Regularization enhanced fast convergence for POR-based multiuser detection

Abstract

Recently, a power of R (POR) technique that is originated from a minimum variance method is proposed for multiuser detection in code division multiple access (CDMA) systems. It raises the power of an inversed data covariance matrix in the MV cost function to a positive integer m to approximate the noise subspace component in order to bridge a gap between the MV method and a widely studied subspace method. The POR method asymptotically converges to the subspace one and also shows more robust performance in practical conditions with finite noisy data record. However, the convergence rate depends on m and signal to noise ratio (SNR). In particular, a wide gap is observed in the low SNR region for a small or moderate m. This work proposes a regularized POR (termed RPOR) technique by introducing a regularization parameter to the POR cost function for fast convergence and better detection performance for low SNR CDMA systems. The parameter aims to compensate the noise effect and concurrently enhance the interference mitigation capability of the receiver.