Reiterative Robust Adaptive Thresholding for Nonhomogeneity Detection in Non-Gaussian Noise
- PDF / 617,810 Bytes
- 11 Pages / 600.05 x 792 pts Page_size
- 37 Downloads / 181 Views
Research Article Reiterative Robust Adaptive Thresholding for Nonhomogeneity Detection in Non-Gaussian Noise A. Younsi,1 A. M. Zoubir,2 and A. Ouldali1 1 Department 2 Signal
of Elctronics, Ecole Militaire Polytechnique, BP 17, 16112 Bordj El Bahri, Algiers, Algeria Processing Group, Darmstadt University of Technology, Merckstraße 25, 64283 Darmstadt, Germany
Correspondence should be addressed to A. Younsi, [email protected] Received 24 October 2007; Revised 4 April 2008; Accepted 23 June 2008 Recommended by Satya Dharanipragada A robust and data-dependent adaptive thresholding algorithm for nonhomogeneity detection in non-Gaussian interference is addressed. The algorithm is to be used as a preprocessing technique to select a set of homogeneous data from a bulk of nonhomogeneous compound-Gaussian secondary data employed for adaptive radar. An iterative version of the algorithm is also suggested in situations of multiple outliers in the secondary data. Performance analysis is conducted with simulated data as well as with real sea clutter data. Copyright © 2008 A. Younsi et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
1.
INTRODUCTION
The deleterious impact of nonhomogeneous secondary data in covariance estimation for adaptive radar systems has been widely reported [1–6]. Typically, the unknown interference covariance matrix is estimated from a set of identically and independently distributed (iid) target-free data, which is representative for the interference statistics in a cell under test (CUT). Frequently, the secondary data is subject to contamination by discrete scatterers or interfering targets (outliers). In both events, the training data becomes nonhomogeneous, leading to a nonrepresentative set for the interference in the CUT. Estimates of the covariance matrix from nonhomogeneous training data result in under-nulled clutter. Consequently, constant false alarm rate (CFAR) and detection performance are greatly degraded. This has led to the development of improved training data selection techniques, seeking to discard bins that are nonhomogeneous from the bulk of training data, and using the resulting outlier-free data for covariance matrix estimation. Recently, several algorithms for outlier removal have been proposed [7–9]. The works of [7, 9] addressed the use of the nonhomogeneity detector (NHD) based on the generalized inner product (GIP) measure involving Gaussian interference scenarios. However, in most practical situations, the Gaussian model is no longer valid. In particular, for high-resolution radars operating at low grazing angles,
or in a maritime environment, a satisfactory fit of the clutter amplitude probability density function (apdf) can be achieved through families of non-Gaussian distributions [10, 11]. For this non-Gaussian interference, the corresponding nonhomogeneity detection problem has received lim
Data Loading...
