Recognition for Radar Emitter Signals Based on Bispectral Feature Fusion
Jundi Wang, Wang Xing, Pengyu Dong, You Chen, Yuanrong Tian · 2022
Radar radiator signal recognition is a key component of electronic reconnaissance system. In order to improve the accuracy of the signal recognition of low probability of interception(LPI) under the condition of low signal-to-noise ratio, this paper proposes an algorithm for feature extraction and recognition in the high-order spectral transform domain. This method overcomes the shortcomings of previous recognition algorithms that rely heavily on experience and cannot adapt to waveform changes. First, the bispectral transformation is used to form a three-dimensional physical representation of the radar signal. Then the bispectral diagonal slice of the radiation source signal is extracted. On this basis, Robust Principle Component Analysis (RPCA) is used to reduce the dimensionality of features. RPCA not only reduces redundancy but also reduces noise. Finally, LSSVM is used to analyze the feature vector to realize signal classification and recognition.