Radar intrapulse modulation classification based on deep learning

Idenfication of threat radar signals is extremely important for electronic support systems. Radar intrapulse modulations are used by electronic support systems to identify radar systems. In this thesis, radar intrapulse modulations are classified using signal processing and deep learning techniques. A computer based radar signal simulator is designed to generate radar signals which are used to train and test a convolutional neural network. Short Time Fourier Transform, Fourier Synchrosqueezed Transform, Smoothed Pseudo Wigner Ville Distrubution, Choi-Williams Distrubution and Cyclostationary Signal Processing are applied to radar signals for feature extraction. After feature extraction, transformed signals are used to train and test convolutional neural network. Simulation results show that, %98 modulation classification performance is achieved at 0 dB signal to noise ratio.