Reduced-Dimension Single Data Set Detection Algorithms

The detection of a signal in coloured Gaussian interference is relevant in many fields such as radar, communications and biomedical technology. In practice, the interference covariance matrix is estimated from training data, which must be target free and statistically homogeneous with respect to the test data. These conditions are often not satisfied, which degrades the detection performance. Single data set algorithms have been proposed to circumvent this problem. In this paper, the issues associated with applying reduced-dimension techniques to them are studied and these reduced-dimension detectors´ probabilities of false alarm and detection are derived. They have the highly desirable CFAR property and theoretical results are verified by simulations, which also show that they are comparable to traditional detectors