Novel Application of Projection Approximation Subspace Tracking Algorithm for Whitening in Wireless Communications
Ronnie Landqvist, Abbas Mohammed · 2005
In this paper we propose two new algorithms that can be used to pre-white signals as a pre-processing step before Independent Component Analysis (ICA). The first algorithm is denoted PASTD for Whitening (PASTD-W) and is based on the Projection Approximation Subspace Tracking with Deflation (PASTD). The second algorithm is a straight forward approach based on exact eigendecomposition, denoted as the Exact Eigendecomposition (EE) algorithm. Computer simulations are performed in order to compare PASTD-W, EE and analytical exact whitening. In these simulations, Binary Phase Shift Keying (BPSK) modulated data are transmitted by two antennas on a channel that acts as linear mixer. The receiver consists of a pre-whitener (PASTD-W, EE or analytical exact whitening) followed by Nonlinear PASTD (NPASTD) for ICA. The simulations show that EE converges faster than PASTD when applied to whitening. Both algorithms show good performance when used together with N-PASTD.