Adaptive reduced-rank interference suppression with adaptive rank selection

An adaptive reduced-rank filter for interference suppression in direct-sequence (DS)-code-division multiple access (CDMA) is presented. The projection subspace is constructed by successively multiplying an initial basis vector (estimated steering vector or matched filter) by the sample covariance matrix. It is shown that this technique is equivalent to adaptive reduced-rank filtering based on the multi-stage Wiener filter of Goldstein and Reed (see IEEE Trans. Inform. Theory, vol.44, no.7, 1998). This method has the important property that the rank needed to achieve a target performance does not scale with the system size (number of users and processing gain). Methods for rank selection are presented along with numerical examples illustrating performance.