Noise and Gaussian Clutter Background Descending Dimensional Subspace Signal Detector

This paper deals with the detection of subspace signals in the context of noise and Gaussian clutter. Classical detectors usually treat clutter and noise as a whole, assuming a set of independent identically distributed (IID) training data without signal components to estimate the covariance matrix of clutter and noise. However, in practical applications, it is usually difficult to obtain sufficient training data, hence, we propose a generalized likelihood ratio test (GLRT) detector based on a one-step design strategy in the context of noise and Gaussian clutter, assuming that the Gaussian clutter covariance matrix has a presymmetric structure, and using a dimensionality reduction method of subspace transformation. Theoretical analysis shows that the detector has a constant false alarm rate property for the covariance matrix of clutter plus noise. Simulation analysis shows that the detector can work properly in the absence of training data, and the detection performance is better than existing comparison detectors.