Rank-deficient prewhitening with quotient SVD and ULV decompositions (Q1387236)

Prewhitening is a technique frequently used in signal processing when dealing with a signal degraded by colored noise. In this paper, based on the generalized singular value decomposition, a method to prowhiten a signal contaminated by an interfering noisy signal whose covariance matrix is rank deficient, is proposed. It is also demonstrated that a quotient-version of the ULV decomposition can be used to provide an efficient updatable implementation.