EMULATING RANDOM PROCESS TARGET STATISTICS (USING MSF)

Authors
Citation
Th. Kerr, EMULATING RANDOM PROCESS TARGET STATISTICS (USING MSF), IEEE transactions on aerospace and electronic systems, 30(2), 1994, pp. 556-577
Citations number
61
Categorie Soggetti
Telecommunications,"Engineering, Eletrical & Electronic","Aerospace Engineering & Tecnology
ISSN journal
00189251
Volume
30
Issue
2
Year of publication
1994
Pages
556 - 577
Database
ISI
SICI code
0018-9251(1994)30:2<556:ERPTS(>2.0.ZU;2-A
Abstract
The topics of matrix spectral factorization (MSF) in conjunction with results from realization theory are applied here in simulating a stati onary multi-input/multi-output (MIMO) linear system from a specified p ower spectral density matrix. MSF provides the appropriate transfer fu nction matrix and Realization Theory specifies the corresponding param eters of a linear system having this transfer function. This approach can be used to correctly capture the cross correlations that exist in a multichannel vector random process (representing a particular radar target signature). Aspects of the solution to this problem are illustr ated using an original representative example problem with a closed-fo rm answer. Existing software programs for accomplishing MSF are identi fied and one has been successfully validated using the known closed-fo rm solution mentioned above. A streamlined realization algorithm (offe red here as the primary theoretical contribution) can be used along wi th the MSF computer program and can now be applied to actual radar dat a. Besides multichannel spectral estimation, several other important a pplications (some old, some new) of this same MSF solution methodology are summarized in Appendix B including extending applicability to the complex case (to handle radar polarization issues related to coherent phase processing), and finally in reexpressing second order statistic s of a multichannel autoregressive moving average (ARMA) process as th ose of a simpler but mathematically equivalent autoregressive (AR) pro cess of slightly higher dimensions (as another original application of the same major result offered here).