Specific Emitter Identification based on Multi-Domain Features Learning

Specific emitter identification (SEI) is a technology to extract the subtle fingerprint features of the received electromagnetic signal, and identify the emitters to which the signal belongs. It has important applications in military and civil occasions. Traditionally, expert-experience is used for feature extraction, which is time-consuming and unstable. In order to overcome this shortcoming, this paper proposes an Intelligent Radiometric Identification algorithm base on Time and Frequency domain feature Fusion (IRI-TFF) which uses deep learning technology. The algorithm designs a new multi-domain fused one-dimensional complex-valued densely connected convolutional network (DenseNet) model after the accurate "calibration" preprocessing of the received signal and the combination of time and frequency domain data as training examples. Meanwhile, three fusion strategies are proposed. The experimental results show that the proposed algorithm is superior to the traditional expert-experience based SEI algorithm or other similar deep learning based SEI algorithm, and is robust to noises.