An Application of Analytic Wavelet Transform and Convolutional Neural Network for Radar Intrapulse Modulation...

Area of Science:

• Radar Signal Processing
• Machine Learning
• Wavelet Analysis

Background:

• Automatic recognition of radar intrapulse modulation is crucial for electronic warfare and signal intelligence.
• Traditional methods face challenges with complex modulations and low signal-to-noise ratios (SNR).

Purpose of the Study:

• To investigate the efficacy of Analytic Wavelet Transform (AWT) combined with Convolutional Neural Networks (CNNs) for radar intrapulse modulation recognition.
• To evaluate the impact of generalized Morse wavelet parameters on classification accuracy.
• To introduce the generalized Morse wavelet (GMW) as a novel approach for intrapulse modulation classification.

Main Methods:

• Utilizing the generalized Morse wavelet (GMW) for time-frequency image (TFI) generation.
• Employing a SqueezeNet-based Convolutional Neural Network (CNN) as the classifier.
• Testing with various intrapulse modulations: Linear Frequency Modulation (LFM) and Phase-Coded Waveforms (PCW) including Frank, Barker, P1, P2, and Px.
• Comparing AWT with other time-frequency methods like Short-Time Fourier Transform (STFT) and Wigner-Ville Distribution (WVD).

Main Results:

• The proposed AWT-CNN method demonstrates successful recognition of various intrapulse modulations.
• Classification accuracy is influenced by the parameters of the generalized Morse wavelet.
• Simulations were conducted in MATLAB across a signal-to-noise ratio (SNR) range of -6 to 0 dB.

Conclusions:

• The combination of AWT and CNN, particularly with the generalized Morse wavelet, offers a promising approach for radar intrapulse modulation recognition.
• The GMW-CNN framework provides a robust solution even at low SNRs.
• Further research can explore optimizing GMW parameters for enhanced performance.