Autonomous Cropping and Feature Extraction using Time-Frequency Marginal Distributions for LPI Radar Classification

E.R. Zilberman and P.E. Pace (USA)


Choi-Williams, time-frequency, autonomous, LPI, feature extraction, multi-layer perceptron


An autonomous (no human operator intervention) feature extraction algorithm that can be used for classification of low probability of intercept (LPI) radar modulations using time-frequency (T-F) images is presented. The approach uses the marginal frequency distribution from a Choi Williams time-frequency image to crop the modulation energy autonomously. A new adaptive binarization algorithm is used to preserve the high resolution detail that emphasizes the differences between modulation classes without overwhelming the classifier. The binary feature vector is used as an input to a multi-layer perceptron classification network. Classification results for five simulated low probability of intercept (LPI) radar modulations are shown for variations in the modulation parameters (difficult case) and variations in signal to noise to demonstrate the feature extraction approach and quantify the performance of the algorithm.

Important Links:

Go Back