A Rcommunication Signal Recognition Method Based On Deep Learning

In the 21st century,wireless communication has been inseparable with our social development and daily life.Due to the openness of wireless communication,the electromagnetic environment around us is very complex and changeable.So the detection technology of wireless signal has been a hot issue.With the rise and rapid development of artificial intelligence technology,the signal detection and recognition technology has changed from relying on the professional ability of technical personnel to the computer to complete the detection of the target signal.This method not only saves a lot of manpower,but also greatly improves the detection speed and accuracy.Firstly,this paper studies the development history and current situation of signal detection and deep learning technology.It analyzes and summarizes its advantages and disadvantages,and introduces the basic principles of signal modulation and deep learning network.Then,this paper constructs a semi-automatic labeled COCO data set including 18 signal types,such as 4G,5G,Bluetooth,WLAN,radar signal,etc.The signal data is transformed into time-frequency diagram by means of Fourier transform.In this paper,a method of automatic detection and intelligent analysis of electromagnetic signal based on computer vision target detection technology is proposed,and experiments are carried out on existing data sets by using YOLOv3 and MASK RCNN.The mAP can reach 0.79 in the noisy environment.The validity of the data set and the feasibility of the method are proved.Then,based on the latest research results of Cornernet model,an improved Signal Cornernet model is proposed.Experiments are carried out on 10 basic types of signal data,and its mAP reaches 0.89.Compared with other methods,it has some advantages.Finally,the research results and shortcomings of this paper are summarized,and the future development of this technology is prospected.