Implementation of Adaptive Short Time Fourier Transform and Sigmoid based Kernel Support Vector Machine for Radar Signal Identification

Distinguishing between simple pulsed and Low Probability of Intercept (LPI) radar signals is crucial for various applications, including radar signal identification and electronic warfare. Existing methods often struggle with achieving high accuracy, and bridging computational complexity, particularly in low signal-to-noise ratio (SNR) environments. This paper proposes a novel approach that merges two powerful techniques: Adaptive Short-Time Fourier Transform (ASTFT) and Sigmoid Kernel Support Vector Machine (SVM). ASTFT offers exceptional time-frequency resolution, allowing for detailed signal decomposition, while the Sigmoid Kernel SVM provides robust classification capabilities. The research investigates the effectiveness of this combined approach for radar signal classification. Mathematical models for various radar signals and design of ASTFT optimized for accurate time-frequency analysis are presented. The Sigmoid Kernel SVM is then employed to leverage the extracted features and achieve precise signal classification. The proposed method demonstrated significant improvement in classification accuracy, particularly in low SNR conditions. At -5dB and above, the performance of 100% classification accuracy for all higher SNR values is achieved. These findings highlight the promise of the ASTFT-Sigmoid Kernel SVM approach for robust radar signal identification, even in challenging environments.