Detection and Classification of Pulsed Doppler Radar Signals Using Deep Learning

Pulsed Doppler radars are commonly used in military applications, especially in aircraft and missile systems, as they are able to detect and track fast-moving targets, such as enemy aircraft, while filtering out ground clutter and slower-moving objects. Classification of modulation types in airborne radar signals is essential for ensuring efficient and reliable communication, target identification, and interference rejection. This research paper presents a comprehensive study on modulation classification techniques for radar signals on an airborne module. The transmitter identity message signal is coded with pseudo-random noise codes and modulated using five modulations, namely, BASK, BFSK, BPSK, QAM and QPSK. Additionally, it is also subjected to Additive White Gaussian Noise(AWGN) and Rayleigh distribution for multipath fading in wireless transmission environments. The paper proposes a modulation recognition scheme based on multi-class classification, hierarchical feature extraction using a convolutional neural network (CNN) and residual neural network (ResNet50) architecture. The simulated result indicates that the proposed method demonstrates adaptability to modulational recognition even in low Signal-to-Noise Ratio (SNR) environments.