Distributed Few-Shot Learning for Intelligent Recognition of Communication Jamming

Effective recognition of communication jamming is of vital importance in improving wireless communication system's anti-jamming capability. Motivated by the major challenges that the jamming data sets in wireless communication system are often small and the recognition performance may be poor, we introduce a novel jamming recognition method based on distributed few-shot learning in this paper. Our proposed method employs a distributed recognition architecture to achieve the global optimization of multiple sub-networks by federated learning. It also introduces a dense block structure in the sub-network structure to improve network information flow by the feature multiplexing and configuration bypass to improve resistance to over-fitting. Our key idea is to first obtain the time-frequency diagram, fractional Fourier transform and constellation diagram of the communication jamming signal as the model-agnostic meta-learning network input, and then train the distributed network through federated learning for jamming recognition. Simulation results show that our proposed method leads to excellent recognition performance with a small data set.