Research On Signal Separation Technology Based On Subspace Method

With the development of small unmanned aerial vehicle(UAV)technology,small UAVs are becoming more and more widely used in electronic warfare battlegroun d.And the concept of UAV swarm combat comes out in such condition.In combat of UAV swarms,reconnaissance with UAV swarm is a key link.In the reconnaissance with UAV swarm,the number of UAVs is numerous.Hence the dimension of the observed signals is v ery large.And the UAVs are moving relatively in the reconnaissance process.How to separate the source signals from the reconnaissance signals by the computing units of UAVs is a problem.This paper stud ies the mixed signal separation technology in the UAV swarm reconnaissance.And this paper mainly studies the mixed signal separation technology using subspace methods and its parallel computation structure.And this paper build s a signal mixed model of UAV swarm reconnaissance to simulate the mixed signal separation technology.Firstly,the basic theory of subspace is introduced.And the method of minimum noise subspace is studied to estimate the noise subspace.On this basis,the minimum noise subspace method is generalized to fast estimate the signal subspace and noise subspace with the number of K fixed computing units.The subspace estimation performance(SEP)and eigenvector estimation performance(EEP)are introduced to evaluate the accuracy of the estimation.Compared with the global eigenvalue decomp osition(EVD)method,the total computational complexity of the method is reduced to the 1/ K of EVD algorithm.Although the performance of the calculation is reduced,the performance index parameters are at the same order of magnitude compared with the global eigenvalue decomposition method.Then,the mixed signal separation algorithm using subspace method is studied.The model of the signal receiving system is established to generate the communication signal and radar signal,which are mixed by a random matrix.A nonlinear projection approximation subspace tracking(PAST)algorithm and a nonlinear orthogonality projection approximation subspace tracking(OPAST)algorithm of signal separation are used to separate the mixed signals.On this basis,the paralle l computing mode of the OPAST algorithm of signal separation is proposed by using the generalized minimum noise subspace method.With the number of K computing units,the amount of single computing unit's computing is reduced to 1/K.Finally,a simple UAVs' motion model and a simple signal receiving model are established,on which the signal receiving in the unmanned aerial vehicle swarm reconnaissance environment is established.And the OPAST algorithm of signal separation and its parallel operation method are used to separate the mixed signals of the UAVs.The simulation experiments results show that the mixed signal of UAV swarm reconnaissance can be effectively separated.And in the iterative updating of the separation process,the performance of the crosstalk error is fine.And the bit error rate is close to the theoretical value when demodulating the BPSK signal in the separated signals.