LPI Radar Waveform Recognition Based on Features from Multiple Images.

Area of Science:

• Electronic intelligence systems
• Radar signal processing
• Signal modulation classification

Background:

• Real-time detection and classification of noisy low probability of intercept (LPI) radar signals is crucial for electronic intelligence.
• The increasing prevalence of LPI radar necessitates advanced LPI radar waveform recognition techniques (LWRT).
• Existing methods struggle with the non-stationary characteristics of LPI radar signals, particularly in low signal-to-noise ratio (SNR) environments.

Purpose of the Study:

• To propose a novel Multiple Feature Images Joint Decision (MFIJD) model for enhanced LPI radar waveform recognition.
• To address the challenges posed by non-stationary LPI radar signals and low SNR conditions.
• To improve the universality and accuracy of LPI radar signal classification.

Main Methods:

• Developed a MFIJD model incorporating two distinct feature extraction structures.
• Utilized short-time autocorrelation feature images, double short-time autocorrelation feature images, and time-frequency images (TFI).
• Employed a hybrid model classifier suitable for non-stationary signals, combining multiple feature inputs.

Main Results:

• The MFIJD model demonstrated high universality across 11 simulated LPI radar signal types.
• Achieved superior adaptation to LPI radar waveform recognition in low SNR environments compared to existing literature.
• Attained an overall recognition rate of 87.7% at a signal-to-noise ratio (SNR) of -6 dB.

Conclusions:

• The proposed MFIJD model offers a robust solution for LPI radar signal detection and classification.
• The method exhibits excellent performance in challenging low SNR conditions.
• MFIJD provides a versatile and accurate approach for electronic intelligence applications requiring LPI radar waveform recognition.