DocumentCode :
1929137
Title :
Knowledge-aided covariance matrix estimation: A MAXDET approach
Author :
Landi, L. ; De Maio, A. ; De Nicola, S. ; Farina, A.
Author_Institution :
DIET, Univ. degli Studi di Napoli Federico II, Naples
fYear :
2008
fDate :
26-30 May 2008
Firstpage :
1
Lastpage :
6
Abstract :
In this paper, we consider the problem of knowledge-aided covariance matrix estimation and its application to adaptive radar detection. We assume that an a-priori (knowledge-based) estimate of the disturbance covariance M, derived from a physical scattering model of the terrain and/or of the environment, is available. Hence, starting from a set of secondary data, we evaluate the maximum likelihood (ML) estimate of M assuming that it lies in a suitable neighborhood of the a-priori estimate. We formulate this ML estimation in terms of a convex optimization problem which falls within the class of MAXDET problems. Both the cases of unstructured and structured disturbance covariance are considered. At the analysis stage, we assess the performance of the new knowledge-aided covariance estimators in terms of detection probability achievable by a class of adaptive detectors. The results highlight that, if the a-priori knowledge is reliable, satisfactory levels of performance can be achieved with considerably less training data than those exploited by conventional algorithms.
Keywords :
adaptive radar; convex programming; covariance matrices; maximum likelihood estimation; radar detection; radar signal processing; MAXDET approach; a-priori estimation; adaptive radar detection; convex optimization problem; detection probability; knowledge-aided covariance matrix estimation; maximum likelihood estimation; physical scattering model; radar signal processing; Clutter; Covariance matrix; Detectors; Jamming; Maximum likelihood estimation; Performance analysis; Radar detection; Radar scattering; Statistics; Training data; Covariance Matrix Estimation; Knowledge-Aided Radar Signal Processing; MAXDET;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Radar Conference, 2008. RADAR '08. IEEE
Conference_Location :
Rome
ISSN :
1097-5659
Print_ISBN :
978-1-4244-1538-0
Electronic_ISBN :
1097-5659
Type :
conf
DOI :
10.1109/RADAR.2008.4720823
Filename :
4720823
Link To Document :
بازگشت