Multi-resolution convolutional neural network for specific emitter identification
Specific Emitter Identification(SEI)distinguishes specific emitters among identical radar types by inspecting hardware fingerprints.Deep learning-based SEI faces numerous challenges due to the large number of model parameters and complex computations,which limit its practical utility in real-time scenarios.So a light-weight multi-resolution convolutional network is proposed,which enriches the features of signals with different frequency domain resolutions,thereby improving recognition accuracy without increasing parameters or computa-tional complexity.Through experiments on a radiation source individual recognition dataset containing 10 naviga-tion radar signals,the results show that the proposed method improves recognition accuracy by about 8%com-pared to the standard convolutional network without increasing parameter and computational complexity.Com-pared with multi-scale network,the proposed method only uses 3/4 of the parameter and computational complexi-ty,achieving comparable accuracy.This method is suitable for real-time and near real-time application application scenarios such as aerospace.
NETL
NSTL
万方数据
