Implementing ResNet for Real-Time Radar Signal Classification in Electronic Warfare

Real-time signal processing is crucial to responding to rapidly evolving threats. Traditional methods relied on exten- sive preprocessing, creating bottlenecks and delaying detection. This research uses the ResNet algorithm to directly process in- phase and quadrature-phase (I/Q) signals with specialized hard- ware, eliminating preprocessing bottlenecks. The ResNet models were trained using a Certified tool for generating high-fidelity data. These ResNet models were modified for compatibility with AMD/Xilinx Versal deep learning chips that achieve precision greater than 90% and inference times of sub100 milliseconds. This approach improves efficiency by directly processing raw I/Q data, addressing a critical gap in real-time signal processing. Research hardware-compatible high-performance algorithms for defense applications.