Research On Interference Detection And Recognition Technology In BDS

Beidou Navigation Satellite System(BDS)is the third mature satellite navigation system after GPS in the United States and GLONASS in Russia.Its navigation accuracy is higher than other navigation systems,and it can provide users with more accurate time and location information.With the progress of science and technology and the vigorous development of electronic equipment,the wireless electromagnetic environment in which satellite navigation systems are located is complex and variable,and most of the signals received by the receiving end cannot be directly used.Therefore,it is necessary to detect the existence of interference and identify interference signals in the received signals in the system to take corresponding antiinterference measures to improve the stability and reliability of the Beidou Navigation Satellite System.This paper first classifies and analyzes the interference signals received by the BDS system,focusing on the study of six common types of suppressed interference signals,completing mathematical modeling of them,and mixing them with Gaussian white noise as input signals to facilitate the research on interference presence detection and recognition technology.In terms of interference presence detection algorithms,in order to address the problem of low detection probability of Energy Detection(ED)algorithms at low signal-to-noise ratios,one is to introduce noise uncertainty coefficients,dynamically estimate noise power,and propose a Double Threshold energy detection algorithm to effectively improve the detection probability at low signal-to-noise ratios;Secondly,from the perspective of mathematical statistics in binary detection,it is proposed to construct an error probability function with a threshold value as the independent variable,and calculate the extreme value of this function to obtain a new threshold value,thereby improving the detection probability under low signal-tonoise ratio.In response to the issue of the significant impact of noise power uncertainty on the detection performance of non blind detection algorithms,combined with the knowledge of Random Matrix Theory(RMT),the blind detection algorithm was studied.Due to the high computational complexity of Covariance Absolute Value Detection(CAVD)algorithm,the eigenvalues of the received signal matrix were calculated,We studied the Maximum Minimum Eigenvalue Detection(MMED)algorithm and the New Maximum Minimum Eigenvalue Detection(NMMED)algorithm respectively.To address the issue of MMED and NMMED algorithms only considering the distribution of maximum or minimum eigenvalues,a weighted fusion detection algorithm is proposed by introducing a weighted factor sum,which further improves the detection probability.In terms of interference recognition algorithms,this article starts from machine learning and studies interference signal recognition algorithms based on decision trees,support vector machines,and deep learning networks.In response to the problem of low recognition probability and slow recognition process caused by the method of manually selecting thresholds in Decision Tree(DT),the CART criterion is adopted for threshold selection,which not only improves recognition speed but also improves recognition accuracy;In response to the sensitivity of Support Vector Machine(SVM)to parameter adjustment,Grid Search(GS)was used for parameter optimization,effectively improving recognition accuracy;In response to the problem of slow training of large-scale datasets using Grid Search(GS),combined with deep learning network knowledge,the GoogLeNet network is used to classify and recognize interference signals.This algorithm has higher accuracy and robustness.