Research On Individual Recognition Of Multirotor UAV Remote Terminal Based On Signal Fingerprint

In response to the increasing number of "illegal flying" incidents of UAVs,regulating the safely and orderly flying of UAVs has attracted attentions from various fields.The identification of the remote controller for UAVs has become an important part of regulating UAV flights.Due to the subtle signal differences between different remote controllers,the traditional identification methods are no longer applicable.It is urgent to propose a new identification method for UAV remote controller based on the individual signal characteristics of the UAV.This paper carries out the research on the individual identification of the UAV remote controller from both theoretical and experimental directions.In this research,we construct the UAV signal transmitter model,study the formulation of the UAV signal fingerprint,and collect the remote control signal of the UAV.By extracting different fingerprint characteristic parameters of remote control signals under different SNR conditions,we create one set of identification system for UAV remote controllers through experimental analysis.The main research progress of this article are as follows:(1)Through the construction of the UAV signal transmitter model,the formulation mechanism of the UAV signal fingerprint feature is discussed.Due to the component differences between different UAV signal transmitters,subtle differences in the UAV signal are inevitable.Those differences,which are specifically manifested in phase noise,carrier frequency deviation,and nonlinearity of the power amplifier circuit,are the cause of signal fingerprints.(2)According to the characteristics of the individual signal transmitted by UAV remote controller,four UAV signal fingerprint extraction methods are proposed.The transient steadystate feature fusion method fits the amplitude curve of the transient signal and extracts the instantaneous frequency of the steady-state signal.The method of constellation diagram first obtains the constellation diagrams of the remote control signals,and then obtains the coordinates of two clustering center points through the K-means clustering algorithm.The bispectrum feature method estimates the bi-spectrum value of the UAV remote control signals,and then reduces the estimated bi-spectrum value to a one-dimensional vector through the rectangular integral.The empirical wavelet transform(EWT)method decomposes the UAV remote control signal into a series of filtering components through the empirical wavelet transform.(3)An experimental platform for remote controller identification was built to verify the four fingerprint extraction methods under different SNR settings.The experimental result shows that,using the method of transient steady-state feature fusion can achieve 90%recognition accuracy when SNR=10d B.Using the method of constellation diagram features,the recognition accuracy of four remote controllers is more than 95% under the same SNR settings.By using the bi-spectrum feature method,95% recognition accuracy is achieved for four remote controllers even when the SNR=0d B.Finally,when the EWT method is adopted,the recognition accuracy rate reaches 95% for four remote controllers when SNR=8d B.(4)A real-time identification system for UAV remote controllers has been conducted.The system uses the USRP B210 as the wireless signal acquisition device to extract the signal constellation features,it can identify four different UAV remote controllers in real time.The remote controllers of the UAV can be identified at different distances from 5m to 50 m.The experimental result shows that the recognition accuracy decreases as the distance increases.