LPI Radar Waveform Recognition Based on Multi-Resolution Deep Feature Fusion

Deep neural networks are used as effective methods for the Low Probability of Intercept (LPI) radar waveform recognition. However, existing models’ performance degrades seriously at low Signal-to-Noise Ratios (SNRs) because features extracted by insufficient under noise jamming. In this paper, we propose a multi-resolution deep feature fusion method LPI First, apply enhanced Fourier-based Synchrosqueezing Transform (FSST), which shows good SNRs, to convert signals into time-frequency images. Then, construct convolutional network extract more from each resolution channel. Next, explore an interactive strategy fusion. By some down-sampling or up-sampling blocks, different fused generate new features. Finally, algorithm fully connected layer achieve classification better performance. Simulation experiments on twelve kinds waveforms show that overall recognition accuracy our can reach 95.2% SNR ?8 dB. It is proved approach does indeed improve effectively SNRs.