A Novel Approach of Feature Extraction for Advanced Radar Emitter Signals Using Time-Frequency Atom Decomposition
Gexiang Zhang · Xi'an Jiaotong Daxue xuebao · 2010
A novel approach,called TFAD,is proposed to extract time-frequency atom features to effectively recognize the intra-pulse modulation types of advanced radar emitter signals. The method decomposes emitter signals based on matching pursuit in Gabor and Chirplet atom dictionaries using a modified differential evolution algorithm. Then both the energy of the first decomposed atoms and the frequency parameters of the decomposed Gabor atoms are utilized to extract two correlation ratio features and a frequency variance feature. These features are used as the classification features to recognize different intra-pulse modulations of emitter signals by constructing a directed acyclic graph support vector machine classifier. The computational complexity of TFAD is O(n),while the computational complexity of the fractal approach is O(nlogn). Simulation results conducted on various signal-to-noise ratios and a wide range of modulation parameters of five typical radar emitter signals show that TFAD achieves an average correct recognition rate of 98.3%.