Radar Signal Modulation Recognition With Self-Supervised Contrastive Learning

Excellent performance in supervised learning-based radar signal modulation recognition (RSMR) techniques relies on the quantity and quality of labeled datasets, while the high cost and difficulty involved in analyzing and labeling radar signal samples limits its development. An RSMR algorithm using self-supervised contrastive learning (SSCL) methodology is proposed to address this issue. Specifically, within the classical contrastive learning (CL) framework MoCo V2, a customized data augmentation method is devised to capture time-frequency features of the radar signal. In addition, the feature extraction network ResNet50 is improved by decoupling spatial and channel filters, resulting in greater sensitivity to the time-frequency features. To enhance the recognition accuracy, two loss functions, alignment and uniformity, are used in place of the info noise contrastive estimation (InfoNCE) loss, and both loss functions are optimized directly. The recognition accuracy of the proposed method can reach 97.66% at a signal-to-noise ratio (SNR) of 4 dB.