Design And Implementation Of Communication Equipment Identification System Based On Deep Learning

In recent years,the individual identification of communication equipment in the real electromagnetic environment has been an important subject in the field of communication and computer.This technology has broad prospects in wireless network communication,civil communication security and military electronic countermeasures.Because the signal fingerprint of communication equipment is stable and unique,as long as the hardware features of the communication equipment contained in the transmitted signal are accurately obtained,and the recognition accuracy is improved through the classifier,the algorithm of communication equipment recognition can be completed.It is aimed at the demand of fingerprint identification of communication equipment in practical application in this thesis,and studies it from two aspects: algorithm and system construction.In terms of algorithm,the key technology of signal recognition is studied,focusing on the multiple feature extraction methods of communication signals,the recognition method of communication equipment based on deep learning,and the realization of individual recognition in the case of multiple equipment aliasing.In engineering,the design and implementation of the fingerprint identification system of communication equipment has been completed.Solve problems such as the joint use of the system and the USRP equipment,the application deployment of the depth model,and the coordinated use of the communication equipment identification algorithm.It provides a scheme for the application of fingerprint recognition technology of communication equipment based on deep learning in practice.The work of this thesis mainly includes the following aspects:(1)It is completed the signal fingerprint identification of similar communication radiation sources,and considered complex situations in this thesis,such as single communication device signal,communication device working simultaneously,signal mixing with each other,receiving unknown communication devices,and so on.First,estimate the number of sources of the collected signal,then extract the subtle features of each communication device hardware,and finally identify the individual types of communication devices.The feature extraction part of this scheme can improve the signal recognition accuracy of known devices,but it takes a long time and is not convenient for using device signals without prior knowledge.Therefore,the possibility of directly classifying and recognizing the collected timing signals is considered,simplifying the algorithm steps,and obtaining recognition results faster.(2)It is completed the design and implementation of a fingerprint recognition system for communication equipment based on deep learning in this thesis.The system combines the use of GNURadio and USRP software radio platform in the design process.It integrates the feature extraction and feature recognition algorithm based on depth model in this thesis,and completes the software interface development that is convenient for users to collect signals and view the communication equipment to which the signals belong.It also allows users to add the classification and identification methods they want to use.(3)Finally,It is seted test cases according to the real scenarios faced by the system in this thesis.The test verifies that the system can complete the classification of communication equipment and get corresponding results in different scenarios.The accuracy rate of individual identification of communication equipment can reach over 93%,and the number of signals is estimated accurately.Each module of the system meets practical requirements.Therefore,the design and implementation of the communication device fingerprint identification system has certain theoretical significance and practical value.