An efficient preprocessor for nonlinear adaptive algorithms

In many signal processing applications it is common to preprocess the input data before they are fed into various adaptive algorithms used to separate original signals from a noisy mixture. The preprocessing transformations not only reduce the dimension of the data matrix but also improve the convergence rate of subsequent nonlinear algorithms. In this paper we propose a new preprocessing transformation based on the ULV decomposition (ULVD). We also show how to efficiently modify the transformation in order to track nonstationary signals.