radiation sources in the received pulses through the network. The algorithm achieves good results in both matching known radar pulses from the radar source library and deinterleaving unknown radars, and achieves the purpose of blind source deinterleaving.