Combining Convolutional Neural Networks and Smoothed Distribution Pseudo Wigner Ville in the Classification of Low Probability of Intercept Radar Signals

• Edgard B. Alves Military Institute of Engineering http://orcid.org/0009-0009-2138-2929
• Jorge A. Alves Naval School https://orcid.org/0009-0004-0766-8588
• Ronaldo R. Goldschmidt Military Institute of Engineering https://orcid.org/0000-0003-1688-0586

DOI: https://doi.org/10.5753/sbbd.2024.240789

Abstract

Radar Electronic Warfare has a fundamental role in the defense of the nations. To adapt it to current threats, it is necessary to use automatic recognition algorithms for intrapulse modulations (ATR) of Low Probability of Interception (LPI) radar signals. The main existing LPI signal ATR combine the Choi-Williams Distribution with Convolutional Neural Networks (CNN). This work proposes a combination based on the Smoothed Pseudo-Wigner-Ville distribution (SPWVD) and the CNN SqueezeNet, to obtain a better-performing ATR. The proposed combination achieved 97,8% classification accuracy for 13 types of modulations and 806,000 samples generated. The datasets with such samples are available for research.

Keywords: CNN, Automatic LPI Radar Signal Recognition, SPWVD

References

Choi, H.-I. and Williams, W. J. (1989). Improved time-frequency representation of multicomponent signals using exponential kernels. IEEE Trans on Acoustics, Speech, and Signal Processing, 37(6):862–871.

da Defesa, M. (2004). Política de ge de defesa. Portaria 333/MD. Art. 4.

Faceli, K., Lorena, A. C., Gama, J., Almeida, T. A. d., and Carvalho, A. C. P. d. L. F. d. (2021). Inteligência artificial: uma abordagem de aprendizado de máquina. LTC.

Huynh-The, T. et al. (2021). Accurate lpi radar waveform recognition with cwd-tfa for deep convolutional network. IEEE Wireless Com. Letters, 10(8):1638–1642.

Iandola, F. N. et al. (2016). Squeezenet: Alexnet-level accuracy with 50x fewer parameters and 0.5 mb model size. arXiv preprint arXiv:1602.07360.

Kong, S.-H. et al. (2018). Automatic lpi radar waveform recognition using cnn. Ieee Access, 6:4207–4219.

Liu, Z. et al. (2024). A method for lpi radar signals recognition based on complex convolutional neural network. International Journal of Numerical Modelling: Electronic Networks, Devices and Fields, 37(1):e3155.

Ma, N. and Wang, J. (2013). Dynamic threshold for spwvd parameter estimation based on otsu algorithm. Journal of Systems Engineering and Electronics, 24(6):919–924.

Milczarek, H. et al. (2023). Automatic classification of frequency-modulated radar wave-forms under multipath conditions. IEEE Sensors Journal.

Niranjan, R., Rama Rao, C., and Singh, A. (2021). Fpga based identification of frequency and phase modulated signals by time domain digital techniques for elint systems. Defence Science Journal, 71(1).

Pace, P. E. (2009). Detecting and classifying low probability of intercept radar. Artech house.

Walenczykowska, M., Kawalec, A., and Krenc, K. (2023). An application of analytic wavelet transform and convolutional neural network for radar intrapulse modulation recognition. Sensors, 23(4):1986.

Wan, C., Si, W., and Deng, Z. (2023). Research on modulation recognition method of multi-component radar signals based on deep convolution neural network. IET Radar, Sonar & Navigation.

Willetts, B., Ritchie, M., and Griffiths, H. (2020). Optimal time-frequency distribution selection for lpi radar pulse classification. In 2020 IEEE Int Radar Conf (RADAR), pages 327–332. IEEE.