Regularized MMSE multiuser detection using covariance matrix tapering

The linear minimum mean-squared error (MMSE) detector for direct-sequence code-division multiple-access (DS-CDMA) systems relies on the inverse of the covariance matrix of the received signal. In multiuser environments, the matrix ill-conditioning may produce large performance degradation. In order to cope with this effect, we propose a modified MMSE detector based on the covariance matrix tapering (CMT). This regularization technique modifies the estimated covariance matrix, thereby reducing its eigenvalue spread. Two different tapering matrices are introduced. The performance comparison with other regularization techniques is carried out by simulations, which show the effectiveness of the proposed technique.