Using Joint Generalized Eigenvectors of a Set of Covariance Matrix Pencils for Deflationary Blind Source Extraction

摘要: In this paper, we develop a new deflationary blind source extraction (BSE) algorithm that extracts signals in sequential fashion via the joint generalized eigenvectors of set covariance matrix pencils. The concept eigenvector is defined. We prove these vectors can be made unique and identical to with properly selected To resolve open problem estimating eigenvectors, an approach based on deflation operation proportional property eigenvectors. Specifically, property, show estimation formulated as optimization involving quadratic cost function unit-rank constraint. An efficient iterative then developed by applying gradient search, shrinkage, deflation, symmetry-preserving vectorization techniques. This estimates conducts BSE sequentially. Its computational complexity convergence are analyzed. Simulations demonstrate outperforms many typical or separation algorithms. particular, more robust both heavy noise ill-conditioned mixing matrices.