Fourth-order tensor method for blind spatial signature estimation

In this paper, we consider a wireless communication scenario where M sources simultaneously transmit towards a base station equipped with an array of K sensors. A new method is proposed to solve the spatial signature estimation problem without resorting to training sequences and without knowledge of sources' covariance structure. By assuming that the sources' amplitudes vary between successive time blocks, a fourth-order tensor decomposition of the multimode spatio-temporal data covariance is proposed, from which an iterative algorithm is formulated to estimate sources' spatial signatures. A distinguishing feature of the proposed tensor method is its efficiency in treating the case where the sources' covariance matrix is non-diagonal and unknown, which generally happens when working with sample data covariances computed from a reduced number of snapshots.