Radar Emitter Identification with Bispectrum based LBP and Extreme Learning Machine

Radar Emitter Identification (REI) has been a long-standing topic in military and civil fields. In this paper, we present a novel REI based on the local binary pattern (LBP) feature extracted from bispectrum of radar signal and the extreme learning machine (LBP+ELM). Comparing with conventional radar features, the high-order spectral analysis method is robust and effective in handling the issues of increasing electromagnetism signal density and system complexity in recent REI. But the bispectrum of radar signals suffers from high feature dimensionality. To address this issue, we extract the local binary pattern based on the bispectrum image. Then, the efficient extreme learning machine (ELM) is applied for classifier training. Experiments on six most frequent radar signals recognition are carried out in this paper for performance validation. Comparisons with the original bispectrum based REI are provided to show the superiority of our proposed algorithm.