Fast Fourier Transform with Multi-head Attention for Specific Emitter Identification

With the development of wireless communication, specific emitter identification (SEI) is important for the management and security of instrumentation and smart devices. Given that the signal differences between different devices of the same type are caused by hardware damage and are mainly concentrated in the high frequencies, the high-frequency component of the signal is reconstructed by Fourier transform, attention mechanism, and inverse Fourier transform in this paper. The reconstructed high-frequency component of the signal is then fed into a recurrent neural network (RNN) to extract features from the time dimension. The frequency attention module and the time attention module are connected serially, which on the one hand allows the overall network to pay attention to both the frequency and time characteristics without increasing the amount of data, and on the other hand ensures that the results of the frequency attention must facilitate the subsequent RNN for feature extraction. The parameter sizes of many SEI methods are measured. The results show that the model proposed in this paper has the highest parameter efficiency and low storage costs. The results on a real-world dataset show that the proposed model has the highest accuracy. These advantages have essential significance for deploying the model in practical applications.