Radar Signal Sorting Based On Distributed Fuzzy Clustering

In the radar reconnaissance system,radar signal sorting is one of the important components,which is the basis of signal analysis and the premise of identification.The result of radar signal sorting is directly related to the performance of the radar reconnaissance system.In the past radar electromagnetic environment,the signal was single and the pulse flow density was low.With the development of technology and the wide application of radar,the electromagnetic environment has become more and more complex,the radar signal is complex and changeable,and the pulse signal density is getting higher and higher.The traditional radar signal sorting method is really difficult to complete the sorting effectively.Aiming at the problem of low sorting efficiency of traditional signal sorting methods,this paper proposes a distributed feature reduction sorting algorithm.Radar signals usually contain five-dimensional features,some of which are of low importance,and too many features involved in the sorting will even affect the subsequent sorting effect.Therefore,in order to reduce unimportant features,this paper increases the process of feature reduction by setting a threshold,mainly by calculating the weight of each feature to eliminate features with small weights.In this way,after iteratively updated several times,the number of features can be reduced,and the number of iterations of the algorithm and the running time of the program can be reduced at the same time.The simulation results show that the algorithm has a good effect on multiple evaluation indicators.Subsequently,in order to further improve the sorting effect of the algorithm,the distributed feature reduction sorting algorithm was improved,and a distributed feature reduction sorting algorithm based on transfer learning was proposed.The algorithm adds the transfer learning item between adjacent nodes in the objective function.In each iteration update,the adjacent nodes learn the cluster center and weight knowledge from each other,which can accelerate the convergence of the algorithm and improve the sorting effect of the algorithm.The experimental results show that the sorting accuracy of the improved algorithm is higher.Finally,an adaptive distributed feature reduction sorting algorithm is proposed,because in the feature reduction sorting algorithm based on transfer learning,there is a learning factor that represents the mutual learning degree of adjacent nodes,which is artificially set before the algorithm starts.It is difficult to meet different radar signal datasets,and it is not universal.Therefore,this paper automatically adjusts the learning factor by calculating the offset distance of the adjacent node cluster centers.The simulation results show that the improved algorithm can reduce the influence of random initialization learning factors on the algorithm and make the sorting results of the algorithm more stable.