Radar Intrapulse Modulation Classification with Deep Learning

Identification 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 work, radar intrapulse modulations are classified using signal processing and deep learning techniques. Short Time Fourier Transform, Fourier Synchrosqueezed Transform and Smoothed Pseudo Wigner Ville Distrubution are used to extract time-frequency images of radar signals. After feature extraction, convolutional neural network is used as classifier and %98 modulation classification performance is achieved at 0 dB signal to noise ratio.