Radar Signal Recognition Based on Dual-Channel Model With HOG Feature Extraction

Objectives: To improve the recognition accuracy of radar signals under a low signal-to-noise ratio (SNR). Technology or Method: We propose a novel radar signal recognition method based on a dual-channel model with the histogram of oriented gradients (HOG) feature extraction. Specifically, multisynchrosqueezing transform (MSST) and Choi–Williams distribution (CWD) transform are adopted individually to obtain the time–frequency distribution images of radar signals, and HOG feature extraction is performed on the preprocessed time–frequency images of each channel, respectively. Then, the features of the two channels are fused and dimensionally reduced by the principal component analysis (PCA). Finally, the compact feature parameters are fed to the support vector machine (SVM) classifier to identify radar signals. Clinical or Biological Impact: The experimental results demonstrate that the proposed model achieves a high recognition performance with a small computational complexity, especially in low SNR. When the SNR is −12 dB, the recognition accuracy can reach more than 92%, which is over 6% higher than that of single-channel models and related convolutional neural network-based models.