Research On Multi-Channel Parametric Adaptive Signal Detection Based On Krylov Subspace Methods

Multichannel signal detection is one of the main research topics in radar,communication and medical filed.The main troubles of classical signal detection methods are that there are large computational costs in space-time adaptive processing,and there are few independent and identically distributed samples in the actual non-uniform environment,which leads to difficulty of real-time operation and detection performance degradation.In the thesis,the Krylov subspace method is used to attain space-time adaptive weight vector iteratively for the problems existing in the classical multichannel signal detection.The clutters is approximated to the AR model by using the smoothness of clutter data in the pulse domain.On the basis,the Krylov subspace adaptive matched filter and Krylov subspace multichannel parametric adaptive signal detection methods are discussed.The main work and conclusions are as follows:1.The Krylov subspace method is applied to the adaptive matched filter.The performance of a series of KAMF detectors generated during the iterative process is analyzed.For the structured disturbance covariance,CG algorithm converges in r(10)1 iterations.KAMF has better detection performance in r(10)1 iterations too.The false alarm probability of the KAMF detectors are analyzed by the fast convergence of the extreme Ritz value,?R-orthogonal projection theorem and the Wishart distribution property.The theoretical expression under the number of iterations k is given.The value of k is between 1 and r(10)1.Based on the simulation data and realistic clutter,the false alarm probability and the detection probability are simulated.The theoretical analysis and simulation results show that the false alarm probability of the KAMF detectors is between MF and AMF.under the same number of training samples,the detection performance behaves better than AMF too.2.Conjugate Gradient algorithm is used to solve the Wiener-Hopf equation in multichannel parametric adaptive matched filter.A series of KPAMF detectors are generated.The iterative process is consistent with the PAMF detectors in detection performance.In most cases,the KPAMF detectors converge in fewer iteration,and the amount of computations is further reduced.At the same time,in the case of large disturbance covariance condition number,the preconditioned conjugate gradient algorithm is used to reduce the condition number and improve the convergence rate.The order and parameters of the AR model are related to the number and the frequency of distribances.The autocorrelation matrix of the prediction vector has an low rank correction structure.The related properties and methods of the KPAMF detectors are demonstrated based on the simulation data and realistic clutter model.3.Non-uniformity clutter environment of power spectrum and statistical distribution is simulated.The KAMF detectors and KPAMF detectors are applied to the target detection in the above non-uniform environment.The simulated results show that the clutter suppression effect of KPAMF detectors and PAMF is better than that of KAMF detectors and AMF due to the lack of effective training samples in non-uniform environment.In some specific iterations,the KAMF detectors perform better than AMF.The detection performance of KPAMF detectors is close to the PAMF in fewer iterations.The related properties and conclusions of KAMF detectors and KPAMF detector are demonstrated in non-uniform environment.