Adaptive detection of point-like targets embedded in Gaussian disturbance has represented an active field of research in the last decades. A variety of different solutions have been explored in the open literature. In particular, Kelly E.J. from the MIT Lincoln Laboratory, U.S. in his work “An adaptive detection algorithm” resorts to the generalized likelihood ratio test (GLRT). It aims to derive a constant false alarm rate (CFAR) test for detecting signals known up to a scaling factor. As well, Robey F. C. from the Washington University in St. Louis, U. S. with several other scientists in their paper “A CFAR adaptive matched filter detector” derive another CFAR test called adaptivematched filter (AMF), resorting to the so-called two-step GLRT-based design procedure.

All the above solutions assume a homogeneous environment. Unfortunately, the secondary data are often contaminated by power variations over range, clutter discretes and other outliers. In these situations, the sample covariance matrix of the secondary data yields significant performance degradations and the CFAR property is no longer ensured. A possible means to circumvent the lack of a sufficient number of homogeneous secondary data is to exploit the structural information of the noise spectral properties.

HAO Chengpeng, MA Xiaochuan and HOU Chaohuan from the State Key Laboratory of Acoustics, Institute of Acoustics, Chinese Academy of Sciences, Beijing, China and Orlando Danilo from Elettronica S.p.A., Roma, Italy considered the problem of detecting a point-like target in the presence of partially-homogeneous Gaussian disturbance with unknown but persymmetric structured covariance matrix.

In this work, two adaptive detectors have been devised relying on the Rao test and Wald test design criteria, respectively. Since there is no particular a priori reason to exploit the GLRT rather than the others, it is worth investigating the Rao and Wald tests for the same problem. In particular, resorting to the above alternatives to the GLRT it could yield that decision schemes more robust/selective than the GLRT in the presence of mismatched environments (usually present in real operating situations). Moreover, receivers require a smaller computational complexity than the GLRT.

The performance assessment, conducted by Monte Carlo simulation, shows that the proposed detectors exhibit the same performance of the P-GLRT for perfectly matching conditions. On the other hand, in the case of mismatched signals, the Wald test provides an enhanced robustness whereas the Rao test is the detector in a two-stage configuration to obtain an overall detector capable of ensuring a wide range of directivity. Finally, it is shown that both detectors are less time consuming than the existing persymmetric GLRT.

This research result was published on the recently issued IEEE SIGNAL PROCESSING LETTERS (VOL. 19, NO. 9, SEPTEMBER 2012, Pages 587-590).