Research On Offshore VHF Signal Recognition Based On Machine Learning

VHF communication system is a commonly used maritime close-distance communication system.The equipment includes ship-borne telephones and wireless intercoms.The characteristics of semi-dual transmission lead to the illegal occupation of emergency channels or special channels,which is directly related to the safety of life and property of marine personnel and the safety of maritime traffic.Because the communication protocol of the VHF(Very High Frequency)signal transmitted lacks identity information,thus cannot be effectively controlled.There is little research on individual recognition of offshore VHF signals,most of which focus on modulation recognition including the VHF frequency band.Based on the individual fingerprint identification technology of wireless signal physical layer,this paper realizes multiclassification detection and individual identification of maritime VHF interphone and further satisfies the all-round demand of maritime communication for security control.This paper combines Gabor features extraction with feature selection algorithm at first,and uses SVM classifiers to identify,the result shows that the recognition rate improves obviously.the recognition rate of the indoor signal can reach 99%at 30dB.The feasibility of VHF signal recognition can be verified.VHF signals on the sea are vulnerable to fading and environmental noise in actual use,which is a major obstacle to the identification process,to further enhance the intelligence,accuracy,and stability of the identification system.In this paper,an intelligent classification algorithm based on machine learning is proposed.For the fading of VHF signals on the sea,CCGAN(Cycle Convolution-GAN)is used to generate enhanced signals and recognize them.In both indoor and outdoor environments,CCGAN can obtain a better classification result.At a distance of 230 meters on the lake,the VHF signal classification rate can reach 94.4%.For signal noise,we use DnCNN(Denoising Convolutional Neural Network).The signal classification rate can still reach 98%in Gauss white noise with a signal-tonoise ratio of 5dB.On this basis,the paper realizes VHF signal acquisition,fading processing,denoising,and signal classification,obtaining a complete maritime VHF signal detection and classification method.