Automatic Classification of Low Probability of Interception Radar Signals: A Deep Learning Approach Utilizing the Smoothed Pseudo Wigner Ville Distribution

Radar Electronic Warfare is crucial for national security. To adapt it to contemporary challenges, it is essential to apply automatic intrapulse modulation detection algorithms (ATR) in Low Probability of Interception (LPI) radar signals. Most ATR for LPI signals combine the Choi-Williams Distribution with Convolutional Neural Networks (CNN). This study suggests a solution that uses the Smoothed Pseudo-Wigner-Ville (SPWVD) distribution and the GoogLeNet CNN, aiming to enhance the ATR performance. The suggested solution achieved a classification accuracy of 99,06% for 13 modulations and 806,000 produced samples. The data sets containing such samples are accessible for studies.