Research On Specific Communication Emitter Identification Based On Complex-Valued Network

Special Communication Emitter Identification is not only a critical research problem in the field of communication countermeasures,but also one of the technologies that are crucial to target reconnaissance systems and intelligence analysis systems,which play a vital role in military communications,e-reconnaissance,battlefield network situational awareness and other fields.In complex electromagnetic environments,traditional methods for Special Communication Emitter Identification have the limitations of weak robustness and over-dependence on expert knowledge,however,deep learning-based methods can overcome the limitations of traditional methods by leveraging their powerful characterization capabilities and self-adaptability,most of these deep learning network frameworks originate from the area of images,so it is difficult to adapt to the unique scenarios of electromagnetic signals,especially in the area of Special Communication Emitter Identification that includes low signal-to-noise ratio and insufficient number of effective samples.This dissertation aims to address the above problems in the field of individual identification of communication radiation sources by initially introducing complex-valued neural network,while making corresponding improvements,and the major study contents and innovations of this dissertation are as follows.1.Aiming at the problem of existing algorithms such as residual networks that lose I/Q two-way data information when processing input signal data,an individual recognition algorithm of Special Communication Emitter Identification based on complex-valued residual networks is proposed.The algorithm uses complex-valued convolution operation to mix I/Q two-way information in each channel to deeply explore the intrinsic correlation between signals and achieve better recognition results.2.Aiming at the problem that current recognition algorithms are affected by the low signal-to-noise environment,Special Communication Emitter Identification algorithm with an embedded attention mechanism is proposed.The proposed algorithm inserts the attention unit into the end of the backbone network as an independent network layer,which improves the robustness of the algorithm with a small increase in the number of network parameters.The actual communication radiation source data show that the attention mechanism can enhance the recognition effectiveness of the algorithm under low signal-to-noise ratio conditions.3.Aiming at the problem of insufficient number of labeled samples,a method for Special Communication Emitter Identification based on complex-valued dilation residual networks is proposed.The proposed algorithm combines the dilation convolution with the residual network to capture the subtle features of the signal by utilizing the wide sensory field of the convolution kernel.In addition,the information mining capability of limited data is enhanced by the use of integrated learning methods on the basis of complex dilated residual networks,and the effectiveness of the proposed algorithm is verified by the actual Special Communication Emitter Identification experiments.