| Summary: | This report presents a comprehensive investigation into the application of Deep
Learning (DL) techniques for the classification and recognition of Pulse Repetition
Interval (PRI) modulated signals, with a focus on radar and communication systems.
The research involves the signal data simulation, network model implementation,
and comparison of Convolutional Neural Networks (CNNs) and Recurrent Neural
Networks (RNNs) for the automatic detection and classification of diverse PRI
modulation techniques across various environmental conditions and interference
scenarios. Our findings demonstrate the superiority of DL-based methods over
traditional signal processing approaches, shedding light on their interpretability and
real-world applicability. In this study, enhancements targeting loss function, network
architecture refinement, and the integration of advanced signal processing techniques
were proposed, collectively leading to notable performance improvements in the
classification and recognition of PRI-modulated signals.
Keywords: PRI modulation, Deep Learning, Data Signal Simulation, Convolutional
Neural Networks (CNNs), Squeeze-and-Excitation Networks, Focal Loss.
|
|---|