Classification of radar signals using time-frequency transforms and fuzzy clustering

A method based on Smoothness Pseudo Wigner-Ville distribution and kernel principle component analysis is proposed to extract features of radar emitter signals. Then, these discriminative and low dimensional features achieved were fed to the classifier which is designed based on fuzzy Support Vector Machines (SVMs). In simulation experiments, the classification of two-class LFM signals was compared with four kernel functions. And the classifier attains over 83% overall average correct classification rate for five radar signals. Experimental results show that the proposed methodology is efficient for complex radar signals detection and classification.