Low Probability of Intercept Radar Waveform Recognition Based on Dictionary Leaming

Low probability of intercept (LPI) radar waveform recognition is a challenging task in modern radar and electronic warfare (EW) systems. To solve the problem of incomplete information and the need for human experience in the existing feature-based radar recognition methods, a robust and automatic LPI radar waveform recognition method based on Choi-Williams time-frequency distribution (CWD) and dictionary learning in the complex electromagnetic environment is proposed. First, the received signals are transformed to obtain the time-frequency matrix by CWD. Next, bilinear interpolation technique and row-orthogonal Cauchy random matrix are used for Iossless compression. Next, the label consistent k-singular value decomposition algorithm (LC-KSVD) is used to learn an over-complete dictionary and obtain the structure parameters of a linear classifier jointly. Finally, with the sparse code and the linear classifier, the type of test signals can be estimated. The superiority of the proposed method is universally applicable and does not need to rely on any human experience. Simulation results demonstrate that the proposed method has an excellent recognition rate at a low signal-to-noise ratio (SNR).