A specific radiation source identification method and device based on a deep residual network

Abstract

The invention belongs to the technical field of radiation source identification, and particularly relates to a specific radiation source identification method and device based on a deep residual network, and the method comprises the steps: carrying out the time-frequency analysis of a received signal, and converting an obtained Hilbert time-frequency spectrum into a grayscale image; And extractingradio frequency fingerprint characteristics reflected in the image by using a depth residual network with the gray level image as input, and obtaining an identification result of the radiation source. Aiming at the characteristics of non-stability and non-linearity of communication signals, the gray level image of the Hilbert time-frequency spectrum is used as the representation form of the signals, the radio frequency fingerprint characteristics of the radiation source are extracted by using the deep residual network, and the classification recognition is completed; Deep learning is appliedto the field of communication signal processing, the powerful self-learning capability is fully exerted, the artificial understanding limitation is overcome, and the processing efficiency is improved;A simulation experiment verifies that the recognition effect under the complex communication system and the complex channel condition has very high robustness, and the method has important guiding significance for the development of a radiation source signal recognition technology.

Description

technical field [0001] The invention belongs to the technical field of radiation source identification, in particular to a specific radiation source identification method and device based on a deep residual network. Background technique [0002] Specific Emitter Identification (SEI, Specific Emitter Identification), that is, by extracting subtle features that can reflect individual differences in radiation sources on radio frequency signals, in order to achieve identification of specific targets. Since radio frequency fingerprint features do not depend on communication content and are difficult to forge; therefore, SEI technology has important application value in civil and military fields such as wireless network security and communication reconnaissance and countermeasures. The core technology of SEI is to discover and extract accurate and effective RF fingerprint features. According to different sources, the existing features can be mainly divided into the following two ...

AI Technical Summary

Problems solved by technology

It should be pointed out that the network architecture of CNN has stronger pertinence and applicability to images, but it cannot give full play to the huge advantages of CNN's deep self-learning image information, which affects the recognition effect to a certain extent; The bispectral image is used as input, and CNN is used to extract features and recognize them. The recognition effect is significantly better than traditional methods.

However, by comparing the existing published features, it can be found that the recognition effect of bispectral features is always unsatisfactory, and using bispectral images as image representations of received signals is not an ideal choice; in addition, the bispectral image Compression and dimensionality reduction will also cause the loss of some feature details while removing redundant information, which further affects the recognition effect

Images