Interference mode open set identification model and method based on zero sample learning

Abstract

The invention discloses an interference mode open set identification model and method based on zero sample learning. A wireless communication anti-interference scene is considered, and in order to avoid interference, a background user selects an idle channel for communication based on spectrum sensing. A sensing device continuously senses the spectrum and identifies an interference pattern with the assistance of an agent. In order to solve the identification problem of an unknown interference mode, an open set identification scheme of supervised training and unsupervised classification is designed, and in the supervised training stage, an encoder is trained to learn potential feature representation of a known interference mode. In the unsupervised classification stage, the known interference mode and the unknown interference mode are classified in the feature space according to the classification criterion based on the distance, the unknown interference mode can be effectively dealt with, and the high open set recognition accuracy is obtained. The model is complete, the physical significance is clear, the algorithm design is reasonable and effective, and the interference mode open set recognition scene in the wireless communication anti-interference scene can be well described.

Description

technical field [0001] The invention belongs to the technical field of wireless communication, in particular to an open-set recognition model and method of an interference pattern based on zero-sample learning. Background technique [0002] Due to the characteristics of open channels, wireless communication is vulnerable to malicious interference attacks. Therefore, anti-jamming communication capability is very important in the field of wireless communication. As the first link of anti-jamming means, jamming pattern recognition plays an important role in anti-jamming communication. [0003] With the increasing development of machine learning technology, deep learning is gradually being applied to the field of interference pattern recognition. At this stage, related work has studied interference pattern recognition based on deep learning. Some scholars proposed to use the spectrogram to describe the spectrum environment, and introduce convolutional neural network to identi...

AI Technical Summary

Problems solved by technology

As a result, the performance of existing methods degrades severely when faced with the open-set recognition problem

Images