Research On Spectrum Monitoring Node Deployment And Wireless Signal Recognition Algorithm

Electromagnetic spectrum monitoring and wireless signal identification are important links for conducting wireless communication services,maintaining airborne radio order,and improving spectrum utilization.However,electromagnetic spectrum monitoring and wireless signal identification face the following challenges: First,due to factors such as topographic feature distribution,monitoring node deployment and transmitter power,the dynamic range of the received signal-to-noise ratio of the monitoring signal is wide;Second,the time dispersion and multipath effect of the wireless channel cause blind areas in the monitoring area,and the recognition performance is low.The third is the spatial mutual interference caused by the coexistence of heterogeneous multi-network and the spatial and temporal uncertainty of the random noise of the receiver.This paper studies two main parts of the spectrum monitoring system,the spectrum monitoring node deployment strategy and wireless signal identification algorithm.The deployment of spectrum monitoring node is aimed to target specific monitoring areas,optimize the deployment of spectrum monitoring nodes,and realize the full coverage of electromagnetic spectrum data in the monitoring area.The radio signal identification algorithm aims to identify and classify the communication mode of the radio signal by analyzing and feature extracting of the radio reception signal,and provides support for judging the interference type and interference traceability.(1)In the research of spectrum monitoring node deployment strategy,by analyzing the spectrum monitoring node deployment strategy in a large number of documents,especially for the shortcomings of the virtual force-based electromagnetic spectrum monitoring node deployment optimization algorithm proposed in the literature [8-9],this paper proposes a class of electromagnetic spectrum monitoring node optimization deployment strategy,which combines the virtual forces from the coverage blind zone,Voronoi polygon vertices and edges,and the monitoring nodes,aiming to further reduce the monitoring coverage blind zone,using as few monitoring nodes as possible to cover as large a monitoring area as possible.Based on the evaluation indexes such as effective coverage rate and monitoring node moving distance,the algorithm is compared and simulated,and the rationality and effectiveness of the algorithm improvement are verified.(2)In the research of wireless signal recognition algorithm,the feature of wireless signal recognition technology based on decision tree and the wireless signal recognition technology based on feature learning is analyzed and compared,this paper studies wireless signal recognition technology based on deep learning.Five deep neural network structures(CNN2,CNN4,CLDNN,LSTM2 and RESNET3)were selected and designed.The I and Q characteristics of the wireless signal were combined with the amplitude and phase characteristics to obtain eight eigenvector combinations.Using the RadioML2016.10 a dataset,11 types of wireless signals were identified based on five deep neural networks.The test results show that different wireless signal characteristics and their combination order,form and neural network structure will affect the recognition rate of wireless signals,and the proper combination of deep neural network and feature combination can improve the accuracy of wireless signal recognition algorithm.