Intelligent Specific Emitter Identification Using Complex-Valued Convolutional Neural Network

Specific emitter identification (SEI) is a promising technology for the authentication of internet of things (IoT) devices. With the rapid increase in the number and types of emitter signals and the growing complexity of the electromagnetic environment, traditional emitter identification methods face issues like declining identification performance. Furthermore, convolutional neural network (CNN) cannot fully exploit the coupling information between the in-phase and quadrature components of complex baseband signals. In contrast, complex-valued convolutional neural network (CVCNN) can better utilize the correlation between the phase and amplitude of complex signals. To address these issues, this paper introduces CVCNN to carry out SEI task. CVCNN is trained to learn effective features of emitter signals, improving identification performance. The method is evaluated on two datasets, automatic dependent surveillance-broadcast (ADS-B) and global system for mobile communications (GSM), comparing the performance of different neural network in identifying these emitter signals. The results demonstrate that the CVCNN method achieves higher identification accuracy. Our code can be downloaded from https://github.com/longlong217/CVCNN.