Research On Key Technologies Of Frequency Hopping Radio Sorting

Since the advent of frequency-hopping communication,it has been widely concerned by various countries.Frequency-hopping has high concealment and strong anti-interference ability due to its unique communication method.High-bandwidth,high-hopping speed frequency-hopping radio stations have become an indispensable part of the military wireless communication data link.This puts forward the requirements for the detection and collection of frequency-hopping communication signals in non-cooperative environments.The research of frequency-hopping reconnaissance is helpful in electronic warfare.Thus,it is an important factor in the victory of modern warfare.Frequency hopping network station parameter estimation and network station sorting are the core contents of frequency hopping communication reconnaissance.This article studies from the following aspects:1.First,analyze the performance of various non-stationary signal time-frequency transformations in analyzing frequency-hopping signals,mainly through two performance indicators of information entropy and algorithm complexity,compare the performance of multiple time-frequency transformations,choose a Video analysis method of complexity and high energy focusing performance.2.Secondly,on the basis of time-frequency transformation,the parameters of single-antenna single-frequency hopping signal and multi-frequency hopping signal are estimated.The time-frequency ridge algorithm is used to estimate the frequency hopping period and time hopping of a single frequency hopping signal.The stepwise difference histogram algorithm is introduced to estimate the hop period and hop time of the multi-hop asynchronous network station signal.Then,through the array antenna,the multi-signal classification algorithm is introduced into the frequency-hopping signal,and the direction of arrival of the multi-frequency-hopping signal is estimated with it.3.Thirdly,on the basis of frequency hopping parameter estimation,each hop signal is sorted.Among them,the "elbow method" in mean clustering is used to determine the number of sources,and an improved K-means algorithm with high convergence performance,low iteration number,and insensitivity to outliers is proposed to classify signals.Then construct a multi-classification SVM machine,classify the network station signals by high-dimensional mapping,and select different kernel functions and adjust the parameters to make it have good classification performance for frequency-hopping signals,Then compare and separate the two classification algorithms and their respective advantages and disadvantages.4.Next,without feature selection and extraction,an independent component analysis algorithm is introduced.The signal is blindly separated without prior knowledge through model limitations by independent component analysis algorithm.Two non-Gaussian metrics,kurtosis and negative entropy are introduced,a mixed matrix is constructed for simulation,and the separation performance is quantitatively evaluated using the Amari index.The effects of the convergence performance and the number of sampling points on the algorithm are analyzed.5.Finally,signals are generated by Matlab,sent through a vector signal source,and combined with the noise generated by USRP.The effectiveness of the proposed algorithm is verified by means of semi-physical simulation.