Robust Identification of Communication Radiation Source Individuals Based on Transfer Learning

This paper proposes a transfer learning-based communication radiation source individual identification algorithm to mitigate the practical limitations of deep learning-based communication radiation source identification technology and address the performance degradation of deep learning networks in complex and dynamic electromagnetic environments. This algorithm constructs a new metric function founded on feature fusion and designs a subdomain alignment loss function to quantify the discrepancy in the distribution of source and target domain data in the feature space. By integrating the subdomain alignment loss function into the network training process, this method can effectively reduce the variability in the data distribution. Furthermore, a transfer learning strategy based on the model parameters is introduced to accelerate the training process of the model. The experimental results demonstrate that the proposed algorithm exhibits superior classification performance.