An Improved Matrix Information Geometric Detector with Parameter Selection

In this paper, we propose an improved matrix information geometric detector with a strategy parameter selection in nonhomogeneous noise. In particular, the correlation or power of sample data is captured by a Hermitian positive-definite (HPD) matrix. Then, the problem of signal detection is treated as distinguishing the matrices of noise and target signal on the matrix manifold. A strategy to choose the initial input matrix and step-size of the iterative matrix equation is presented to accelerate the convergence for the computation of Riemannian mean. Unlike the classical detectors that is designed by resorting to the known (assumed) statistical characteristics of noise, our proposed detector only considers the underlying geometry of Riemannian manifold of HPD matrices. Numerical results show that the proposed detector achieves a detection performance improvement over the conventional detector in nonhomogeneous noise.