Specific Emitter Identification under Dynamic Noise Based on Domain Adaptation

Existing deep neural network (DNN) based specific emitter identification (SEI) algorithms are limited by training scenarios, and the identification accuracy of the network is seriously degraded when the signal to be identified is not consistent with the channel noise of the training data. In order to solve this problem, this paper proposes a SEI algorithm based on transfer learning. Combining with the idea of domain adaptation, this method established an optimization model to align the features of signals under different signal-to-noise ratio (SNR), so that the neural network trained under a specific SNR can learn the radio frequency fingerprint (RFF) features which are independent of channel noise, and realize the identification of signals under other SNR conditions with high accuracy. Simulation results show that the proposed algorithm improves the accuracy of the SEI algorithm based on DNN under the interference of dynamic noise. When the SNR of the signal to be identified decreases by 4dB, the identification accuracy can be improved by 45.18%.