Cross-domain specific emitter identification based on domain adversarial multi-constraint
Wenqiang Shi, Yingke Lei, Hu Jin, Fei Teng · 2024
Specific emitter identification technology can detect and identify the target signal by extracting the subtle features of the radiation source, and associate it with the target radiation source, so it has an important position in the field of information countermeasure reconnaissance. However, in the actual task of specific emitter identification, the phenomenon that the pre-collected data and the data to be identified are not in the same channel often occurs, which usually leads to a drastic decrease in the recognition rate. To solve this problem, we propose a cross-domain identification algorithm for radiation source individuals based on domain adversarial multi-constraints. The main idea of the algorithm is to make the feature distribution of the pre-collected data and the data to be recognized is similar through domain adversarial training, achieving the goal of high precision recognition of the data to be recognized treated by the classifier. In addition, constraint parameters are introduced to constrain domain adversarial errors and inter-domain difference errors, and effectively improve performance. In the experiment, in order to verify the effectiveness of the algorithm, we trained and tested on two data sets, and compared with multiple models, and the experimental results proved the superior performance of our model.